) and B;r•
there are but 6 independent jumps, as expected. Conditions of this kind, being
pure identities, merely rearrange the variables. Only when they are used in
connection with some further hypothesis, such that some particular quantity
is continuous, may fruit be gotten from them.
In the important special case when both 'V and 'V k are continuous1, so that
A = B = 0, ( 176.8) reduces to the form 2 '
['V,k",] = [oko",'V] = Cnkn",, } (176.10) C = [nPnq 'V,pq] = [nPnq op oq'V] = ['V,j,P].
1 If we suppose merely that ~ ,k is continuous and [~];r= O, we obtain
[~,kml = Cnknm, C = [nknm~,km],
but not all the other forms included in ( 1 76.10) remain necessarily valid.
2 HADAMARD (1901, 8, § 1], (1903, 11, ~ 74].
Handbuch der Physik, Bd. III/1. 32
498 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 177.
More generally1, if \V and its derivatives of orders 1, 2, ... , p -1 are continuous,
we have
[w,,.,,. •... ,."J = [8,., 8,., ... 8,." w],
= [n"'' n1111 ••• n"'P 81111 8m, ..• 81111> \V] nk, nk, ... nkp>
= [n""n"" ... n"'P \V, 11111111 ••• mp] nk, nk, ... nkp, (176.11)
{
[\V•"'•,..,,'"'',m, .. .'"'~'r',..."r,] n,., n,., ... n,." if p is even
= [nq \V•"'• ·"'• •"'IP-•>I•q] n,. n,. ... n,. if p is odd. ,m1 •"'•··· 1 t p
That is, the jumps in all the p! derivatives of p-th order are determined uniquely
by the jump in the completely normal p-th derivative and by the unit normal n.
By using an argumentsuch asthat leading to (175.8), we see that (176.11) holds
in a metric space of any dimension; a corresponding generaliiation of (175.6)
holds in any kind of space, irre~pective of what geometrical structute it may
have.
A second iteration would enable us to derive a general resolution of [IV,,...,p]
in the Euclidean three-dimensional case, but the formal complexities encountered
in deriving ( 176.8) render the details of such an analysis forbidding and the
result too complicated to be useful.
II. The motion of surfaces.
177. The speed of displacement and the normal velocity of a moving surface.
Consider a family of surfaces given by
:xJ = :xJ (V, t) , (177.1)
where V stands for a pair of surface parameters VLI identifying what we shall
call a surface point. V, in general, is not to be confused with a material particle
of any motion that may be occurring; indeed, the considerations in this section
should be regarded as independent of the motion of substances, although as an
aid to visualization it is often convenient to picture the moving surface as consisting of identifiable particles. The representation (177.1) gives the places :xJ
occupied by the surface point V as the time t progresses; thus it describes the
motion of a surface. The velocity of the surface point V is defined by
az I
U = Tt V=const' (177.2)
If we eliminate the parameters V, we may write (177.1) in the form
f(:IJ,t) =0. (177.3)
Conversely, however, from a spatial representation (177.3) it is not possible
to calculate a unique form {177.1). This is easy to understand: Given a moving
surface, there are infinitely many ways of identifying the points on its successive
configurations in such a way that all those configurations are swept out smoothly
by the surface points constituting any one of them.
Supposing, now, that we have any one parametrization (177.1), by differentiating (177.3) with respect tot we get
at ae+u·grad/=0. (177.4)
1 CouLON [1902, 3, § 46], HADAMARD [1903, 11, '1174]. Contrary to the implication of
THOMAS [1957. 15, § 1], the analysis of HADAMARD is not restricted to any special choice of
CO-ordinates.
Sect. 177. The speed of displacement and the normal velocity of a moving surface. 499
Writing n for the unit normal to the surface, by (177.4) we have
Of
grad t ot
Un=U·n=U· !gradfl =- V f,k ,,k ' (177-5)
since the right member is determined by the spatial equation (177-3) alone, it
is independent of our choice of the parametrization (177.1). That is, alt possible
velocities u of the moving surface have the same normal component un, which is
called the speed of displacement of the surface1. Cf. Sect. 74.
For some purposes it is convenient to make the particular choice of surface
points implied by requiring u to be normal to the surface;
( 177.6)
This velocity will be called the normal velocity of the surface. The identification
of surface points may be visualized by erecting normal vectors of magnitude un dt
from each: point on the configuration of the surface at some one time t; the
termini of these vectors then sweep out the configuration at time t +dt. When
un = f(t), the surfaces so generated are parallel surfaces.
Suppose now that we have any parametrization
X =X (v, t) ( 177.7)
which is consistent with (177.3). Then both of these equations may be used simultaneously, so that (177.3) becomes f(x(v, t), t) =0. Differentiation with respect
to t yields of oxk .
Te+ l,k8t = o. (177.8)
From (177.5) it follows that
{177.9)
Conversely, (177.5) follows from (177.9), so that these two equations furnish
equivalent definitions of the speed of displacement according as the representation (177.3) or (177.7) for the surface is preferred.
The parameter v in (177.7) may, but need not, be identified with what was
called a surface point V above. In any case, given a particular parametrization
(177.7), an observer moving with the velocity (177.6), which we have called the
normal velocity of the surface, will encounter points on the surface (177.7) having
surface co-ordinates v which vary in time. Their rates of change ur, which
we shall call the tangential velocity of the parametrization, may be calculated 2•
Such a velocity must satisfy
oxk r . .k - k 81 + u X";"r- unn . (177.10)
Taking the scalar product first by nk and then by gkmx7'LI yields (177.9) and 3
(177.11)
l This quantity was introduced by STOKES [1848, 4, p. 353], who called it "the speed of
propagation"; cf. also KELVIN [1848, .5]; for general surfaces it first appears, unnamed,
in the work of CHRISTOFFEL [1877, 2, § 1]; also HUGONIOT [1885, J, p. 1120] first called it
"vitesse de propagation" but immediately thereafter [1885, 4, p. 1231] distinguished "deux
vitesses de Propagation", the other being that we consider in Sect. 183. The term "vitesse de
deplacement de l'onde" was introduced by HADAMARD [1901, 8, § 1], [1903, 11, ~ 100]. 2 THOMAS [1957, 1.5, Eqs. (52) to (54)].
3 Cf. ( 177.2). Eq. ( 177 .10) thus furnishes a resolution of the particular choice of surface
velocity specified by ( 1 77 .6).
500 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 178.
Conversely, if we choose to define un and ur by (177.9) and (177.11), we have
by (App. 21.4h
oxk r k - ()zk rLl oxP m _ _k )
-----;)~ + u X; r - -----;)~ - a gp m ----a"t X; Ll x;-r
(177.12) - oxk oxP ( mk m k) - -----;)~- gpm ----a"t g - n n ,
whence (177.10) follows. A necessary and sufficient condition that the parameter v correspond to a surface point V for the normal motion (177.6) is Ur=O.
178. Differential description of a moving surface1• Supposing that a, b, un,
and ur are given as functions of v and t, we ask if there exists a representation
(177.7) suchthat these quantities belong to it. In other words, in order to define
a moving surface does it suffice to assign as functions of t and the parameters v
its first and second forms, its speed of displacement, and the tangential velocity
of the parametrization? The answer, in general, is negative. In addition to the
Eqs. (App. 21.8) of MAINARDI-CODAZZI and GAUSS, it is necessary that other conditions of compatibility be satisfied.
First we differentiate (177.11), obtaining
( ox~Ll m + oxk m ) Ur;Ll =- gkm Be X;r -----;)~ X;rLJ · (178.1)
By (App.19.7) and (App. 21.6h it follows that
oarLl b -o-t-+2u(r;Ll>=- 2un rLJ· (178.2)
From the definition of the total covariant der~vative it is easy to verify that
( oxk;r) = ox~rLl ~ ~{A} ot ;Ll ot + ;A ot Llr · (178.3)
By (App.19.6h we then have
(178.4)
We now differentiate (177.9) and by use of (178.4), (App.21.6) 2 , (App.21.7h,
(App.21.6)1 , (177.11), (177.9), and (App.19.7) obtain
oxk OX~Ll) OX~rLl Un;rLJ = nk;I'Ll-----;)1 + 2nk;(r-8-t-+ nk-8-t -,
OXk (bA m +bAb m) = -gkm -----;se r;LI X; A r ALl n -
A 0 X~ Ll) 0 ( k b ) - 2gkmx~Abcr- ~+nkßt n rLI,
(178.5)
- bA bA b - bA oaLI)A obrLI -UA r;LI-un r ALl (r-o-t-+-8-t-.
From (App. 21.8)1 and (178.2) it follows that
obrLI Ab bA bAb - 0-t-+u FLI;A+ (I'ULI);A=Un;rLl-Un r ALl· (178.6)
1 The results ( 178.2) and ( 178.6), though not the proofs given here, were disclosed to us
by J.L. ERICKSEN.
Sect. 179. The displacement derivative. 501
Equations ( 178.2) and ( 178.6) are conditions of compatibility to be satisfied
by a, b, un, and ur. ERICKSEN has shown that conversely, if the quantities
a, b, un, and ur satisfy (App. 21.8), {178.2), and (178.6), then they are derivable from
a relation of the form {177.7) with an assigned spatial metric g; that is, the conditions of compatibility here derived are also sufficient for the existence of a
moving surface. Therefore any other condition satisfied by a, b, un, and ur will
be a consequence of the relations already derived.
179. The displacement derivative. Given a function F(v, t) defined upon the
moving surface, its rate of change tJFjtJt as apparent to an observer moving with
the normal velocity (177.6) of the surface is1
!_!___=~+ ro F IJt at u r ' (179.1)
where ur is the tangential velocity of the parametrization, given by (177.11).
Suppose that G(x, t) be a function such that on the surface j we have G(x (v, t) t) =
F(v, t). Then aF ac axk -81 = 81 +-8TokG, 8rF=x7rokG. {179.2)
Hence by (177.10) we have
( 179-3)
If Fand G are tensors, tJFjtJt and tJGjbt as defined by (179.1) and (179-3) 3
generally fail to be tensors. For a double tensor WL:~~·.:·.~(x, v, t), we define
the displacement derivative (Jd 'l! jbt as that double tensor under the group of
transformations x*=x*(x), v*=v*(v, t) which reduces to bWjbt when the spatial co-ordinates arereetangular Cartesian and the tangential velocity ur vanishes.
To calculate (Jd lj! JM, we first introduce the "Lie derivative" f: lj! when any spatial
indices of 'l! or dependence of 'l! upon x is ignored, viz. "
f: lj!k ... mr ... Ll _ u<~> 0 lj!k ... mr ... Ll _ lj!k ... m + .. . p ... q
lj!k ... m r ... Ll lj!k ... m
+ .. . p ... q <~> ... r ,A ·
( 179.4)
Then we have
{179.5)
where both x and v are held constant when olj!jot is calculated. Forthis formula
to be meaningful, it is not necessary to use any equation x = x (v, t) for the
surface whose normal and tangential velocities are un n and ur. To verify its
correctness, however, let us eliminate x; by {177.10) we have
~
bd '*' - aw ( oxk r k ) - 81 +w,k at+u x;r +;w,
= !'*'_ + oxk oklj! + ~_x~ [wq ... { p} + ... ] + t:w +urx~r'l! at at at ... k q .. , , k' {179.6)
= -~j + fW + ~ [wq ... { P} + ... ]. ot V=COnst U Ot ... k q . ------
1 This definition, stated in words by HAYES [1957, 7, p. 595], seems to give the sense
intended also by THOMAS [1957, 15, § 4].
502 C. TRUESDELL and Ro TouPrN: The Classical Field Theorieso Secto 179°
where ~ IV is obtained by replacing " ~" by "0 ~" in ( 179.4) 2 o When ur= 0 and .. ' '
the spatial co-ordinates arereetangular Cartesian, by (179.6)3 and (17901) we have
d~: = ~~~v=const = ~~ · (179.7)
Since the right-hand side of (179o5) is a double tensor under transformations
~· =~* (~). v* =v* (v, t), it fumishes the required expression for the displacementderivative in all co-ordinates systems.
Fora spatial tensor IV~:::;'(~. t), (17905) reduces to
Öd't' i.l't' " -Ü- = -fJt- + IV,k Un n , (179°8)
and it is this form that is most useful.
From (179.8) we have at once
ddgkm = 00
t5t ' (179°9)
hence raising and lowering of spatial indices commutes with tJdjllto Not so, however
with surface indices, for the conditions of compatibility (17802) and (17806)
assume the forms
(179o10)
By differentiating the relation ar.:~ a.:!A =ll'J. we see that (179o10h is equivalent to
ddar.d - + 2 br.:! dt - Un • (179°11)
Also Öda - Te=-2UnaK, (179.12)
whence it follows that tJdajtJt =0 is a necessary and sufficient condition that a
moving surface be and remain a minimal surface.
From (179.11) and (179o10) 2 we have
~ ddb~ --Un ;r ;.1 + Un br.:! b A.d> l
Ödbr.d - ;r.d+ 3 v.:~ b.:! - 15-t-- Un Un A•
(179°13)
From these results and (Appo 21.10) it is easy to show that
~ ödR (K-2 K) + or =Un -2 Un;r' ,
~
ddK- KK-+ (K- r.:~ br.:!) -Un a - Un;r.:J, (179.14)
ddb - - ~ =-UnaK K + a(K ar.:~_ bF.:!) Un;r.d·
ddnk We now calculate 1 ~· From the relations (Appo 19.6)1 and (App.21.3)1 it
follows that
(179.15)
1 The result is stated by HAYES [1957, 7, Eqo (22)] without proof; a proof different from
ours is given by THOMAS [1957, 15, Eqo (60)]o
Sect. 180. Kinematical condition of compatibility. 503
By differentiating (177.9) we obtain
(179.16)
where we have used (177.10), (App.19.6) 1 , and (179.15) 3 . Since (177.9) presumes
that X =X (v, t), the time derivative ojot is taken Oll the understanding that X
is eliminated. Hence
Ar k A q p bE r A k ( kq k q) 0 n,l ) a X;AUn;r=U gqpX;AX;E ra X;A- g -n n 81-,
~ k ( 179.17) - A bA k un -U AX;A-7)t• .
where we have used (App.21.6b (App.21.4) 1 , (App.19.7), and (179.15) 1 . By
(A pp. 21.6h it follows that
( 179.18)
where, as stated above, x has been eliminated before onkjot is calculated. By
use of (179.6) 2 we thus obtain the desired formula:
(179.19)
For an alternative derivation, we may suppose that ur= 0 and the spatial Coordinates are reetangular Cartesian; most of the terms in the above calculations
are then absent, and we quickly obtain a formula recognizable as the appropriate
special case of ( 179.19), which is a tensorial equation.
From (179.19) and (App.19.7) we have
( 179.20)
a result that might have been expected directly from the definitions, since it
asserts that the length of the projection of the displacement derivative of the
normal onto a given direction on the surface is just the negative of the gradient
of the speed of displacement in that direction. In particular, a necessary and
sufficient condition for parallel propagation is un = f (t), as was already remarked
in Sect. 177.
III. Kinematics of singular surfaces.
180. Kinematical condition of compatibility. We now consider a moving
surface d (t) which divides a varying region Bi+ (t) from another, :?~- (t). The
moving surface is assumed to satisfy the conditions stated in Sect. 177 and at
each instant t to be a singular surface with respect to a quantity \V, as defined
in Sect. 173; the conditions laid down for \V are now supposed to hold for each t.
Assuming also that the limiting values \V+ and \V- are continuously differentiable
functions oft in Bi+ and &l-, respectively, we derive a condition that the discontinuity in \V persists in time rather than appearing and disappearing at some
particular instant.
In a general space this temporal persistency is expressed by superficial conditions analogaus to those discussed in Sect. 175 but applied when one more
dimension, that of t, is added both to the surface and to the space in which it
504 C. TRUESDELL and R. TOUPIN: The Classical Field Theories. Sect. 180.
lies, i.e., we need only regard f(~, t) =0 as a single surface in the ~-t-space. In
this degree of generality, nothing new results.
In a metric space, however, the existence of a definite speed of displacement
un for the moving surface makes possible results of a more concrete kind. The
essential step, again, is fumished by HADAMARD's lemma (174.1), butthistime
we apply it to a particular tangential path on the n-dimensional surface f (~, t) = 0
in the n + 1-dimensional ~-t-space, namely, the path tangent to the vector un n, 1.
The derivative 1 on the + side of the surface is the quantity öw+Jöt as defined
by (179.1). Thus
(180.1)
where, as in previous formulae, okw+=(okw)+. Writing a similar equation for
the other side of the surface and subtracting the result from ( 180.1), we obtain
the kinematical condition of compatibility 2 :
[~~] = -un[nkokW] + :t [W]. (180.2)
The jumps occurring on the right-hand side are those occurring also in the geometrical conditions (175.11). Thus the fumps of the derivatives 8kW and 8Wf8t
across a persistent singular surface are determined by the quantities un, [ nk 8k W],
and [W].
A condition equivalent to (180.2) but expressed in terms of tensors is easy
to obtain by using the displacement derivative (179.5). Considering a spatial
tensor field W k.. · m we ha ve p ... q '
[o'ii] _ k t:5d 8t --un[n W,k] +Tt[W]. (180.3)
This is so because ( 1) it is a tensorial equation and (2) when the spatial co-ordinates
are reetangular Cartesian, it reduces to (180.2).
In the important special case when W is continuous, (180.2) reduces to the
form
( 180.4)
Thus, in particular, across a stationary surface that is singular with respect to
8k W but not with respect to W, the time derivative 8W Jot is · continuous, as is
evident also directly from the definitions.
We may write (175.8) and (180.4) as the system
[W,k] = Bnk, [~~] =- unB. (180.5)
1 There are n independent paths at any one point of d (I). but we use only a particular
one. If all n are employed, we may derive the geometrical and kinematical conditions of
compatibility simultaneously, as is done in a special case by HADAMARD [1903, 11, "if97]
and more generally by CouLON [1902, 3, § 46]. We prefer separate treatment of the two sets
of conditions so as to separate the underlying ideas. In many cases useful in continuum
mechanics, the geometrical conditions are satisfied when the kinematical is not; e.g., at the
instant a portion of material splits in two, or two parts are joined together. 2 The essential content of this condition seems to be contained in the "phoronomic
conditions" of CHRISTOFFEL [1877, 2, § 7], but these are not easy to use, and certainly the
general concept of compatibility is due to HUGONIOT. First [1885, 3, p. 1119] he used "propagation" to mean compatibility, but soon thereafter [1887, 1, §§ 3. 5] he introduced the terms
"compatibilite" and "conditions de compatibilite". HADAMARD [1903, 11, "if97] and LovE
[1904, 4, § 7] obtained the system (180.5); HADAMARD's term is "conditions de compatibilite
cinematique". The full condition (180.2) is implied by the results of CouLON [1902, 3, § 46];
cf. also KorcHINE [1926, 3, § 1]. Our presentation follows THOMAS [1957. 15, § 4].
Sect. 181. Iterated kinematical conditions of compatibility. 505
Of all the forms of the conditions of compatibility, it is these, which presume W
itself to be continuous, that are most often used. If we square (180.5) 2 and use
(175.13)1 , we obtain1 ~] 2 _ 2 2 _ 2 ,k
[ at - un B - un [W,,.][W ], (180.6)
whereby the magnitude of the speed of displacement is shown to be the quotient
of the magnitude of the jump in 8Wj8t by the magnitude of the jump in W k·
This last result is expressed in terms and notations appropriate to the case wh~n
W is a scalar, but generalization is easy.
Since HuGONIOT's time 2 it has been stated that a singular surface upon which
the kinematical condition of compatibility does not hold will instantly split into
two or more singular surfaces or will become singular with respect to a different
quantity, such as a derivative of W. To substantiate a statement of this kind,
the theory of some particular material is needed. In the generality maintained
here, all that can be said is that the singular surface will not persist.
181. Iterated kinematical conditions of compatibility. Amplifying the notations (176.1), set ·
=[~;], B1=[n,.a;;k]. (181.1)
We may then write (180.3) in the form
A l
=- B 6dA ( ) Un +~· 181.2
By replacing W by 8Wj8t in (176.2h and (181.2) we obtain
[ ----a"t aw, k] -
_ BI n,. + g,.ma 4 X;4 ;r• r m A 1 l
[ß2'l/] 1 ddA (181.3) ---ai2 = - Un B + (ft .
By (181.2), the quantity A 1 is already expressed in terms of A and B. We
now obtain a like expression for B1 • To this end, we differentiate (176.1) 3 and
use ( 176.2)1 , obtaining
ddB _ 4 r m A '6dnk + k 6d['ll,k] (181.4) 6t - g,.ma X;4 ;r 6t n 6t •
Now by another application of the argument leading to (180.1) we can show that
:t (o,.w+)=(:t akwf+unnmomakw+, (181.5)
since the various derivatives are assumed continuous in f!Jt+; consistently with
our practice, we have put 8m8kW+=(8m8kW)+. Hence
6d - [aw,k] m[ Tt[W,k]- 8t- +unn W,mkl· (181.6)
Substitution of this result into (181.4) yields a result which when simplified by
use of (176.1) 6, (181.3) 1 and (App.21.3) 2 becomes
BI_ C 6dB 6dnk 4r m A ) --Un +~-gkm~a X;4 ;F•
=-unC+~~ +un;rA;r•
where the latter form follows by use of (179.19).
1 DUHEM [1900, 3].
2 [1887, 1, § 14]. Cf. HADAMARD [1903, 11, 'if 108].
(181.7)
506 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 182.
The identities (181.3), with A' and B' replaced by right-hand sides of (181.2)
and (181.7), are THOMAs's iterated kinematical conditions of compatibility 1•
In the case when \V is continuous, we have A = 0, and the conditions ( 181.3)
reduce to the simpler forms
( 181.8)
When not only \V but also o\Vfot and ok\V are
still simpler, for (181.8) and (176.10) yield 2
continuous, the results are
where
[okom\V] =[\V,km] =Cnknm, l [a~kt\V] = [a;;k] = -unCnk,
[~~~] = u~C,
(181.9)
(181.10)
More generally3, if \V and all its derivatives of orders 1, 2, ... , p- 1 are
continuous, we have
[ok, ok, ... ok, otP-s 0P-s \V] = [ otP-s 0P-s \V,k,k, ... k,] l
(181.11) = (- un)P-s [nm, nm, ... nmp om, o~, ... omp \V] nk, nk, ... nk,,
= (- un)P-s [nm, nm• ... nmp \V ,m,m, ... mp] nk, nk, ... nk,,
the result being valid in any metric space. In particular, choosing s =0 we have
(181.12)
interpretation of which yields the Hugoniot-Duhem theorem4 : The speed of
displacement of a singular surface across which \V and its derivatives of orders
1, 2, ... , p -1 are continuous but at least one p-th derivative of \V is discontinuous
is determined up to sign by the ratio of the jump of (}P \V/ o tP to that of the fully normal
p-th derivative, dPWfdnP.
IV. Singular surfaces associated with a motion.
182. Material and spatial representations of a surface. So far in this chapter
our considerations have been independent of the motion of any material medium.
We now suppose that a medium consisting of particles Xis in motion through
the space of places x according to (66.1). For the time being, we shall assume
1 [1957, 15, § 6]. THOMAS obtains also an alternative form for orA'; see his Eq. (51) as
corrected. As regards the history of these conditions, remarks similar to those at the beginning
of Sect. 176 may be made. 2 HUGONIOT [1885, 4, p. 1231], HADAMARD [1901, 8, § 2]. 3 DuHEM [1901, 6], CouLoN [1902, 3, § 46], HADAMARD [1903, 11, '1!97].
'Given in the special cases P=1 and P=2 by HUGONIOT [1885, 3, p.1120] [1885, 4,
p. 1231], in general by DuHEM [1900, 3] [1901, 7, Part II, Chap. li, § 2], but expressed by
means of Laplacians (cf. (176.11) and (181.10)) rather than normal derivatives.
Sect. 182. Material and spatial representations of a surface. 507
that the functions occurring in (66.1) are single-valued and continuous; modifications appropriate to motions suffering discontinuities will be given in Sect. 185.
We consider a surface d(t), given by a representation of the form (177.3). and
we set F(X, t) = f(x(X, t), t), so that f(x, t) = F(X(x, t), t), (182.1)
identically in x, X, and t. Alternative representations of the moving surface
are thus 1
f(x,t)=O, F(X, t) = 0. (182.2)
In the latter representation, which we denote by Y (t), we may conceive the
particles as stationary and the surface Y (t) moving amongst them, being occupied by a different set of particles at each timet. The two representatives (182.2)
are the duals of one another in the sense of Sect. 14. The analysis given earlier
in this chapter did not presuppose any particular choice of Co-ordinates, so long
as they be independent of time, and is equally applicable to both representations.
It is easier to visualize in terms of the spatial variables x, t, and from now on
we agree to regard all the foregoing equations as so expressed; where we wish
to employ a material counterpart, we shall invoke the principle of duality.
Thus in the special case when ( 182.2)1 reduces to the form
f(x) = 0, (182.3)
weshall say that the surface d is stationary; when (182.2) 2 reduces to the form
F(X) = 0, (182.4)
that Y is material 2 [cf. (73.4)]. In the former case, the surface consists always
of the same places; in the latter, of the same particles.
Although (182.2)1 and (182.2h are but different means of representing the
same phenomenon, the two surfaces so defined are, in general, entirely different
from one another geometrically. The surface f(x, t) =0 is a surface in the space
of places, while the surface F(X, t) = 0 is the locus, in the space of particles, of
the initial positions of the particles X that are situate upon the surface f (x, t) = 0
at time t. Such connections as there are must be established by use of the transformations (182.1). For example, from the assumption that f(x, t) =0 has a
continuous normal it follows that F(X, t) =0 also has a continuous normaP.
We assume, in fact, that (182.2) are sufficiently smooth astopermit any number
of differentiations and functional inversions. The theory we construct is local.
Some aspects of the foregoing theory, whil,e not losing their validity, lose
their intuitive appeal when applied to the material variables. The shape of
Y (t), including its first and second differential forms, and its unit normal, have
no immediate interpretation, for they do not correspond to any geometrical
properties that an observer of a singular surface in space would perceive.
The material representation, rather, is of the nature of a diagram for the
moving surface. It is only one of many such diagrams, for by choice of the initial
instant, or of the co-ordinates or parameters X corresponding to given initial
positions, the particular functional form that results from ( 182.1 )1 will differ 4•
As an example of the above remarks we consider the unit normals n and N to f = 0
and F = 0, given by
1 HUGONIOT [1885, 4, p. 1231].
2 French: stationnaire.
(182.5)
3 This is proved under weaker assumptions by LICHTENSTEIN [1929, 4, Chap. 6, § 1]. 4 The effect of such changes is discussed somewhat by HADAMARD [1903, 11, ~~ 79
to 84].
508 C. TRUESDELL and R. TouPrN: The Classical Field Theories. Sect. 183.
where, as usual, f•m=gmqf,q and p,ß=gßYF,y. The former has an immediate geometric
significance as a unit vector normal to the surface d (t) in the space of places x. The latter,
dual to the former, applies only to one of many possible diagrams for the moving surface
in various possible spaces of particles X and has no immediate interpretation.
Since F,a. = f,k x~ rx.• by the dual of (17.3) we have 1
F,rx. p,a. = I xfXI 2 gßY. t ekrp eß6e f,k x~, Xfp. t emsq eycr,l,m xfs X?q· (182.6)
a result which in a common frame assumes the form
VF.a.F,a. =I zfZI ( 8(/, Y,Z) )2 + (8(X,f,Z))2 + (8(X, Y,"7))2 a (x, y, z) a (x, y, z) a (x, y, z) .
For the unit normals themselves we have the relations
( 182.7)
F a.X'\ VFpF.ß n = · · = Na.xrx. · • (182.8) k Vt,mf'm ;k Vt,mf'm'
where VF.pF,ß and X~k are thought of as expressed in terms of spatial gradients by means
of ( 182.6) and ( 17 .3), respectively.
It is a natural requirement that the moving surface d (t) shall have a continuous and nonvanishing gradient vector f,k· Such a requirement if put upon the gradient vector F,rx. of .'7'
has no immediate appeal. From (182.7), we see that F,rx. if continuous can vanish at a point
if and only if either the radical or the Jacobian on the right-hand side vanishes. For the
radical to vanish in a neighborhood, it is necessary and sufficient that f be functionally
independent of X, Y, Z: that is, by (182.1), that F=F(t), and such an equation cannot
represent a surface. Thus F rx. vanishes only with the Jacobian I zfZI. By the Axiom of Continuity in Sect. 65, it follows that for a material diagram F(X, t) = 0 obtained from a surface
f(z, t) = 0 in a continuous motion, Fa. does not vanish except possibly at isolated points
or lines. '
183. Speeds of propagation. Waves. The dual of the speed of displacement,
defined by (177.5), is the speed of propagation 2 UN:
aF
at UN== ---==·
VF.a.F•"
( 183 .1)
This speed is a measure of the rate at which the moving surface 9'(t) traverses
the material. In particular, a necessary and sufficient condition that F = 0 be
a material surface in an interval of time is that UN = 0 throughout the interval.
A surface that is singular with respect to some quantity and that has a nonzero speed of propagation is said to be a propagating singular surface or wave 3 •
Now it is evident that the value of UN for a given surface f(x, t) =0 in a
given motion, unless UN = 0, depends in general upon the choice of the instant
regarded as the time t = 0 for the motion of each particle 4• Thus there are infinitely many different speeds of propagation. Often it is most convenient to
take the instant t=O as the present instant. Then Xrx.=6~xk, 8f8Xa.=6~8f8xk,
and the functional forms of F and /, at this one instant and qua functions of x
or X, are the same, but of course the time rates calculated with x held constant
do not generally coincide with those calculated when X is held constant. In
particular, with this choice of X the speed of propagation UN wiii be written
as U and called the local speed of Propagation 5 of the surface. This speed, which
1 HADAMARD [1903, 11, '\[82, footnote], TRUESDELL [1951, 35].
2 This quantity, for a general surface, first appears in the work of CHRISTOFFEL [1877, 2,
§ 1]; it is the second of the "deux vitesses de Propagation" introduced by HuGONIOT
[1885, 4, p. 1231]. Cf. also HADAMARD [1901, 8, § 1] [1903, 11, '\[98]. 3 HADAMARD [1901, 8, § 1] [1903, 11, '\[91]. 4 HUGONIOT [1887, 2, Part2, § 7], HADAMARD [1903, 11, '\[99]. 6 HuGoNroT [1885, 3, p. 1120], "vitesse de propagation rapportee au fluide lui-meme".
RANKINE [ 1870, 6, § 2] in dealing with a one-dimensional case called U "the linear velocity
of advance of the wave".
Sect. 184. Boundaries. 509
is the normal speed of the surface with respect to the particles instantaneously situate
upon it, is related to the speed of displacement as follows:
at oF
U 7ft= UnTt• {183.2)
More generally, we have from {182.5), {183.1), (177.5). and (74.1) the alternative forms
UN V F,a.F;a. = Un Vf,k /•k- ik /,k l
= (un- Xn) Vf,k /•k
=-I
(183-3)
(d. (74.6)). If we choose the present configuration as the initial one, {183.3)3
becomes
U=---i
Vt,k f.k '
while {183.3) 2 reduces to the moreelegant form1
U = Un-Xn,
(183.4)
{183.5)
expressing the evident fact tha:t the normal speed at which the particles now
comprising d (t) are leaving it is the excess of their normal speed over the normal
speed of the surface.
184. Boundaries. Recalling that a body P4 is a set of particles X having
positive mass, we define its boundary 2 at time t as the set of places x whose
every neighborhood contains two places distinct from x, one of which is
occupied by a particle of P4 and one is not. In kinematical terms, the bounding
surface is adjacent to P4 but not crossed by any particle of P-4. In general, it is
a moving and deforming surface.
While an axiom of continuity was laid down in Sect. 65, we now replace it
by the weaker requirement that the motion (66.1) be a topological transformation,
i.e., a transformation that puts open sets in a region of X-t-space into one-to-one
correspondence with open sets in x-t-space. In particular, for each fixed t the
transformation of X into x will then be topological. Hence the boundary of a
set in X-space is mapped, at each t, into the boundary of the corresponding set
in x-space. Therefore, the boundary surface of every body in a topological motion
is a material surface 3• Conversely, any material surface permanently divides
the material (if there is any) on one side from that on the other, and thus consists in boundary points of the bodies (if any) upon each side.
From these results and LAGRANGE's criterion in Sect. 74, it follows that
in a continuous motion, a necessary and sufficient condition that a surface f = 0
be a portion of the boundary of the material ( if any) instantaneously lying upon
either side o f it is
i=o. ( 184.1)
When the motion fails to be topological, these results hold no longer, and
boundaries may be instantly created or destroyed. General transformations
1 HADAMARD [1903, 11, ~ 100] but given in effect by HUGONIOT [1885, 3, p. 1120] in a
sentence which is confusingly mispunctuated. 2 Same properties of boundaries have been given in Sect. 69. 3 HADAMARD [ 1903, 11, ~ 48] uses the differentiability of the motion to prove this result.
The result was asserted by LAGRANGE [1783, 1, §§ 10-11] (cf. Sect. 74), but his discussion
of a bounding surface is insufficient.
510 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 184.
are not of interest in field theories. There are important cases, however, when
motions fail tobe topological at one certain instaut only, or upon certain isolated
surfaces, curves or points.
An example of the former is fumished by the fracture or welding of a solid
(Fig. 25) or by the formation or coalescence of drops in a fluid. In these cases,
0 0
0 0
Fig. 25. Fracture and welding. Fig. 26. Tear.
interior particles suddenly find themselves on the boundary of two disjoint
motions, or boundary particles suddetlly become interior ones. A still more
complicated singularity is a tear such as that shown in Fig. 26, where the singular
line .'!/ is propagating into the material, splitting each particle X in its progress
Anyconlinuous
moltun
a
c Prisms rolling upon
one onoll!er
Anyconlinuous
motion
b
d
Fig. 27 a-d. Examples of non·topological motions.
into two particles x+ and
x-, one of which stays upon
the boundary fJB+ and the
other upon the boundary
!!4-. Singularities located
upon lines seem not to have
been studied from a general
viewpoint.
The latter type of nontopological motion allows
us to represent motions in
which particles enter and
depart from a boundary
surface, possibly quite
smoothly along tangential
paths1 . Let Fig. 27 (a) represent, say, a rigid cylinder
or a fluid vortex spinning
just below the plane free
surface of a fluid and tangent at the top, and let the
region outside the cylinder be endowed with any topological motion such that
the cylinder and the plane constitute its boundary 2• The combined motion
fails to be topological at the line of tangency, and we may say that particles on
the cylindrical stream surface continually rise into the plane boundary and fall
away from it again. In this example, however, as in all others formed by
piecing together topological motions upon portions of their boundaries, the
1 PorssoN [1831, 2, § 12] [1833, 4, § 652]. 2 This example and that in Fig. 27 (d) were given by KELVIN [1848, 5].
Sect. 184. Boundaries. 511
complete boundary, consisting of the union of the constituent boundaries, is
again a material surfacel, though a material surface on which the motion fails
to be topological. Other examples are shown in Fig. 27.
In the case of motions that fail tobe continuous in the sense defined in Sect. 65,
the condition ( 184.1) is neither necessary nor sutficient that I= 0 be a material
surface or consist in boundary points of the material, if any, that it instantaneously separates.
From (183.3) it is plain that UN =0, for all choices of the initial configuration, is equivalent to f=o provided F "F·"=I=O. Now F "F·" is not determined
by the instantaneous shape of the mo~ing surface alone, but is influenced also
by the motion. The two effects are in some measure separated in the identity
{182.7). The quantity under the root sign on the right assumes the values 0 or oo
if and only if the equation F(X, t) = 0 reduces to the form F(t) = 0, which does
not represent a surface. Thus F,"F·", for a surface F =0, is singular only with
Jz/ZJ; if the Axiom of Continuity in Sect. 65 holds, from {183.3) we thus read
off a formal proof of LAG RANGE' criterion. More generally, by ( 156.2) we have
VF.(XF·" oc ~, and hence UN eo oc (! f. (184.2)
We consider particles suchthat eo=I=O, oo. Then from (184.2) 2 we conclude thatl:
1. II (! = oo upon the surlace I =0 and UN=I= oo, then (184.1) is satisfiea, but
the surface may or may not be material.
2. II e =0 upon the surface I =0 over an interval of time, then the surface is
material, whether or not (184.1) is satislied. The condition {184.1) remains sullicient, but not necessary, that I= 0 be a material surlace.
3. Bothin Case 1 andin Case 2, the condition (184.1) is necessary, but not
sullicient, that the surlace I =0 be an admissible boundary.
The case when e = oo upon f = 0 is illustrated by the following example 1 :
x=X-ct, y=(Y}-kt)3 , z=Z. (184.3}
Since
( 184.4)
the velocity field is plane, single-valued, steady, and irrotational 2, and the stream lines are
the similar cubical parabolas
y = [: (x- X)+ y~r z = z. {184.s)
which cross the y = 0 plane tangentially (Fig. 28). The density is given by
_I!Q_ = o(x, y, z) = (1-)i e o(X, Y,Z) Y '
so that I!= oo upon the plane y = 0. For this plane we have the equations
I = y = F = ( y!.- k t) 3 = o,
and hence
(184.6}
(184.7}
{184.8}
Although (184.1) is satisfied, the plane y=O is neither a material surface nor a boundary,
since the particles are continuously crossing it. Neither is y = 0, while indeed a persistent
surface of discontinuity, a singular surface in the sense defined in Sect. 180, since no jump
discontinuity occurs across it.
1 TRUESDELL [1951, 35].
2 Hence this motion, despite its artificial appearance, is dynamically possible (in theory)
for an ideal gas subject to no extrinsic force; the motion is tobe conceived as occurring in a
channel bounded by cylinders erected upon two of the curves shown in Fig. 28.
512 C. TRUESDELL and R. TouPrN: The Classical Field Theories. Sect. 185.
The speed of displacement, Un, is zero, since the plane y = 0 is stationary. The speed of
propagation UNis given by
(184.9)
Thus the surface y = 0 is suffering propagation with respect to all the particles except those
instantaneously situate upon it. The local speed of propagation is zero, U = 0, but for no
other choice of the initial instant or of material
!I variables can UN vanish. These observations are
X
Fig. 28. The plane y = 0 is not a material surface even though y = 0 upon it.
also evident from Fig. 28.
The nature of the discontinuity of the motion
is made clearer if we calculate the distance d at
time t between the two particles X, Y, Z and X,
(Y!+n!) 3, Z, whereD>O. From (184.3) we have
d = D + 3(YL kt) n! + 3 (Yl- kt) 2 n!. (184.10)
Thus D is the distance between the two particles
at the instant they cross the plane y = o. From
(184.10) it follows that no matter how small is D,
there exists a time t0 such that both after t = t0
and before t = - t0 the distance d is arbitrarily !arge. Thus no material volume remains
bounded.
185. Slip surfaces, dislocations, vortex sheets, shock waves. The first kind of
singular surface tobe used in continuum mechanics was the slip surface of HELMHOLTZ1. On such a surface, the inverse motion X =X(~. t) is discontinuous.
Each place ~ is simultaneously occupied by two particles, X+ and x-. Two
different masses thus slip past one another without penetration. The surface
f (;r, t) = 0 is a material boundary of each motion.
Fig. 29. Vortex sheet of order o. Fig. 30. Dislocation.
A surface across which the velocity suffers a transversal discontinuity,
[ :i:] =f= o, [in] = o , (185.1)
is called a vortex sheet.
Slip surfaces are vortex sheets; sometimes they are called vortex sheets of
order 0. Such vortex sheets are easily constructed by placing adjacent to one
another two motions having a common boundary. Unless it happens that
[x] =0, that boundary will be a vortex sheet of the composite motion. Thus,
alternatively, a slip surface may be regarded as a material surface on which
the functions occurring in (66.1)1 are double-valued (Fig. 29).
The dislocations of VaLTERRA 2 are singular surfaces intended to represent
the deformation corresponding to removal or insertion of one mass within another,
or the welding of boundaries (Fig. 30). There results a surface upon which X(~. t)
is double-valued. Such surfaces need not be material; they may propagate, and
they may bear any kind of discontinuity in the velocity.
1 [1858, 1, § 4], "diskontinuierliche Flüssigkeitsbewegung". 2 [1905, 6], "distorsioni". Cf. the example given in Sect. 49.
Sect. 185. Slip surfaces, dislocations, vortex sheets, shock waves. 513
A shock surface1 is one across which the normal velocity is discontinuous:
(185.2)
Dislocations may be shock surfaces.
Little of a generalnature may be said regarding such discontinuities, especially
since they may represent removal or insertion of material.
The considerations of Sect. 182 regarding the material diagram must now be
modified, since to a single spatial surface f (:.c, t) = 0 there correspond two distinct
diagrams F+=o, F-=o, viz.,
0 =f(:.c,t) =F+(X+(:.c,t),t) =F-(X-(:.c,t),t), (185.3)
the functions X+(:.c, t) and x-(:.c, t) being the two inverse functions to the singlevalued equation :.c =:.c(X, t) defining the motion of the medium. Thus the principle of duality fails to hold for these singularities.
Moreover, we cannot apply the conditions of compatibility given in Sects. 175
and 180, since the function :.c (X, t) is not defined, in general, upon both sides
of either one of the diagrams F+=o, F-=o in the space of particles. HADAMARD's
lemmastill holds, however, and may be used to derive some meager information 2•
For example, we can calculate the normal n in terms of the normals N+
and N-, using (17.3) and (182.8) with appropriate limiting values from one side
only. Thus we may put the condition (185.1) 2 for a vortex sheet into the following
form 3 : [xkl e ea.fly F± xm± xP± = 0. J kmp ,a. ,{J ,Y ( 185 .4)
Each of the diagrams F+=o and F-=o has its own speed of propagation, u,:
and U;J, given by (183.1) applied to F+ and to F-. Each of these speeds of propagation has the indeterrninacy described in Sect. 183. By using (185.3) we may
repeat the analysis leading to (183-3) and so obtain
ti+ J<+.a: · V"F-::-p-. a: • u,:Vfk ,k+x;=un=U;JVt~f.k +x~. (185.5)
For this identity to hold, the derivatives X~a. and xk need not exist upon the
singular surface; by HADAMARD's lemma, limit values from the + side of F+=o
and the - side of F-=o, respectively, are employed throughout, it being assumed, as usual, that these limit values are continuously differentiable functions
of position upon the surfaces.
This relation connects the two different speeds of propagation, whatever be
the choices of the initial configurations corresponding to the two diagrams
F+=o and F-=u. We are at liberty to choose the present configuration as the
initial one, both for the particles on the + side of F+ = 0 and for those on the -
side of F-=o. By (185.5), the corresponding local speeds of propagation, U+
and u-. satisfy the relations 4
(185.6)
1 RIEMANN [1860, 4, §§4-5], "Verdichtungsstoß"; cf. STOKES [1848, 4], "a surface
of discontinuity ".
2 E.g., the Counterpart of (175-5) is
(xk a. Xa.LI)+ = (xka. Xa.LI)-, , J J J
where the curves vr = const on F+ = 0 and F- = 0 are selected by identification of the points
X+ and x- that occupy the same place x. lt does not seem possible to make any use of this
relation.
3 BJERKNES et al. [1933, 3, § 18]. . 4 CHRISTOFFEL [1877, 2, § 1], HADAMARD [1903, 11, '1\102].
Handbuch der Physik, Bd. lll/1. 33
514 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 185.
The jump in the speed of propagation of a singular surface is the negative of the
jump in the normal velocity of the material, and a necessary and sufficient condition for the speed of propagation to be continuous is that the normal speed of
the material be continuous, i.e., that the singular surface not be a shock surface.
From (185.6) we read off the following theorems concerning shock surfaces
and vortex sheets, subject to the assumption that x (X, t) be single-valued
and continuous, though X(x, t), in general, is double-valued:
1. The local speed of Propagation cannot be continuous across a shock surface.
2. There are no material shock surfaces, i.e., shock surfaces are always waves 1 •
3· The local speed of Propagation of a vortex sheet is always continuous. In
particular, a vortex sheet which is material with respect to the particles on one side
is also material with respect to those on the other.
In the case of singular surfaces where the motion itself, or the velocity, is
singular, the dual of the kinematical condition of compatibility (180.3) must
be modified. Retracing the steps leading to that condition, we first write the
dual of (180.1) for paths on each of the two material diagrams that may represent
the singular surface f (x, t) = 0, thus obtaining
~+w+ ='f++U.+N""8 w+ !5_w- =w-+U.-Nß8 wM N "" ' !5t N ß ' (185.7)
where (J+j(Jt and (J_.((Jt as defined with respect to the normal velocities of the two
diagrams in the space of particles X. Even when the two diagrams coincide,
the speeds of propagation appropriate to the two sides are in general different,
as already noted. We now choose the present configuration as the initial configuration, both for the + and for the- sides, and subtract the second of (185.7)
from the first. Thus follows 2
['f] = [~~]- [Unk8kW],
= [~~]- u+[nkokw] -[U]nkokw-,
= [~~]- U+[nk 8kW] + [.iJ nk okw-,
( 185 .8)
where we have used (185.6). In these formulae it must be remernbered that on
the two sides of the surface 'f is calculated on the basis of ~+ and ~-. while the
normal velocities used in calculating (Jj(Jt are U+ n and u- n, respectively.
1 A shock surface may be material with respect to the particles on one side but not with
respect to those on both sides.
2 A result of this kind is asserted by KoTCHINE [1926, 3, § 1], but his statement has
!5[W]f!5t rather than [!5W/!5t] and thus is incorrect unless the operator !5/!51, referred to the
material diagram, is continuous, i.e., unless the singular surface is a vortex sheet.
Fora check on (185.8), put W =ik. The dual of (179.3) yields
~-X
!5+xk - 'k+ + u+ '"" k+ N1V X,rx •
When the present configuration on the + side is taken as the initial state, this formula
reduces to
o xk _+_=ik++ M u+nk· ,
hence
[ ~
!5xk] = [.~k] + [U]nk.
Therefore (185.8) is satisfied.
Sect. 186. Material vortex sheets. 515
Whenx is continuous, (185.8) reduces to the dual of (180.2). Another special
case will be discussed in the next section.
186. Material vortex sheets. In the case of a material singularity, irrespective
of whether or not m(X, t) is continuous, we have U+= U-=o, andin virtue of
the duals of {179.1) we may reduce (185.8) to the form
(186.1)
where d+fdt and d_jdt are the material derivatives calculated with x+ and xheld constant, respectively. Hence1
['i!l = d+J:l +[x"]\V~",
= d_[w]_ +[xk]\V+ dt ,k,
= __!_ (~+_ + d-_) [\V]+~ [x"J (\V\+ \V-"). 2 'dt dt 2 • •
(186.2)
It was suggested by HELMHOLTZ2 that the velocity ol the vortex sheet be defined
as the mean of the velocities on each side:
(186.3)
Only when the sheet is steady is this velocity tangent to it. Writing dfdt for
the material derivative following this velocity u, we have
(186.4)
and (186.2) becomes
(186.5)
where we write
grad \V == i (grad \V)++ i {grad \V)-, ( 186.6)
etc. Eqs. {186.1) and {186.5) arealternative forms of the kinematical condition of
compatibility lor material singular surlaces.
In particular, if \V and grad \V are continuous, from (186.5) we have
[ 'i!] = [x] . grad \V. (186.7)
Since the singular surface has been assumed material, we have d+lfdt =d-lfdt =0,
and hence [j] =0. Putting \V =I in (186.7) yields
[x] · grad I= o. (186.8)
Hence it follows that il the velocity is not continuous across a singular surface
that is persistently material with respect to the motion on each side, that surface is
necessarily a vortex sheet 3• Thus we have another proof that shock surfaces are
always waves.
1 The first of these forms is given by KOTCHINE [1926, 3, § 4].
2 [1858, 1, § 4]. 3 HADAMARD [1903,11, ~94].
33*
516 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 186.
For any material derivative, we have identically T;. = i k-x"' k Im.
d+lfdt = 0, this equation yields d+f.ddt = -x"'+,k l,m, ~nd hence ' ·
Since
d-f,k _ 1 ( 'm+ 'm- ) I ~dt- - -2 x ,k +x ,k ·"'' (186.9)
or
dgrad/ -~. ~(jj- = - grad;v · grad I. (186.10)
Therefore,
dn -d. . -d. ) -d1-=-gra X·n+n(n·gra ;v.n. (186.11)
This result shows that the manner in which the unit normal to the material
singular surface changes as viewed by an observer moving with the mean velocity
u is definitely determined.
The same is true of the jump in the acceleration, for if we put \!! =;r, in (186.5),
we obtain
(186.12)
a result which bears a formal similarity to (98.1) 2 for the acceleration of a continuous motion. ( 186.12) is an iterated kinematical condition of compatibility
for the special case of a material singular surface. It is a simple corollary that
[or] =0 implies [.Z.] =0; that is, il the velocity is continuous across a material
singular surlace, so is the acceleration1•
We now interpret (186.12) more closely by resolving it in directions normal
and tangential to the surface. From (186.12) 1 and (186.9) we have
I ,k [xk] =I ,k ä[ dt xk] + _1_ 2 [x"'] (xk+ ,m + xk-) ,m I ,k' l - k - (186.13) =I d[i 1 _ [ 'k] df,k • k dt X dt '
By (186.8), we may write this result in the alternative forms 2
I ,k [xk] = -2 [ik] ät.k dt = + 21 . k aukl dt ' l
[x] = 2n. ä[x]_
" dt '
(186.14)
whence it follows that the normal acceleration is continuous if and only if the
time rate of change of the velocity as apparent to an observer moving with the
mean velocity is tangent to the surface.
The more difficult reduction of the tangential component of ( 186.12) has been
achieved by MüREAU 3• By (186.12) 2 and (186.11} we have
1 HADAMARD [1903, 11, '\[ 9+] has noted an interesting variant: By differentiating twice
the equation i = 0, we conclude that [x nl = 0 if [x] = 0. That is, if at some particular instant
the velocity is continuous across a material singular surface, then at that instant the acceleration may suffer a transversal jump but not a longitudinal one. The italicized result in the
text above (due also to HADAMARD) refers to persistent continuity of ;i;.
2 KorcHINE [ 1926, 3, § 4].
3 [1949, 19] [1952, 13, §Sc].
Sect. 187. General classification of singular surfaces.
-fe (n X (;r]) = (x] X (grad X· n) - n X ([x] · grad x) +
+ (nx[x]) (n·gradi:·n) +nx[x],
= (nx[x]). gradx + [x] (n. w)-
- (n x [i:]) (div x - n · grad ;i: · n) + n x [x],
517
(186.15)
where we have used vector identities and ( 186.8), and where w == curl x = ! (w+ + w-), the mean of the vorticities on the two sides. Now by (App. 21.4)1 we
have
( 186.16)
Thus the quantity on the left-hand side is the divergence, calculated intrinsically
upon the surface, of the projection of the mean velocity onto the surface. If,
therefore, we imagine on the surface itself a fictitious motion with the velocity
field n x u, an element of area da that is carried by this motion will change
according to the formula
1ft dda (-d. . -d. ) d = 1v x - n · gra x · n a (186.17)
[ cf. (76.6), which holds in any metric space]. Putting ( 186.17) into ( 186.15) yields
~ (nx[i:] da)= nx[x] da+ (nx[x] da). gradx+[x] (n. w) da }
[ =n '']d [ '] -. ( [. ) (186.18) x a+nx x da·gradx-(n·w) nx nx x] da.
This is MoREAu's result. Its significance is easier to assess when we use EMDE's
notation (footnote 3, on p. 492): W= Curl x, W* = Curlx, for then we have
d(':tdt1_ = W* da+ W da. grad i: - (n · w) n X W da. (186.19)
Except for the last term, this equation has the same form as BELTRAMr's vorticity Eq. {101.7) 3 , establishing an analogy with between the convection and
diffusion of the vorticity w dv of a material element of volume in a continuous
motion and the transport of surface vorticity W da in an element of area
which is material with respect to the mean motion on the vortex sheet. In
many cases of vortex sheets in fluid mechanics, the spatial vorticity w vanishes
or is tangent to the sheet on each side; in such cases, n · w = 0, and the analogy
becomes precise. The analogue of the circulation preserving case is W* = 0. For
this case, as MOREAU remarks, superficial analogues of all the classical theorems
such as the Helmholtz theorems on spatially circulation preserving motions exist.
187. General classification of singular surfaces. In Sect. 173 a singular
surface was defined with respect to an arbitrary quantity \II. DuHEM1 proposed
to regard all quantities associated with a motion as functions \II (X, t) of the
material variables X, t and to define the order of a singular surface with respect
to \II as the order of the derivative ( ~J aa, aa, ... aCI.p \j! of lowest order p + q
suffering a non-zero jump upon the surface. Here, as in all that follows, we
assume that in regions fll+ and f]l- on each side of the singular surface !/'(t) in
1 [1900, 3].
518 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 187
the space of material variables X, the function \1! (X, t) and all its derivatives up
to the highest order considered exist and are continuously differentia ble functions
of X and t, while on f/(t) they approach definite limits which are continuously
differentiable functions of position. Thus we may apply the geometrical and
kinematical conditions of compatibility ( 175.11) and ( 180.2) and use differential
manipulations freely.
There is no compelling reason to allow only discontinuities of this special
type. J ump discontinuities upon surfaces are not the only ones that occur in
physical problems; e.g. in Sect. 184 we have examined a simple and otherwise
smooth motion in which the Jacobian I :x:fXI increases steadily to oo as a certain
surface is crossed and decreases steadily thereafter. Boundaries, studied in
Sect. 184, and slip surfaces, dislocations, and tears, studied or mentioned in
Sects. 185 to 186, are excluded as not being defined by sufficiently smooth jump
discontinuities in functions of the material variables. Singularities at isolated
lines or points are common; some of these are described in works on potential
theory. In the case of jump discontinuities on surfaces, there is no a priori
ground to expect that the limit values on each side of the surface be continuously
differentiable on the surface, as we have assumed. The reasons for considering
here only singularities of this kind are, first, that for more general singularities
other than those analyzed above, scarcely any definite results are known except in
very particular cases, and, second, that singular surfaces of the above types are
frequently found useful in special theories of materials.
Clearly the definition of the order of a singular surface may be expressed
1 . 1 . f h . d . . (q) N d'f' . a ternahve y rn terms o t e covanant envahves W;a.,a., ... a.p· o mo 11cahon
in the results of Sects. 175 to 176 and 180 to 181 is needed to allow us to substitute double tensors of the type TL:~ ~:J in the various jump conditions.
Many of the singularities of greatest interest are included in the case when
\1! = :x: (X, t), (187.1)
i.e., are surfaces across which the motion itself, or one of its derivatives, is discontinuous. By the order 1 of a singular surface henceforth we shall mean, unless
some other quantity is mentioned explicitly, that we are taking \1! =:1:. Thus
surfaces across which at least one of the functional relations (66.1) defining the
motion itself is discontinuous are singularities of zero order; those across which
some of the derivatives ik and x~a. are discontinuous are of first order, etc.
In the classification of LICHTENSTEIN 2, the definition of the order is based
upon the derivatives of the velocity field, ..C. Since i~m=X~a.X~m· a singular surface
of order p in LICHTENSTEIN's scheme is also one of order p in that of DuHEM
and HADAMARD, but the converse does not hold, for it is possible that a gradient
such as x~a. may be discontinuous without there being any discontinuity in ik,
etc. Thus the DUHEM-HADAMARD scheme includes a greater variety of singularities.
Most researches on singular surfaces in fluids follow LICHTENSTEIN's classification, since in hydrodynamics it is possible largely to avoid consideration of the
material variables. In retaining the DuHEM-HADAMARD classification we recognize
its more fundamental scope 3 and its necessity in contexts such as the theory of
1 HADAMARD [1901, 8, § 1] [1903, 11, ~ 75]. 2 LICHTENSTEIN [ 1929, 4, Chap. 6, ~ 2]. 3 E.g., LICHTENSTEIN [1929, 4, Chap. 6, ~ 2] concludes that there are no material singularities of first or second order. While this is true according to his definitions, it is true only
because those definitions offer no possibility of considering discontinuities in xka. and x\.ß,
unaccompanied by discontinuities in time derivatives of x. This example illust~ates th~ insufficiency of LrcHTENSTEIN's scheme.
Sects. 188, 189. Singular surfaces of order 1: Shock waves and propagating vortex sheets. 519
elasticity, but we take care to derive from it, among other consequences, the
spatial formulae that have found use in hydrodynamics.
In the following sections we find the kinematical properties of singular
surfaces of finite order1.
At a singular surface of order 0, the motion x = x (X, t) suffers a jump discontinuity. This must be interpreted as stating that the particles X upon the
singular surface at time t are simultaneously occupying two places x+ and xor jump instantaneously from z- to x+. Such discontinuities have not been
found useful in field theories up to the present time. Therefore, in what follows,
we study singular surfaces of orders 1 and greater.
188. Material singular surfaces. Material vortex sheets have been studied in
Sect. 186. The results derived there remain valid for material singular surfaces
of all orders. Fora singularity of order 1 or greater, X (z, t) is continuous, d+fdt =
d_fdt, and (186.1) reduces to
[Ii!] =['V]. (188.1)
This is the generat kinematical canditian af campatibility far material singularities
af arder greater than 0.
Its major use is to show that ['V] = 0 implies [ W] = 0: Continuity of 'V implies
continuity of '!'. In other words, the derivative af lawest arder that is discantinuaus
acrass a material singularity is always a purely spatial derivative, never a time
derivative 2• This is the dual of the theorem stated just after (180.4).
In particular, across a material singularity of first order, since x (X, t) is
continuous, so is ~. That is, not only shock surfaces but also vartex sheets af first
arder are waves, while for the material vortex sheets described in Sect. 186 it is
impossible that the motion itself be continuous across them. Vortex sheets are
thus divided into two distinct categories: those of order 0, which are material,
and those of order 1, which propagate. Across a material singularity of first
order, ~ is continuous, but at least one of the deformation gradients x~" suffers
a discontinuity.
189. Singular surfaces of order 1: Shock waves and propagating vortex sheets.
Forasingular surface of order 1, we put 'V = x!' in the duals of (180.5) and obtain 3
[~ .. ] = sk N .. , sk =[Nß ~p]. [X"]=- UN sk. (189.1)
1 A singular surface of infinite order is defined by HADAMARD [1903, 11, ~ 76] as one such
that on each side, the function occurring in (66.1) aredifferent analytic functions, yet all
their derivatives are continuous across the surface. Such singularities seem not to have been
studied. They offer interesting possibilities. For example, a one-dimensional motion starting
very smoothly from rest at t = 0 is furnished by x =X for t ~ 0, x =X+ f (X) e-c/1' for t > 0,
where c > o.
In works on physics we often encounter discontinuous solutions regarded as limits of
continuousones; cf., e.g., the definition of adiscontinuity given by MAXWELL [1873. 5, §§ 7-8],
[1881, 4, § 8]. Perhaps seeking to justify such a treatment, LicHTENSTEIN [1929, 4, Chap. 6,
§§ 6-7] proves that any motion with a singular surface of order 1 or 2 may be obtained as a
limit of analytic motions. Such theorems, however, do not reflect the physical situation. In
electromagnetic theory, for example, physicists are wont to regard the solution of a problern
for a material whose dielectric constant changes abruptly from K 1 to K 1 upon a certain surface
as the limit of the solutions of "the same" problern for a dielectric suchthat K varies smoothly
from K 1 to K1 in a thin layer containing the surface. In gas dynamics, a flow of an ideal gas with
a shock wave is regarded as the limit of solutions of a corresponding boundary value problern
for viscous, thermally conducting fluids as the viscosity and thermal conductivity tend to
zero. Problemsofthis kind presuppose some definite theory of materials and have no meaning
in the generality of the present treatise. Even in relatively simple definite cases, they offer
the highest mathematical difficulty. Cf. Sect. 5. 2 HADAMARD [1903, 11, ~~93, 104). 3 HADAMARD [1903, 11, ~ 101].
520 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 189.
The vector s is the singularity vector; while (189.1) 2 shows it to be parallel to
the jump of velocity, its magnitude varies with the choice of the initial state and
thus does not furnish a measure of the strength of the singularity. Rather,
guided by the result given in Sect. 188, it is convenient to divide singular surfaces
of order 1 into two classes:
1. Material singularities, which affect only the deformation gradients ~"'.
2. Waves, including both shock waves and propagating vortex sheets.
For the former, the choice of the initial state is of prime importance. For the
latter, it is not, and the nature of the waves is best specified in terms of the jump
ofvelocity itself, [:V], which may be arbitrary bothin direction andin magnitude.
Indeed, if we adopt a strictly spatial standpoint, we may say the only geometrical
and kinematical requirement is that discontinuities in velocity be propagated, both
the amount of the discontinuity and the speed of propagation being arbitrary.
Even here the adherents of a strictly spatial standpoint are closing their eyes
to one of the phenomena occurring, since from (189.1) it follows that a fump ,·n
velocity is impossible unless it is accompanied by fumps in the deformation gradients ~«·
The results derived in Sect. 185, since they presume less regularity than is
here assumed, remain valid for singularities of order 1. In particular, the speed
of propagation satisfies (185.6), and the first and second theorems derived from
it remain relevant. The third theorem is irrelevant, since, as just shown, vortex
sheets of first order are always waves.
Since the functions occurring in (66.1) are continuous and single-valued by
hypothesis, there is only one material diagram corresponding to a given initial
state, and the principle of duality now holds. However, it is necessary to proceed
with caution, since the fact that xk« experiences a jump on " implies that if we
regard a succession of initial states X corresponding to different initial times t0 ,
these states jump discontinuously at t0 = t ± 0. In particular, choosing the states
on the + and - sides of the singular surfaces leads to different local speeds
of propagation, u+ and u-, satisfying (185.6), as already stated. Writing the
corresponding vectors s as s+ and s-, we have
hence
By (185.6) follows
[Us]=O.
un[s] =[ins].
In the case of a vortex sheet, this relation reduces to
in the case of a stationary shock wave, to
[ins] = 0.
(189.2)
(189.3)
(189.4)
(189.5)
(189.6)
The nature of first order shock waves is illustrated by the very simple example1 furnished
by the one-dimensional motion defined by the equations y = Y, z = Z, and
X= l
X+ 2vt when X ~0,
l X + v t when X :::;; 0,
X + v t when X ~ 0,
2X + 2vt when X ~0,
1 LICHTENSTEIN [1929, 4, Chap. 6, § 8].
- oo 0, remain a distance D apart until
the latter encounters the shock. After the shock has passed, their distance apart is t D.
The shock thus effects not only a sudden drop in velocity but also a sudden condensation.
We now prove that this is representative of the general case.
To determine the jump experienced by a volume element through which a
wave of first order passes, we calculate the jump in the Jacobian V?J/VG = xjX.
By {189.1h we have
By the dual of (17.3) follows
_E_ -1 + PN XY- 1- - s y ,p· (189.11)
Now we may apply (182.8) and (189.1) 3 , writing Uii to recall that the derivatives
X~"k are used in the calculation2 ; using also (185.5) 2 we thus obtain
(189.12)
or, by (185.6),
Xri-un u+
Xn -Un = U~' {189.13)
Thus the passage of a shock wave of first order 3 causes an abrupt change of volume,
the ratio being that of the local speeds of Propagation.
1 We are not obliged to use initial positions or reetangular co-ordinates as material
co-ordinates.
2 The calculation may be shortened by taking the state on the - side of the surface as the
initial state from the beginning.
3 The qualification, "of first order", refers to the possibility of shocks in which :r(X, t)
is discontinuous. About such shocks, mentioned in Sect. 185, little definite can be
said.
522 C. TRUESDELL and R. ToUPIN: The Classical Field Theories. Sect. 189.
This result is usually presented in terms of the density e; from ( 156.2) then
follows the Stokes-Christoffel condition:
(189.14)
These forms are of frequent use1 .
From (189.14) we see that shock waves offirst order areimpossible in an isochoric motion, and that the passage of a vortex sheet offirst order leaves the volume
unchanged.
By combining (185.6)a and (189.14)1 we obtain the identity
e±U±[xn]=-[eU2]=(U-) 2 ~= [e]. (189.15)
Since, by hypothesis, J > 0, from ( 189.13) it follows that in- un is of the
same sign on each side of the singularity. That is, if to an observer situate upon
the surface the material seems to approach on one side, it departs upon the
opposite side. The side on which the material approaches is called the front
side, and the sign- will be assigned to it; the other side is the rear. The strength ö
of a shock wave of first order is defined as
.ll = _hl = _[:_- 1 = [in] u- (F (F u+ . (189.16)
If ö > 0, the shock effects condensation; in the contrary case, rarefaction 2 •
By putting \V= xk in the duals of (176.8) and (181.8) we may calculate the forms of the
jumps of x:rxß• i:rx• and xk across a shock wave. The results are complicated. Weshall give
them only for the quantity of greatest interest, the acceleration fi. Putting
(189.17)
from (181.8) and (189.1) 2•3 we obtain
2 t5ct sk t5ct UN ) [xk] = UN ck - 2 UN --- sk --- M M '
_ U? k <5d[ik] _ [ 'k] <5ctlog UN - N c + 2 t5t X 0t '
(189.18)
where the displacementderivative t5ct/M is defined in terms of the motion of the samematerial
diagram F(X, t) = 0 as is used to calculate the speed of propagation, UN· Since the quantity
ck, the jump of the fully normal second spatial derivatives, is essentially arbitrary, there
is no immediate interpretation for the result ( 189.18).
Indeed, since the jump in velocity is arbitrary, so are the jumps in its derivatives, and
there is no restriction in general upon [ik,m] or [xk]. It may be convenient, nevertheless,
to resolve these jumps by means of the identities (175.11) and (180.3). The results are 3
[ik,m] = [nPik,p] nm+ gmpa.dr xf.d [ik];r,l
[ 8 i k] - [ p • ] + t5ct [ i k] 8t -- Un n xk,P _t5_t_'
(189.19)
1 The direct proof given in the text simplifies and generalizes that of KoTCHINE [1926,
3, § 3. ~ 1]; the customary proof, resting upon an integral form of the principle of conservation
of mass, will be given in Sect. 193. For one-dimensional motion, the result is due to STOKES
[1848, 4, Eq. (2)], RIEMANN [1860, 4, § 5], and RANKINE [1870, 6, §§ 2-4], the general case,
to CHRISTOFFEL [1877, 2, § 1] and JouGUET [1901, 9]. The classical treatment isthat of
HADAMARD [1903, Jl, ~~ 109-110). 2 There is no kinematical or dynamical reason why rarefaction shocks cannot occur, although special thermodynamic conditions, including those usually assumed in gas dynamics,
may forbid them. 3 The somewhat different results obtained by HADAMARD [1903, 11, ~~112, 113 bis,
119-120] are equally inconclusive.
Sect. 190. Singular surfaces of order 2: Aceeieration waves. 523
where the displacement derivative 15d/15t now refers to the spatial equation of the singularity,
f(:JJ, t) = o.
190. Singular surfaces of order 2: Aceeieration waves. For singular surfaces
of order 2, the analysis is simpler, since the speed of propagation UNis continuous.
Substituting 1,1! = xk into the dual of ( 176.1 0) yields
[~cxlll = [8px~"'] = [8"'x~11] = sk N"'N11 , }
sk =[NY N~ o~x~y] =[NY N~ x~y.,]. (19°·1)
From the dual of (181.9) we have similarly1
(190.2)
These formulae show that a singular surface of order 2 is completely determined
by a vector s and the speed of propagation, UN. In particular, material discontinuities of second order affect only the derivatives ~"' 11 , while discontinuities
in the acceleration and in the velocity gradient are necessarily propagated, and
conversely, every wave of second order carries jumps in the velocity gradient
and the acceleration. Waves of second order are therefore called acceleration
waves.
Since X~m is continuous, by (190.2h and (182.8) we have
[i~m] =- UNskNcxX"',m• l =- UN sk VVt.t>_t__t:,_ nm.
F,pF.II
(190.3)
The left-hand side is independent of the choice of the initial state; therefore so
also is the coefficient of n upon the right-hand side. Thus if we put
Us0k= UNsk Vt,pt:P , (190.4)
VF.pF.II
the vector Us0 is independent of the choice of the initial state. We may choose
to regard U as the local speed of propagation and s 0 as the value of s corresponding to a choice of the material co-ordinates as being equal to the spatial Coordinates at the instant the wave passes. In this notation, (190-3) and (190.2) 3
become 2
[~m] =- U s~ nm, [xk] = s~.
From (190.5) we have a number of corollaries. First,
- U [nm i~m] = [xk],
[()ik]- ["k]- [ ·~ ] •m ot - X X ,m X '
= s~ + U S~Xn,
= Uunst
(190.5)
(190.6)
where we have used (183.5). Thus the local acceleration is continuous across
an acceleration wave if and only if the wave is stationary. Second,
[Ic~J = [divi:] =- Us0 • n, } (190.7) [w] = [curli:] =- Us0 X n;
l Eqs. (190.1) to (190.2) are due essentially to HUGONIOT [1885. 4, p. 1231] [1887, 2,
Part II, § 9)]. Cf. HADAMARD [1901, 8, § 2] [1'903, 11, ~ 102].
2 These results and (190.6) were given by HUGONIOT [1885, 3, p.1120] [1887, 2, Part I,
§ 9] in forms with U s~ eliminated.
524 C. TRUESDELL and R. TOUPIN: The Classica] Field Theories. Sect. 191.
interpretation of these identities yields Hadamard's theorem1 : A longitudinal
acceleration wave carries a fump in the expansion but leaves the vorticity unchanged,
while a transverse acceleration wave carries a fump zn the vorticity but does not
affect the expansion.
By (156.5) 2 , we may put (190.7) 2 into the alternative form
[lo~ e] = U s 0 · n = -& [xn]. (190.8)
Since log e is continuous across an acceleration wave, by putting lj! =log e in
the dual of (180.5h we have
[lo~el =- u [a 1
;!e]. (190.9)
By ( 190.8) follows the important relation
[Xn] =- U2 [d~!e]. (190.10)
191. Singular surfaces of higher order 2• For a singular surface of order p,
the results are easy generalizations of those for the case when p = 2.
Putting lj! = xk into the dual of ( 176.11) yields
[xk•"''"'•···"'P] = [8"', 8"', ... 811Pxk] = sk Na, Na, ... N"'P' }
sk = [Nß, Nß• ... Nßp 8p, 8p, ... 8ppxk], (191.1)
while the dual of (181.11) yields
[xk ] = - U. sk N N N ,cx1a:2···!Xp-1 N a1 a2 • •• cxp-I'
( 191.2)
[(P_;Ilk] = (- U. )P-I sk N ,o:l N CXt'
[~lk] = (- UN)P sk = (- U)P s~.
Thus a singularity of order p carries a jump in the p-th acceleration if and only
if it is a wave.
Now by (156.5) we have
-·- 'k "' 'k - log !? =X , a. X, k =X, k, (191.3)
so that
_(h_)_ (h)
-(log e),a.,a., ... a.p-1-" = x k,a.a., a., ... a.p-1-" x~k+ ... , (191.4)
where the dots stand for a polynomial in derivatives of orders less than p. By
( 191.2) we thus obtain
_l!L_
-[(loge),a.,a, ... ap- _,.]=~- UN)hNa.,Na.,···Na.p-1_,.Na.xa..•sk,} (
191.5)
h - 0, 1' ... 'p - 1 .
Choosing the initial state as that on the singular surface yields
_l!L_
-[(log e),k,k, ... kp-I-A] = (- U)hson nk, nk, ... nkp-1-h' ( 191.6) ------
1 HADAMARD [1903, 11, ~~111-115]; announced in part in [1901, 8, § 3]. In part, these
results are equivalent to WEINGARTEN's formulae (175.10) and are foreshadowed by a theorem
of HuGONIOT [1887, 2, Part I, § 12].
2 HADAMARD [1903, 11, ~~88, 103,111-111 bis].
Sect. 192. The transport theorem for a region containing a singular surface. 525
In particular, from this result and (191.2) 4 we have
__lE::!L 1 ( p)
[(logg)] = u [xn]. (191.7)
This relation enables us to attach a meaning to the sign of a discontinuity. [Cf.
(189.16) and (190.8).] If the propagating singular surface of p-th order carries
(p-1)
with it an increase in log g, it may be said tobe a compressive wave; in the contrary case expansive. From (191.7) we see that a wave is compressive or expansive
d. . b . . d . h 1 (p) accor mg as It nngs an mcrease or ecrease m t e norma component Xn of
the p-th acceleration 1• In particular, in an isochoric motion, acceleration waves
of all orders are necessarily transversal, and, conversely, material singularities
and transverse 'li/aves of all orders leave the density and all its derivatives continuous.
In the case of a surface which is singular with respect to 'iJ and also a singular
surface of order 2 or greater with respect to the motion itself, the principle of
duality when applied to ( 180.3) yields
['f] =- UN[N" 'il,)d•+~pxei>i>·da- J(l+pxf)d!JJl. I
-r .9' -r
Thus iJ.9' and ~.9'. the force and torque exerted by the material outside !/ upon
that inside, may be calculated from the fields 0 ~(! p) ' ~1!_ X e p, I and l within r
and the field e p p on !/. öt t
First, in steady motion within a stationary outlet
vessel of any form, by (69.1) wegetfrom (202.1)
iJ.9'=- fldiDl,
~.9'= _!"J(l +pxl)diDl. (202.2)
-r
Thus the steady motion of a materialfilling a inlet
closed stationary vessel has no reaction upon the Fig. 33. Material in steady llow through a pipe.
vessel 2•
Fora less trivial example 3, consider a material which is being forced in steady
flow through a stationary pipe (Fig. 33). Suppose that l =0 and that I is the
field of uniform gravity. If we take the normal to the inlet cross section .9{
inward, retaining that on the outlet cross section ~ as outward, then (202.1)
reduces to
il.9' = f e PP· da- f e PP· da-~. ) ~ .Y'i
~.9' = f p x e p p · da - f p x e p p . da - m c x ~,
9;, .Yj
(202.3)
where $ is the weight of the material in the pipe. Thus, no matter what the
shape of the pipe or the nature of the material in it, from a knowledge of Im
and of the density and velocity of the material at the inlet and outlet only, we
can determine the total reaction on the material instantaneously occupying the
pipe. Writing g;, for the surface constituting the pipe itself, by (202.3) and (202.1)
1 In principle, these formulae are old. Apparently the first general discussion is that of
V. MISES [1909, 8, § 9). A clear treatment is given by CISOTTI [1917, 4, §§ 1-4). Cf. LELLI
[1925, 8 and 9], MüLLER [1933, 8].
2 CrsoTTI [1917, 4, § 7].
3 CISOTTI [1917, 4, § 8].
540 C. TRUESDELL and R. TOUPIN: The Classical Field Theories. Sect. 202.
we obtain also
Jt(n)da = J (t(n)da- e p p. da)- J(t(n)da- e PP· da)-~.
9p .9'i .9"0
f (m(n) + pXt(n))da = f [m(n)da + p X (t(n)da- (! p p ·da) - .9"p .9"1 (202.4)
- f[m(n)da + px (t(n)da- e p p. da)- 9Jl cx ~ . .9"0
Thus a knowledge of ~ and (!, p, t(n), and m(nl at the inlet and the outlet alone
suffices to determine the reaction the pipe exerts upon the material.
Consider next a single finite closed rigid boundary submerged and fixed in
a continuous medium, and let .5I;; be an imagined control surface sufficiently
large as to include all of [/ (Fig. 34). For simplicity, take f = 0, l = 0. So as to
calculate the reaction exerted by the rigid obstacle on the material, we apply
/-----.........
/ ' / ' (D y \ I
'- I ' / ............... ____ ,.../
Fig. 34. Reaction on a submerged
rigid object.
(202.1) to the material between [/ and 9';;, thus obtaining for the force and torque exerted by the obstacle the
expressions1
iJ = :t J pd!Dl +~e p p ·da- ~t(n) da,
" .9'C .9'C
f= Jpxpd9Jl+~PXepp·da-
" .9"c
-~pxt(n)da.
.9'C
(202.5)
The first terms represent the force and torque arising from local changes in velocity. In steady motion, they vanish, and (202.5) shows that then the reaction
of a submerged body may be determined from the values of the stress vector, the velocity, and the density on a control surface far from the body.
To calculate this reaction in the steady case, let V be any constant vector;
then by mere algebraic identity we have
~ e p p · da = ~ e (P - V) (P- V) · da + l ~ ~
+ ~e(P- V) da· V+ V~eP·da.
~ ~
(202.6)
Since e p is solenoidal in a steady motion, and since p is normal to the fixed
boundary !/, application of GREEN's transformation to the integral of div(e p)
over the region between [/ and .5I;; shows that the last summand vanishes.
Thus (202.5h may be written
iJ = ~ e(P- V)(p- V)· da+~ e(P- V) da· V-~ (t(nl + Pn)da, (202.7)
~ ~ .9'j;
where P is an arbitrary constant. This formula may be applied to the case
when at great distances from the obstacle the material is moving at a uniform
velocity V and the stress is hydrostatic. Indeed, from (202.7) we read off the
1 For the classical special case of isochoric irrotational motion subject to hydrostatic
pressure, see e.g. MILNE-THOMSON [ 1938, 9, § § 17.10- 1 7. 51] and the more general result of
RASKIND [1956, 12].
Sect. 202. The reaction on bounding surfaces. 541
generalform of the Euler-D'Alembert parado~ : Let a stationary rigid body be
immersed in an infinite material in steady motion past it; if there exist constants V
and P such that
e (p - V) (p - V) = o (p-2), e (p - V) = o (p-2), t + Pn = o (p-2). (202.9)
The result is extremely general: lf the velocity approaches its constant limit
at oo sufficiently quickly, and if the stress vector approaches a uniform pressure
at oo sufficiently quickly, the submerged body exerts no force. However, the
"paradox" has limited application, as the order conditions (202.8), which are
the very essence 2 of the assumptions underlying it, are rarely fulfilled in particular theories of materials. The classical example is homochoric irrotational flow
of a perfect fluid, where (202.9) follows easily from theorems of potential theory.
It is often asserted that in fact it is the adherence of materials to solid boundaries
that accounts for the resistance to steady motion in real fluids. This may be so,
but it does not enter the present argument directly. Indeed, if the stronger
condition (69.4) were to replace (69.2) here, the most we could expect would
be the annulling of further surface integrals, not the addition of extra terms.
Rather, so far as is now known, taking account of adherence in conjunction
with any dissipative mechanism has the effect of transmitting stronger disturbances to oo, sufficient to violate the order condition (202.8).
There are various generalizations. If the material is confined by a stationary
canal, from (202.6) we see that instead of integrating over all of ~, for the first
two integrals in (202.7) we need consider only the cross sections of the canal
at great distances, and again (202.8k 2 are sufficient to make these integrals
vanish in the limit. To make the third integral vanish also, in addition to (202.8) 3
we supply an appropriate assumption regarding the value of the stress vector
on the canal walls 3• lf there are surfaces on which p is discontinuous, provided
they do not extend to infinity, the result still holds 4 ; if, however, there are
infinite shocks or slip surfaces, the resultant force in general is not zero 5•
To calculate the torque, we note first that
da · p (p Xe p) =da · (p- V)[p X(! (p- V)] + }
) . • (202.10) +(da· V [pxe(P- V)]+ da· e ppx V.
Now by GREEN's transformation and (69.2) and (156.6) we have
f da· ePP = fdiv(epp)dv = Jpd!JR = 14§ (202.11)
.9"c -r -I'"
1 The result was asserted and proved correctly by EuLER [1745. 2, Satz I, Anm. 3] for
steady plane flow of a perfect fluid, on the assumption that the stream lines straighten out
at oo. D'XLEMBERT [1752, I, §§ 66-69] rediscovered or appropriated it; later [1768, 2, §I]
he reasserted it in sensational terms and proved it by an argument assuming the body to
consist in eight congruent parts. It has given rise to misunderstandings lasting over centuries.
For the hydrodynamical case, a correct proof based on integral transformation was given
by CtSOTTI [1906, 3]; we present, essentially, the reformulation by BOGGIO [1910, 2]. A
variant argument based partly on the energy balance was suggested by DUHEM [1914, 2]
and worked out by B. FtNZI [1926, 2]. 2 In the older treatments this is glossed over, but it is made clear by CtSOTTI [1917,
4, § 9].
3 CISOTTI [1909, 4].
& DUHEM [1914, 2 and 3], PICARD [1914, 9], MANARINI [1948, 14] [1949, 17]. 5 Examples were given by HELMHOLTZ and others; a general discussion is presented by
JouGUET and Rov [1924, 7].
542 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 203.
where !f3 is the momentum of the material between !I' and ~. Substituting
(202.10) and (202.11) into (202.5) 2 yields
)! = p p Xe ('p- V) (p- V) ·da +V· f da p Xe (p- V) + l
9: 9: c c (202 12) + !f3x V- pdapx (t(n; +Pn). · 9'c
Therefore if we strengthen (202.9) to read
p-V=o(p-3), e=0(1), t+Pn=a(p- 3),
we conclude not only the Euler-D'Alembert paradox but also1
(202.13)
)! = !f300 X V, (202.14)
where !f3oo is the relative tnomentum 2 of the material exterior to the obstacle. Under
the severe conditions (202.13), then, a stationary rigid body will generally exert a
torque perpendicular to the direction of motion. In certain cases of symmetry
we shall have lf300 = 0; the body then exerts no torque.
CISOTTI 3 extended these results to the case of an obstacle spinning at angular
velocity w, provided the motionrelative to the obstacle be steady. He found that
then
(202.15)
the expression for )! is more complicated.
All our results are expressed in terms of the reaction of the obstacle on the material. That this is equal in magnitude and opposite in direction to the reaction
of the material on the obstacle is to be expected and follows from CAUCHY's
lemma, to be proved in the next section.
203. The stress tensor 4• The field of stress vectors t(n) is not an ordinary
vector field. Rather, since the stress vectors across two different surfaces through
Fig. 35. CAUCHY's construction to prove bis fundamental theorem.
the same point are generally
different, at any given time
t(n) is a function both of the
position vector p and of the
direction n. The first problern is to delimit the dass of
t(n) forfixed p as n varies.
We apply (200.1) to a
tetrahedron (Fig. 3 5), three
.Xz sides of which are mutually
orthogonal, the fourth having outward unit normal n.
Let the altitude of the tetrahedron be h; the area of the inclined face, s; the
projections of n onto the orthogonal faces, n1, n2, and n3, so that the areas of
the inclined faces are snl, sn2, and sn3. Assurne that the fields eP and e/ are
1 CISOTTI [1910, 4], BoGGIO [1910, 2]. 2 I.e., $ 00 =o f(p- V) d'.m, where by (202.13)1 the integral, taken over the infinite space
exterior to the obstacle, is convergent. 3 [1910, 4]; a vectorial derivationwas given by BoGGIO [1910, 2]. 4 The idea and the results here are due to CAUCHY [1823, 1] [1827, 1]. In a sense, the
fundamental theorem (203.4) is contained in a memoir written by FRESNEL in 1822 but not
published until much later [1868, 7, §§ 3-4]; however, FRESNEL did not disentangle stress
in general from purely elastic stress. Cf. also HoPKINS [1847, 1, § 2].
Sect. 203. The stress tensor. 543
bounded, that t(n) is a continuous function both of p and of n. We may then
estimate the volume integrals in {200.1) and apply the theorem of mean value
to the surface integral:
{203 .1)
where K is a bound and where t(n) and t~ are the stress vectors at certain points
upon the outsides of the respective faces. We cancel 5 and let h tend to zero,
so obtaining
(203.2)
where all stress vectors are evaluated at the vertex of the tetrahedron. From
their definitions, the quantities t 1 , t 2 , t 3 do not depend upon n.
In the case when n,.=1, n 2 =n3 =0, we have t 1 =t(-n)• since the outward
normal to the face 1 is -n. Although the construction of the tetrahedron fails
for this case, we may nevertheless suppose nc~1, ~0, ~0, and by the
assumed continuity of t(n) as a function of n infer from {203.2) that
{203.3)
This is Cauchy's lemma: The stress vectors acting upon opposite sides of the same
surface at a given point are equal in magnitude and opposite in direction.
Now select a reetangular Cartesian co-ordinate system whose planes are the
three orthogonal faces. With the convention that tkm is the k-th component of
the stress vector acting upon the positive side of the plane zm = const, by {203. 3)
we infer that -t1 x=l:.x, -t1 y=tyx• -t1 z=tzx, etc., and hence (203.2) becomes
t(n)k =tkmnm, where the quantities tkm are independent of n. By the quotient
law of tensorsl, the quantities tk m form a Cartesian tensor of second order. The
tensor equation
tk - tkmn (n)- m> {203.4)
having been established in a special co-ordinate system, is valid in all co-ordinate
systems. This proves Cauchy's fundamental theorem: From the stress vectors
acting across three mutually perpendicular planes at a point, all stress vectors at
that point are determinate; they are given by {203.4) as linear functions of the stress
tensor tkm.
Application of a parallel argument to (200.3) yields m )
,h J (205.18) = 'j'tk1 ... knqda + fk 1 ... /nqdiJJl.
f/' j/'
The corresponding boundary conditions and differential equations are, respectively,
tk .... knq= tk .... knqPnP, } (205.19)
(! Zk 1 ... kn Zq = tk1 ... kn q p, p +e /k" ... kn q ·
They may be recast by writing tk .... k.q as the sum of the n-th moments of all
stresses of lower order plus a tensor of excess; e.g., in the notation introduced
above,
(205.20)
The assigned moments fk .... knq are subject to a law of transformation such as
to render {205.20) invariant with respect to change of the origin of Co-ordinates.
205A. Appendix. Notations for stress. The following table, shortened from that of
PEARSON [1886, 4, § 610], presents the notations for stress found in the classical works and
in some cases still in use today. The blocks stand for the matrix of tkm in reetangular Cartesian co-ordinates, or possibly for physical components in curvilinear Co-ordinates.
CAUCHY's earlier work
A F E
F B D
E D c
F. NEUMANN,
KIRCHHOFF,
RIEMANN
Xx Xy xz
Yx Yy Yz
Zx Zy Zz
PoisSON
Pa Qa Ra
~ Q2 R2
~ Ql R1
KELVIN
p V T
V Q s
T s R
CoRIOLIS,
CAUCHY's later work,
ST.-VENANT,
MAXWELL
Pxx Pxy Pxz
Pyx Pyy Pyz
P.x Pzy Pzz
CLEBSCH
tll 112 113
121 122 123
131 Ia 2 l33
~ Ta I;
Ta ~ :z;_
I; :z;_ Na
PEARSON
XX .iY fi
jiX yy yz
Zi iY Zi
German engineering works usually employ ~. N,;, or r1x for lxx and :fxy, :Z:,, l'xy• or Tz for
txy• etc. French authors usually follow LAME, though some adopt the more luminous notation
of CoRIOLIS. The definitions sometimes differ in sign. The notation lkm may be remernbered
by the word tension; Pkm• defined by (204.5), by the word pressure.
In this work we always use PEARSON's notation k;n, already introduced in Sect. 204,
when employing physical components in orthogonal curvilinear co-ordinates. By (App. 14.6}
and (App. 14.7), CAUCHY's first law (205.2) then assumes the explicit form
ex = ef + ± {-v1 a!m (liiJ;;n). + m~l g V gmm
+-·-------·----- ~ 1 ayg,;;. mm 1 ayg;;.~}
Vfkl. yg:~ oxm Vgmm yg,.-,; oxk I {205A.1)
Sects. 206,207. The equivalence of stress, extrinsic Ioads, and transfer of momentum. 549
(LAME [1841, 4, §§VII-VIII] [1859. 3, § 14, Eqs. (14) (15)]; cf. MoRERA [1885, 6], ANDRUETTO [1931, 1]; for generat curvilinear Co-ordinates, cf. Rrccr and LEvr-CrvrTA [1901, 14,
Chap. 6, § 3]. ToNOLO [1930, 7]).
206. Stress impulse. For the laws of impulse in continuum mechanics, we
apply to ( 198.1) the general results given in Sect. 194. The influx of im pulse
is the negative of the stress-impulse tensor i; the supply of im pulse is a vector s;
and as the counterpart of CAUCHY's first law (205.2) we get from (194.11)
(206.1)
while the balance of moment of momentum assumes exactly the form (205.10)
with i replacing t and with appropriate new symbols replacing m and l.
In the special case when i is spherical and s = 0, the covariant form of
{206.1) is
(206.2)
Hence follows trivially a celebrated theorem of LAGRANGE and CAUCHY1 : The
mass flow produced by impulsive hydrostatic pressure is lamellar; in particular, a
homochoric motion initiated by impulsive hydrostatic pressure is irrotational in the
first instant.
More generally, by comparing (205.2) and (206.1) we derive the following
analogy: The correspondences
(206.3)
at a given instant enable us to construct a dynamically possible impulse, Stressimpulse, and supply of impulse from a dynamically possible acceleration, stress,
and force, and conversely.
207. The equivalence of stress, extrinsic loa:!s, and transfer of momentum.
As follows from the remarks in Sects. 15 7 and 199, in any given problern the
stress may be replaced by equipollent extrinsic forces, or the extrinsic and mutual
forces by a stress. Indeed, CAUCHY's first law (205 .2) asserts that the resultant
force exerted by the stress is tkm m per unit volume, while (205.10) asserts that the
stress and the couple-stress exe~t a resultant couple of magnitude mkPq,q_tlkPl per
unit volume.
Conversely, to replace mutual and extrinsic loads by stresses we must
integrate tkm,m=efk with jk given. Any solution of this equation may be added
to the stresst in (205.2), resulting in a combined stress under which the material
moves as if subject to no extrinsic or mutual force. In general, to calculate an
equivalent stress is elaborate, but it turns out to be simple in the special case
when ef is lamellar:
For then
(207.1)
(207.2)
so that when the assigned force per unit volume is lamellar, it is equipollent to an
appropriate hydrostatic pressure 2•
This equivalence of various loads is generally artificial and useless, however.
The course of the discipline of mechanics is to prescribe functional dependences
for fand t (cf. Sect. 7). With a single prescription fort, a host of different motions
1 The statement and proof of LAGRANGE [1783, 1, § 20], motivated perhaps by earlier
remarks of D'ALEMBERT [1752, 1, §§SO- 51], were corrected by CAUCHY [1827, 5, Part 2,
Sect. 1, §§ 4- 5].
2 In effect, this observation is due to EULER [1757. 1, §§ 25-31].
550 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 207.
of the body, corresponding to different initial and boundary conditions, are
possible; for each of these different motions, a different equivalent f will result.
There is a simple and natural equivalence of stress to transfer of momentum,
obtained1 by substituting (156.7) into CAUCHY's first law (205.2):
:t (e .i") = (t""'- e .i" x"') ,". + e I". (207.3)
This equation asserts that if the tensor - e .i" .im is added to t""', the true rate
of change of momentum may be replaced by the local apparent rate 8(e~)f8t;
therefore - e~ ~ is called the tensor of apparent stress due to trans/er of momentum 2•
I t is the negative of the momentum transfer, which we have discussed in Sect. 170.
Very little is known regarding the transformation of systems of stresses appropriate
to one problern into those appropriate to another. An elegant result of this kind is due to
ÜLDROYD and THOMAsa and NoLL 4 • We present NoLL's form of the argument. For an incompressible medium, we consider two velocity fields d:1 and re, which differ only by a rigid rotation at constant angular velocity w (cf. Sect. 143). We assume further that, apart from
arbitrary hydrostatic pressures p1 1 and p2 1, the stress tensors in the two motions are the
same functions of place and time 5 :
p11 + tl = p,1 + t,. (207.4)
Let the frame in which the velocity is d:1 be inertial, so that CAUCHY's first law (205.2) is
satisfied. By the results in Sect. 197. the second motion, if it is tobe dynamically possible,
must then satisfy an equation of the same form but including the apparent forces arising
from the rotation with respect to an inertial frame. Thus we have
(! fi1 = div t1 + (! f, } (207.5)
(! fi1 = div t2 + (! (f + g),
where - g is the sum of the CoRIOLIS and centrifugal accelerations. Subtracting (207.5) 1
from (207.5) 2 and using (207.4), we see that
(! g = grad (P2 - P1 ). (207.6)
In order that the motion d:2 be dynamically possible, it is thus necessary and sufficient
that g possess a single-valued potential. From the results in Sect. 143, a sufficient condition
for g to be lamellar is that the motion be plane and that w be normal to the plane of motion 6 •
Conditions sufficient to ensure single-valuedness of the function Q' in (143.10) have been
stated in Sect. 161. Directly from (207.6) we see that the reaction per unit height exerted
upon any cylinder perpendicular to the plane of motion and of cross-sectional area ~ differs,
in the two cases, only by the reaction of a hydrostatic pressure equalling the potential of
the centrifugal and CoRIOLIS accelerations. The moment is thus zero, and the resultant
force is easily shown to be that requisite to impel a mass-point of mass (! ~ located at the
centroid to move with the velocity of the centroid. Alternatively, the difference of the forces
is that which would be exerted upon the region occupied by the cylinder if it were filled with
the surrounding material. Hence if a rigid homogeneaus cylinder immersed in an incompressible substance of the same density as the cylinder is made to undertake any motion in a plane
normal to its generators, then a uniform rotation of the whole system about any axis parallel
to the generators will not alter the motion of the cylinder relative to the surrounding substance.
This interesting discovery of G. I. TAYLOR, in the special case of a fluid, has been verified
by experiments.
1 Given for the hydrostatic case by GREENHILL [1875. 2, § 22], for the general case by
MATTIOLI [1914, 7, § 2].
2 While equations of the form (207.3) in the case when t is hydrostatic were given by
earlier writers, the tensor (!d: re is often named after REYNOLDS; it was discussed by LORENTZ
[1907, 5, § 11]. 3 [1956. 16]. 4 [1957. 12]. 5 There is reason to believe that all material constitutive equations for incompressible
media must satisfy this requirement; cf. NoLL [1955, 18, §§ 4, 10].
8 This result, for a perfect fluid, and also the result below for the force exerted upon an
immersed cylinder, are due to TAYLOR [1916, 6, pp. 100-104]. Casesofintermediate generality
are treated by DEAN [1954, 4] and }AIN [1955, 15].
Sect. 208. Principal stresses and stress invariants. 551
208. Principal stresses and stress invariants. In the non-polar case, CAUCHY's
secondlaw (205.11} asserts thattissymmetric. Bythe firsttheorem in Sect. App. 37,
it has real proper numbers, called the principal stresses 1, and real orthogonal
principal directions, which define the principal axes of stress. Elements normal
to the principal axes of stress are free of shearing stress, being subject to normal
tension or pressure according to the sign of the corresponding principal stress;
these are extremal values of the normal stresses. The scalar invariants of stress,
It, Ilt, Illt, are symmetric functions of the principal stresses.
When t2=t3=0, t 1=f=O, the state of stress is called simple tension, and the
principal direction corresponding to t1 is called the axis of the tension. In the
engineering literature, when one and only one principal stress vanishes, the
state of stress is often called bi-axial; similarly, if no principal stress vanishes,
the state of stress is tri-axial.
A state of non-vanishing plane symmetric pure shearing stress (Sect. 200)
is called simple shearing stress; in a suitable reetangular Cartesian system
0 txy 0
lltkmJI= 0 0 , txy=f=O; (208.1)
0
an invariant condition is
(208.2)
so that txy = t 1 = - t2 =V-Ilt, and the principal axes of stress lying in the
plane of stress bisect the angle between the elements suffering pure shearing
stress.
In the non-polar case, the condition (204.4) that all stress vectors have equal magnitude
reduces to (t1 ) 2= (t2) 2= (t3) 2, as is obvious.
An exhaustive study of the general state of stress at a point was made by
KLEITZ2• Some of his results, as weil as the fundamental theorem of CouLOMB
and HoPKINS on the maximum shearing stress, have been given in more general
form in Sect. App. 46. In particular, (App. 46.14} shows that the magnitude of the
maximum shearing stress for a pair of directions, one of which is kept fixed as
the other swings perpendicularly about it, is a function only of the differences
of the principal stresses, and hence is independent of the mean pressure. For
the overall maximum and minimum shears, this independence holds a fortiori.
From results of BoussrNESQ3 given in more general form in Sect. 23 it follows that among
the planes through the principal axis corresponding to the principal stress t2 when t1 ;;; t2 ;;; t3
occur
1. The planes across which the magnitude of the stress vector is greatest and least;
2. The planes across which the magnitude of the normal stress is greatest and least;
3. The planes on which the magnitude of the shearing stress is greatest;
4. The planes across which the angle subtended by the stress vector is greatest and least.
Since theorems of this kind are no more than verbal adjustments of theorems
on an arbitrary symmetric matrix, for economy we refer the reader to Sects. App. 3 7
-App. 50. Perhaps the most important function of the present section is to give
warning that these results follow only from the assumption that there are no
extrinsic or mutual couples or couple stresses.
1 The theory is due to FRESNEL (1821-1822) [1868, 5, § 28] [1868, 6, §§ 1, 8-9] [1868,
7, § 1].
2 [1872, 2, § 6] [1873, 4, Chap. II]. 3 [1877. 1. § 2].
552 C. TRUESDELL and R. ToUPIN: The Classical Field Theories. Sect. 209.
209. Geometrical representations of stress. Geometrical representations for
the stress tensor and stress vector may be read off from the results of Sects. 21
to 24. For the non-polar case, the following quadrics were introduced by CAUCHY
and by LAME and CLAPEYRON 1 :
1. The quadric oft, called Cauchy's stress quadric. The normal stress across
any plane through its center is inversely proportional to the squared length of
that radius vector of the quadric which is normal to the plane.
2. The quadric of t-1, called Lame's stress director quadric. The radius vector
from the center to any point of the surface is in the direction of the stress vector
across a plane parallel to the tangent plane at the point.
J. The quadric of t 2, called Cauchy's stress ellipsoid. The central radius vector
in any direction is inversely proportional to the magnitude of the stress vector
across the plane at right angles to that direction.
4. The quadric of t- 2, called Lame's stress ellipsoid. The magnitude of the
stress vector across any plane is proportional to the central perpendicular on
the parallel tangent plane of the ellipsoid.
The asymptotic cone of LAM:E's stress director quadric, called Lame's cone
of shearing stress, is the locus of elements subject to pure shearing stress. When
it is real, it separates the planes across which the normal traction is a tension
from those across which it is a pressure. When the cone is imaginary, the normal
traction across allplanes is a tension or a pressure according as lt>O or lt= ef + ---=c- -k + -----.
vgkk ox r!kq (!km
(209.1)
where the three indices k, m, q are unequal, and where the (!km are the principal radii of
curvature of the surface xk = const. While these equations have limited correctness in three
dimensions, their specialization to plane stress is always valid and often useful. In view
of the results in Sect. 208, in the case when x = 0 and f = 0 these formulae express the
rate of variation of a given principal stress along its trajectory in terms of two principal
shearing stresses. Further results concerning the lines and sheets associated with the stress
field may be read off from the analysis in Sects. App. 47 to App. SO.
1 See Sect. 21 for references. In 1821-1822 FRESNEL [1868, 5, § 28] [1868, 6, §§ 1-7]
[1868, 7, §§ 1-4] constructed but did not publish a theory of elasticity based on postulating
the ellipsoidal distribution of stress and the coincidence of the principal axes of stress and
of infinitesimal strain.
For plane stress, a method of representing both the magnitude and the direction of the
principal stress fields on a single diagram was constructed by TESAR [1933, 11].
2 [1948, 19]. 3 While the existence of these trajectories is weil known, we have been unable to trace
their history or to find Iiterature concerning them except in very special cases. 4 LAME [1841, 4, §XI] [1859. 3, § 149]. Cf. also WARREN [1864, 5], MoRERA [1885, 6, § 3].
Sect. 210. The equations of motion expressed in terms of a reference state. 553
Other trajectories can be associated with a state of stress. Rather arbitrarily, VaLTERRA 1
chose to consider the trajectories of the field of stress vectors across the elements normal
to x. These trajectories, which he called lines of tension, are indeterminate for static problems.
210. The equations of motion expressed in terms of a reference state. We
now give forms of CAUCHY's laws in which the independent positional variable
is not the place ;I! where the stress is experienced but rather is a reference point X
functionally related to ;I! through (66.1) [or (16.2)]. The apparatus of Subchapter BI is at our disposal. Thus far we have considered tkm as a function of ;I!
and t only; for such a tensor field, the definition (App. 20.2) yields tkm,m=tkm;m·
In this section we prefer to consider t as a double tensor field. CAUCHY's laws
(205.2) and (205.11) then assume the forms
(210.1)
Since tkm;m=tkm;MX~m• by substituting (17.8) into (210.1h we obtain 2
(!xk=tkm;Mofloxm;M+efk, (210.2)
where e == (! ]. When the differentiations are carried out, the first term Oll the
right becomes a sum of three J acobians 3 :
TkK = ]tkmXfm,
From (203.4) 2 and (20.8) we have
tfnl da= TkK dAK.
(210.3)
(210.4)
(210.5)
Hence II TkK II gives the stress at ;I! measured per unit area at X; the quantity
TkK is the component along the XK co-ordinate of the component of the stress
vector along the xk Co-ordinate, multiplied by the ratio of the area at ;I! to the
area at X. Thus the quantities TkK, sometimes called pseudo-stresses, are awkward
to interpret. Moreover, in terms of TkK CAUCHY's second law (210.1) 2 assumes
the elaborate form 5
(210.6)
By (210.4) 1 and (18.2) we have
TkK;K = (] Xfm);K tkm + J X~mtkm;K' }
=Jtkm;m• (210.7)
1 [1899, 4]. VaLTERRA calculated the flux of mechanical energy across vector tubes of
this field.
2 BouSSINESQ [1872, 1, §I, Eq. (3)]. The result may be read off from a transformation
given by CLEBSCH [1857. 1, § 2]. Cf. E. and F. CossERAT [1896, 1, § 17]. 3 Due essentially to E. and F. CossERAT [1896, 1, § 43, Eq. (117)]; cf. TRUESDELL [1952,
21, § 26]. The result in the hydrostatic case was given by EuLER [1770, 1, § 119].
4 [1833. 3, Eq. (42)] [1836, 1, Eq. (73)] [1848, 2, ~m 34-38]. Cf. also KIRCHHOFF [1852,
1, pp.763-764], C.NEUMANN [1860, 3, §§2, 4-5]. E. and F.CossERAT [1896, 1, §15,
Eq. (33)]. KIRCHHOFF remarked that the matrix IITkKII is not generally symmetric, as indeed
is manifest from the present notation. Although PoiNCARE [1892, 11, § 40] gave a clear
explanation of this fact, based upon (210.5) and the observation that three orthogonal
elements daa at :r do not in generat arise from three orthogonal elements dAa at X by (20.8),
nevertheless his elaborate, confusing, and unnecessarily restricted manner of stating [ 1892,
10, § 35] that for infinitesimal displacement gradients the matrix II TkKII is approximately
symmetric gave rise to an unnecessary discussion of this "paradox" and an incorrect notion
that it is connected with the presence of initial stress (E. and F. CossERAT [1896, 1, § 26],
COLLINET [1924, 3], }OUGUET [1924, 6]).
• The corresponding form of the more general law (205.10) is obtained by E. and F.
CossERAT [1909. 5, §53].
554 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 211.
Substituting into (210.1h yields an expression1 for CAUCHY's first law in terms
of T"K:
!!x" = PK;K +et".
PIOLA 2 introduced also the tensor TK M defined by
TKM == xK;k T"M = J X5,Xf'!,tkm,
In terms of it, CAUCHY's laws (210.1h 2 assume the forms 3
-""-( k TKM) +-!" TKM=TMK. (!X - X;K ;M (! '
(210.8)
(210.9)
(210.10)
While the expression of the first law is simple in terms of T"K, the second
law is complicated; for TK M, the reverse holds. There is no simple and exact
form 4 with X as independent variable.
211. CAUCHY's laws in space-time; the stress-momentum tensor. The results
of Sect. 205 express the balance of momentum and moment of momentum in an
inertial frame. Apparent forces and torques on a body were calculated in Sect.197,
but for the local equations in a frame which is not inertial it is easier to use the
transformations (143.3) and (143.6). Since neither :il nor i: occurs in (205.10)
or (205.14), we see that the equations for local balance of moment of momentum
are valid for all observers. This follows only because, as emphasized in Sect. 205,
linear momentum is assumed already in balance and the quantities m and l
represent couples only, not torques combined with forces. In the condition
(205.3) we find not:il buti:; however, from (143.3) follows [i:] =[i:'], and hence
the condition for conservation of linear momentum at a surface of discontinuity,
when written in terms of the velocity of propagation of the surface relative to the
material, is valid for all observers. Thus among all the dynamicallaws, only the
differential equation (205.2) for balance of linear momentum is affected by the
apparent forces.
To obtain a form of CAUCHY's first law valid in a rotating frame, we have
only to replace :il by the expression on the right-hand side of (143.6). The result
shows that in order to calculate the balance of linear momentum in an arbitrary
frame, in addition to the metric tensor we must know the linear acceleration b
and the angular velocity w of that frame with respect to an inertial frame.
From the discussion of Sect. 197 it is clear that such dependence on the observer's
motion relative to an inertial frame is unavoidable.
1 PIOLA [1833, 3, Eqs. (22), (29)] [1836, 1, Eq. (56)] [1848, 2, ~ 36, Eq. (26)].
2 [1833, 3, Eq. (45)] [1836, 1, Eq. (132)]. Cf. also KIRCHHOFF [1852, 1, p. 767], E. and F.
CossERAT [1896, 1, § 15, Eq. (31)]. L. BRILLOUIN [1928, 2, § 11] [1925, 2, § 7] [1938, 2,
Chap. X, § X] and RIVAUD [1944, 10] have remarked that (210.9) is a statement that the
quantities TK M and tk m are the components at X and at a: of a tensor density under the deformation (16.2). Since (210.9) is a mere definition, motivated only by the relative simplicity
of some of the resulting forms of CAUCHv's laws, this observation does not seem to have
mechanical significance.
3 PIOLA [1833, 3, Eqs. (33), (35)]. Cf. also SrGNORINI [1930, 5, § 4] [1943, 6, Chap. Il,
§ 4], TOLOTTI [1943, 7].
4 Other forms are given by SIGNORINI [1930, 5, § 5] [1930, 6, § 2], ZELMANOV [1948, 39,
Eq. (4)], and CASTOLDI (1948, 5, § 7]. The form given by DEUKER (1941, 1, Eq. (8.7)] was
shown tobe false by TRUESDELL (1952, 21, § 2612]. The forms given by PLATRIER (1948, 20,
Eq. (13)] and GLEYZAL [1949, 12, Eq. (5)] also seem dubious. Various approximate forms
of (210.10) or (210.8) occur in the literature; e.g., NovozHILov [1948, 18, §§ 21-22]. Explicit
forms for (210.8) in orthogonal curvilinear co-ordinates are worked out by YosHIMURA
[1953, 37].
Sect. 211. CAUCHY's laws in space-time; the stress-momentum tensor. 555
However, it is possible to obtain a more elegant formalism1 for the balance
of momentum by using the ideas and methods of Sect. 152. First we recall that
the stress tensor t was introduced through use of the balance of linear momentum
in an inertial frame; the contravariant components t"m occurring in (203.4)
therefore are defined in all inertial co-ordinate systems. We now introduce the
agreement that what we shall mean by the stress components t"m in any frame
obtained from an inertial frame by a time-preserving transformation (154.8) is
"'k' <> m' t"' m' = _u_x __ u_x- tP q
- oxP oxq • (211.1)
That is, in the terms used in Sect. 152, the components t"m may be regarded
as the non-vanishing components of a space tensor tD.d having the canonical form
(211.2)
in any frame related to an inertial frame by a time-preserving transformation.
The agreement (211.1) is consistent with the intuitive notion that assignable
forces (including the stress vector 2) are invariant under the Euclidean group
of transformations (152.1). The action of one part of the material upon another
is thus assumed to be the same to all observers.
From (211.2) it follows that the quantities T 04 defined by
(211.3)
where e is the mass density, taken as a world scalar, and where v is the world
velocity (153.6), also constitute an absolute contravariant world tensor, which
we shall call the stress-momentum tensor. The name is suggested by the result
(207) 3, since in a Euclidean frame
T"m = t"m _ e i" zm, )
P" = T"4 = - e i".
T44=- f!·
(211.4)
Similarly, we define a space vector F suchthat in every Euclidean frame F n = (jk, 0).
N ow consider the world tensor equation
(211.5)
where V(/) denotes the covariant derivative based on the Galilean connection r. r
(Cf. Sect. 152.) In every Galilean (inertial) frame, (211.5) with .Q =k is equivalent to CAUCHY's first law; the equation that results from putting .Q =4 is the
equation of continuity (156.5h- Thus Eq. (211.5) expresses the balance of mass
and of linear momentum in world-invariant form.
From the world-invariant form (164.2) of the continuity equation and from the
definition (154.1) of the world acceleration vector, it follows that an alternative
form of (211.5) is
(211.6)
1 Special cases have often been noted; e.g. LEVI-CIVITA [1928, 5, pp. 67-81], FINZI
[1934, 2], KILCHEVSKI [1936, 5, §§ 1-2] [1938, 5, Part!,§ 10], PAILLOUX [1947, 11], MANARINI [1948, 13], ARZHANIKH [1952, 1]. Our approach differs basically from that of CARTAN
[1923, 1, §§ 15-17], who makes the connection F!p depend upon the dynamicallaw.
2 This concept is developed and emphasized as a postulate by NoLL [1957. 11, § 9].
556 C. TRUESDELL and R. TouPIN: The classical Field Theories. Sect. 212.
By substituting the components of the Galilean connection ( 154.1 0) into (211.6),
we easily obtain an explicit form for CAUCHY's first law in an arbitrary rotating
and deforming co-ordinate system 1 :
(211.7)
where uk - 0 xk ~~m, t) , ~ = ~ (z, t) being the transformation giving the general
co-ordinates xk in terms of the co-ordinates zk in a Galilean reference frame
(cf. Sect. 154), where xk is the acceleration as apparent to an observer in the
~-system, and where the comma denotes covariant differentiation based upon
the time-dependent metric g (~. t) in the rotating, deforming ~-system.
212. Stress and couple resultants for shells. I. Direct theory. From the dynamical standpoint2, a shell may be regarded in one of two ways: as a surface il,
or as a region between two surfaces il1 and il2 • In both cases, the shell is subject
to normal forces as well as to tangential forces, and therefore its theory is that
of a surface or body in three-dimensional space, not a merely intrinsic theory.
To follow consistently the second view, in which the shell is regarded as a region
in space, we must derive from the general theory of three-dimensional stress the
nature of the forces and couples acting upon the shell. To follow consistently
the first view, we cannot use the momentum principle and the stress principle
as stated in Sect. 200, since the forms given there are appropriate only to bodies
of non-zero volume. Rather, for the first approach we must Postulate new forms
of the stress principle and the momentum principle. In the older work 3 on shells
as models for solid bodies the two approaches are often confused, while theories
of shells as two-dimensional models for soap films, water bells, etc., are so specialized as hardly to be typical. We present the two alternatives independently,
beginning with the first.
Consider a portion of a surface il bounded by a circuit c, and let this surface
be in equilibrium subject to forces F and couples L per unit area. Fand L are
three-dimensional vectors which may point in any direction in space. We Postulate 4 a stress principle for shells : The action of the part of the shell outside c
1 Inability to read Ukrainian prevents us from following in detail the related work of
KILCHEVSKI [1936. 5, §§ 1-4] [1938, 5, Part I, § 10]; like the corresponding result of
McVITTIE [1949, 18, § 4], it seems to rest upon the unnecessary andin general unjustified
assumption that there is a four-dimensional metric. The meteorologicalliterature abounds in
special cases obtained by laborious transformation. 2 Just as the theory of stress in three-dimensional bodies is independent of kinematics,
so also for the dynamics of shells we do not need to mention the theory of deformation given
in Sect. 64. 3 While the dynamical equations are implicit in the pioneer work of LovE [1888, 6, '\[ 8],
he did not disengage them from special elastic hypotheses and approximations, and they
were first given in the forms (212.13), (212.14), (212.20), in special co-ordinates, by LAMB
[1890, 6, § 4]. Later, LovE [1893. 5, § 339] remarked that LAMB's dynamical equations
are valid for finite deformations if referred to the strained shell. Among the numerous repetitions of essentially the same argument as LAMB's we cite only the efficient vectorial
derivation of E. REISSNER [1941, 5, §§ 3-4].
Fundamentally sharper reasoning has been applied in the two special cases when 1. there
are no cross forces or moments, the shell being then called a membrane, or 2. the surface is
plane, the shell being then called a plate. For the former, a geometrically intrinsic theory
is easy; for the latter, the trivial geometry makes a rigorous treatment easy. The history
of membranes and plates, which goes back to the eighteenth century, we make no attempt
to trace; cf. TRUESDELL [1960, 4, §§ 48-49]. 4 Our treatment follows ERICKSEN and TRUESDELL [1958. 1, §§ 24-26], who shortened
the argument of SYNGE and CHIEN [1941, 9, pp. 104-111]. SYNGE and CHIEN were the first
to obtain the equations of equilibrium in the fully generalform (212.6).
Sect. 212. Stress and couple resultants for shells. I. Direct theory. 557
on the part inside is equipollent to a field of stress resultant vectors S(n) and couple
resultant tensors M(n) located on c. The subscript n refers to the unit outward
normal to c; of course, n is a vector defined intrinsically in ~. but S(n) and M(n)
are three-dimensional fields. In analogy to (200.1) and (200.3), a mathematical
expression of this postulate is
~S(n)·ds+ JFda=O,) c d
~(M(nJ+P X S(nJ)ds+ f(L+p X F)da=O, c d
(212.1)
where d s is arc length along c and p is the three-dimensional position veetor.
According to the convention of Sect. App. 13 these vector integrals are understood
to be written in reetangular Cartesian co-ordinates.
The argument in Seet. 203 can be adapted to two dimensions if we replace
the tetrahedron by a curvilinear triangle on ~. The results, analogous to (203.3),
(203.4}, and (203.5) are
S(nJ =-S(-nJ•
Sk _ Sk6n
(n)- 6•
(212.2)
(212.3)
In (212.3) the quantities n6 are covariant components of the unit normal to c
in any curvilinear co-ordinate system vl, v2 on ~- Greek minuscule indices run
from 1 to 2 in this section and the next one. By hypothesis, S(nJ is an absolute
vector and M(n) is an axial vector, for given n; while to derive (212.3) we employed
reetangular Cartesian co-ordinates, the results are tensorial equations within
the scheme of double tensors of Seet. App.15, and hence arevalid in all co-ordinate
systems. The double tensors Sk 6 and Mk~ are the fields of stress resultants and
stress couples. Wehave
dim F = [M L -l r-2], dim L = [M 7-2], }
dimS = [MT-2], dimM = [ML r-a]. (212.4)
There are six components Sk~ and six components MH; the latter we may
sometimes wish to replace by the components of the equivalent skew-symmetric
tensor MkP 6• In the classical treatrnents of the theory of shells the vectors L
and M(n) are assumed tangent to ~; this assumption, which is analogous tothat
defining the non-polar case in three dimensions, reduces the number of nonvanishing components Mk 6 from six to four.
Again we suppose the space co-ordinates reetangular Cartesian, we consider S
and M as funetions of v only, and we substitute (212.3) into (212.1). Byreasoning strictly parallel to that in Sect. 205 we obtain as analogues of (205.2) and
(205.10) the differential equations
SH, 6 +Fk=O,}
MkP6 + z[k SPJ6 + LkP = 0 (212.5)
,d ,d '
where Mkpd and LkP are absolute alternating tensors equivalent to the axial
veetors Mk 6 and Lk. In these equations the subscript comma indicates covariant
differentiation with respeet to the surface metric a, except that z~ ozkjov6,
where z = z(v} is a reetangular Cartesian equation of the surface ~. Now consider the equations
Sk~ +Fk=O,}
MkPIJ;d + x[k; 6 SPl d + LkP = 0 , (212.6)
558 Co TRUESDELL and Ro TouPIN: The Classical Field Theorieso Secto 2120
where the space co-ordinates and the surface Co-ordinates are arbitrary independently selected generat curvilinear systems, where x =x(v) is an equation of ~
referred to these two systems, where x':6 = oxkjov6, and where other occurrences
of the semi-colon indicate the total covariant derivative (Appo 2002) 0 These equations
are in double tensor form; when the space co-ordinates arereetangular Cartesian,
they reduce to (21205), which have been proved valid for such co-ordinate systems;
therefore (21206) are the generat differential equations of equilibrium for shellso
We may continue to regard Sand M as functions of v only, or we may consider
them to be functions of x also, as we pleaseo
The elegant simplicity of this derivation should not conceal the complexity
of the resulto When (21206h is written out, it assumes the form
asH + { k } xf 5m 6 + { ; } 5k6 + p = o
ovd mp ·6 t5; ' (21207)
where {n:p} and {la} are Christofiel symbols based upon the space metric g
and the surface metric a, respectively, the two metrics being related as usual:
a6 E = gkm x':6 x~ o When {21206) 2 is written explicitly in terms of the axial vector
Mk 6, it assumes the form
(21208)
These formulae involve doubly contravariant tensors considered as functions
of v onlyo In terms of physical components, or of components allowed to depend
on x as weil, they would take on still more complicated formso
To assume L and JU tangent to ~ is equivalent to assuming the existence
of surface fields U and Me~ such that
{21209)
To obtain equations of motion instead of equations of equilibrium, we may assign momentum to d and express its rate of change in terms of a surface density Ak, then replace
pk by pk- Ak in (212o6h 0 However, the vector A does not bear any simple relation to the
accelerations of points on d, and we prefer to postpone determining the effect of inertial
force until the general discussion of relations between three-dimensional and surface variables
in Secto 2130
We now resolve all quantities into components normal and tangent to 11:
Fk=F6 x~ +FNk, Lk=L6 +LNk, }
5k6 = 5Yd ~7" +56 N"' Mk6 = MYd.x7" +Md Nk' (212o10)
where N is the unit normal to l1o The following table connects the components
occurring in (212010) with the terms usually employed inshell theory:
F, L = normal components of specific applied force and coupleo
F6, L6 = specific applied force and couple tangent to the shello
56 = cross force resultanto
5Yd = membrane stress resultanto
M6 = cross moment resultanto
M"d = couple resultanto
According to the usual assumptions (21209), L =0 and Md =Oo The normal
and shear components of 5" 6 in an orthogonal co-ordinate system are called
normal and shear membrane stress resultants; the normal and shear components
of MY 6 in such a system are called bending and twisting couple resultants, respectivelyo
Sect. 212. Stress and couple resultants for shells. 559
To express the equations of equilibrium in terms of tangential and normal
componentsl, we use the identities (App. 21.6), (App. 21.2), and (App. 21.4) 2 to
obtain from (212.1 0) 3 the following resolution:
5k6;6 = (5Y~;ß- aar ba6 56) x7r + (5f6 + br6 5Y 6) Nk. (212.11)
Substituting this result and (212.10) 1 into (212.6)1 yields
(5Y6;6- aarbad 5 6 + F6) X~y + (56;6 + by6 5Y~ + F) Nk = 0. (212.12)
Taking the scalar product of this equation by N yields as the condition for
equilibrium of normal forces
5°;o+byo5r~+F=O; (212.13)
taking the vector product by N, the condition for equilibrium of tangential
forces:
5Y0;0 -araba656+FY=O. (212.14)
Similar resolution of (212.6) 2 yields as the condition for equilibrium of bending
moments 2
(212.15)
for twisting moments,
MYO;o- ara ba6 Mo+ aro e~a 5a + LY = 0. (212.16)
In these formulae, it is legitimate and natural to regard all fields as functions
of v only and to interpret "; r5" as covariant differentiation based on a. Under
the usual assumption L = 0, M 0 = 0, the total number of independent components
of S and M is reduced from 12 to 10, the second term in (212.16) vanishes, while
(212. t 5) reduces to an algebraic equation expressing the difference of shear
resultants 5[121 as a linear combination of the four couple resultants Mr~.
In works on shell theory it is customary to use in place of M the dual tensor B,
defined as follows:
(212.17)
so that the physical components of these tensors in an orthogonal co-ordinate
system satisfy the relations
M=-B<21), M<12>=-B(22), M(21)=ß, M(22)=ß(12). (212.18)
Eqs. (212.15) and (212.16) representing the balance of moments may be expressed
in terms of B:
M;oo_ erob~ Boa+ ero5ro + L = 0, }
Bvo;o- ava.ea.r b~Mo- 5" + C = 'o, (212.19)
where C=ava.ea.rLY; when the classical assumption (212.9) is adopted, these
equations reduce to
(212.20)
the former of which is especially interesting because it shows the tensors S and B
to be non-symmetric except in special circumstances.
1 This resolutionwas effected by SYNGE and CHIEN [1941, 9, p. 109] by use of the special
co-ordinate system we explain in Sect. 213. 2 The fully general equations (212.15) and (212.16) were given by E. and F. CosSERAT
[1909, 5, §§ 35-37]. who derived also forms in material co-ordinates. Cf. also HEUN [1913,
4, § 20].
560 C. TRUESDELL and R. TouPrN: The Classical Field Theories. Sect. 213.
The dynamics of shells is not a mere two-dimensional analogue of the dynamics of threedimensional bodies (cf. Sect. 238). As is evident from the ideas used to formulate it and
from the occurrence of the second fundamental form b in every one of its equations, shell
theory concerns properties of two-dimensional idealizations of three-dimensional bodies.
For an intrinsic analogue to the three-dimensional case, we should have to have a tensor p/; 6
satisfying
(212.21)
where G is assigned. From (212.14), we see that an equation of this form emerges from
shell theory if the cross force SY is known. In particular, when SY = 0 we obtain from
(212.14) and (212.13) the equations
St; 6,6+Ft;=O, b~; t;+F=O. (212.22)
When, in addition, the couple resultants and assigned couples are zero, (212.15) reduces to
(212.23)
Eqs. (212.22) and (212.23) are said to describe a state of membrane stress. The four
membrane stress resultants satisfy a system of two linear partial differential equations and
two linear algebraic equations with coefficients which are determined by the surface 6. There
exists an extensive theory of integration of this determined system. See also Sect. 229.
213. Stress and couple resultants for shells. II. Derivation from three-dimensional theory. If we choose to regard a shell as a portion of material between
.... .... 4 ........ .... ' Fig. 37. Shell regarded as a three-dimensional body.
two surfaces !l 1 and !l2 , the theory
of equilibrium and motion of a shell
is derivable as a consequence of the
three-dimensional theory. According
to this view, the stress and couple
resultants, instead of being introduced through a postulated twodimensional stress principle as in
Sect. 212, should be defined in terms of the three-dimensional stress tensm: t.
Moreover, a shell need not now be a body by itself: All results we are now going
to derive hold equally for a shell which is but a part of a three-dimensional body,
though this interpretation is unlikely to be useful. ·
The resultants Sk 6 and MM are defined by the condition that their action
upon a curve c lying on a reference surface !l shall be equipollent to the action
of the three-dimensional stress tensor tkm upon a finite surface h ( c) intersecting !l
along c. E.g.,
(213.1)
where n* is the unit outward normal to h ( c), and where our usual convention regarding integrals of vectors in curvilinear co-ordinates is understood (Sect. App. 17).
For each curve c, the surface h ( c) is to be fixed once and for all, subject to the
understanding that to a curve which is a part of c there corresponds a surface
which is apart of h. The requirement (213.1) then defines SH uniquely.
In the practice of shell theory, h is always taken as a surface swept out by
the normals to !l along c, and it is always assumed that the surfaces !l1 and !lz
are given by equations x0 = h1 (v) and xD = h2 (v), where x0 is the normal distance
from !land where v1, '1!8 are curvilinear co-ordinates upon !l (Fig. 37). The quantity
1 h2 - h1 1 is then the thickness of the shell at the point v on !l. In many applications the two surfaces are supposed given by the equations xD = ± h (v); in this case
!l is called the middle surface of the shell. For the general theory, however, no
such restriction is necessary, and the reference surface !l need not even lie within
the shell.
Sect. 213. Stress and couple resultants for shells. 11. Derived theory. 561
It is natural to use a co-ordinate system1 in which one family of co-ordinate
surfaces consists in the surfaces x" = const, which are parallel to .1. If a (v) and
b (v) are the fundamental tensors of .1, then the spatial metric g (v, x0) assumes
the form
goo =t' 0 = 1, goa: =t'"' = 0, }
ga:ß = aa:ß + 2x0 ba:ß + (x0) 2 ba:r b;;
(213.2)
i.e., the superficial components of g are of the form g = a · (1 + x0 b) 2• In most
of the older researches, the Iines of curvature on .1 are chosen as co-ordinate curves, so that
gn=an1+x ( OK )2 1 - ( 0 K )2- 1 ( ) 1 =11· g12=0, g22=a221+x 2 -22, 213.3
g g
where K 1 and K 2 are the principal curvatures of .1. From (213.2) we have
where 2
and also
{oko} = {k0o} = 0 •
{cx0
ß} = [O, since the parameters v, x 0 are not admissible as co-ordinates if (1 + x° K1) ~ 0
or (1 + x°K )~o.
Handbuch der Physik, Bd. 111/1. 36
562 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 213.
tensors are the quantities usually employed in shell theory; e.g., if we make the
special choice of co-ordinates leading to (213.3), we have 1
hz
5(11) = «n 5 11 = «n J t11 (1 + x° KJ) 2 (1 + x° K2) dxO'
hJ
ht~
= J 11 (1 + x° K2 ) dx0 ,
hJ
h2--..
5(12) = J 21 (1 + x° K1) dx0 ,
hl
hz~
5(21)= J 12(1 +x°K2)dx0 ,
hl
hz~
5(1) =- f01(1 +x°K2)dx0 ,
h,
hz ~
ß(ll)=- J x0 11(1 +x°K2 )dx0 ,
hJ
hz ~
ß(I2)= - J x0 21 (1 + x° K1) dx0 ,
kJ
etc., where, as usual, k~ is the matrix of physical components oft.
(213 .8)
We now find the reflection of CAUCHY's second law, in its narrower form
(205 .11), upon the stress resultants and couples. Integrating the algebraic identity
(213.9)
across the shell and expressing the result in terms of the definitions (213.7),
we obtain a formula identical to (212.20) 1 . This should not be surprising, since
CAUCHY's second law is itself a condition for the balance of moments. The interesting thing about this result is its indication that the classical assumptions (212.9)
of shell theory are consistent with the three-dimensional non-polar case, for it is the
absence of three-dimensional applied couples and couple-stress that CAUCHY's
second law in the narrower form (205.11) asserts.
We now derive Eqs. (212.13) and (212.14) for a shell by integrating the
corresponding components of CAUCHY's first law (205.2), thus showing that the
balance of momentum for a shell as a whole is a consequence of the balance of
momentum of the three-dimensional body with which we identify it, as indeed
is physically plain 2• Since there are some formal difficulties in using general
co-ordinates on the surface ~' weshall use the special metric components (213.3),
one advantage of which is the form assumed by the Mainardi-Codazzi identities,
VIZ.,
8Vgu - 0 [11-( OK )] - ( OK) o}'all ----a;2- - [}V2 yall 1 + X 1 - 1 + X 2 ov2 ' (213.10)
and a similar formula obtained by interchanging 1 and 2, 1 and 2.
1 The curvature factors were mentioned by LAMB [1890, 6, § 2] and were used in special
cases by BASSET [1890, 1, §§ 5, 18]; the full set (213.8) was given by LovE [1893, 5, § 399],
and the general equations (213.7) were formulated by GREEN and ZERNA [1950, 9, § 3].
For discussion of the mechanical significance of the resultants, cf. ZERNA [1949, 39, §§ 3-4].
Definitions not obviously equivalent to these were given by KILCHEVSKI [1938, 5, Part II, § 3]. 2 Üur derivation follows that of NoVOZHILOV [1943, 4] and of NOVOZHILOV and FINKELSTEIN [ 1943, 6, § § 1, 4 ], which is more general than that given independently by TRUESDELL
[1945, 6, § 8]. Derivations in general co-ordinates have been sketched by CHIEN [1948, 7] and
GREEN and ZERNA [1950, 9, § 3].
Secto 2130 Stress and couple resultants for shellso Ho Derived theoryo 563
As the normal component of CAUCHY's first law (205 o2), from (205 Ao1) we have
,, \1 { <>81 cva22 (1 + X° K2) 01] + -:.8 fall yaz2 vV uV -2 [yall (1 + X° KJ) ozJ} -l - K 1 (1 + x° K 2) f.i- K2 (1 + x0 K 1) 22 + (213°11)
+ a~o (K* 00) + (! K* (f - x) = 0 0
Integrating this equation from xO =h1 to x0 =h2 and taking account of (21308)
yields
lau , __ 1l'c.={ a 22 uv
<>81 (ya;; 5<1>) + uv
<>82 (yall 5<2>)} + l
+ K1 5<11> + K2 5(22) +F = 0,
providing we put h2 ~ h
F = - f (! K* (f - x) oxO - K* 00 I 2 0
~ ~
(213012)
(213°13)
The result (213012) is identical in form with (212013), in the co-ordinates employed
andin terms of physical componentso
As the tangential component of CAUCHY's first law (20502) corresponding to vl,
from (205Ao1) we have
, \,- {-;~dVa22 (1 + xO K2) 11] + -<>8 2 [Va11 (1 + x° K1) 12J} +
tau ya22 vV uV
+ 1 8 logV"ll (1 + xO K) 21- ~ 8 logj'a;; (1 + xO K) 22 + (213°14) Va22 8v2 2 Vau 8vl I
+ Kl (1 + x° K:~) 01 + 8~0 (K*01) + (! K* (/(1)- x(l)) = 0,
where (213o10) has been usedo lntegrating this equation from xO =h1 to xO =h2
and taking account of (21308) yields
(213°15)
provided we put
(213°16)
The result (213015) is identical in form with (212014), in the co-ordinates employed
and in terms of physical components, when y = 1o
To complete the identification of the results obtained by integration of the
three-dimensional equations with those of the direct shell theory of Sect. 212,
we need to replace (213 013) and (213 016) by invariant formulae valid in general
co-ordinates on the surface ~, as follows:
36*
564 C. TRUESDELL and R. ToUPIN: The Classical Field Theories. Sect. 214.
Thus far, except in deriving (213.9) and its direct consequence (212.20)1, we
have not assumed the three-dimensional stress tensor to be symmetric. To obtain
the equation of moments, however, we rest content with the non-polar case and
simply multiply the three-dimensional equation (213.14) by x0, then integrate
across the thickness of the shell1• Thus follows
l' a11 1v~{ a22 uv "o I(ya;; B) + uv "o2 (~B)} + l
+ 1 o log Vau B< 1 o log ya;; B< -v- n>- -~- "vl 22)- a22 ov2 lall u
- S<1> + C<1> = o,
(213.18)
where, in general co-ordinates upon 6, we have put
C"' =-f2
e K* xo (c5~ + xO b~) (f -xY) dx0 - K* xO (c5~ + xO b~) t0Y1::. hl
(213.19)
The result (213.18) is identical in form with (212.20)2 , in the co-ordinates employed
and in terms of physical components, when ct = 1.
In comparing the results of this section with those in the preceding one, we
must understand that it is impossible to prove that the quantities entering the
two systems are identical, since the difference of basic assumptions and definitions in the two cases makes a statement of isomorphism the best that can be
hoped for. What we have shown isthat no error can result if we choose to regard
the surface fields S"'fl, SY, and B6 •, defined in terms of the three-dimensional
stress tensor t by (213.7), as equivalent to the fields denoted by the same symbols
in Sect. 212, where they were defined in terms of the double tensors Sand M,
introduced a priori.
If we accept this identification, then the results of this section show how the
equilibrium theory of the previous section can be generalized to the case of motion.
In (213.17) and (213.19) appears not only the assigned forces f but also the acceleration ~. From (213.17) we see that the effective force of inertia, per unit area
and surface mass on 6, is not necessarily the acceleration of any particle on 6;
rather, at a given point P on 6 it is a certain weighted mean of the accelerations
at all points on the normal to 6 through P. Moreover, inertial forces occur also
in (213.19) and hence affect the balance of moment of momentum-an unusual
phenomenon in mechanics. Finally, even in the static case the effective surface
Ioads F and C which enter the equations of equilibrium for shells are not merely
the vector differences of the Ioads in the interior, but rather areweighted averages
and differences, influenced by the thickness and the curvature of the shell aswell
as by the forces and couples applied.
214. Stress and couple resultants for rods 2• If we consider a rod simply as
a curve c which may be the seat of dynamical actions, by considerations anal1 In the general case, the definitions (213.7)3 are no Ionger adequate, since additional
resultant couples are produced by the couple stress m. Also, instead of simply multiplying
the equations of linear momentum by 0 and then integrating, we should integrate (205.10). 2 For the plane case, the stress principle for rods and the appropriate special cases of
(214.1), (214.2) and (214.7). independent of any hypothesis regarding the constitution of the
material, were first given by EuLER [1771, 2, §§ 1-11, 35-40] [1776, 4, § 17]. For the history
of the earlier special theories of rods and flexible lines by PARDIES, ]AMES BERNOULLI, and
others cf. TRUESDELL [1959, 8, §§ 2-3, 7-14, 20-21, 25).
ST. VENANT [1843, 3,, 3] was the first toremarkthat six equations are needed to express
the equilibrium of rods which are twisted as weil as bent, but he did not succeed in obtaining
them without special simplifying hypotheses. The general equations were given in principle,
but very obscurely, by KIRCHHOFF [1859, 2, § 3), explicitly by CLEBSCH [1862, 2, §50). These
Sect. 214. Stress and couple resultants for rods. 565
ogous to those at the beginning of Sect. 212 we are led to postulate a stress
principle for rods: At each point on a rod, the action of the material to one
side upon the material to the other is equipollent to that of a stress resultant
vector Sand a couple resultant M. These quantities have the physical dimensions
[M L T-2] and [M L 2 T- 2], respectively. Properly, we should define them as
acting on the opposite sides, + and -, of a cut through the rod; then analogously
to (212.2) and (212.3), (203.3) we have
(214.1)
as the first consequences of the principle of equilibrium. Dropping the subscripts + and - but adopting an appropriate convention of sign, as the definitive condition of equilibrium we obtain
(214.2)
where "~" is the dual of the intrinsic derivative defined by (63.6), where p is the
position vector with respect to a fixed origin, and where Fand L are the applied force
and couple, per unit length. By substituting (214.2)1 into (214.2) 2 we obtain
M+txS+L=O, (214.3)
where t is the unit tangent to the rod c.
A rod such that M = 0 if L = 0 is said to be perfectly flexible; such rods are
often called strings. By (214.3), a necessary and sufficient condition for perfect
flexibility is that the stress resultant S always be tangent to the rod.
Since the two Eqs. (214.2)1 and (214.3) are in vectorial form, they are valid
in an arbitrary curvilinear Co-ordinate system. It is customary, however, to
refer them to a particular frame defined with respect to the rod c. Retaining
full generality at the start, in the scheme of Sect. 61 let us assign any three linearly independent directors and reciprocal directors da and da to c. With Sk
and Fk as the contravariant components of S and F, ~ and Lk as the covariant
components of M and L, in general curvilinear Co-ordinates, we define corresponding anholonomic components:
sa = d~Sk,
P=d~Fk,
Ma = d~Mk,}
La- d~Lk.
By (214.2)1 and the result dual 1 to the reciprocal of (63.10) 3 , we have
!d~a_ = sa = d~ Sk + J~ Sk' l =- d~Fk- d~wmk Sk,
(214.4)
(214.5)
and other early treatments are difficult to follow, sometimes imparting the impression that
some approximation is made. E.g. LovE [1906, 5, § 254] says "the extension of the central
line may be disregarded". In fact, as was noted by BASSET [1895, 1, § 2] (cf. also [1892,
1, § 4]), Eqs. (214.7), analogaus to CAUCHY's laws, are exact when referred to the actnal
position of the rod; just as in the three-dimensional theory, no question of approximation
appears unless we attempt to refer the equations to a configuration assumed by the rod
prior to its being loaded by the forces under which it is in equilibrium. The derivation given
in the text is that of ERICKSEN and TRUESDELL [1958, 1, §§ 21-23], patterned on earlier
work of E. and F. CossERAT [1909, 5, § 10] and HEUN [1913, 4, § 19]. 1 Note that w is not the F of Sect. 63 but rather the dual of W; neither is it tobe confused
with the vorticity vector, which is denoted by the same kerne! index.
566 C. TRUESDELL and R. ToUPIN: The Classical Field Theories. Sect. 214.
where w is the wryness of the directors along c. Similarly
dMa - M~ - - dk ( tP Sq + L ) + dm k M. -dS- a- a ekpq k a W m k' (214.6)
Hence the statical equations in anholonomic components are 1
dsa + wa Sb + pa = 0 ) ds b · '
dMa b b c _ ds - W a Mb + e0 b c t S + La - 0.
(214.7)
Thus far the director frame has been arbitrary. We now require it to be a
unit orthogonal triad suchthat d1 =t, the unit tangent. By the dual of (63.9) 3 ,
the wryness w then satisfies Wab =-wba and may be interpreted as an angular
velocity 2• The component 5 1, which is the projection of S onto the tangent
to c, is called the specific tension in the rod; the components 5 2 and 5 3, the
specific shearing forces; M 1, the specific twisting couple; M 2 and M 3, the specific
bending couples. This special choice of directors, while not simplifying (214.7) 1 ,
implies that t1 = 1, 2 =t3 =0 and hence reduces the three components of (214.7) 2
to the following explicit forms:
(214.8)
In analogy to the reasoning in Sect. 213, it should be possible to derive the equations
for rods by integrating the three-dimensional equations or the equations for shells; using
power series expansions, GREEN 3 has given a derivation of the former type.
If F = 0, from (214.2h it follows that S = const; (214.3) then becomes a statement that
M is a prescribed function of s. Since M is an axial vector, such a statement is a formal
analogue of (196.3), the general condition for balance of moment of momentum, provided s
and I are made to correspond. Such a correspondence may be carried further by taking the
director frame of the rod as an orthogonal unit triad, so that the wryness w becomes the
analogue of the angular velocity m. This observation forms the basis of KIRCHHOFF's
1 Since by definition
we have
eabc = + Vi cabc det d~. Now
hence
g(det d~)2 = det gk m d~ df' = det ge(
eabc = ± Ydetgef 10abc•
where the sign is to be selected so as to agree with that of det d~. The quantity detgef is
evaluated by the dual of (61.4). In particular, for a right-handed unit orthogonal triad we
have
eabc = 10abc·
2 The classical notation for the component w2 a is T or - r; the other two independent
components are written as ±" and ± "'. lt is important to remernher that in the exact
theory all these quantities refer to the loaded rod. Cf. the footnotes on p. 565.
3 [1959. 7]. Various earlier authors, e.g. KIRCHHOFF [1876, 2, Vorl. 28, § 5] and LovE
[ 1906, 5, § 254 ], had given definitions of the stress resultants and couple resultants in terms
of three-dimensional stresses, but their subsequent arguments rest on unnecessary and unrigorous limit processes rather than exact integration such asthat given for shells in Sect. 213.
The method of power series expansionwas initiated by HAY [1942, 8, § 6].
Sect. 215. Partial stresses in a heterogeneaus medium. 567
celebrated analogy between the motion of a rigid body and the deflection of an elastic rod 1•
The result as usually presented takes on an appearance of greater complexity because (214.8)
rather than (214.3) is used as the starting point.
215. Partial stresses in a heterogeneaus medium 2• To discuss the transfer of
momentum in a mixture, we employ the formalism of Sects. 158 to 159. Each
constituent ~ is regarded .as being subject to partial stress t is normal, the third condition
for both theorems is satisfied in virtue of (69.2). Also, it is easy to formulate
assumptions under which the surface integral in (218.4) or (218.9) may be expressed as the time derivative of another surface integral. For example, if we
apply the stress boundary condition (203.6), assume that the surface Ioad s
1 Suggested by the somewhat vague analysis of GREEN [1839, 1, pp. 248-250] and KELVIN [1855. 4, § 187], given in essentially the above form by LovE [1906, 5, § 125]. 2 HADAMARD [1903, 11, ~ 265].
3 Due in principle to HELMHOLTZ [1858, 1, § 4], though it may be traced back to D'ALEMBERT in restricted cases.
Sect. 219. The virial theorem. 573
satisfies sk=- Bk where B = B(x); then when .9 is stationary we have
p t(n)kxkda = p skxkda = -lh, where 5B = p B da. {218.15)
f/' f/' f/'
In this case, from (218.4) follows
~ + Ü + ~ =-J P dv. (218.16)
r
We leave it to the reader to formulate conservation theorems appropriate to
this case.
While the requirements 1 and 1~ may always be satisfied trivially, and while
there are many cases where requirement 3 is relevant, the other requirements
are satisfied only in restricted elastic and hydrodynamic situations. Our purpose in giving these theorems here is to make it clear that such a conservation
law as (218.10) is not to be expected in any typical situation in continuum
mechanics, where dissipation of energy is the rule, not the exception.
219. The virial theorem1 . Put
IDmk==fzmikdm, 2S'rmk=fzmzkdm, (219.1)
r r
the quantity- 2S'rmk being the total apparent stress due to transfer of momentum
(Sect. 207). Then (216.4) may be written
{219.2)
The skew-symmetric part of this equation was considered in Sect. 216. To interpret the symmetric part, notice that
W(mk) = t ~mk•
where ij is EULER's tensor (168.4) 1 with a =0. From (219.2) follows
t (i:mk = 2S'rmk + P Z(m tk)q daq + J (e Z(m fk)- t(km)) dv,
f/' r
the trace of this equation being 2
(219.3)
(219.4)
(219.5)
where we write G: for G:k k, the polar moment of inertia of the body about the
origin, where S'r is the kinetic energy and where p is the mean pressure {204.7).
The typical application of these results is to obtain time means by integration
with respect to t, often under the added assumption that certain terms are
periodic.
1 The quantity 1: Fa· Pa was introduced into statics by MöBius [1837, 3, § 123] and a studied by ScHWEINS [1849, 2] [1854, 2], who called it the "Fliehmoment" of the forces.
Its introduction in the dynamics of mass-points is due to }AcOBI [1837, 2, § 6] [1866, 2,
Vierte Vorl.]. Cf. also LIPSCHITZ [1866, 3] [1872, 3], CLAUSIUS [1870, 1], VILLARCEAU [1872,
4]. Herewe follow a generalization due to FINGER [1897, 3, §I] and elaborated by PARKER
[1954, 18, § § 1, 3]. Cf. also the hint of MAXWELL [1874, 2, p. 410]. 2 CISOTTI [1923, 2, § 3] [1940, 6, § 6] [1942, 3].
574 C. TRUESDELL and R. TOUPIN: The Classica! Field Theories. Sect. 220.
220. SIGNORINI's theory of stress means. A more useful application of FINGER's
virial formula (216.4) has been found by SIGNORINI 1• Set
'oakm-PZmtkqdaq- fzm("ik-fk)dim, (220.1) d V
where 'o is the volume of v. If we use a superposed bar to denote a mean
value over the body, (216.4) is equivalent to
{220.2)
In the static case, the quantities ak m may be calculated from the applied loads
only, so that (220.2) furnishes the mean stresses directly in terms of known quantities.
Simple as is the reasoning used to derive (220.2), the result is important: While
CAUCHY's first law (205.2) in itself constitutes an underdetermined system and
is thus insufficient to yield unique values for the stresses, its corollary (220.2)
- T - T
Fig. 38. Body subject to intemal
and extemal normal pressure.
Fig. 39. Body subject to tensile Ioad.
enables us to calculate the mean stress uniquely and independently of the physical
constitution of the material.
We now give SIGNORINI's examples, all for the static case.
1. Torque. If m=O on ~ and l=O in v, a[kml is the torque acting on v. If
a[kmJ=O, (220.2) yields tkm=tmk· This is consistent with CAUCHY's second law
(205.11), which holds under the stronger assumption that m =0 throughout v.
If akm = 0, the load on v is said to be astatic. From (220.2), a necessary and sufficient condition for astatic load on v is
(220.3)
2. Hydrostatic pressure. Suppose v is the region between a surface ~o subject
to hydrostatic pressure Po and a surface ~; subject to hydrostatic pressure P;
(Fig. 38), and suppose f = 0. Then if we write c for the volume of the cavity,
from (220.2) follows
(220.4)
This shows that hydrostatic loading always gives rise to a stress system which is
hydrostatic in mean 2• Moreover, if Po~P; and Po>O, the mean normal stress
is a pressure.
3· Tensile loading. Let a body be subject to two equilibrated concentrated
forces T and -T, acting at points a distance L apart (Fig. 39). Choosing the
1 [1932, 13, §§ 1-2]. In [1939, 11], SIGNORINI applied these results to the motion of
rigid bodies.
2 NARDINI [1952, 14] has shown that when a body is subject to equalloads applied at the
vertices of a regular polyhedron and directed toward its center, the mean stress is hydrostatic. He has obtained a similar result for Ioads applied at the vertices of a regular polygon.
Sect. 220. SIGNORINI's theory of stress means. 575
axis of z1 parallel to T, from (220.2) we get
- LT
tn = -b-, (220.5)
while all other tk m vanish. Thus tensile loading gives rise to simple tension in mean.
4. Body rotating steadily about a principal axis of inertia. Consider a body
in steady rotation at angular speed w about the z1-axis, so that z1 =0, z2= -w2 z2 ,
z3 = -w2z3 • In order that the force and torque acting on the body be zero,
we must have
[zkdm =O l
and J zl zk diDl = 0, (220.6)
k =2,3;
that is, the axis must pass
through the center of mass and
coincide with a principal axis
of inertia. From (220.2) we get
t _ ~kk ) kk- b ' (220.7)
k = 2, 3 (unsummed),
where Cfkm is EULER's tensor
(168.4h with a=O, and all Fig.40. Heavysolidonanaxle.
other tk m vanish.
5. Heavy solid on an axle 1• Consider a heavy solid on an axle which makes
an angle 'P with the horizontal, being supported by a hinge at one end, a bearing
at the other (Fig. 40). Choose the zcaxis along the axle, the z2-axis normal to
it, the origin at the hinge, and the plane of zcz2 vertical. The loads are equilibrated; the reaction S of the hinge, being located at the origin, makes no contribution to (220.1); the reaction B of the bearing, being directed along z2 at a
point where z2=z3=0, contributes only to a21 , but need not be mentioned in
the non-polar case, since then a21 = a12 , and ll:t 2 can be calculated without knowledge of B. By (165.1), from (220.2) we calculate
tJ t11 =- jffi c1 sin 1p, tJ t12 =- jffi c2 sin 'P = tJ t21 , tJ t22 = jffi c2 cos 'P, (220.8)
where c is the vector from the hinge to the center of mass and jffi is the weight
of the body and axle. All other mean stresses vanish. For the mean value of
the mean pressure p, we have
(220.9)
where h is the height of c above the hinge. Thus p is a pressure or a tension
according as c is above or below the hinge. When the axle is horizontal and c2 =f= 0,
t2 2 is the only stress which does not vanish in mean; when the axle is vertical
and the center of mass lies upon it, we have the case of a body balanced upon
a single point, and the resulting expression for t11 , the only non-vanishing mean,
agrees with (220.5).
1 This example and the next are due to TEDONE [1942, 13, §§ 2-3].
576 C. TRUESDELL and R. TOUPIN: The Classical Field Theories. Sect. 221.
6. Pendulous body. Consider a body swinging rigidly about a fixed point
(Fig. 41) located at distance L from its center of mass. So that the motion be
possible, we assume the support pin is parallel to a principal axis of inertia,
while the motion takes place in a plane containing the other two; we take these
axes as co-ordinate axes. Since the co-ordinates of the center of mass are c1 =L,
c2=0, the only contribution made by the weight Q9 to the _,...".Zz
- bak". is made to ba11 and is of amount ~L cos cp, where cp is
the angle of inclination to the vertical. From (143.6), for
the components of acceleration in an inertial frame instantaneou~lycoincidingwith the co-ordinate frame we get- ~ z2,
-~ +~z , 0. From (220.2) then follows
bt11 _ ~L coscp + 11 • •• bt22 22 ,} (220_10)
bt12 - cp 22 • bt21 -- cp (~11 • t3k- 0.
Fig. 41. Pendulous body.
That t12 =I= t21 except when 11 = 2 2 = 0 or when the angular
acceleration vanishes is because a pendulous body not devoid
of rotary inertia must be provided with a restraining torque
from the support if it is to execute an accelerated rotation.
221. Moments of stress. Setting
bbpqr= ~zpzqt,".da".- J zpzq(z,-f,) diDl = ~bqpr•
f "
in (216.2) we put \II =ZpZq and obtain1
Therefore
t,p Zq + t,q Zp = bpqr•
------1
t(pq) z, + t[pr] Zq + t[qr] Zp = 2 (bqrp + bprq- hpqr} = Cpqr•
(221.1)
(221.2)
(221.3)
Henceforth we consider only the non-polar case. From (221.3) follows then
(221.4)
so that the second moments of the load determine the mean values of all the
first moments of the stresses.
Since bpqr=c,pq+c,qp in the non-polar case, the first moments of the stress
determine the second moments of the Ioad. It is impossible that this one-to-one
correspondence between moments of the Ioad and moments of the stress can
continue indefinitely, for if it did, the stress would be determinate from the Ioad,
while in fact CAUCHY's laws form an underdetermined system. lndeed, we
have tpq=apq from CAUCHY's first law alone; to obtain (221.4), we use the second
law as weil, reducing the number of independent stress moments tpq z, from 27
to 18, the number of independent Ioad moments. For higher moments, the
number of independent Ioad moments of a given order is much less than the
number of independent stress moments of the next order, and thus the full set
of Ioad moments is insufficient to determine all the stress moments.
To calculate the higher moments 2, set
tJ b!:'bc == ~~ z~ 4 t,".da".- f ~ z~ 4 (z, -/,) diDL (221.5) f "
1 SIGNORlNI [1933, 10, § 2]. A special study of the bpq is made in [1932, 12]. 2 GRIOLI [1953, 12, § 1] [1952, 8, § 1].
Sect. 222. Estimates for the maximum stress. 577
The bn, for which a + li + c = n are the load moments of ordern, being-! (n + 1) x
(n+2) in number; for n=1 they reduce to the a,p; for n=2, to the bpqr· By
putting W =z~z~z~ in (216.2), we obtain an equation satisfied by the )n(n+1)
stress moments of order n:
at, z~ zgz~+lit, z~z~ z~+ct, z~zgz~ =li~~'' (221.6)
on the understanding that any term in which the exponent " - 1 " appears is
to be annulled. From the choice a = n, li = c = 0, we get
t,sz~- 1 = ~ Pi'~ (s unsummed, n ~ 1), (221.7)
where p'f~ denotes b~~c with exponent n for zs and with the other two exponents
taken as 0. From the choice a = n-1, li = 1, c =0, we get
(n- 1) t,k z~ 2 zs + t,sz~- 1 = qJ:Jn (k unsummed), (221.8)
where qJ:J n denotes b~~, with n -1 for the exponent of zk, 1 for the exponent of zs,
and 0 for the third exponent. When k =r, (221.8) and (221.7} yield in the nonpolar case
t zn 2 z = - 1
- [q''l - ..!_ prsl] (r unsummed) . " , s n- 1 rsn n rn (221.9)
From (221.7) and (221.9) we see that when n ~ 3, the values of the load moments
of order n determine unique values for at least 15 of the stress moments of order n
in the non-polar case.
222. Estimates for the maximum stress. SrGNORINI was the first to observe
that a lower bound for the maximum stress is determined by the loading. Here
we present generalizations and extensions of his work by GRIOLI 1•
Considering only the non-polar case, write
ti-tn, t2 = t22• t3 = taa. }
t4=t2a=ta2• t,=ta1=t13• t6=t12=t21;
(222.1)
let Q'l1, W. = 0, 1, ... , m be a set of functions orthogonal over v, with mean norms
W1~ gi ven by b W1~ = f Qfu d v ; let kb, , li, c = 1, 2, ... , 6 be the symmetric coeffi- V
cients of any constant positive semi-definite form; and let Cb 21 , li=1, 2, ... , 6,
W. = 0, 1, ... , m, be any constants. Then
0~~ J L kb,(tb-Cb 21 Q~ )(t,-C,!BQ!B)dv,) V b, m:,~
= I. kb, tbt, + I. kb, cb'll (W?fu c,'ll- 2 Q'l1 t,) . b, c b, c, ~~
(222.2)
Choosing the constants C, 21 so that
c,'ll = Q2>"\' ~(lb lbmax=lb=L_, 2
~( rnl'll (222.15)
Other estimates for the maximum stress 1 follow immediately from (221.7) and (221.9):
lBIP1')1
l
1
rslmax ~ nJiz.r s~ dv' I
m I q\!A- PrMn I (222"16)·
ltrrlmax ~ ( n-1 )JI 2 1 · z~ z5 dv V
From the identity
L: km ~~ rs max = -=L;=----Jcc-l-k~~Qi8l dv •
!llv
(222.18}
GRIOLI 2 has shown that in each case there exists a particular choice of the constants k~(
rendering the bound (222.18) sharper than (222.15). In fact, with r and s held fixed, put
k
the means and first moments of the stress, and vice-versa. By (223.2), the conditions (223.1} become
(223 .4)
The former of these is equivalent to
(223.5)
Conversely, if (223.5) and (223.4) 2 are satisfied, the linear stress defined by (223.3)
and (223.2) will be a null stress such that the stress vector on 6 iss. Thus the
conditions (223.4} 2 and (223.5), along with the condition e(z -I) =const, constitute necessary and sufficient conditions to be satisfied by the loads in order that
a linear solution to Cauchy's laws, with a prescribed load on the boundary, may exist.
In Sect. 208 we have derived a condition that all stress vectors at a point have the same
magnitude. The hydrostatic special case, t = - p 1 and x = 0, is statically determinate (cf.
Sect. 208). The corresponding problern for plane stress yields t3= 0 and (t1) 2= (12) 2 and is
statically determinate even in the non-hydrostatic alternative 1, since t1=-t2 implies
tl = - t~ in all co-ordinate systems in the plane of stress, so that the equations of equilibrium
become
tl,l + ti,2= et1. ~~.1- tl,2= f.!/2.
with (1-gl2g12) t~= (gllg21-gl2gn) tl+gllg22ti.
(223.6)
Statically determinate problems also follow from assumptions regarding the stress
trajectories. For example, from (209.1) we see that in plane stress with f- x = 0, if one
family of stress trajectories consists in straight lines, the other principal stress does not
vary along its trajectories 2 . In this same case, the angle in (App. 6.3) is of course constant
along the straight trajectories.
TRUESDELL 3 has found the most general stresses compatible with the contravariant Velocity components given in cylindrical CO-ordinates r, (), Z by
r=rR(t), 0 = z A (t), .i = z Z (t); (223.7)
this motion is a simple type of torsion, expansion, and extension of a circular
cylinder. For the physical components of d we have
R 0 0
d = R trA
z
By solving (156.5} 2 we obtain
e = eo S(t), S = exp [- J (2R + Z) dt].
The contravariant components of acceleration are
We restriet attention to stresses such that
1 THEODORESCO [1937, 9].
2 HEYMANS [1924, 4].
3 [1955, 28, § 11]
okm
871 =0, rz =0,
(223.8)
(223.9)
(223.11)
582 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 224.
and we put /==0; the dynamical equations (205A.1), in cylindrical co-ordinates,
become
a
or
rr + rr -
r ifo = <::O
n s r [R + R2 - z2 A 2] ' l
a;{) + aifz + 2 ;o = Sr z[i + A(2R +Z)] or oz r (!o ' I
an · a-;- = (!o S z [ Z + Z2J .
(223.12)
The general solution is
fTz = - r~ :r (r2 J r7J d z) + ~ (!o r z2 S [ Ä + A (2 R + Z) J + Q,
- 0 • () () = 7iY (r ,-,) + (!o r 2 S [ z2 A 2 - R - R2J , (223.13)
1 •
zz = 2 (!o z2 S [ Z + Z2] + P,
where P and Q are arbitrary functions of r, t. The terms proportional to (!o
express the effect of the inertia of the material.
The most interesting statically determinate problems arise in the theory of
shells, where a state of membrane stress Ieads to the determined system (212.22),
(212.23). The general solution of these equations will be developed in Sect. 229.
IV. General solutions of the equations of motion.
224. Steady plane problems. So as to make clear the approach to stress functions in general, we begin with the siruplest case, when the motion is steady and
plane, the stress is plane 1, and the assigned forces satisfy (207.1). Then in reetangular Cartesian co-ordinates CAUCHY's first law (205.2) assumes the form
(txx- ex2 - V), ... + (t..-y- ex:V).r = o, } (224.1) (tyx- e:Vx),x + (tyy- (!Y2- V),y = 0.
Since each of these formulae is a condition of integrability for a differential form,
the pair is equivalent to the existence of functions F and G such that
txx-ex2-V=F.r· t..-y-ex:V=-F,,..} (224.2)
lyx- (!YX = G,)'' tyy- (!Y2- V=- G,x•
When the stress is not symmetric, no simplification is possible, but in the nonpolar case, to which this subchapter is restricted, the left-hand sides of (224.2) 2
and (224.2) 3 are equal, so that
F,,. + G,y = o. (224.3)
This, in turn, is a condition of integrability for the existence of a function A
such that
F=-A,y, G=A,r (224.4)
Substitution in (224.2) yields
txx- ex2 - V=- A.rr• trr- e:Y2 - V=- A,xx• t,.y- exy = A,xr. (224.5)
When t-(!XX- V 1 is twice continuously differentiable, the theorem on exact
differentials implies that the existence of a function A satisfying (224.5) is
1 These results hold also when 1zzoF 0 but lzz z=O, as for example when the stress is a
hydrostatic pressure independent of z. '
Sect. 224. Steady plane problems. 583
necessary and sufficient that (224.1) hold. Accordingly, (224.5) gives the general
solution of the equations of motion for the case considered. The function A
is the celebrated stress function of AIRY1. Our argument, since it rests upon the
theorem of the exact differential, implies that A is single-valued in simply connected regions, generally multivalued in multiply connected regions 2•
Since a unit normal to the element of arc i dx + j dy is - i dyfds + j dxfds,
from (203.4) and (224.5) we obtain the components of the stress vector across
the arc in the forms
t (ra)x - (! i. Pn + V !!_t ds -A
- ,yy !:J'_ ds + A
,xy !!.!_- ds - dA,y ds ' l
(224.6) t _ · · _ V dx __ A !_}'-_ _ A dx __ dA,x
(raJy eYPn ds - ,yx dx ,u ds - ds ·
Therefore the normal stress and shear stress on the element of arc are given by
t - p'2- V=- dA,y !.!.._- dA,% ~ n (! n ds ds ds ds '
__ !____ (A dx + A !_}'-_) + A !____ (dx) + A _!_ (dy) - ds ,x ds ,y ds ,x ds ds ,y ds ds '
- - d2A + "(- A !!t + A !!.!_) - ds2 ,x ds ,y ds ' (224.7)
d2A dA
= - ds2 + "!in' • • d2A dA
tt- (! Pn Pt = ds dn + "ds'
where Pn and Pt are the normal and tangential components of velocity. These
results are due to MICHELL 3, for the case of equilibrium. If the element is a
stationary boundary, the momentum transfer vanishes upon it, and (224.7) yields
the stress vector directly.
To find the dynamical significance of the intermediate function F, we integrate
along a curve c from ;x:1 to ;x:2 , obtaining
Fl:~: Fi:r:1 = f dF = f[(- txy + eiy) dx +(tu- ei2 - V) dy ], l c c (224.8)
= f[i (tu- ei2 - V) + j (txy- ei:Y)] [i dy-j dx].
"'
If we include the apparent stress due to assigned force and to transfer of momentum, by (203.4) the integrand is the x-component of the stress vector actinc
upon a cylinder of unit height based upon the curve c. Thus the difference of
1 AIRY [1863, 1] considered only the case when e:i::i:- V 1 = 0, and he did not prove
necessity; the generality of the solutionwas asserted by MAXWELL [1870, 4, pp. 192-193].
The first fully satisfactory treatment for the case of equilibrium was given by MICHELL
[1900, 6]; among other things, he included V [ibid., p. 100] and gave an explicit form of
(224.5) in orthogonal curvilinear co-ordinates [ibid., p. 111]. E.R. NEUMANN [1907, 6, § 1]
obtained (224.5) for motion in which t is hydrostatic and V=O. Cf. also BRAHTZ [1934, 1],
BATEMAN [1936, 1, § 2], CRocco [1950, 5]. VorGT [1882, 4, pp. 297-298] remarked that
(224.1) continues to hold when all components of stress are constant in the z-direction; in
this case, which is often appropriate to torsion of a cylinder, the z-component of CAucHv's
first law yields a function W such that ty:- e yz = l~x· t,..- exz =- ~Y· when (V+ ez2),z
=0.
2 After making this observation, MrcHELL [1900, 6, p. 103ff.] determined the nature
of this multi-valuedness for the case of linear elasticity theory.
a [1900, 6, p. 110]. An incorrect formula of this kindwas given by NEUMANN [1907,
6, § 3].
584 C. TRUESDELL and R. TouPIN: The C!assical Field Theories. Sect. 225.
the values F at ;r1 and x 2 equals the x-component of the force 1, including inertial
force, acting upon c. When c is a stationary boundary, the momentum transfer
terms contribute nothing to the integral, which then yields the force alone. A
similar interpretation holds for G.
To interpret2 the function A, we see that for a curve c connecting x 2 to x 1
we have
Alor2 - Alor1 = f dA= f (A,xdx + A," dy) = f (Gdx -F dy),
" " "
= (x2- xl) Gior2 - (Y2- YI} Fizz+
+ J {(x- x 1} [(t"y- (}'y2 - V) dx- (tx"- eiy) dy J + (224.9)
" + (y- y1) [(tu- ei2 - V) dy- (tx"- eiy) dx ]},
= (xz- xJ} Gior2 - (Y2- YJ} Flor2 + 2
where 2 is the torque about x 1 exerted by the stress, including the apparent
stress due to rate of change of momentum and to applied force, acting upon the
right-hand side of c, thought of as directed from x 1 toward x 2 •
While we have used reetangular Cartesian co-ordinates, the extension of
(224.5) to arbitrary curvilinear Co-ordinates is immediate 3 :
tkm- eikim- V gkm =- ekP emq A,pq• (224.10}
where gkm is the contravariant metric tensor in the plane, ekP is the absolute
alternating tensor, and "," denotes covariant differentiation based on g. Since
(224.10} is a tensor equation, and since it reduces to (224.5) when the co-ordinates
arereetangular Cartesian, it is valid in all co-ordinate systems.
225. Generalized Iineal motion. Consider a motion in which t, iJ, (!, and V
all depend only upon x and t. Using (161.22) and the x-component of (205.2),
we readily infer the existence of a function U such that 4
o(t;,, L;xl
U i= _ Z:.:x 1 t =V+ o(x,t)- e = ,XX' L;xx ' XX L;xx • (225.1)
The second of these equations may be written in the form
i =- (t;~l,t . (225.2) (L; rl,, '
thus the velocity equals the slope of the curve U.x = const in the t-x plane. The
remaining components of (205.2) yield the existence of functions V and W such
that
Z = W xfUxx,
In the lineal case, V= W = 0.
(225.3)
1 This result is due to MAXWELL [1870, 4, pp. 192-193], who considered only the static
case; he called the vector F, G "the diagram of stress ". His derivation was criticized and
corrected by MrcHELL [1900, 6, p. 107]. In rediscovering this result for the case of hydrostatic stress, BATEMAN [1938, I, § 1] called Fand G the "drag and Iift functions". 2 PHILLIPS [1934, 4] based his proof of the existence of AIRY's function on this interpretation. Cf. also SoBRERO [1935, 7]. An energetic interpretation for A and for stress functions of all types considered below has been given by L. FINZI [1956, 7, § 6]. 3 B. FINZI [1934, 2, § 1]. 4 The result was worked out by W. KrRCHHOFF [1930, 2, § 1], following a suggestion of
E.R. NEUMANN [1907, 6, § 6]. It is rediscovered by McVITTIE [1953, 18, § 3].
Sects. 226,227. General solution for a flat space. 585
226. Conventions for the remaining general solutions. All general solutions
are obtained by essentially the same reasoning as in Sect. 224, although the details
may be more elaborate. At bottom, we are to integrate1
skm =0 ,m ' (226.1)
that is, to find the most general symmetric null stress. It is the additional condition of symmetry that makes the problern interesting and different from that
solved in Sects. 161 to 164. The variations in the answers from case to case arise
because of different numbers of dimensions and different metrics on which the
covariant differentiation is based.
For the Euclidean case, whatever the number of dimensions, the entire
analysis consists in repeated use of the fact that a continuously differentiable
tensor whose divergence vanishes may be expressedas the curl of another tensor.
The tensor potentials so determined exist subject to various conditions that may
be found in the literature on the theory of the potential.
We do not remind the reader again that when all co-ordinates are space Coordinates we may take skm as tkm- exkxm with an additional term -V gkm
when f satisfies (207.1). For general f, the linearity of CAUCHY's laws enables
the general solution to be gotten by adding to any particular solution the general
null stress.
The problern of finding a particular integral is relatively trivial in a flat space. In a
convex region, quadratures suffice. More generally, for any Riemannian space, Iet u be
any solution of the linear differential equation
(226.2)
where R::O is the Ricci tensor. Then it is easy to verify that a particular solution pskm of
the system skm,m+efk=o, skm=smk is given by2
(226.3)
We now turn our attention to determining the most generalnull stress 3.
227. Generalsolution for a flat space. We suppose s tobe twice continuously
differentiable. In order that skm m = 0 in a flat space of any number of dimensions,
it is necessary and sufficient tliat there exist a tensor b such that 4
(227.1)
The condition skm =smk may now be written in the form
(bkmP _ bmkP),p = 0. (227.2)
1 Cf. the treatment of this class of problems by FINZI and PASTORI [1949, 11, Chap. IV,
§ 7].
2 This result is suggested by the special case for flat spaces given by ScHAEFER [1953,
28, § 5].
3 Most of the material in the rest of this subchapter is taken from a work of TRUESDELL
[1959, 11].
' As observed by GWYTHER [1913, 2], for a stress tensor which is not symmetric the
analysis breaks off at this point. In the more general viewpoint allowing couple stress as weil
as ordinary stress, all differential conditions of equilibrium are prescriptions of divergences,
as shown at the end of Sect. 205. Stress functions for the system (205.2}, {205.10) are given
by GüNTHER [1958, 4, § 3].
586 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 227.
This condition, in turn, is equivalent to the existence of a tensor c such that
where
ckmpq = _ ckmq p = _ cmkPq 0
Therefore
2bkmP = (cPkmq + cmPqk + ckmPq),q,
so that (227o1) becomes
where
skm = kPkmq ,pq•
kPkmq = i (cPkmq + cmqPk), }
kPkmq = _ hkPmq = _ kPkqm = hmqpk 0
(22703)
(227.4)
(227o5)
(227.6)
(227o7)
The foregoing derivation, given by DORN and SCHILD1, shows that (227o6) furnishes the gener•al solution of Cauchy's laws in a flat space of any dimension.
In the three-dimensional case, set
a - 1 e e hPkmq rs = 4 rpk sqm •
so that
hPkmq = ekrp emsq a, 50 a,. = a.,.
Then (227.6) becomes the Gwyther-Finzi general solution2 :
skm = ekrp emsq a,s,pq.
(227o8)
(227.9)
(227o10)
B. FINZI noticed that the tensor a is indeterminate 3 to within an arbitrary
symmetric tensor· a0 satisfying ekrp emsq a~s,pq =0. As we have mentioned in
1 [1956, 4]. The basic idea was suggested by BELTRAMI [1892, 2]. Cf. also MORERA
[1892, 10].
A variant ofthis derivation had been given earlier by GüNTHER [1954. 8, § 1]. He begins
by introducing the skew-symmetric dual tensor of fourth order,
1/.mpq = Bkmn Bpqs T"s' (*)
which he interprets as a "transversal stress tensor". An explicit solution for 1/.mpq in terms
of stress functions is simply obtained. In DoRN and ScHILD's proof, the duals appear in
(227.8)0 While GüNTHER's solution for the dual tensor is indeed valid, as he says, in a flat
space of any dimension, to derive from it the ordinary stress tensor we must use the inverse
of (*) and hence presume the nurober of dimensions to be three. Of course GüNTHER's proof
can ·be adjusted to the n-dimensional case also, but DoRN and ScHILD's proof is equally
simple and natural in all cases. 2 While the Cartesian tensor form of (227.10) is almost obvious from an equation of
KLEIN and WIEGHARDT [1905, 3, Eq. (33)], they did not infer it. GWYTHER [1912, 3]
obtained (227.10) in orthogonal curvilinear co-ordinates, writing out the special cases appropriate to reetangular Cartesian, cylindrical polar, and spherical polar co-ordinates (cf. also
[ 1911, 5]). His steps are such as to imply the necessity of the result; the sufficiency is immediate. B. FINZI [1934, 2, § 3] observed that (227.10) yields a symmetric tensor satisfying
skm ". = 0; for a proof of completeness he was content to refer to the previously known fact
that the special cases (227.12) and (227.13) are complete. While KRÖNER [1954, 11, § 1] does
not prove completeness, he begins from an invariant decomposition of symmetric tensors
that might be made the basis of a rigorous proof. 3 This fact is used by PERETTI [1949. 23] to show that it is possible to choose a in such
a way that from its components may be obtained simple expressions for the resultant force
and moment of the stresses on a surface. Cf. also BLOKH [1950, 2]. GüNTHER [1954. 8, § 3]
gives simple expressions for the resultant force and torque on a body in terms of integrals of
stress functions around particular curves. ScHAEFER [1955. 22] discusses the nature of
null stress on this basis. In a later work, SCHAEFER [1959. 10] interprets the components
of a as dynamical actions upon the bounding surface, conceived as being that of a plate and
of a slab simultaneously.
Sect. 228. Two applications: rotationally symmetric case, and plane unsteady motion. 587
Sects. 34 and 84, such a tensor is of the form a~, = b(m,r)• where bis an arbitrary
veetor. For a given a, in reetangular Cartesian co-ordinates we may choose b
so as to satisfy one or the other of the conditions
b(m,r) = - amr>
bm,m =- amm
r=f=m }
(unsummed). (227.11)
These two choices of b show that for reetangular Cartesian co-ordinates there
is no loss in generality in assuming in the first place that a is diagonal, or that
the diagonal components of a are zero. The former alternative yields
etc. (227.12)
with a1 a,."'"' a3 = aYY, a3- az z; the latter alternative yields
(227.13)
with a4 a23 , a5- a31 , a6- a12 • These two forms of the general solution were
obtained by MAXWELL and MoRERA1, respectively. As follows from the special
choices of a made to derive them, these special forms are not invariant under
transformations even of reetangular Cartesian co-ordinates. The explicit form
for (227.10) in reetangular Cartesian co-ordinates, with no restrietions on the
six potentials, may be obtained by adding together the right-hand sides of (227.12)
and (227.13). Other special choices of the potentials are possible 2, but it by no
means follows that a solution obtained by imposing three arbitrary conditions
on the six potentials akm remains complete 3 •
An attempt to adjust the stress funetions so as to satisfy the stress boundary
condition (203.6) 1 on a reetangular parallelepiped has been m.ade by FILONBNKOBoRODICH4.
228. Two applications: the rotationally symmetric case, and plane unsteady
motion. If we write (227.1 0) explicitly in cylindrical polar co-ordinates, at the same
time supposing that all derivatives with respect to the azimuthangle are zero, we
1 [1868, 12], [1870, 4]; [1892, 9]. Using results given by BELTRAMI [1892, 2], MaRERA
[ 1892, 10] modified his derivation so as to yield (227 .12) alternatively to (227 .13). Cf. also
GwYTHER [1913, 2]. A Iiterature devoted mainly to rediscovery of known results regarding
this subject has arisen recently; cf. KuzMIN [ 194 5, 3], WEBER [ 1948, 38], MaRINAGA and
NöNa [1950, 19], ScHARFER [1953, 28], LANGHAAR and STIPPES [1954, 12], ÜRNSTEIN [1954,
16]. 2 Cf. MaRINAGA and NöNa [1950, 19, § 3]. BLOKH [1950, 2] lists the essentially different
forms which result from such special choices: 5 in reetangular Cartesian Co-ordinates, 20 in
general co-ordinates, 18 in cylindrical co-ordinates, 10 in cylindrical Co-ordinates for rotationally symmetric problems, 19 in spherical Co-ordinates. 3 What reductions are possible is not obvious. In writings on stress functions there is a
deplorable custom of inferring completeness by merely counting the number of arbitrary
functions. Apart from the logical gap in such inference, its danger is illustrated by the solution written down without proof of completeness by PRATELLI [1953, 25, § 1]:
(A)
While indeed a solution for any choice of the three arbitrary potentials F, H, K, it is not
general. In fact, by using the expression for the product skpq smrs as a determinant of t5~'s,
we may put (A) into the form
5km= p.q,qgkm_ p,km, (B)
where P ~F- H + K, and it is easy to exhibit solutions of skm m = 0 which cannot be expressed in the form (B). '
4 [ 19 51, 7 and 8] [ 19 57, 4]. The work rests on trigonometric series or special functional
forms.
588
obtain1
C. TRUESDELL and R. TouPIN: The Classical Field Theories.
--8+13 25 rr-a,. -a,--a,, r 1' ' r ,
- 8 2 8 1 1 zz=a"+-a,--a" ' r ' r ,
rz=- a -j--a --a - [ ll 1 8 1 1] •' r r ,z'
0-- 4 1 4 1 4 6 + 2 z--a"--a,+- 6 2 a +a., -a., • r ' r ' r •
Sect. 228.
(228.1)
where commas denote partial derivatives, and where we have set a1 = a,,,
a8=a66fr 2, a3=a .. , a4=a61fr, a5=a," a6=a,6fr. The potentials a1, a8, a3, and
a5 occur only in the first four members of Eq. (228.1); the potentials a4 and a6 ,
only in the last two. Rotationally symmetric stress distributions in which rO = 0,
Oz = 0 are often called torsionless; the most general stress of this kind is obtained
by setting a4 = 0, a6 = 0 in (228.1). In any case, the six potentials may be reduced
to three in a variety of ways 2• For example, if we set
L = a8 + __1_ a8 - __1_ a1 ) ,r •' r r '
M =a,.+-a,--a,- - 8 1 .~ 2 5 L .,, , r ' r ' , ..
(228.2)
then the first four members of (228.1} become 3
;"Y =L, .. +M, Oo = (rM),,+L,..,)
- 1 - zz=L"+-L,, rz=-L", , " . '
(228.))
furnishing the generat solution for torsionless rotationally symmetric stress. Similarly,
if we put
(228.4)
the last two members of (228.1} become 4
- 1 rO=----.-W,, r• •
- 1 0z=-2 W,, r ' (228.5)
fumishing the generat solution for purely torsional stress.
A more interesting application, resting upon a device to be used more strikingly in Sect. 229, begins with the observation that in the case of plane motion,
1 BRDil:KA [1957. 2, § 6]. A more symmetrical expression is given by MARGUERRE [1955.
16], but his potentials are connected by a condition of compatibility, and there is no proof
of completeness.
2 Cf. the work of BLOKH [1950, 2].
3 This solution, whose completeness is easy to prove also directly from the equations of
equilibrium, was obtained by LovE [1906, 5, § 188]. Variants are given by BRDil:KA [1957.
2, § 4].
' This isanother form of the solution of VoiGT mentioned in footnote 1, p. 583. Cf. also
MICHELL (1900, 7, p. 133], PHILLIPS [1934, 4].
Sect. 229. Stress functions for membranes. 589
Eqs. (211.5} with F=O are of the form skm,m=O in a flat space of three
dimensions with reetangular Cartesian co-ordinates x, y, t. Restriding attention
to these special Co-ordinates, we apply the solution (227.10). After time differentiations and time components are written explicitly, the result turns out
to be in tensorial form under transformation of general Co-ordinates in the plane:
) (228.6)
where a prime denotes 8j8t and where the range of indices is 1, 2. In the steady
case, this reduces to Amv's solution (224.10} with A =-a33 . The six potentials
may be reduced to three in various ways. For example, if we choose a to be
diagonal in a particular reetangular Cartesian co-ordinate system, we obtain 1
- (! = a~yy + a~:w (!X= a~xt• (!Y= a~yt• )
·2- A 2 ·2 _ A 1 txx=(!~.--=_- ,yy+a,tt• tyy-(!Y -- ,xx+a,tt>
(,~, (!XY-A,xy•
(228.7}
generalizing Amv's solution to the case of arbitrary plane motion.
229. Stress functions for membranes. From the result at the end of Sect. 212,
the problern of equilibrium of a membrane leads to a system of the form (226.1);
explicitly, we are to find the most general symmetric tensor s~ ~ satisfying
os~~ { ~ } M { .; } 6A _ Tue-+ ;..; s + ;..; s -o, (229.1)
where the {;.~.;} are Christofiel symbols based on the positive definite surface
metric tensor a6 ;, and the range of indices is 1, 2.
There have been several attacks upon this intrinsic problem, which turns
out to be more difficult than those considered in the preceding sections because
forcurvedspaces an analogue of STOKES's representation (App. 32.9) of a solenoidal
field, which yields (227.1) in a Euclidean space of arbitrary dimension, is not
known. A different approach is needed. STORCH! 2 has obtained a solution in
geodesie co-ordinates by a direct and laborious reduction. At the present writing,
a general invariant solution is not yet known, but we present what results are
available.
The similarity in form between the conditions of compatibility (84.3} and
the general solution (227.10} has been remarked for half a century. This simi1 A similar but not obviously identical solution is obtained by KILCHEVSKI [1953, 14].
2 [1950, 28].
STORCH! [1950, 27] observed that if for an arbitrary surface we choose co-ordinates
x, y so that ds2 =). (dx2 + dy2), then a solution is furnished by the formulae ).2sxx = o2Afcy2,
).2sxY = ).2sY"' = - o2Afoxoy, ).2sYY = ()2Afox2, provided 82 Afox2 + 82 Afoy2 = o. For such
solutions the mean pressure vanishes: Ä(s"'"'+sYY)=s~=O, but it is not shown that all
solutions such that s~ = 0 are included.
STORCH! treated the case of a surface of revolution in [1949, 28]; in [1952, 18] [1953, 29],
the case when a general solution involving derivatives of orders no higher than the fourth
is possible. Cf. the counterpart expressed in terms of the conditions of compatibility which
we have mentioned in Sect. 84. A solution for minimal surfaces is initiated by CoLONNETTI
[1956, 3]. We do not discuss the older solutions for special cases defined by conditions of
in extensi bility.
For Riemannian spaces of 2, 3, and 4 dimensions, a special solution containing an arbitrary function of the total curvature is obtained by STORCH! [1957, 13].
590 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 229.
larity may be used to yield a direct solution1 of (226.1) based upon the classical
principle of virtual work. While that principle will be developed in Sect. 232,
here we need only a special case which may be verified at once, namely, that s
satisfies (226.1) if and only if
f skmd dv-O km - (229.2)
"
for all fields d suchthat dkm= c(k,m) for some vector c vanishing upon the boundary.
Here, as henceforth in this section, we systematically neglect surface integrals
as having no effect on the differential equations we seek; then the condition
(229.2) follows at once from GREEN's transformation and hence is valid in any
space where covariant differentiation is defined. Now suppose the conditions
of compatibility (Sect. 84) for the system dkm=c(k,m) have the explicit form
0 _ kmd + ßkmrd + kmrsd + S:kmrstd + -(X km km,r Y km,rs U km,rst · · ·' (229-3)
where the tensors IX, ß, ... , are symmetric in the first two indices. Introducing
a multiplier A, we replace (229.2) by
J [skmdkm- A (1Xkmdkm + ßkmr dkm,r + .. ·)] dv = 0'
"
where the variation of d is now unrestricted. Equivalently,
J[skm_IXkmA +(ßkmrA),,- ···]dkmdv =0.
Hence for any s satisfying (226.1) there exists a function A such that
skm = IXkm A - (ßkmr A),, + (ykmrs A),,.- ....
(229.4)
(229.5)
(229.6)
In this general solution, the stress function A appears as a multiplier for the
constant expressing the compatibility of the field d with virtual displacements
in the space considered.
To apply to the two-dimensional case the local result obtained, consider first
a surface of constant Gaussian curvature K. By comparing (229. 3) with {84.13)
we have
(229.7)
Substitution into (229.6) yields the generat solution for surfaces of constant curvature 2 :
s60 = e6a eO'~' A,aq; + KA a60 . (229.8)
Second, consider a surface applicable upon a surface of revolution. By comparing
(229.3) with (84.15) we have
IX6E = K,AK,Aa6E + K·6 K·"' ß6Ea=KK•aa6<, }
b6EarpVJ =- ea (6 e<>'P K·"'. (229.9)
1 Given by TRUESDELL [1957, 17], revising work of L. FrNzr tobe described in Sect.224.
Earlier ScHARFER [1953, 28, § 4] had introduced multip!iers in just the same way, concluding
that "Jeder Verträglichkeitsbedingung ist eine Spannungsfunktion zugeordnet", but his
presentation employs results of a kind valid only in flat spaces, and he did not mention any
further possibilities. GüNTHER [1954. 8, § 2] had in effect noted the method and had remarked
that the tensor of stress functions in a three-dimensional flat space may be interpreted as
Lagrangean multipliers expressing the reactions agairrst the geometrical constraints but had
concluded that "no new point of view results". 2 B. FrNzr [1934. 2, § 2] verified that (229.8) satisfies (226.1) when K is constant but did
not prove the completeness of this solution; his result is rediscovered by LANGHAAR [1953. 15].
B. FINZI [1934, 2, §§ 4, 7] conjectured also corresponding general solutions for spaces of three
and four dimensions with constant curvature; TRUESDELL [1959, 11, § 12] establishes their
completeness by this method.
Sect. 229. Stress functions for membranes. 591
Substitution into (229.6) yields the generat solution for a surface applicable ttpon
a surface of revolution of non-constant curvature 1 :
s ,A ,Aa K·"' JA + (229.10) ·"'" ,q>lp ,;.
+ [2e"'(6 e eA S'F!J Ax." 'PD, A s'
where (227.7) imply the following conditions of symmetry for A:
Al:." 'PD=- A."l:'F!J =- Ax."D'F• A Idl'FD = A 'FDI!f>. (230.4)
Inspection of (230. 3) shows that only the second of these sets of symmetry conditions is essential; the first set may be abandoned without impairing the solution. The tensor Ais indeterminate to within a tensor A 0 suchthat
erA Elf> eA S'FD A'l-."'FD, A x = 0.
The mostgeneralsuch tensorisalinear combination of tensors of the type B Idl 'I', D;
to satisfy the essential symmetry condition (230.4)8 , we may choose Bx."'F D +
B'FDI,."; if, finally, we wish to satisfy also the condition (230.4)1, 2 , we have
A~'dl'F!J: 4B[Idl]['l',!J] + 4B['FS}][I,dl]• l - B l:dl'F, !J- B Edl!J, 'I'- B."l:'F,!J + B."l:!J, 'I'+
+ B'l'Dl.',."- B'F!Jdl,E- Bu'Fx,<~> + BD'Fdl,x·
(230.5)
Since classical space-time need not be regarded as a flat tour-dimensional
space, these formulae do not necessarily have invariant significance for it (but
cf. Sect. 153). However, in reetangular Cartesian co-ordinates in an inertial frame
(211.5) with F=O do reduce to the form
yrA,A = 0, yrA = pr, (230.6)
For these special co-ordinates, then, the solution (230.3) is general. In this
solution, we may write time differentiations and time components explicitly.
The resulting formulae, derived in reetangular co-ordinates, turn out to be of
tensorial form under transformations of the space co-ordinates alone. These
formulae, valid for all curvilinear co-ordinate systemsinan inertial frame, are 2 :
- (! = s'' = e•mnePqr Amnqr,sp• )
_ nxk = 5k4 = _ 6ksmePq•(A' + 2A ) o: smqr,p 4mqr,sp (230.7)
tkm _ exk ,im= 5km = 4ekPq emsn A,p,s, qn +
+2(ekPqem•"+eksnemPq)A' + ekPqem•nA" 4npq,s pqsn•
1 B. FINZI [1934, 2, § 5] wrote down (230.3) and symmetry conditions consisting in
(230.4) and a further requirement which we do not verify; he was content to infer completeness
by counting the number of assignable arbitrary functions. A somewhat involved proof was
given by MaRINAGA and NöNo [1950, 19, § 4]; we do not follow the argument whereby they
claim [ibid., § 5] to establish the alternative form
sl'Lf = el'Adl'FeASI'l'Adls,A:I:;
they give corresponding results in n dimensions.
2 B. FINZI [ 1934, 2, § 6]. In reetangular Cartesian co-ordinates, this result is rediscovered
by ARZHANIKH [1952, 1] (while he uses 21 potentials, obviously one may be eliminated).
KILCHEVSKI [1953,' 14] observes that if RrA is the contracted Riemann tensor based on the
Riemannian metric tensor G l'A, then in any Riemannian 4-space the quantities TrA == RrA -
tGrA R"'." satisfy TrA A = o. Putting GrA = drA + eHrA, he calculates TrA = eQrA+O(e2);
hence follows öQFAfo~A = o, so that Ql'Lf, in reetangular Cartesian co-ordinates, gives a
solution of the type presented in the text above. A similar approach, involving a detour
through relativity theory, is presented by McVITTIE [1953, 18, § 2]; that his solution is
not general is remarked by WHITHAM [1954, 26], who obtains what appears tobe a special
case of FINZI's solution in a special co-ordinate system. Rediscoveries of Gther special -cases
are made by MILNE-THOMSON [1957, 9] and by BLANKFIELD and McVITTIE [1959, 1 and 2).
Handbuch der Physik, Bd. III/1. 3 8
594 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 231.
where primes denote ofot, and where the symmetries of the tensors Apqmn• A 411 Pq•
and A4 p4 q may be read off from (230.4). This is Finzi's general solution of
the equations of balance of mass and momentum in an inertial frame. In the
case of equilibrium, this solution reduces to (227.10) with 4A4 p4m=apm·
In a particular co-ordinate system, by means of (230.5) we may impose 14 conditions
upon the 20 independent potentials occurring in the solution (230.7). For example, we are
tempted to take A,p"'' as diagonal, A 4mqr as zero, and Amnqr as zero when m =!= q andn =!= r.
We have been unable to prove that it is possible to choose Hin (230.5) in such a way as to
justifythis special choice of A; if it is legitimate, then, writing A 1 = 4A4141 , ••. , A 4 = 2A 2323 , ... ,
in this case we reduce (230.3) to the form
- e = A~:u + A~yy + A:.,, ei = A~xt .... , }
Iu- ei 2= A~u+ A:yy+ A~tt • ... , lxy- exy=- A~xy • ... , (230.8)
extending MAXWELL's formulae (227.12) so as to yield a simple yet general solution in terms
of six potentials.
When the field theories were discovered in the eighteenth century, solutions
in arbitrary functions such as those presented in this subchapter were sought
eamestly, but, for the most part, sought in vain. In the nineteenth century,
researches on partial differential equations tumed away from such general
solutions so as to concentrate upon boundary-value problems. When, in the
twentieth century, the general solutions were at last obtained, scarce attention
was paid to them, and to this day they remain virtually unknown. Though
so far they have been used but rarely, they might turn out to be illuminating
in .studies of underdetermined systems, where the conventional viewpoint of
partial differential equations has gained little.
V. Variational principles.
231. On the qualities of variational principles. In favor of variational principles as expressions of physicallaws it is commonly alleged1 :
1. They are statements about a system as a whole, rather than the parts
that it comprises.
2. Since they refer to the extremum of a scalar, they are invariant, and may
be used to derive the special forms appropriate to any particular description.
3. They imply boundary conditions and jump conditions as well as differential
equations.
4. They automatically include the effects of constraints, without requiring
that the corresponding reactions be known.
5. They have heuristic value for suggesting generalizations 2•
No. 2 is outmoded, now that the principles of tensor analysis offer usasimpler
and more direct method, used throughout this treatise, for obtaining invariant
statements. No. 3 is shared by the direct statement of physical laws in integral
form as equations of balance 3, as shown by the development of continuum mechanics given earlier in this chapter (cf. especially Sects. 203 and 205). Moreover,
the boundary conditions ernerging from a variational principle depend upon
what boundary integrals, if any, are included in the statement of the principle,
and the selection of these boundary integrals is not always dictated by the
physical idea which the variational principle is assumed to express. No. 4 is a
somewhat dubious blessing, since only a special kind of constraints is included
1 Cf. HELLINGER [1914, 4, § 1], who speaks of "the pregnant brevity". 1 A sixth, the use of direct variational methods to calculate or prove existence of solutions,
may.be added in cases where a determinate system is considered but in the present treatise,
devoted to underdetermined systems, is not relevant.
a Cf. footnote 4, p. 232.
Sect. 232. Virtual work and the Lagrange-D' Alembert principle. 595
in each case, namely, those constraints having no effect on the quantity being
varied. In mechanics, these are typically constraints which do no work. Not
all constraints are of this kind, and for those which are not, the variational approach requires as direct a statement as does any other. No. 5, while having
a basis in the physics of this century, is largely an expression of taste.
There remains only No. 1, along with the elegance that variational principles
sometimes exhibit. Both these are reduced if not annulled when the variational
principle itself is awkward of unnatural. This is usually the case in continuum
mechanics. The lines of thought which have led to beautiful variational Statements for systems of mass-points have been applied in continuum mechanics
also, but only rarely are the results beautiful or useful.
For completeness, we now present variational principles and related topics,
but we regard them as derivative and subservient to the principles of mechanics
already developed. In particular, no variational principle has ever been shown to
yield Cauchy's fundamental theorem (203.4) in its basic sense as asserting that
existence of the stress vector implies the existence of the stress tensor1 . For
the statements henceforth we take the stress tensor rather than the stress vector
as given. Our purpose, in each case, is to learn the role of the effective force of
the stress, tkm,m, in modifying the classical theorems concerning mass-points 2 •
Our presentation concerns solely the formal problern 3 of setting up expressions
such that the vanishing of their first variation is equivalent to CAUCHY's laws.
Analytical questions and, except in Sect. 23 5, the existence of minima are not
discussed.
232. Virtual work and the Lagrange-D' Alembert principle. The principle of
virtual work, which when the reaction of inertia is included may be called the
Lagrange-D'Alembert prin,ciple, is the oldest general variational form of the equations of mechanics. We begirr by following but extending the traditional
development 4 of the principle. Some variants are presented afterward.
Corresponding to a sequence of independent variations ~xk, ~xkm• ~xkmp• ... ,
that is, a set of arbitrary covariant fields, we define the virtual work done on a
1 The derivation given by HELLINGER [1914, 4, § 3a] fails through petitio principi, since
the stress components appear in the original variational principle. We do not understand
the remark attributed to CARATHEODORY by MüLLER and TrMPE [1906, 6, footnote 32].
Existence of the stress tensor can be proved from variational principles which assume the
existence of an internal energy having a special functional form. Such results are presented
in Sects. 232 A, 236, and 262.
2 It is possible simply to transcribe the theorems for systems of mass-points, written as
Stieltjes integrals over material volumes, and then add tkm, m to the contravariant force vector.
Cf. EICHENWALD [1939, 7].
a We do not attempt to discuss variational principles from the axiomatic or conceptual
standpoint. For the difficult question of the contrast between the momentum principle
and the Lagrange-D'Alembert principle, cf. HAMEL [1908, 4, Kap. 2, §§ 1, 3].
4 HAUGHTON [1855, 2, pp. 99-100], KIRCHHOFF [1876, 2, Eilfte Vorlesung, § 5], BELTRAM! [1881, J, pp. 385-388], HELLINGER [1914, 4, §§ 3a-b]. A variant, bringing in explicitly the dependence of general curvilinear Co-ordinates on reetangular Cartesian co-ordinates, was given by MoRERA [1885, 6, § 1]. We do not Iist the many sources that formulate
a principle of virtual work for special systems such as rods or membranes, nor do we trace
the origin of the principle for continuous media in general through the studies of LAGRANGE on
perfect fluids, of GREEN and KELVIN on elastic bodies. Thc firsttreatmentvalid for general
continua isthat of ProLA (1833), tobe given below. The "general formula" of LAGRANGE
[1788, J, Seconde Partie, Seconde Sect., ~ 7] is valid only for systems of mass-points and
certain other special systems. For D' ALEMBERT's principle, see above, p. 532, footnote 1.
All the foregoing references concern only the classical case, where the terms in Oxkm•
oxkmp, ... , are absent from (232.1). The general theory given here is suggested by an intermediate case due to HELLINGER [1914, 4, § 4b].
38*
596 C. TRUESDELL and R. TOUPIN: The Classica] Field Theories.
body "f/ as the linear form 1
~ = j [sk~xk + skm~xkm + skmP ~xkmp + .. ·]da+ I
+ J {e [jk ~xk + fk"' ~X km+ fkmP ~Xkmp + · · ·] - r
- [tkm ~xk,m + tkmp ~Xkm,p + .. ·]} dv.
The quantities sk, skm, ... ,er. efkm, ... , tkm, tkmP, ... are defined simply
coefficients of the form. By GREEN's transformation follows
~ = p [(sk- tkmn") ~xk + (skm_ tkmPnp) ~Xkm +···]da+ )
+ j [(e fk + tk"',m) ~Xk + ((> fk"' + tkmP,p) ~X km+···] dv ·
The principle of virtual work is the assertion
f (> [xk t5xk + ixmt5xkm +pkpmf:P Oxkmp + · · ·J dv = ~. r
Sect. 232
(232.1)
as the
(232.2)
(232-3)
p being, as usual, the position vector from an arbitrary origin. Eq. (232.3)
is to hold for all variations consistent with the constraints. For the time being,
we leave aside the effect of constraints and assume that the virtual fields may
be varied arbitrarily, Then, by (232.2), the principle of virtual work is equivalent2
to the system (205 .19), with boundary conditions sk = tkm nm, etc.
The classical special case of (232.3) is
Jexkoxk=~ pskoxkda + f[efkoxk-tkm()xk,m]dv. (232.4)
r f/' r
This special case, for unconstrained variations, is equivalent to CAUCHY's first
law (205.2).
We now consider some alternative variational formulations for one or both
of CAUCHY'S laws.
If we restriet the variations considered, we may avoid using the stress components directly in the principle of virtual work. Set
)ffi - p sk oxk da +Je jk oxk dv. (232.5)
f/' r
The theorem oj Piola 3 asserts that the condition
Je xk ~xk dv = )ffi (232.6)
r
1 Here and the sequel we set o:rk m == (o:rk) .,., etc.
2 For the non-polar case, PIOLA [1848, 2, '1[48] and HELLINGER [1914, 4, §3d] provcd the
equivalence by adjusting the variations so that (232.2) reduces to (200.1). We prefcr the simpler argument (232.3) ~ (205.19) ~ (205.20). The result we have established showsalso that
the common claim that symmetry of the stress tensor does not follow from the principle
of virtual work is misleading. lndeed, it does not follow naturally. The principle must be
adjusted so as to imply that the virtual work of the torques is exactly the virtual work of the
moments of corresponding forces. As follows from (205.21), this may be done by adding
a priori the assumption that the coefficients in (232.1) satisfy t[km]p =P[k tm]P• i[kp] =P[k /p],
and only skew-symmetric o:rkm need be considered.
3 The pioneer work of PIDLA [1833, 3] [1848, 2, '\['\[ 34-38, 46- 50] is somewhat involved.
-1
First, PIOLA used the material variables, and his condition ofrigidityis oC KM =0 or oCKM = 0,
so that the outcome is (210.8) or (210.10) rather than (205.2); (205.11) and (205.2) are then
proved by transformation. Second, he seemed Ioth to confess that his principle employed
rigid virtual displacements; instead, he claimed to establish it first for rigid bodies only. In
the former work, he promised to remove the restriction in a later memoir; in the latter, he
claimed to do so by use of an intermediate reference state. He was also the first to derive the
stress boundary conditions from a variational principle [1848, 2, '\[52], and he formulated
an analogaus variational principle for onc-dimensional and two-dimensional systems [1848,
2, Chap. VII].
Sect. 232. Virtual work and the Lagrange-D'Alembert principle. 597
for virtual translations is equivalent to Cauchy's first law (205.2); for rigid virtual
displacements, to Cauchy's second law (205.11) as well. In these statements, a
virtual translation is a field (Jx suchthat bxk.m =0, while a rigid virtual displacement is a field (Jx suchthat (Jx(k,m)=O. To prove ProLA's theorem we first set up
the nine side conditions (Jxk m=O, and we write -tkm for the corresponding
multipliers1• Then (232.6) is 'equivalent to
f (! 3(k (Jxk d V = ~ sk (Jxk da+ .r ((! jk (j Xk- tkm (Jxk, m) dv (232.7)
j' Y' j'
for arbitrary variations bx. By applying GREEN's transformation, we derive
both (203.6) and (205.2); conversely, these latter imply (232.6). To derive the
second statement in PIOLA's theorem, we set up the six side conditions (j xq,q) ~x[k,Pl dv +
+f (tx satisfying
m f o ck ... m " pn ... q __(l_n . ._._q_ (l>x') dv = 0 L.J a k ... m oxr ,s , a~l "f" ,s
(233·3)
Consider only those
(233 .4)
where aPL:~. is a tensor of multipliers. The same procedure leads to boundary
conditions and equations of motion in which tkm is replaced by
m 0 cu ... s
tk m + gk r " pn .. . q __(1____1<_..:: • L.. a u ... s oi' a~l ,m
(233.5)
The fact that in general there exist no motions satisfying the fully general constraints (233.3) does not invalidate the procedure.
Sometimes there are constraints depending on the deformation gradients x7K
from a reference state 1 . For simplicity, consider a single equation of the form
c ( x\ K, t) = 0 . (233 .6)
The variations are now subject to the constraint
oc (~ k) - -k- uX ;K-0.
OX;K (233.7)
To apply this side condition, it is convenient to transform (232.2) by introducing
the variables appropriate to the description in terms of a reference state. By
(210.4), (210.6), and (20.9), in the classical non-polar case we get
12! = p (skda- PK dAK) l>xk + J (PK;K +?! jk) l>xk dV. (233.8)
g' "f"
Now introducing a multiplier p corresponding to the constraint (233.7) and proceeding as before, we obtain the boundary condition corresponding to (210.5)
and the equations of motion in ·ProLA's form (210.8), except that in both the
double vector PK is replaced by
TkK -pgkm~. (233.9) OXm;K
Equivalently, the stress tensor t is replaced by
(233-10)
From the results (233.5) and (233.10) it is evident that general constraints
such as (233.3) or (233.6) yield a non-symmetric contribution to the stress 2•
As is shown by the example of isochoric motion, certain special constraints may
be maintained by symmetric stresses.
1 PoiNCARE [1889, 8, § 152] [1892, 11, § 33]. Cf. HELLINGER [1914, 4, § 4c], who considers constraints involving higher derivatives. ERICKSEN and RIVLIN [1954, 6, § 4] work
out the explicit form of the result which is implied because c is a scalar. 2 This was remarked by ERICKSEN and RIVLIN [1954, 6, § 3].
602 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 234.
All the constraints considered in this section are of the type called holonomic1•
Non-holonomic constraints will be mentioned in Sect. 237.
234. Converse of the principle of virtual work. DoRN and ScHILD2 have formulated an elegant converse to the principle of virtual work in the static
and non-polar case: Suppose that, corresponding to any vector u given on Y, there
exists a symmetric tensor c such that
JOtkmckmdv = pOtkmukdam (234.1)
-r [/'
for all null stresses 0t; then the vector field u can be extended throughout f in such a
way that
(234.2)
In other words, if u =X it follows that in fact c = d. The statement of the theorem
is limited to null stresses only for simplicity, the extension to general equilibrated
stresses being easy, but for motion rather than equilibrium the formulation is
awkward. The proof of the theorem is divided into two stages: (a) we show that
c satisfies the conditions of compatibility for d, and thereafter (b) we prove
(234.2).
Proof of (a). By (227.10), we may write the hypothesis (234.1) in the form
J ekrs e'lmn a,m,sn Ckq dv = p ekrs eqmn a,m,sn UR daq, (234.3)
-r [/'
where a is an arbitrary symmetric tensor. Two applications of GREEN's transformation put (234.3) into the form
-r (234.4)
J ekrs eqmnckq,snarmdv )
= j ekrs eqmn {ckq,n a,m da5 - Ckq a,m,s da"+ a,m,sn Uk daq} ·
We choose a as zero outside a small region surrounding a given point, thus annulling the surface integral on the right; since a may be arbitrary, to within
requirements of smoothness, we conclude that ekrseqmnckq,sn• since it is symmetric,
must vanish. By the remarks following (84.3), or by those at the end of Sect. 34,
there exists a vector v such that
(234.5)
Proof of (b). From (234.5), GREEN's transformation, and the fact that 0t
is a null stress, we have
Comparison with (234.1) yields
p 0tkm (uk- vk) dam = 0
[/'
(234.6)
(234.7)
for arbitrary null stresses 0t. This condition asserts that the virtual work of an
arbitrary equilibrated stress in the virtual displacement uk- vk is zero. Hence
the motion u- v is rigid. That is, there exist constants wkm and bk such that
1 HELLINGER [1914, 4, § 4c] discusses also holonornic constraints applied only on surfaces or lines, as weil as constraints which are inequalities. 2 [1956, 4]. The staternent (a) and its proof are due to LocATELLI [1940, 15 and 16],
who considered also the case when the assigned force does not vanish.
Sects. 235,236. Lagrangian and Haroiltonian principles. 603
(234.8)
on f/'. Now v is defined through "f/". Therefore we may take (234.8) as extending
the definition of u throughout "f/". Since u(k,m) =v(k,m)• (234.2) follows from
(234.5). Q.E.D.
L. FINZI 1 has remarked that the process leading from (234.1) to the conditions of compatibility for (234.2) is entirely general and may be applied in any
space where covariant differentiation is defined. For example, let the general
null stress have the form
Otkm = akm A + bkmr A,, + ckmrs A,,. + .... (234.9}
Substitution in (234-3) and proceeding as in the lines following yields
0 = akm ckm- (bkmr ck".},, + (ckmrs ck ... ),,.- . . . (234.10}
as the conditions of compatibility. The reader may use this result to derive
(84.5) from (229.8), and to derive (84.4) from (229.10). The process is easily
modified so as to apply to the case when the general null stress is given in terms
of a tensor of stress functions Akm or Ak,.Pq• etc.
235. Principle of minimum stress intensity. Given a symmetric tensor field skm
in "f/", we shall say it is given a potential increment when it is replaced by skm +
bck,m)• where bis a vector field which vanishes upon the boundary f/'of "f/". If
sk"',".=O, we have
[skm + b(k,m)] [skm + b(k,m)] = Skm skm + (2bk sk"'),m + b(k,m) b(k,m). (235.1)
By GREEN'S transformation follows
I [skm + b(k,m)J [skm + b(k,m>] dv =I [skm skm + b(k,m) bCk,ml] dv, )
..".. ..".. (23 5 .2) ~ Isk".skmdv,
.."..
where equality holds if and only if b(k,m) = 0. The analysis is valid in a Riemannian
space of any number of dimensions, so long as the metric be positive definite.
We apply the foregoing result to a body subject to no assigned force, both
in the three-dimensional and the four-dimensional cases, so obtaining PRATELLI's
theorems of minimum stress intensity 2 :
1. In steady motion, the total intensity of a stress tensor t- (!~~ satisfying
Cauchy's laws is a minimum with respect to potential increments.
2. In a reetangular Cartesian co-ordinate system in an inertial frame, the total
intensity of a world stress-momentum tensor (211.2) satisfying the equations of
continuity and momentum balance is a minimum with respect to potential increments.
From the analysis, it is evident that no result of this kind can be expected to
hold for general variations of stress.
236. Lagrangian and Hamiltonian principles. In this section and the next we
derive principles related to (232.4). Thus we are restricting attention to principles
equivalent to CAUCHY's first law.
1 [1956, 7, § 10]. Indeed, the stateroent above roay easily be inferred froro the work of
LOCATELLI [1940, 15 and 16]. 2 [1953, 25]. PRATELLI, who uses variational calculus, does not obtain an absolute
roinirouro in space-tiroe because he uses the energy-rooroenturo tensor of special relativity
rather than the classical tensor (211.2).
604 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 236A.
By (170.4), we may write (232.4) in the form
:t J ex,.~x"dv = ~~ + ~. (236.1)
"'" where ~ is the kinetic energy and where the variation of density satisfies ~ (ed v) = 0.
Eq. (236.1), which is merely a rewriting of the principle of virtual work, is
called the Lagrangian central equation1•
Integrating (236.1) from t=t1 to t=t2 , we impose the condition that ~:JJ=O
at t=t1 and at t=t2 so as to obtain Hamilton's principle 2 :
tz f (~~ + ~) dt = 0' (236.2) tl
und conversely, by the identity (236.1), if (236.2) holds for every pair of times t1
and t2 , (232.4) for all admissible ~:JJ follows.
By further restriction of the variations, it is easy to derive the principle of
least action from (236.2).
The elegant and useful forms that these principles assume for systems of masspoints do not carry over to general continuum mechanics.
236A. Appendix. HAMILTON'S principle in the case when there is a strain energy. In the
special case when the formula (232A.S) holds, HAMILTON's principle (236.2) may be written
in the more familiar form
tg
t5 J 53dt = 0, tl
(236A.1)
where
53= f J.dV+ ~s,.ukda, ;. =e0 [!x2 - •].
"'" 9'
(236A.2)
Here the tildes have been dropped, for the reference configuration must be a fixed one.
While the equivalence of this variational principle to CAUCHY's first law, the associated
stress boundary condition, and the stress-strain relations (232A.4) follows from results already
given in Sect. 232A, (236A.t) is our only example of a variational principle of the conventional simple type to which the Euler formulae may be applied, so we give an independent
verification 3 • Recalling that in the material description_ik = 8 xkf8t, we think of the time integral
of the volume integral in (236A.t) as an integral over a four-dimensional manifold, and to
this integral we apply the usual rules of the calculus of variations. Hence, when there are
no constraints, the spatial differential equations following from (236A.t) are
(236A.3)
equivalent to CAucHv's first law when there is a strain energy and a potential energy.
1 For continuum mechanics, this equation was introduced in a somewhat more general
form by,HEUN [1913, 4, § 21] and HELLINGER [1914, 4, §Sb]. These authors consider also
an alternative form in which the time as weil as the path is varied. 2 According to HELLINGER [1914, 4, §Sb], the extension of HAMILTON's principle to
continuous mediawas first obtained byWALTER [1868,15]; it was given also by KIRCHHOFF
[1876, 2, Vorl. 11, § S]. 3 KIRCHHOFF [1859. 2, § 1]. While in [1852, 1] it was assumed that the strain energy
is a quadratic function, as far as concerns the variational formulation this restriction was
not used.
The second variational principle of CLEBSCH, which we have presented in a purely kinematical form in Sect. 137, may be interpreted as a special case of HAMILTON's principlc.
ZEMPLEN [190S, 7] [1905, 9, § 3] used (236.1) to obtain (236A.3), (210.5). and (20S.3). The
general formula has been rediscovered by DE DONDER and VAN DEN DUNGEN [1949, 5] and by
E.HöLDER [19SO, 11, §§ 1-3], who discuss alternative forms.
Sect. 237. The principle of extreme compulsion. 605
In the above formulation of HAMILTON's principle, l?o is taken as an assigned function
of X, and any additional parameters upon which the function T may depend are kept fixed
in the variation. We may set eo= ef by ( 156.2), and then, as has been remarked by HERIVEL1,
we may vary I? and the additional parameters, at the same time setting up as side conditions
the continuity equation and the constancy of the additional parameters for each particle.
The result is unchanged. It is also possible, though more elaborate, to formulate HAMILTON's
principle in terms of the spatial variables and to vary the velocity field rather than the displacement2.
237. The principle of extreme compulsion 3• In the principle of virtual work
(232.4), the variation IJ;r may be selected as any field satisfying the constraints,
presumed holonomic. Precisely, as we have seen in Sect. 233, this means that
if there are constraints 4
aC(x,X,x~K,x7KM• ... ,t) =0, (237.1)
then (j;r ranges over the dass of vectors v satisfying
8aC vk+~~vkK + _Jla~vkKM + ... = 0.
8xk 8x7K ' 8x7KM ' (237.2)
For the actual motion, differentiating (237.1) materially yields
8aC xk+ 8aC X~ +_Jlo~_xk + ... + 8aC =0 8xk 8 k ,K 8 k ,KM 8t ' x;K x;KM (237-3)
8aC "k+ 8aC "k + 8aC "k + +F _ --X ··~·-X.K ---X·KM ··· -0 8xk 8x7K ' 8x7K M ' ' (237.4)
where Fis a function of JJ, X, x7K' x7K M' ... '.Xk, x7K' .... This identity suggests
a means of constructing variational fields which conform to the constraints.
Consider the dass of variationssuch that, at each fixed instant, the varied motion
has the same displacement and the same velocity as the original motion, but not
necesarily the same acceleration:
(j;r = 0, I'Jx = o. (237.5)
For such variations, by (237.1) we have IJ(oaCfoxk)=O, ... , and IJF=O. From
(237.4) follows
--uX 8aC ~"k+ --uX.K 8aC ~"k + ... - .
_0
8xk ox7K ' (23 7.6)
This is an equation of the form (237.2). What we have proved is that if IJx is
any variation satisfying (237.5), then IJx is an admissible virtual displacement
field. We may therefore replace IJ;r by I'Jx in (232.4), obtaining
f exk IJxkdv = 9i sk IJxkda + J [elk IJxk- tkm I'Jxk,m] dv. (237.7)
r Y' r
Hence, supposing IJ(e dv) =0, we get
1Jfiex2 dv= pskiJxkda+ f[efkiJxk-tkmiJxk,mJdv. (237.8)
r Y' r
1 [1955. 13, § 1].
2 Unsatisfactory special cases are given by EcKART [1938, 4, § 3] and HERIVEL [1955.
13, § 2]; their result is corrected by LrN in a work not yet published. We make no attempt
to cite the !arge hydrodynamical Iiterature on special variational principles, some of which is
discussed by SERRIN, Sect. 15 of The Mathematical Principles of Fluid Mechanics, this Encyclopedia, Vol. VIII/1. 3 For systems of mass-points, this principle is associated with the names of GAuss, LrPSCHITZ, GrBBS, and APPELL. For continuum mechanics, we follow the development of BRILL
[1909, 2, § 18] and HELLINGER [1914, 4, §Sc]. 4 The tensorial character of the equations of constraint is irrelevant here.
606 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 238.
If we add the convention that at points and times in the varied motions the same
force f is assigned as at the corresponding points in the actual motion, so that
()j=O, we may replace (237.8) by
(j f l e(~- /) 2 dv = ~ s" ()x"da- ftkm(jxk,mdv. ~ ~ ~
(237-9)
The integral whose variation stands on the left is the totalsquared effective force,
or the compulsion, of the motion, and the variational equation (237-9) is the
principle of extreme compulsion1•
The principle may be extended so as to hold for non-kolonomic constraints of
the more general form
C ( X k .. k "k "k ) - a :JJ, ,X;K•X";.KM•···•X,X;K•···•t -0. (237.10)
For the actual motionundersuch a constraint, we have
oaC "k + oaC ··k + + oaC "k + + oaC _ --X --X.K · · · --X · · · ---0 oik oi~K ' oxk ot . (237.11)
In order that varied motions satisfying (237.5) be compatible with (237.11), it
is necessary and sufficient that
(237.12)
Thus for constraints of the type (23 7.1 0) we may still infer the principle of extreme
compulsion, but the variations besides satisfying (237-5) are subjected to (237.12)
as an additional condition.
238. Remarks on mechanics in generalized spaces. There have been many
discussions of mechanical principles appropriate to non-Euclidean spaces 2• Except
for relativistic mechanics, which is outside the scope of this treatise, these developments seem to consist mainly in observations that certain parts of the theory
do not require Euclidean three-dimensional space, but may be carried over
bodily to more general ambients.
For example 3, while the momentum principle in the form {196.3), since it
requires that the integrals of vector fields over bodies enjoy invariance, is restricted
to Euclidean spaces, or at least to spaces with distant parallelism, CAUCHY's
laws (205 .2) and (205 .11) are meaningful in any space where covariant differentiation may be defined 4• Sometimes they are derived for Riemannian spaces by
assuming that the momentum principle applies to infinitely small volumes, but
in essence such a derivation is no more than a direct postulation of the desired
result. There are infinitely many possible "laws" of mechanics in generalized
spaces if we demand no more than that they be invariant and intrinsic equations which in the Euclidean case reduce to CAUCHY's laws. For example, if
R",.. is the contracted curvature tensor of an affine space, we may replace x" by x" +KR" ,..x"' in (205.2), and for all values of K the resulting equation reduces
in Euclidean spaces to the usual form of CAUCHY's first law.
Another formal generalization, the analogue of common practice in masspoint mechanics, replaces sr in (236.1) or in (236.2) by
(238.1)
1 As if English vocabulary were insufficient to supply two different words to translate
"Nebenbedingung" and "Zwang", (237-9) is often called "the principle of least constraint". 2 See Special Bibliography M at the end of this treatise. 3 VAN DANTZIG [1934, 10, Part IV, § 1].
' At bottom, thisis the content of BELTRAMI's observation [1881, 1, p. 389] that Eq. (232.4)
is meaningful and may be applied in curved Riemannian spaces.
Sect. 239. Scope and plan of the chapter an energy and entropy. 607
where a""' di" dx"' is an arbitrary quadratic form, not necessarily reducible to
a sum of squares, and not necessarily the metric tensor of space.
A third generalization1, more interesting from the mechanical point of view,
replaces the principle of virtual work (232.4), after converting it as in {232A.2)
to an expression in terms of a material reference state xcx, by a tour-dimensional
equation
/dt { fsk t5x" da+ f [eo {f" t5x" + bk t5~")- Tkcx t5x7cx] dv} = 0, I I f/' 'lVg
(238.2)
putting the impulsive coefficients b" on a par with force and stress as coefficients
in a linear variational form. As in HAMILTON's principle, it is supposed that
t5~=0 at t=t1 and at t=t2 • Equivalent to {238.2), in the case when there are
no constraints, are the equations
skda=T ... cxdAcx on Y,}
eobk=eoi ... +T,.cx;cx in -r. {238.3)
The case when b = ;i gives the classical equations in the form (210.5), {210.8);
the case when bk =akmx"' gives the generalization described in the paragraph
preceding. It is easy to extend (238.2) to a mechanics of moments which generalizes that following from {232.3).
E. Energy and entropy.
239. Scope and plan of the chapter. We attempt to collect here everything
of a general nature concerning the balance of energy in continuous media and the
mathematical properties of entropy. There can be no doubt of the relevance
of the first subchapter, which defines the specific internal energy and derives
differential equations and jump conditions expressing the balance of total energy
in such a way as to reflect the interconvertibility of heat and work.
The second subchapter concerns the more special and more dubious subject
of thermodynamics. The specific entropy is regarded as a static defining parameter entering a caloric equation of state which is supposed to regulate the
specific internal energy, regardless of deformation and motion. The formal content of this subchapter coincides with that customarily said to describe "reversible" processes in a substance obeying an equation of state depending upon any
finite number of parameters, except that our considerations are phrased in terms
of a particle in a general continuum and that we do not take up the special thermodynamic properties distinguishing fluids from solids. The subchapter ends by
deriving a differential equation for the production of entropy and by discussing
relations between the rate of change of total entropy and inequalities governing
local changes of entropy.
The last subchapter considers several definitions of equilibrium and mentions
connections between stability of equilibrium and inequalities restricting thermodynamic equations and quantities.
In historical origin the balance of energy and the theory of the equation of
state, logically independent and in fact relevant to different levels of physical
1 Due to HELLINGER [1914, 4, § 5d], generalizing a form proposed by E. and F. CossERAT [1909, 5, §§ 61-67, 76-80].
608 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 240.
generality, are intertwined not only with each other but also with a third concept,
that heat and temperature are mean manifestations of molecular motion. To
CARNOT (1824-1832) is due not only a general understanding and statement
of the equivalence of heat and work, formulated independently by JOULE ( 1843)
and WATERSTON (1843), but also the concept of entropy. While owing much to
numerous earlier researches, the first thermodynamic writing to achieve a modicum
of clarity is that of GIBBS (1873-1875), upon which our treatment is based,
though we do not enter many of the details and applications he developed, nor
do we fail to elaborate the mathematical structure rendered possible and natural
by his approach.
There exists no other comparably general and complete exposition of the
material given in this chapter1•
I. The balance of energy.
240. The law of energy balance. In Sect. 217 we have seen that kinetic energy
st' generally fails to be conserved. Using the apparatus of Sect. 157, we could
set up an influx and a supply of kinetic energy and so obtain a general equation
of balance for it 2• Such a balance, while not incorrect, does not lead to a fruitful
theory, because it fails to reflect the physical principle of interconvertibility of
heat and mechanical work. This principle, which is too broad and too vague for
us to attempt a general formulation, as well as too old for us to include its history3
in this treatise, suggests that any equation of energy balance should contain
terms which can be identified with non-mechanical transfer of energy. These
terms may, but need not be dependent upon changes of temperature. To keep
full generality, we refrain from defining temperature specifically until the next
subchapter, but its effects remain in our minds to motivate the introduction of
the non-mechanical power, 0.
At the same time, the special theorems on conservation of energy in Sect. 218,
in the circumstances when they hold, should not remain outside the general for1 Wehave been unable to derive much help from any of the numerous treatises except
that of PARTINGTON [1949, 22, Sect. II], distinguished for its concise and clear statements of
practice and for its critical references.
2 Such is the approach of MATTIOLI [1914, 7].
3 That heat is a mode of motion was widely believed in the eighteenth century, and both
EuLER ( 1 729, 1782) and DANIEL BERNOULLI ( 1738) constructed kinetic molecular models
in which temperature may be identified with the kinetic energy of the molecules. The generat
and phenomenologi<;al principle, independent of a molecular interpretation, is more recent.
That it was known to CARNOT by 1832 is proved by his notes [1878, I], which calculate the
mechanical equivalent of heat and project the porous plug experiment. The contention of
CLAUSIUS, still reproduced in textbooks, that CARNOT's celebrated treatise [1824, I] obtained
correct results from an incorrect axiom, is shown tobe false, as PARTINGTON [ 1949. 22, Part II,
§ 33. last footnote] has remarked, by the presentation of LIPPMAN [1889, 5, pp. 76-78];
to render CARNOT's work in accord with later views, translate "calorique" as "entropy"
(cf. CALLENDER [1910, 3, §§ 16, 20, 22], LAMER [1949, I6]), but is unlikely that anyone would
grasp the principle of conservation of energy from reading CARNOT's treatise.
In our opinion, the first clear statements to be published are those of JouLE [1843, 2]
[1845, I and 2] [1847, 2] and WATERSTON [1843, 5], the latter's being more restricted in
scope because derived exclusively from a kinetic molecular model. Forerunners are MoHR
[1837. 4], SEGUIN [1839. 2, Chap. VII, § 1], and J.R. MAYER [1842, 2]. The history of the
"first law of thermodynamics" is discussed by JouLE [1864, I], TAIT [1868, 14, Introd. and
Chap. I] [1876, 6, §§ Il, III, and Introd. to 2nd ed.], CHERBULIEZ [1871, 3], HELMHOLTZ
[1882, I], MACH [1896. 3, PP· 238-268], SARTON [1929. 8], EPSTEIN [1937,I. § 11], BOYER
[1943, I]; the most nearly complete account is given by PARTINGTON [1949. 22, Part Il,
§§ 10-12].
Sect. 241. The equation of energy balance for continuous media. 609
malism. The existence of a strain energy in some cases suggests that in the most
general motions we introduce an internal energy G:, an additive set function
such that the total energy Sl' + @ is balanced.
The fundamental energy balance is then 1
(240.1)
where ID3 is the mechanical power or total rate of working of the mechanical
actions upon the body. This basic law, sometimes called the "first law of thermodynamics"2, is tobe set alongside the laws of momentum (196-3).
241. The equation of energy balance for continuous media. For a continuum,
the mechanical power ID3 is the rate of working of the stress vector and the couple
stress on the boundary, plus the rate of working of the assigned forces and couples
in the interior (cf. Sects. 217, 232); the non-mechanical power 0, as tobe expected
from Sect.157, is expressed in terms of an efflux of energy 3 h and a supply of energy
q. Thus (240.1) becomes 4
~ + ~ = p (tP' ip- mPqrwpq) da,+ J (fP.iq-lPq wpq) dl)R + ) [/' -r
+ p hP d ap + J q d>JR. [/' -r
(241.1)
The internal energy @, being an additive set function 5, may be expressed in terms
of a specific internal energy e:
(241.2)
Thus (241.1) is an equation of balance (157.1). In regions where t.~. m, w, and h
are continuously differentiable, while ä-!, i,f, l, and q are continuous, we may apply
(157.6) and obtain
exPip+ ei =tP',,xp + tq dpq + k~p + eq, (241 A.1)
as follows at once from (218. 7). In particular, ifthe right-hand side is zero, we get e = 11 + f (X),
showing that when there is no dissipative stress, no flux of energy, and no supply of energy,
the internal energy and the strain energy differ by a constant for each particle.
Further formal simplification results when we assume 3
an 0 tPq = --
odpq • (241A.2)
side of the shock ensures conservation of energy also at the shock. This is not so. That a
distinct additional condition is needed was first seen by RANKINE [1870, 6, §§ 7-9], who
obtained a spl(cial case of (241.9) for lineal motion, as did HuGONIOT [1887, 1, § 149]. According to a note added in the 1883 reprint of [1848, 4], KELVIN and RAYLEIGH were aware of
the matter. Cf. the·discussion by HADAMARD [1903, 11, 'i[ 209]. A special case for general
motionwas derived by JouGUET [1901, 9] and discussed by DuHEM [1901, 7, Part 2, Chap. I,
§§ 7-8] and ZEMPLEN [1905, 9, § 8]. A fairly general case was obtained by CouRANT and
FRIEDRICHS [1948, 8, §§54, 118].
We are unable to follow the physical arguments sometimes adduced to infer (241.9)
directly from remarks concerning the enthalpy. Indeed, if p = n, the thermodynamic pressure
(Sect. 247) in a fluid obeying a caloric equation of state of the forme= e(1j, v), then (241.9)
assumes the form [X+ l U2] = 0, but in more general circumstances (241. 7) bears no apparent
relation to the enthalpy defined by (251.1)3 . Indeed, to derive (241.7) no thermodynamic
formalism is used, and in the degree of generality maintained here, the thermodynamic
tensions need not be defined, and hence an enthalpy need not exist.
1 Note that the condition [J:1] = 0, which follows from (205.6) 2 in the present case, has
been used here; cf. the remark after {241.7). 2 Contrary to the implication of McVITTIE [1949, 18, § 7]. solution of this problern requires no commitment as to the form of the energy equation in relativistic theories. McVITTIE
obtains a special case of (241.11). in which e has a special functional form.
3 RAYLEIGH [1873, 6, § Il].
39*
612 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sects. 242,243.
where Dis a homogeneaus function of degree r in the components dpq• since (241A.1) then
becomes
e(e-ä)=htp+eq+rD. (241A.3)
This result is an equation of balance ( Sect. 1 57) : If we set ). == e- a, then h is the efflux
of ). and e q + r D is the supply of ).. The function D is called a dissipation function. This
same result, with D == o, holds a fortiori when the dissipative stresses vanish.
Thus it is clear that energy equations of the type encountered in classical elasticity,
hydrodynamics, and thermal conduction, while consistent with the general scheme of energy
balance, result only from the special assumptions peculiar to those disciplines and would be
false guides in any general approach to energetic theory.
242. Energy impulse. Equations for energy impulse may be gained by considerations parallel to those of Sect. 198 and Sect. 206, using the results of Sect.194.
Apparently, however, this has never been done, and when we set about it, we
find an unexpected complication. Suppose a stress impulse given by the tensor iP'
on a boundary .9 gives rise to a velocity im pulse -{ .ip}· Then there certainly
results an energy impulse on Y, but it is not~ iP' {.ip} da,. Indeed, since iP' da,
[/'
is a diflerence of momenta, while -{ .ip} is a difference of velocities, the difference
of energies resulting will not in general bear any simple relation either to i or
to {x}.
Thus the causes of energy impulse in general we cannot easily separate into
a purely energetic portion plus a portion arising from the impulses of velocity,
momentum, and stress. In full generality we have
(242.1)
where k and n are the total influx and the total supply of energy impulse, but for
the reasons given above we cannot generally write k and n partially in terms of i,
s, and {x}, and thus analysis parallel to that leading from (241.3) to (241.4)
is not possible.
An exception is the case when the motion is generated from rest, for then
we have -{ x }- = i:, {e ;i:} = e i:, and in the non-polar case
kP = f iqP .iq + OkP, n = f sq .iq + 0n, (242.2)
where 0k and 0n are the non-mechanical influx and supply of energy impulse.
In this case (242.1) may be reduced by use of (206.1), so that
(242.3)
243. Balance of energy in a heterogeneaus medium 1• To discuss the transfer
of energy in a mixture, we employ the formalism of Sect. 158, and we proceed
1 Equations of the type (243.1). (243.2), and (243.3), with special forms for eme and m
arise in MAXWELL's kinetic theory of mixtures of monatomic gases [1867, 2, pp. 47-49].
There, however, it is customary to define all quantities in terms of molecular motion relative
to :i:, not to i-m; thus the partial heat flux vector qm of HIRSCHFELDER, CuRnss and BIRD
[1954, 9, Eq. (7.2-25)] is not to be identified with our -hm but rather, subject to the
special assumptions of the kinetic theory, with the negative of the whole expression in brackets
on the right-hand side of (243.2). Our ef'm includes as a special case what CHAPMAN and
CowLING [1939, 6, § 8.1] denote by nm L1 Em- Cm · nm L1 mm Cm; our ee, as defined by (243.1),
- ~ - includes their nE, our Eq. (243.6) corresponds to the sum of their equations L nmLl mm Cm = 0 ~ _ m=I
and L n'MLl Em = o, etc.
m=l
The continuum theory, while more general and simpler in concept, developed only later,
imperfectly, and apparently in oblivion of what had been done long before in the kinetic theory.
Differential equations for the internal energy of a mixture of continua have been given
Sect. 243. Balance of energy in a heterogeneaus medium. 613
as in Sects. 159 and 215. Each constituent m has its own partial internal energy e'll
and is subject to flux of energy h 111 and supply of energy q111 • The total internal
energy e is the sum of the partial internal energies plus the kinetic energies of
diffusion:
R
e= L c'1!(e'1!+-!u~). 111=1
(243-1)
The total flux of energy h arises from three sources: The constituent non-mechanical fluxes of energy h'Jl, the rates of working of the partial stresses against diffusion, and the fluxes of total constituent energies by diffusion:
(243.2)
The total energy of the constituent m need not be balanced in itself, as
energy may be transferred from one constituent to another. Thus, restricting attention to the non-polar case, so that ~"' = t;tk, we define the supply of energy e~1 by
(243-3)
the condition e111 = 0 is then necessary and sufficient that the energy of the constituent m be in balance by itself.
for various special cases and under various special hypotheses by REYNOLDS [1903, 15, § 39].
}AUMANN [1911, 7, §IV], HEUN [1913, 4, § 24c], LOHR [1917, 5, Eqs. (108), (109)] [1924, 10],
VAN MIEGHEM [1935, 9, § 2], MEISSNER [1938, 7, § 3], EcKART [1940, 8, p. 272], MEIXNER
[1941, 2, Eq. (12)] [1943, 2, Eq. (2,8)], VERSCHAFFELT [1942, 14, §§ 15-16] [1942, 15,
§§ 7-8], PRIGOGINE [1947, 12, Chap. VIII, § 3], KIRKWOOD and CRAWFORD [1952, 12,
Eqs. (12) and (18)], and later writers. These authors write down their differential equations
essentially by inspection, without derh.•ing them from the equations governing the constituents,
and without unequivocal specification of the total internal energy in terms of constituent
energies. ECKART [1940, 8, p. 271] hinted at the definition (243.1) but in the end neglected
the kinetic energies of diffusion; they are included by KIRKWOOD and CRAWFORD [1952, 12].
An equation of the form (243.3) with EIJl = 0 was proposed by LEAF [1946, 7, Eq. (14)]; from
the kinetic theory it is clear that such an assumption is usually false. Thermodynamic writers
often pass over mechanical aspects rather cavalierly; typically (e.g. DE GROOT [1952, 3, § 44])
they replace or supplement the mechanical power term t~ "'dk ". by expressions containing
some or all of the thermodynamic power terms in (255.15). Thus it is not surprizing that
the results do not always agree with one another and do not contain all the terms in (243.9)
or any Counterpart of (243.6). A discussion similar to ours is given by PRIGOGINE and MAZUR
[1951, 21, § 3] [1951, 17, § 2], but their definitions differfrom (243.1) and (243.2) in including
terms depending on potential energy and in employing the velocities ~'l! rather than the
diffusion velocities Ul]l; thus their definitions of e and h do not reduce to e1 and h1 when
S't =I. Also, they do not derive any counterpart of the condition (243.6), but it may be
implied in the equation they write down for the balance of potential energy. Cf. also the
special case considered by NACHBAR, WILLIAMS and PENNER [1957, 10, §V].
Our treatment follows TRUESDELL [1957. 16, § 7]. In view of the divergences among thermodynamic writers, it appears necessary to emphasize that all results in the text stand in
detailed consistency with their counterparts in the kinetic theory. In particular, that the
correct energy equation for the mixture should be of just the sameform asthat forasimple
medium, viz., (243-7), has long been known; cf., e.g., the careful derivation of HIRSCHFELDER,
CURTISS and BIRD [1954, 9, Eqs. (7.2-49) and (7.6---7)]. The difference in results obtained
by thermodynamic writers arises only partly from their apparent use of e1 rather than e
(cf. our analysis in Sect. 259) but seems tobe inherently unresolvable because of their failure
to define t and h in terms of quantities associated with the constituents.
614 C. TRUESDELL and R. TouPrN: The Classical Field Theories.
Summing (243.3) on ~. we obtain
where
.R .R ' '\' A ~{ (' 12) "k (! .wE~= 2..,. (!'l( Eil(+ 2U'll - (!'llU~kX~I- (!'l(q'l(- 'll~1 ~~1
- (t~m- (!'llU~u~i)xk,m-
- [ht + t~m u'llm- (!'ll (EIJl + i- Ufu) utJ.k +
+ t~':'ku'llm- [eiJl(EIJl + i- Ufu) u~J,k},
= e i-tkmxk,m- hk,k-
.R
- 2::{e c~ (E~+i-ut) + u~k(e'l(x~- ttml + e~q~}, ~~1
= (!E- tkmdkm- hk,k_ (!q-
.R
- e ~ [c'll(E~ + i- ut) + Nru'llk],
'll~1
.R
q ==:= L Cll( (q'll + f;t UIJlk) · &~1
Sect. 244.
(243.4)
(243.5)
To derive (243.4), at the first equality we have used only algebraic rearrangement,
(158.7), and (158.11); at the second equality we have used (243.1), the fundamental
identity (159.5), and (215.1) and (243.2); the last equality follows by (215.2).
From (243.4) we see that a necessary and sufficient condition for the non-polar
case of (241.4) to hold for the mixture is
.R '\' [A A k A ( 1 2 )] .w E~+P'llu'llk+cm E~+2u~ =0. (243.6) &~1
This result, analogous to (159.4) and (215.5), asserts that the energy supplied
by an excess internal energy rate, plus the energy supplied by the work of the
excess inertial forces against diffusion, plus the energy supplied by the creation
of mass, must add up to zero for the mixture.
Wehave shown that for the non-polar case the condition (243.6), along with
the definitions (243.1), (243.2) and (243.5), Ieads to an energy equation of the
form
(243.7)
for the mixture. For later use we require this same equation put in terms of
the inner parts Er, hr, and qr of E, h, and q:
.R
er= LC~E~,
'll~1
.R
qr == L c~ q~. ~1~1
(243.8)
Either by transforming (243.7) or by summing (243.3) on ~ and then using
(159.5), (215.7), and (243.6), we obtain
(!Er=t1m dkm + .R 1 t~m U~:,m + (h1- 'll~ .R 1 (!& E~ u~) ,k + I
+ e qr - e L [Pt u2tk + c~. t Ufu]' 'll~1
(243.9)
where tr is defined by (215 .6).
244. Remarks on the generat energy balance. The interconvertibility of heat
and ~ mechanical work is expressed only indirectly in (241.4) and (241.6), or, for
Sect. 245. Thermostatics and thermodynamics. 615
that matter, in the basic equation (240.1). In the interpretation, the variables
:0, h, q, and s are to be associated at least in part with thermal phenomena, yet
they are measured in mechanical units. In fact,
dim :0 = dim ~ = [M L 2 T-3], l
dim s = dimi2 = [L 2 T-2],
dim h = [M T-3], dim q = [L q-3].
(244.1)
In order that h may be connected with a temperature gradient, for example,
the mechanical equivalent of heat must be used. Only in its tacit assumption,
necessary for its intended applications, that all thermo-energetic phenomena may
be measured in mechanical units, does the energy balancebring in any new physical
idea beyond those used in pure mechanics.
Indeed, with an equation such as (241.4) we appear to be further from solving
any problems concerning energy than we were in Sect. 217. To secure balance
of energy, we have introduced three new quantities s, h, and q, and only one
condition connecting them. In any given motion, (241.4) may be used for each
particle X as a definition of s to within an additive constant, as a definition of h
to within an arbitrary solenoidal field, or as a definition of q, when two of
these quantities are assigned arbitrarily. But this is not the way in which
(241.4) is used in practice. Rather, its new variables correspond to physical
ideas, and the simple structure set up in this subchapter serves as a framework
(cf. Sect. 6) within which more special considerations concerning changes of
energy may conveniently be expressed. Some of these special assumptions will
be discussed in the remainder of the chapter.
II. Entropy.
a) The caloric equation of state.
245. Thermostatics and thermodynamics. The reader who has no preconception of thermodynamics may pass over this section, entering at once into the
theory in Sect. 246.
1. The classical difficulties. Thermostatics, which even now is usually called
thermodynamics, has an unfortunate history and an unfortunate tradition. As
compared with the older science of mechanics and the younger science of electromagnetism, its mathematical structure is meager. Though claims for its breadth
of application are often extravagant, the examples from which its principles
usually are inferred are most special, and extensive mathematical developments
based on fundamental equations, such as typify mechanics and electromagnetism,
are wanting. The logical standards acceptable in thermostatics fail to meet the
criteria of most other branches of physics; books and papers concerning it contain
a high proportion of descriptive matter to equations and results. The obscurity
of its concepts is witnessed by the many attempts, made alike by engineers,
physicists, and mathematicians and continuing today in greater number, to
reformulate them and to set the house of thermostatics in order.
The difficulty of the subject lies partly in its task of comparing different
equilibria without describing the intermediate states whereby bodies may reach
equilibrium. At the outset, the reader is told to imagine a system changing so
slowly as to be in equilibrium at all times, for such paradoxical "quasistatic
processes" are to furnish the main subject of the theory. The critical student
must long have realized that some kind of linearization is involved; in the shadows
behind classical thermostatics must stand a better theory including motion and
616 C. TRUESDELL and R. ToUPIN: The Classical Field Theories. Sect. 245.
change of motion, and from this theory the classical quasistatic process should
result when kinetic energy, diffusion, and inhomogeneity are neglected. In addition to quasistatic or "reversible" processes, the classical theory attempts to
deal with certain "irreversible" processes, but rather than describing their course
in its formal structure, it merely lays down prohibitions regarding their outcomes1•
Here too the traditional obscurities arise from an incomplete description 2 of the
phenomena the theory is envisioned as representing.
2. "Irreversible" thermodynamics. Both these difficulties are cleared by a
simple expedient: The basic equations of classical thermostatics are applied to
elements of volume in a moving material, or in a mixture of materials. This
device of a local thermostatic state 3, when employed in conjunction with the
general principle of energy given in Subchapter I, leads to a theory in which
equilibrium is but a special case of motion and change, and through which many
processes considered "irreversible" in thermostatics are described easily and
naturally, at least in principle. This theory, usually called "irreversible thermodynamics ", we prefer to call simply thermodynamics 4•
That so great a difference in scope can result from transferring long familiar ideas to
a local scale should not be surprizing. It is typical of the success of the jield viewpoint (cf.
Sect. 2). Similarly, problems of fluid motion which could be treated only grossly by the
mechanics of "bodies" (i.e., mass-points) were solved successfully 200 years ago by hydrodynamics, which applied the principles of mechanics to the continuous field. What is surprizing is that, granted occasional exceptions, the dynamical theory of heat should have
remained for nearly a century in a stage of dElvelopment analogaus to the theory of masspoints in mechanics.
Even though both theories employ parallel concepts on different scales, beyond
the first few steps we cannot expect simply to read off results for the thermodynamic field from counterparts in thermostatics, any more than hydrodynamics
can be read off from mass-point mechanics.
). The concept of entropy. Much of the wordiness of the traditional presentation grows from its insistence on justifying the basic assumptions by experience,
and in particular on developing the concept of entropy in terms of heat and
1 Cf. the remark of PARTINGTON [1949, 22, Sect. li, §51]: "The thermodynamics of
irreversible processes is entirely qualitative and of little interest in physical chemistry."
This remark applies to the traditional view, to which the text alludes, but not to the more
recent studies mentioned in footnote 1, p. 618, andin Sect. 306.
2 Cf. the remarks of DuHEM [ 1904, 2]: "La Thermodynamique ne possede pas de moyens
qui suffisent a mettre completement en equations Je mouvement des systemes qu'elle etudie ... ",
etc. Hence arise the peculiar difficulties of the theory of thermodynamic stability (Sects. 264
and 265).
3 The earliest examples are cited in Sect. 248 in connection with the thermal equation
of state for gas flows. The first systematic treatments of the energy and entropy fields in a
deformable medium were given by J AUMANN [1911, 7] [1918, 3] and LoHR [1917, 5] [1924, 10];
their work is difficult to study because its main object, the explanation of a set of linear
constitutive equations intended to describe all physical phenomena known to the authors,
has lost what interest it may have had (cf. Sect. 6), and from the maze of calculation, which is
highly condensed despite its length, the reader can scarcely disengage the physical principles.
It is clear, however, that J AUMANN and LOHR deserve great credit for realizing the nature
and importance of the production of entropy and for being the first to derive differential
equations for it. Cf. also the early exposition of DE DONDER [1931, 4, § 6].
4 A description of "irreversible thermodynamics" in classical terms would be: Valurne
elements are assumed to suffer only reversible changes, possibly resulting in irreversible
changes for the body as a whole. Such terms can be rather confusing, as when PRIGOGINE
[1947, 12, Chap. IX, § 1] interprets (255.1) as an assertion that the mean motion of a mixture
does not produce entropy and is thus a "reversible phenomenon", even if accompanied by
viscosity, diffusion, and the conduction of heat. We prefer to cleave to equations and eschew
verbalisms.
Sect. 245. Thermostatics and thermodynamics. 617
temperature 1• In the more highly developed parts of theoretical physics, such
discussions do not ordinarily form a part of a treatise on the theory itself, but
belong rather to works on the physical foundations and on the connection between theory and experiment. While it is true that the physics laboratory does
not contain an entropy meter, the concept of entropy is not more difficult than
some others, such as electric displacement 2 ; even temperature and mass prove
elusive to critical inquiry.
A glance at the equations of the theory, once the preliminary words are past,
shows that thermodynamics is the science of entropy. This is true even more of
recent works on irreversible processes than of the classics on thermostatics.
4. The nature and scope of our presentation. An axiomatic development, deriving entropy from heat and temperature, would be desirable, but in our opinion
there exists no acceptable treatment of this kind 3. As in the other domains
presented in this treatise, we are content to explain the formal structure of the
theory as it is practised (cf. Sects. 3, 196). Thus we take entropy as the primitive
concept in terms of which thermodynamics is constructed 4• Surely, any future
axiomatic treatment if successful will lead to the same equations as those from
which our presentation begins. For readers who prefer arguments concerning
steam engines, we cite the original memoirs on thermostatics 5• Neither do we
1 Hence results an apologetic tonein many recent works on "irreversible thermodynamics ",
which often find it necessary to discuss whether or not it is "meaningful" to speak of temperature and entropy for systems not in equilibrium. Often included are arguments from
statistical mechanics. While a rigorous development of equations governing entropy and
temperature from general statistical mechanics would be most illuminating (cf. Sect. 1),
all attempts thus far rest on formal approximation procedures in the kinetic theory of monatomic gases or on the theory of small perturbations from statistical equilibrium, so that
their validity is confined a fortiori to physical situations far less general than those which
the results they claim to derive are intended to represent. In any case, it is not right to single
out thermodynamics as the only branch of physics where such arguments are in order. Rather,
the development of field theories from statistical theories constitutes a general program of
inquiry (cf. Sect. S). Such a program is outside the scope of the present treatise, the
purpose of which is to explain the field theories as such. 2 The parallel is good. Entropy cannot be measured except in terms of other quantities,
such as energy and temperature; the same is true of dielectric displacement, but the constitutive assumption ~ = e E (cf. Sect. 308) is so common that we are often led to regard ~
as closer to experience than in reality it is. 3 Special mention must be made of the celebrated work of CARATHEODORY [1909, 3],
[1925, 4]; cf. BORN [1921, 2], EHRENFEST-AFANASSJEWA [1925, .5], LANDE [1926, 4], MrMURA [1931, 7 and 8], IWATSUKI and MIMURA [1932, 7], }ARDETSKY [1939, 9], WHAPLES
[1952, 24], FENYES [1952, 6]. CARATHEODORY succeeded in deriving the concepts of absolute
temperature and of entropy from a suitable formalization of the idea of equilibrium and the
assumption that for any state, there is an arbitrarily near state that the system cannot
reach without work's being expended. Despite the mathematical elegance and success of this
approach, we cannot regard it as fundamental for a theory intended to describe arbitrary
changes of energy, where thermal equilibrium is as little to be expected as is mechanical
equilibrium in dynamics. For thermodynamics, it is not equilibrium that is basic, but entropy production. Cf. also the criticisms expressed by LEAF [ 1944, 8, p. 94]. 4 This method is due to GrBBS [1873. 2, p. 2, footnote] [1873, 3, p. 31] [1875. I, pp. 56, 63]
who did not attempt to justify it; it was recognized at once by MAXWELL [1875. 4, p. 195]
and was adopted by HrLBERT [1907, 4, pp. 435-438]. Among modern authors who follow
it, we cite EcKART [1940, 8] and MEIXNER [1941, 2, § 3] [1943, 2, § 2). 6 The traditional theory was developed by CLAPEYRON [1834, I], KELVIN [1849, 4]
[1853, 3), CLAUSIUS [1850, I] [1854, I) [1862, J) [1865, J), REECH [1853, 2], RANKINE
[1853. J], and F. NEUMANN [1950, 2I] (deriving from 1854 to 1855 or earlier). A particularly
careful discussion of the dozens of assumptions, mostly tacit, on which the traditional development restswas given by DuHEM [1893, 3]. A variant system, Jargely unpublished until 1928,
was devised by WATERSTON from 1843 onward; it is recommended and developed by
HALDANE [1928, 3, Chap. II]. Other variants or extensions are proposed by BRONSTED
[1940, 4 and .5] and LEAF [1944, 7]. Brief histories of thermostatics are given by MAcH
618 C. TRUESDELL and R. ToUPIN: The Classical Field Theories. Sect. 246.
consider it the province of any treatise on mathematical theory to explain how
measurements testing the theory are to be made. Finally, and also in defiance
of the tradition of the subject, we do not consider it any more appropriate here
than in the theory of stress to work out elementary examples based on special
assumptions. This treatise presents the generat theory of the entropy field 1 .
The theory of entropy, like any other theory, has limitations. That there
are many physical occurrences to which it does not apply may be taken for
granted. To justify its inclusion in this treatise, it is enough that there are many
physical situations to which it does apply, and even in the rather limited special
cases now being explored its relevance seems to be waxing.
246. Entropy 2 • At a given place and time in a body, let there be given f
parameters Va which are regarded as influencing the internal energy e. The
assignment of these parameters 3 is made a priori; their totality is the thermodynamic substate. The physical dimensions of the Va are made up of mechanical
and electromagnetic units but are otherwise arbitrary. The most familiar case
is when the thermodynamic substate consists in a single scalar parameter, the
specific volume v. In another common example, the Va are the nine deformation
gradients x7a from a material reference state. In still another, they are the densities or the concentrations of the constituents of a mixture. For all general
[1896, 3, pp. 269-301], DUHEM [1903, 10], CALLENDER [1910, 3, §§ 20-23], and PARTINGTON [1949, 25, Part Il, §§ 23-37] [1952, 15].
In the system of DuHEM [1893,1, Eqs. (43bis) and (56)], [1911, 4, Chap. XIV, §§ 1-2],
the caloric equation of state (246.1) is regarded as only approximate; DuHEM adds a mutual
internal energy resulting from the interaction of all pairs of elements of mass.
The system of FINK will be described in Sect. 253.
Herewe mention the system of REIK [1953. 26] [1954. 19], which replaces the traditional
"second law" by an axiom governing the time change of entropy. This axiom seems to us
tobe a constitutive relation, generalizing the conventionallinear ones (cf. the next footnote),
and thus we do not attempt to present REIK's theory here. We recognize the difficulty of
drawing a firm distinction: Eq. (246.1), on which the rest of the chapter is founded, is itself,
in a strict view, a constitutive equation, and the chapter should stop after Sect. 244. 1 The majority of recent studies on "irreversible thermodynamics" rest on the special
constitutive assumption (cf. Sect. 7) that the affinities arelinear functions of the fluxes. In
the present treatise, these special developments would be as inappropriate as classical linear
elasticity. The Iiterature is too extensive to cite, but we mention the expositians of PRIGOGINE [1947, 12], HAASE [1951, 11], DE GROOT [1952, 3] and MEIXNER and REIK, this
Encyclopedia, Vol. III/2. Although special cases of such linear relations are old, apparently
the first proposal of a general theory was made by DE DoNDER [ 1938, 3]. While great
emphasis is currently laid upon the so-called "Onsager relations ", we are unable to see
in them, at least for the present, anything more than an indication of a special choice of variables; cf. CoLEMAN and TRUESDELL [ 1960, 1 A]. Future analysis may show that they result
by linearization from some as yet undiscovered general principle of invariance.
For the now generally accepted approach to the theory of the entropy field, see the
SUmmary by DE GROOT [1953, 9]. 2 Since the developments of Sects. 246-249 are parallel to those of classical thermostatics, we do not cite referehces beyond those for Sect. 245. except toremarkthat the early
papers concern only the case f = 1, corresponding developments for equations of state with
arbitrarily many variables having been given by GIBBS [187 5. 1], SCHILLER [1879. 4] [1894, 8],
HELMHOLTZ [1882, 2, § 1], DuHEM [1886, 2, Part II, Chap. II] [1894. 2] [1891, 2], and
OuMOFF [1895. 3]. Some of the results of GIBBS are special in that they rest upon a condition
of homogeneity (Sect. 260). The general view of the subject is due to HELMHOLTZ: "Der
Zustand des Systems sei durch () und eine Anzahl von passend gewählten Parametern Pa
vollständig bestimmt." Some of the identities are derived anew and interpreted in linear
thermo-elastic contexts by TING and Lr [1957. 14].
That X may enter the equations of state, so that the thermodynamic behavior of one
particle may differ from that of another, was noted by HuGONIOT [1885, 4] and emphasized
by DUHEM [1901, 7, Part II, Chap. IV, § 1]. 3 In classical thermostatics they are often divided into two classes, called "extensive"
and "intensive", but this distinction is not necessary here. See, however, Sect. 260.
Sect. 246. Entropy. 619
developments, the parameters Va are left unspecified. They are tensor fields of
arbitrary order, functions of place ~ and timet, or, if we prefer, functions of time
for each particle X.
In following the more recent custom of the subject, where, except in the case
of fluids, the variables Va are arbitrary, we feel compelled to caution the reader
that the physical meaning of the results is likewise left uncertain. Results depending only on the possibility of differentiation are in the main rather insensitive
to the choice of variables, but results following from inversion of functional
relations are applicable, in most cases, only to particular choices of the variables va
in any given physical system. Our aim here is to present certain mathematical
features common to all the simpler thermodynamic theories. Compared to the
contents of the other chapters of this treatise, the matter here is not concrete;
in a satisfactory treatment, the variables va should be identified with definite
physical quantities, as they always were in the sturlies of GIBBS.
We have said that the thermodynamic substate is regarded as influencing e.
The basic assumption of thermodynamics is: The substate plus a single further
dimensionally independent scalar parameter sulfices to determine e, independently
of time, place, motion, and stress. That is, we assume that it is possible to assign
a priori a function f such that
e = f('fJ, v1 , v2 , •• • , vr. X) = e('fJ, v, X). (246.1)
The parameter 'fJ is called the specific entropy 1• Its physical dimension, postulated
to be independent of [MJ, [L ], [T] and the electromagnetic units, is traditionally left unnamed, but the dimension given by
dim e [L 2 r-2]
[G] = dim 7J = -dim 7J (246.2)
is called the dimension of temperature.
In any given motion, of course we have e =g(X, t). In a different motion,
a like functional relation of different form, e =h(X, t), will hold. The first
implication of the postulate (246.1) is that we can determine e without knowing
the particular motion occurring, and without regard to the time. In other words,
the value of the internal energy can be ascertained from information which is
static and universal. This information consists in
1. The value of the substate, v.
2. The value of the entropy, 'f/·
3· The functionalform of the relation (246.1).
Thus the role of entropy is that of a specifying parameter. The mechanical and
electromagnetic information expressed by the substate is in itself not enough,
but, for any given substance, the value of the one additional quantity 'fJ suffices
to yield the internal energy of each particle, whatever the motion it is undergoing or has undergone 2• Adjoining the entropy 'fJ to the substate v, we obtain the
1 Entropy is to be identified with the "calorique" of CARNOT (1824). While used by
others, notably by RANKINE (1854), its distinction from the caloric of CLAPEYRON and earlier
writers was emphasized by CLAusrus, who invented the name [1865, 1, § 14]. That thermodynamics is, at bottom, the science of entropy was first made clear by the researches of
GIBBS (1875). 2 This striking property of entropy results only from the field viewpoint. Typically of
field theories, the greater generality obtained by introducing a field which may vary from
point to point is gained at the cost of closeness to experiment. HADAMARD [1903, 11, ~ 107]
remarked that experimental verification of equations of state for !arge masses in equilibrium
gives no indication whatever that local equations of state hold in deforming media. As in
any other field theory, the experimental justification must result indirectly by comparing
the solution of specific problems with measurements.
620 C. TRUESDELL and R. TOUPIN: The Classical Field Theories. Sect. 246.
thermodynamic state: 1], v. The relation (246.1) is the caloric equation of state.
Choice of the form of f in (246.1) defines different thermodynamic substances.
When X does not appear in (246.1), the substance is thermodynamically homogeneous. In the present treatise, the form of the caloric equation of state is left
arbitrary, except for some inequalities to be laid down in Sects. 263 and 265.
The meaning of (246.1) is most easily seen from Gibbs' diagram1, where e
is represented by a height over the space of thermodynamic states. For the case
when f=1, such a diagram is shown in Fig. 43. The relation {246.1) is then represented by an energy surface, whose form is fixed once and for all for each
V I
v*, TJ*
Fig. 43. GIBas' diagram.
substance considered. Suppose we have a
motion in which a particle X is carried from
a state v, 17 at timet to a second state v*, TJ*
at timet*. During the motion, e, v, and 'fJ will
vary: e =e(t), v =v(t), 'fJ =TJ(t). By (246.1),
the motion in space is mapped onto a curve
on the energy surface, and the changes in
energy and thermodynamic state during the
motion must be consistent with this fact. Conversely, however, the energy surface does not
determine the motion, and many different
motions may be mapped onto the same curve
on the energy surface. Moreover, if two motions
carry the particle from the state v, 'fJ to the state
v*, TJ* by different paths on the energy surface and over different intervals of time,
the energy increments are the same, being determined by the form of the energy
surface.
The dimensional independence of 'fJ fumishes the avenue by which thermal
phenomena are to be connected with mechanical and electromagnetic actions.
Were it not for this dimensional independence, 'fJ would be but another parameter in the set of Va. An essential part of the physical content of the caloric
equation of state is that thermal information must be added to mechanical and
electromagnetic information if we are to determine the internal energy without knowing the motion.
Since e and 'fJ are dimensionally independent, any relation between them
must involve constants e0 and 'fJo having the same dimensions as e and 'fJ, respectively. Thus the caloric equation of state {246.1) must really be of the form
.!____ = t (l, v, x). ~>o fJo (246.3)
Dimensional invariance imposes some restriction upon the manner in which v
enters (246.3) also, but until the physical dimensions of the Va have been rendered
explicit, nothing definite can be concluded. Even after the introduction of reduced
variables, it is not clear what geometric structure, if any, is to be adduced for
the thermodynamic space of points B-TJ-V. In particular, there is no reason
to think of it as metric. But until the space itself is fully defined, we are at a
loss to know what properties of the energy surface can. have physical significance
for a given substance, apart from the accidents of its representation 2• Invariance
requirements for thermodynamics have never been established.
1 The method of GIBBS [1873, 3, pp. 33-34] was adopted and made known by MAxWELL [1875. 4, PP· 195-208]. 2 Cf. the remarks of GIBBS [1873. 3, pp. 34-35] and L. BRILLOUIN [1938, 2, Chap. I,
§§VIII-IX].
Sect. 247. Temperature and thermodynamic tensions. 621
247. Temperature and thermodynamic tensions. The temperature () and the
thermodynamic tensions Ta are defined from (246.1) by
(247.1)
Hence for any change whatever in the thermodynamic state of a given particle X
we have1
!
ds = Odrj + LTadva.
a=l
(247.2)
The temperature and the tensions are the slopes of the curves of intersection
of the energy surface with planes parallel to the co-ordinate planes in the diagram
of Sect. 246. The temperature measures the sensitivity of energy to changes
in entropy; the tensions, to changes in the corresponding parameters. When
v1 is the specific volume, - T 1 is called the thermodynamic pressure, :n:. In the
case of a homogeneaus mixture, when the substate includes both the total
volume and the masses of the constituents, and when the caloric equation of
state is taken as referring to the whole mass, then the tension Ta corresponding
to the mass of the constituent 58 is called its chemical potential 2, ,u~ (cf. Sect.
260). When Va is the deformation gradient x~,.. we may set ,Tt"- Ta/eo and
obtain a double vector of elastic stress analogous to (218.6). See Sects. 25 5 to
256A for development of the properties of these coefficients.
In all cases, from (247.1) and (246.2) follow
dim () = [8], . [L2 r-2J d1m Ta = -d-. --. Imva (247.3)
Thus temperature is dimensionally independent from the thermodynamic tensions.
Also, from (247.1) and (246.1) it is immediate that
() = () (rJ, v), (247.4)
The temperature and the tensions are functions of the thermodynamic state 3•
Since (247.2), being the result of differentiating a scalar relation with fixed X,
is valid for all paths on the energy surface for a given particle, as a special case
it is valid along the curve onto which is mapped the actual motion of the particle X. Hence
!
e = ()i; + L TaVa. (247.5)
a=l
1 A special equation of this kindwas derived by GIBBS in bis first work [1873. 2, Eq. (4)]
but is taken as the starting point of bis later work [1873. 3, Eq. (1)] [1875. 1, Eq. (12)] and
hence is often called "the GIBBS equation ". For references to related earlier work, see Sect. 248.
2 GIBBS [1875. 1, p. 63]. 3 The temperature is easier to interpret physically than is the entropy, and for this
reason most treatments of thermodynamics prefer to take the temperature as a primitive
concept and then introduce the entropy as a defined concept. We have remarked upon this
in Sect. 245, No. 4, and further remarks are given in Sect. 250. Our reasons for preferring
the present course are two: ( 1) it is formally much simpler, and (2) in our opinion no existing
treatment along classical lines is logically clear. A possible formal alternative would be to
take free energy 1p rather than internal energy e as primitive, define TJ through (251.4)3 , then
define e through (251.1) 2 , but to us this seems indirect and more difficult to motivate. This
same criticism applies to the work of DuHEM [1893, 1] [1911, 4] who always started from the
free enthalpy {;", taken as a function of 8 and the substate.
622 C. TRUESDELL and R. TouPIN: The Classical Field Theories.
For arbitrary changes, however, we have
f
Tt 06 = () Tt OTJ + a-s '\'
ia Tt OVa + axa.
oe Tt' axa. l
f
e,k = Oru + L ia Va,k + a~a. x~k· a=l
Sect. 247.
(247.6)
In a homogeneaus material, the last terms in these expressions vanish.
We lay down an assumption of regularity: All thermodynamic functional
relations are differentiable as many times as needed and are invertible to yield
any one variable as a function of the others. The caution stated when the substate v was introduced in Sect. 246 must be borne constantly in mind: a strong
restriction not only on the functional forms admissible for the caloric equation
of state (246.1) but also on the choice of the variables Va is implied; in particular,
it is thus assumed that various partial derivatives occurring are of one sign. While
some related inequalities will be derived in Sect. 265, an adequate analysis of
the nature of equations of state and of the singularities they may possess is
not available.
We lay down also a notation for partial derivatives: A subscript denotes the
variables held constant, and
(247.7)
Moreover, we agree to hold X constant in all differentiations unless the contrary
is noted explicitly. In this notation, (247.1) reads
(247.8)
In all such expressions it is understood that the quantity being differentiated
is regarded as a function only of X and of the variables actually written in the
denominator and the subscripts.
Rates of change subject to the condition 'YJ = const are called isentropic. Thus
the thermodynamic tensions are the rates of change of the internal energy in
isentropic changes of the substate. Other isentropic rates are defined by 1
( 07:a) ab- OVb q,ub' (247.9)
As a consequence of the assumption of smoothness, we may invert (246.1)
and obtain 'YJ = 'YJ (s, v). Hence
f
d'YJ = (:~lds+ t;):~J. dva, l
(247.10)
= (:~t[()d'YJ+ t/adva] + at):~J,uadva.
Equating coefficients of differentials yields
(247.11)
1 When v1 =V, we have v2 cjl11 = (onfoe) 11 = yU2 , where V0 is the Laplacean speed of sound.
Cf. (297-13).
Sect. 248. Thermal equations of state. 623
248. Thermal equations of state. From the assumption of smoothness it follows also that we may invert (247.1) 1 and obtain
1J = 'YJ (0, v). (248.i)
Substituting this into (246.1) yields a functional relation of the form
e = e(O,v). (248.2)
But also we may substitute (248.1) into (247.4) 2 and obtain
Ta= Ta (0, v). (248.3)
Similarly,
Va = Va (0, T) · (248.4)
The relations (248.)) and (248.4) are the thermal equations of state 1• They
are conveniently represented by surfaces over the subspace 0-v; a set of such
surfaces may be called Euler's diagrams for the particle or the subspace 2•
The following coefficients may be calculated from the thermal equations of
state:
~ - ( ~~a )T • ßa = ( ~~a )., • ~ab = ( ::: \,.,b V ab = ( ::; )o,Tb • (248.5)
so that
l f
dva=LVabdTb+ocadO, dTa=L~abdvb+ßadO. (248.6) b~l b~l
Since the coefficients (248.5) are rates of change of measurable quantities, they
are useful for inferring the forms of the thermal equations of state by experiment.
When v1 is the specific volume, oc1fv 1 is called the coeflicient of thermal expansion
or isobaric compressibility, -v11fv1 is called the isothermal compressibility, and
-ß1 is called the Pressure coeflicient. Th(coefficient((248.5) are related by the
identities
ßa + L
r
~abOCb = 0, OCa + L
r
Vabßb = 0,
l b~l b~l (248.7)
LVac~cb=t5ab• L~acVcb=t5ab• c~l c~l
As compared with the caloric equation of state, the thermal equations of
state offer the advantage of connecting easily measurable quantities, the disadvantage of being insufficient, despite their number, to determine a11 the
thermodynamic properties of the material. The latter statement will be proved
in Sect. 251.
In a theory in which entropy is not regarded as a primitive, the thermal equations of
state are often taken as the first postulate, and from them an argument leading to the caloric
equation of state is constructed. The heat increment ( Q), the excess of the increment of internal
energy over thermostatic work is defined by3
l
(Q)==de-~Tadva. (248.8) a~l
1 For a perfect gas in equilibrium, the thermal equation of state :n;v = A (0- 00) was
established by combining the experimental results of BoYLE (1662) and of AMONTONS (1699),
rediscovered many times subsequently. This equation, along with the more general equation
n=/(0, v) and other of its special cases, was used regularly by EULER [1745, 2, Chap. I,
Laws 3. 4, 5] [1757, 1, §§ 17-18] [1757, 2, § 21] [1757. 3, §§ 29-31] [1764, 1] [1769, 1,
§§ 24-30, 90-108] [1771, 1, Chap. V], but did not reappear in field mechanics until the
work of KIRCHHOFF [1868, 11], after which it quickly became universal for studies of gas
flows. 2 EULER [1769, 1, § 28]. Cf. also GIBBS [1873, 2]. 3 Special equations ofthisform appear as derived results in the work of CLAUSIUS [1850,
1, Eq. Ila] [1854, 1, Eq. {2)].
624 C. TRUESDELL and R. TOUPIN: The Classical Field Theories. Sect. 249.
( Q) is a differential form in the variables v, 6, since both Ii and the Ta, according to this
approach, are assumed to be known functions of these variables. In general, this differential
form is not integrable. Special arguments and added assumptions are required in order to
conclude that ( Q) is integrable and that an integrating factor is 1/6. Under these assumptions, entropy is defined by integrating
d7J = (~) = ~ [dli- ±Ta dva], (248.9}
a=l
identical with (247.2), and the consequent theory is the same as that we have discussed in
Sects. 247 to 248 and will develop further in the rest of this chapter.
It is also possible to consider a theory in which the thermal equations of state hold but
there is no entropy or at least no caloric equation of state.
249. Thermodynamic paths. Specific heats, I. Treatmentbasedon entropy1• For
a given particle X a tkermodynamic patk is a path 'YJ = 'YJ (Ä), v = v (Ä} in the space
of thermodynamic states 'YJ, v. On any path P, at a given point, any differentiable
function g of the thermodynamic state has a differential, which we may write
as dgp. Isentropic paths are defined by 'YJ=const (Sect. 247}; isothermal paths,
by () = const.
The specific keat "P on the path P is defined by
d7Jp
"P = () (i"O . p
(249.1)
For an assigned path, the ratios of differentials dvapfd'YJp are assigned, and from
(249:1), by use of (247.4}1 , results adefinite value of the specific heat. Wehave
dim "P =dim 'YJ = (L2 T-2 9-1].
By (247.2) we may write
!
dep-1:TadVaP a=l "p= d6p (249.2}
From the assumption of regularity in Sect. 247, we may regard e as a function
of () and u, and thus (249.2) becomes
(~t d6p+ ~[(~)B,u« dvap- TadVaP] l
"P = d6p '
~ (!;t + tJ( :a .. -··I·:::· I (249.3)
In particular, if the path is one on which u = const, "P is called the specific
heat at constant substate and is written "u· Forthis quantity, (249.3) yields 2
_ ( a11) _ ( ae) "u= () 7fii u- 7fii u' (249.4}
Hence (249.3) becomes
!
"' [( ae ) ] dvaP "P = "u + L..J -a- a- Ta ~6 . a=l Va O,u P
(249.5)
From the assumption of regularity, we may regard v0 as expressed in the
form (248.4). Writing ""' for the specific heat at constant tensions 3, by holding 't
1 REECH [1853, 2, Chap. IV], for the case of a fluid. 2 CLAUSIUS [1854, 1, p. 486) for the case f = 1, HELMHOL TZ [1882, 2, § 1) for the general case. 3 DuHEM [1894. 2, Chap. IV, § 7].
Sect. 250. Specific heats and latent heats, II. M. BRILLOUIN's general theory. 625
constant m (249. 5) we obtain
"""- "u = ± [(;-) a- Ta] OC:a, a=l Va O,u
where OC:a is defined by {248.5) 1 •
The ratio of specific heats,
{249.6)
(249.7)
is an important dimensionless scalar. It is connected with the coefficients is the differential form {248.8). Similarly, the latent heats Ara and
Ava are defined by
f f
e- ~ TbVb i.QL e- ~ TbVb (Q) A-ra- b=l
Ava=
b=l (250.2) ia d-ra ' Va dva
In general, such specific heats and latent heats will be functions of time even
for a given path. Definitions equivalent to these were used long prior to the
concept of entropy and the theory of thermostatics. While they serve to record
the results of measurements, they are too generaltobe the basis of a mathematical
treatment.
M. BRILLOUIN 2 proposed assumptions which, while far more general than
those used in the preceding sections, are yet definite enough that the major classical formal properties of specific and latent heats remain valid. In order to
represent the possibility of permanent deformation, he suggested discarding even
the thermal equations of state, replacing them by the differential forms {248.6),
where the coefficients '~'ab•OC:a,~ab•ßa are given functions of V, T, and e. The
resulting differential forms in all 2 f + 1 variables v, T, () are not necessarily
integrable. The identities (248.7) remain valid, but the interpretations {248.5)
are correct only for the integrable case.
The idea is easiest to picture in the case of three variables, d v = v ( v, -r, 0) d -r + IX ( v, -r, 0) d 0.
Consider a closed loop in the -r-0 plane, described by T=-r(t), 0=0(1). As the system is
carried through this closed cycle, the point -r, 0, v traverses a curve on the right cylinder
whose base is the loop. This curve is obtained by integrating
v = v(v, T(l), O(t)) i(t) + ~X(v, -r(t), 0(1)) Ö(t). (250.3)
1 An attempt to study specific heats in something approaching this degree of generality
was made by M. BRILLOUIN [1888, 2], who discussed possible restrictions on the dependence
of" on the substate.
2 [1888, 3, §§ 3, 8-9]. Our treatment is somewhat more compact and general than
BRILLOUIN'S. BRILLOUIN also studied a generalized entropy 1)(V, T, 0) [1888, 3, §§ 12-25].
The criticism of BRILLOUIN's theory expressed by DuHEM [1896, 2, Introd. to Part I] refers
to this generalized entropy, not to the developments presented above. DuHEM [ibid., Chap. I,
§ 1] proposes a generalized free enthalpy.
Handbuch der Physik, Bd. III/1. 40
626 C. TRUESDELL and R. TOUPIN: The Classical Field Theories. Sect. 250.
If the argument v does not actually appear in v and et, as is the case when there is a thermal
equation of state, this curve is closed, and v returns to its initial value when the loop is traversed; indeed, the only possible curve is given by the intersection of the cylinder with
EuLER's diagram. If the form (248.6h is not integrable, however, v will not generally return
to its initial value, and after the system is carried through the closed cycle of values of (}
and T, a change in v will result. But all these changes are reversible to the extent that if
a path may be traversed in one sense, it may also be traversed in the opposite sense, since
the form (248.6h is linear and homogeneous.
We now add the assumption that an internal energy exists, but we replace
(248.8) by allowing the more general assumed functional relation
B = B (t', T, fi). (250.4)
Then from (248.6)1 follows
(250.5)
where the quantities Ba and Ca are arbitrary. With the choices 1 Ba=TJ- 8
8 e , 8e Va
Ca=- -
8-, we get Ta
f
(Q) =a~/TadTa +x"dO, l
= L Ava d Va + Xu d e' a~l
where the latent heats and specific heats are given by
Hence
(250.6)
(250.7)
(250.8)
1 By other choices of the Ba or Ca it is also possible to eliminate d (} and any f- 1 of the
dva and dra. but the result is not illuminating except in the special case f = 1, for which
it is given at the end of this section.
Sect. 251. Thermodynamic potentials and transformations. 627
a result which reduces to (249.6) when (250.5) is replaced by the more special
assumption (248.8). From (250.7), (250.8), and (248.7) we obtain
r r
x .. - "" = L IXa Äva =-L ßo. Ä.To. • (250.9)
0.=1 o.=1
These identities express the difference of specific heats as bilinear forms in the
thermal coefficients 1X0 , ßo. and the latent heats Ä.To, Ä.v0 . Their form is precisely
the same for Brillouin's theory as for the classical special case when there arethermal
equations of state. Thus if one specific heat is known, the other may be calculated
directly from the measurable quantities defined as coefficients in the forms
(248.6) and (250.6). Moreover, the relation (250.9) does not serve as a test for
the existence of a thermal equation of state.
When f = 1, we write ß for ß1 , ct for ct1 , etc., and obtain "T- "v = - ß l. = ctÄ", so that
tke latent keats are proportional to tke differentes of specific keats, and (250.6) become
1 1
(Q)=- ß '"T- "v) dT + "Td0 = -;x- ("T- "v) dv + "vd0, (250.10)
with ß + ct~ = 0, ct + ßv = 0, ~11 = 1. Also we have the alternative form
(Q) ~ ~:d:+T :;T :; ) '' + (:: + t :; ) dT, l ct ß
(250.11)
Therefore
(250.12)
Thus when f = 1 the ratio of specific heats may be calculated from ct, ß, and the ratio dTjdv
in a process where ( Q) = O, or, equivalently, all energy changes are balanced by thermostatic
work. Such a process generalizes the notion of "isentropic path" introduced in Sect. 249.
251. Thermodynamic potentials and transformations. We return to the classical theory based upon the caloric equation of state (246.1). The thermodynamic
potentials1 are named and defined as follows:
internal energy: e ,
free energy : 1p - e - 'Y} () ,
enthalpy:
r
X-e- LTa.Va,
0.=1
r
free enthalpy: C - X - 'Y} () = e - 'Y} () - L Ta Va ·
They are related through the identity
e-"P+C-x=o.
0.=1
(251.1)
(251.2)
Each of the potentials has certain specially simple properties. From (247.2)
and (251.1} it follows that
f f
de = fJ d'Yj + a~f /a dva, d1p _=- 'Y} df) + a~f /a dva, l
(251. 3)
dx = Od'Yj- L VadTa, dC-- 'Yjdf)- L vadTa·
0.=1 o.=l
1 Proportional functions were introduced by MASSIEU [1869, 4 and 5] [1876, 3, §§I
and IV], who was motivated by the desire to express all thermodynamic properties in terms
of functions of 0, u and of 0, 't". Cf. Eqs. (251.7). The theory was elaborated by GrBBS [1875.
1, p. 87]. Cf. also HELMHOLTZ [1882, 2, § 1]. Further comments on the interpretations of
the potentials are made by NATANSON [1891, 4] and TREVOR [1897. 9].
40*
628 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 251.
Hence
o =(~L·
1] =-(::t.
0 = (:~)T, (251.4)
1] =- (!~t.
the first two of these having already appeared as {247.1). From {251.1) and
(251.4) follow the relations
f
u a=l Ta 'I,T {251.5)
'P- e = 0 (::) , X-e= L Ta( :x) a,)
C - e = 0 (~CO) + ± Ta ( !' ) a • u T a=l uTa 8,T
From {251.3) it appears that we may use any one of the alternative sets of
independent variables
1], v; 0, v; 1], T; 0, T (251.6)
and yet be able to calculate all energy changes. The functions appropriate to
the four cases are the four potentials e, 'P· x. C. In particular, (251.4) 3 shows
that from the equations'P='P(O, v) andC=C(O, T), theentropymaybe calculated
as a change of free energy per unit temperature. The name free energy is appropriate for 'P because, as follows from {251.3)2 , it is the portion of the energy
available for doing work at constant temperature. Similarly, by (251.3) 3 it
follows that the entkalpy X is the portion of the energy which can be released
as heat when the thermodynamic tensions are kept constant.
When f =1, the free enthalpy is called the "Gibbs function" or "thermodynamic potential". When f>l, the usage of GIBBS, followed in most texts
today, was somewhat different from ours. In the case when v1 is the specific
volume v, GrBBS set
(251.6A)
even wken f > 1; i.e., in the enthalpy he left out of account the energy corresponding to any parameter other than v1 • This results in a different physical interpretation for X and C if f > 1; in particular, GrBBS' function C is useful in situations
where the temperature is controlled and a uniform hydrostatic pressure is maintained while other thermostatic parameters vary. In adopting here the generalized
enthalpy given by (251.1) 3 , not only do we seek to exploit the neater mathematical
development which will be seen below, but also we recognize that in the general
case in continuum mechanics the thermostatic pressure does not exist or does
not enjoy much importance.
An equation giving one thermodynamic variable as a function of one of the
four sets {251.6) is said tobe afundamental equation1 iffrom it all thermodynamic variables that are not its arguments may be calculated by partial differentiation, functional inversion, and algebraic operations. In this definition, the set of
"thermodynamic variables" is Va, Ta, 1], 0, e. Fundamental equations are
e=e(?J,v), 'P='P(O,v), x=x(?J,T), C=C(O,T). (251.7)
1 GIBBS [1875. 1, PP· 85-92].
Sect. 252. Thermodynamic identities. 629
As will be shown in Sect. 253, an example of a thermodynamic relation which
is not a fundamental equation is a thermal equation of state such as Ta= Ta (0, v).
There is a formal analogy between the equations appropriate to one potential
and those appropriate to another. Examples will be given in Sect. 252. For
the case when f =1, HAYES1 has constructed a systematic and exhaustive method
of permutation to obtain them all.
252. Thermodynamic identities. From the assumption of smoothness in
Sect. 247, we may write down the following conditions of integrability2 for the
four differential forms (251.3):
(252.1)
where T" b stands for the set of all the T's except Ta and Tb. These identities are
called Ma:xwell's relations or reciprocal relations. The expressions for cxa and ßa
are of particular interest in showing that certain rates of change of entropy
can be inferred from the thermal equations of state.
When f =1, the above identities are easily expressed in terms of Jacobians.
For example, (252.1h implies that
o(O, TJ) o(T, v) .
o(v,fi} o(TJ,Vf • (252.2)
hence 3
o(T,v) _ 1
o(TJ, 0) - • (252.3)
This may be regarded as a summary of all the Maxwell relations when f = 1,
since the same procedure applied to any one of the identities (252.1) yields
this same end product, which asserts that a mapping from the plane of 'fJ, 0 to
the plane of T, v is area-preserving.
Again by the assumption of smoothness, we consider both 'fJ and e as functions
of 0 and v, by (247.2) obtaining
OdrJ =de- ±Tadv0 , I
= (::ta;~ + L [( ::Jo,ua- Ta] dva·
a=l
(252.4)
1 [1946, 5]. Earlier SHAW [1935. 6] had given tables based on rather Jaborious calculation.
It has Jong been noted that the relations (251.1) are contact transformations. CoRBEN
[1949. 3] and FENYES [1952, 5] have constructed an analogy to the dynamics of mass-points,
whereby each dynamical formula enables us to write down a thermodynamic identity, certain
new thermodynamical quantities being thus suggested. 2 The results, when f = 1, are due to MAXWELL [1871, 5, Chap. IX] and, for the general
case, to GIBBS [1875, 1, Eq. (272)] and DUHEM [1886, 2, Part II, Chap. II, Eq. (82)]. Cf. also
MILLER [1897. 6]. It was NATANSON [1891, 4, §I] who first remarked that they are conditions
of integrability.
3 Allowing for the fact that CLAPEYRON did not distinguish properly between entropy
and caloric, we may see nevertheless that he gave the essential content of this relation and
of (252.1) 8 [1834. 1, §V].
630 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 252.
Hence we again obtain (249.4) 2, but also1
() (~) - (_!_!_) - T ova 6, ,.a- ova 6, ,.a a • (252.5)
This last result enables us to express the latent heats in terms of entropy changes.
Indeed, specializing the formulae (250.7) 3, 1 to the case when e = e (v, 0), we obtain
Ava = 0 ( :7J ) , uVa 6,ua (252.6)
By the Maxwell relation (252.1) 4 , these results may be put into the forms 2
f
Ava =- 0 ßc., ATa =- 0 L ßb Vba· (252.7)
b=l
While the last formulae express the latent heats in terms only of the quantities
occuring in the differential forms (248.6), the presence of the factor 0 indicates
that the differential properties of the entropy have been used, and in fact test
of the relations (252.7) serves as an experimental check on the integrability of
the form JO.
By differentiating (252.7) 4 and using (252.5), (249.4), and (252.1h we obtain
the following identity of CLAUSIUs 3 :
( o.ilva) -(o~,.) --ß ----e6,. ova 6, ,.a- a ·
From (252.7)1 and (250.9) it follows that
f
""'- "" = - 0 L Clta ßa • a=l
(252.8)
(252.9)
whereby the difference ""'-"u is given in terms of quantities obtainable from
the thermal equations of state. For alternative forms, we use (249.7), (248.7)1 2 ,
and (252.9) to obtain '
0 0 f
Y- 1 =- L ~abotnotb =- L Vabßaßb· (252.10) ~ .. a, b=l ~ .. a, b=l
We now address ourselves to the task of finding an explicit relation between
the coef:ficients a b and m,, defined from the caloric equation of state by (247.9),
and the coefficients ~ab and ß,, defined from the thermal equations of state
by (248.5)s. First we write out the two Hessian matrices whose individual
components we seek to relate:
ll ~~~~=llab m, llj o(u, 71) m, 0/"u '
o(U,7J) Vab Cltc
118(~11 = II ot, ""joll ·
(252.11)
where we have employed only the definitions (247.9), (248.5)1, 4 , and (249.1).
The product of the two matrices on the left-hand side is the unit matrix; there1 CLAUSIUS (1854, 1, p. 486], when f= 1.
2 The first of these relations, for the case when f = 1, is dueto CLAPEYRON [1834, 1, §IV],
with the pr-oviso noted in footnote 3. p.629. It was obtained from the present theory by CLAUsrus,
first for the case of a perfect gas [1850, 1, Eq. IV] and then more generally [1854, 1, Eq. (13a)]. 3 [1854, 1, Eq. (3)], given earlier for the case of a perfect gas [1850, 1, Eq. II].
Sect. 252. Thermodynamic identities.
fore, so also is the product of those on the right-hand side. Thus
f
L ac '~'cb + W"a IXb = Öab,
c~l
f
L UJ"b 'Vb a + () 1Xa/Xu = 0, b~I
f
L abiXb + W"ax-./0 = 0, b~I
!
L W"a 1Xa + Y = 1 • n~I
where for the last identity we have used the definition (249.7).
From (252.12) 4 we have f
Y - 1 = - L W"a IXa.
n~I
Multiplying (252.12h by IXn and summing yields
f
x-.(y-1) =0Lab1Xa1Xb,
a,b~I
631
(252.12)
(252.13)
(252.14)
where (252.13) has been used, while a similar process applied to (252.12) 2 yields
f
Y - 1 = ~u L 'V ab W"a UJ"b ·
a,b~I
Similarly, from (252.12h and (252.12) 4 we have
(J f -1
Y- 1 =-L ab W"aW"b.
;e-. a, b~I
These formulae for y-1 aretobe set alongside (252.9) and (252.10).
(252.15)
(252.16)
Coming now to (252.12) 1 , we multiply by ~be and sum on 6; simplifying the
result by use of (248.7)t, 3 and (252.1) 9 yields
ab =~ab+ W"a ßb ·
From this identity and (252.1) 8, 9 we obtain a reciprocity theorem:
W"a ßb = UJ"b ßa •
If we write
then taking the determinant of (252.12}t yields
(252.17)
(252.18)
(252.19)
v = -:- _= det (Ö1
ab- W"a ocb) ·]
(252.20)
- 1 - L W"a1Xa
a~I
[cf. the steps used in deriving (189.11)]. By (252.13) it follows thatl
=y$. (252.21)
1 For the case when f=I, the identity (252.21) becomes y= (~:)j(~:) . which may
be interpreted as a statement that in a perfect fluid, the ratio of the isentropic and isothermal
speeds of sound is always y, whatever be the form of the equation of state e = e (TJ, v) (cf.
Sect. 297). While this weil known result is often attributed to REECH [1853. 2, p. 414],
he did not derive or state it, although he gave related expressions for ;e,. and ;ev. A different
generalization is given by DuHEM [1903, 10].
632 C. TRUESDELL and R. ToUPIN: The Classica! Field Theories. Sect. 252.
Looking back at (252.11)1 , we take the determinant of each side and expand
the right-hand member according to minors of elements in the bottarn row:
8(-r,O) Ocp f_1
a{u, ") = -;t - L ab IDa lUb, ., " a,b=J
Ocp = "[y-(y- 1)]' ... (252.22)
= _IJ_t = _~}_! "u = () ~ "·· "..- y V'
where we have u:>ed (252.16) and (252.21).
The identities just derived enable information about the caloric equation of
state to be inferred from properties of the thermal equations of state, which are
more easily accessible to measurement. Essential use will be made of them in
Sect. 265 in connection with the theory of stability.
Returning to the system composed of (252.5) and (249.4) 2 , as its condition of
compatibility we derive the following necessary and sufficient condition for the
existence of entropy as an integral of the form fO=drJ:
~) = l"a- () (~) =-()2 ( 8(ra/O)) . (252.23) 8va o,ua 80 u 80 u
Many treatments of thermostatics contain arguments rendering plausible the
validity of (252.23) as a postulate, thus enabling the concept of entropy to be
derived from those of temperature and an internal energy assumed given by an
equation of the form (248.2), supplemented by thermal equations of state (248.3).
While we have developed only the identities following from use of e as a potential, the same processes Iead to corresponding identities if any one of the functions
1p, x. C is used as the starting point. For physical applications, the function C
is often the best suited.
For a concrete example, consider the van der Waals gas, defined by a caloric equation
of state (246.1) given parametrically as follows:
e=Jc(u)du-:. 1J=Jc~u)du+Rlog(v- v), v> 0v (252.24)
where a, R, and 0v are positive constants, and where c (u) is an arbitrary function. From
(247.7) 1 and (249.4) we obtain at once
0 = u, Xv = c (u) = c (0), (252.25)
whereby physical interpretation is attached to the parameter u and the function c (u). From
(247.7) 2 and (252.25) follows the thermal equation of state:
RO a :TC=- T = ---~. (252.26) v- 0v v2
By (248.5). (249.11}, (252.10}, and (252.13) we have
ot= R ß=--R-, :rc _ ~ + 2a 0v ' v - 0v
v2 1fl
a
:rc+vz
V- 0v
2a
v3 '
1
V=T·
y-1 l11=---,
ot
R (y- 1) c(O) = x"- Xv = ~~~~~~ 2a(v- 0v) 2
1 - RO'Ifl
(252.27)
Sect. 253. Thermodynamic degeneracy. 633
The conditions of ultrastability, as shown in Sect. 265, are y > 1 and "" > 0, where
the former is equivalent, alternatively, to v > 0 and ; > 0. Hence with c (0) assumed positive,
a state is ultrastable for this substance if and only if R 0 > 2a (v- 0v) 2 v-3 ; neutral stability
occurs when " >" is replaced by " = ".
When a = 0 and 0v = 0, the van der Waals gas reduces to the ideal gas, which is a
special case of the ideal materials to be defined in the next section. For the ideal gas,
if c (0) = const, u is easily eliminated between (252.24h and (252.24) 2 , so that
(252.28)
where v0 and 1'/o are arbitrary constants connected with the zero of e and the unit of 0. All
states are ultrastable for the ideal gas if 0 > 0, provided, of course, that "v > 0.
253. Thermodynamic degeneracy. When f + 1 is the greatest nurober of
independent variables occurring in any equation of state, either caloric or thermal,
but in one of them less than f + 1 are present, the material is said to be degenerate.
The most familiar example of a degenerate substance is one for which the equation
of state e = e (0, v) reduces to e = e (0). Such a substance is called ideal. From
(252.23) we read off as a necessary and sufficient condition for an ideal material
Ta= 0 fa(v) = Oßa =- Äva' (253.1)
where the steps follow by (248.5) 2 and (252.7h. This example establishes the
contention just following (251.7), since thermal equations of state of the form
(253.1) do not yield the functional form of the caloric equation of state (246.1),
but only the restriction
(253.2)
For an ideal material, inversion of (253.1) yields Va=ga(t:/0) and hence by
(248.5)
f Tb oga
- O(l.a =~I T o(Tb/0) = ha (v). (253-3)
Substitution of these results into (252.9) yields
f f
X,.- Xu =-L OCaTa = L fa(v) ha(v) =F(v). (253.4) a~I a~I
Thus for an ideal material the difference of specific heats, besides enjoying the
special expression (253.4)1 , is a function of the substate only1 . For an ideal
material the identity (248.7h becomes a differential equation for determining ßa
when all the functions - Orxb or hb (v) are known:
f oßa O.l:a-rxb+ßa=O. (253-5) b~I Vb
A second familiar degenerate material is one in which the thermodynamic
tensions are determinate from the thermodynamic substate alone, and conversely.
1 For an ideal gas, rv = - R 0, so that a. = - R/r and ß = - Rfv, and (253.4) reduces to
"T-"v = R [cf. (252.27) 8]. While this celebrated relation is often named after MAYER, the
only possibly related specific statement we have been able to find in his paper [1842, 2,
p. 240] is the unproved assertion, " ... findet man die Senkung einer ein Gas comprimirenden
Quecksilbersäule gleich der durch die Compression entbundenen Wärmemenge ... ". Even
the generaus interpretation of MACH [ 1896, 3, pp. 247- 250], who finds MAYER "so unzweifelhaft klar ... , daß ein Mißverständnis nicht möglich ist," does not impute to him the relation
in question, which seems tobe due in fact to CLAusrus [1850, 1, Eq. (10a)].
634 C. TRUESDELL and R. ToUPIN: The Classical Field Theories. Sect. 254.
Such a material, characterized by the relations
{253.6)
is called piezotropic. From (252.1) 1 follows the necessary and sufficient condition
0=0('YJ)· In other words, the caloric equation of state {246.1) in the piezotropic
case reduces to1
e = eo ('YJ) + eu (v), (253-7)
so that the internal energy may be split into a thermalportioneo and a substantial
portion eu. For the former, by (249.4) we obtain
eo=f~ud(), {253.8)
where ~u=~u (()). By {248.5) and {247.9) we obtain as alternative necessary and
sufficient conditions for a piezotropic material
ota=O, ßa=O, Wa=O ~ab=ab· {253.9)
From (249.10), or alternatively from (252.10) or any of Eqs. (252.13) to (252.16),
follows
y =1, {253.10)
this condition being necessary but not sufficient unless f = 1.
An interesting question regarding apparent degeneracy has been raised by FINK 2 . In
all the formal developments of thermodynamics it is assumed that the f parameters va are
sufficient to describe the physical body under consideration. If this is not so, the theory
may yet be sufficient to describe some, though not all, of the physical behavior of the body.
To see the effect of neglecting the additional parameters necessary for a complete description,
we consider the theory based on
e = f(rJ, V 10 V2 , ••• , Vf, Vf+l• ••• , Vf+m) (25).11)
as exact, and we compare the exact results with those which result from treating the material
as if it were degenerate:
(253.12)
where A 1 , ••• , Am are constants. If f did not in fact depend on its last m arguments, the
material would be really degenerate and the results obtained from (253.12) exact. Since,
by assumption, e does depend on these arguments, quantities such as (J and r 1 , ••. , Tf
calculated from (253.12) will depend on the constants A 10 • •• , Am. That is, different behavior
as an apparently degenerate material may correspond to different values of the unknown
constants A 1 , •.. ,Am.
254. Heterogeneous media I. Compatibility of an equation of state for the
mixture with equations of state for the constituents. We now divide the parameters v into two groups: the set of individual densities e21 and the set of additional
parameters w 1 , .•• , w1, independent of the densities. We assume that each constituent has its own partial entropy 'YJ~r and its own caloric equation of state
~=1,2, ... ,sr. (254.1)
We now define the total specific inner energy er and the total specific entropy 'YJ
of the mixture:
5l
er L c~re'H, 'H=1
1 COURANT and FRIEDRICHS [1948, 8, § 3].
5l
'YJ = L C'H'YJ~l· ~1=1
(254.2)
2 [1947, 5] [1948, 11] [1949, 10] [1951, 6]. Most of FINK's work considers the effect of
adding one additional parameter in the usual thermostatic theory for fluids so as to describe
metastable states, etc.
Sect. 254. Heterogeneous media I. 635
By (254.1) and the definition (254.2)1 , which was given earlier as (243.8)1 , E1 is
defined as a function of all the 1J'11, all the em, and the set w. It is then a definite
mathematical problern to find conditions under which Er depends on the 1J'D. only
through the linear combination 1J defined by (254.2) 2• Thus no thermodynamic
arguments, but only Straightforward use of the theory of functional dependence,
is required to determine whether or not the caloric equations of state (254.1)
for the constituents imply a caloric equation for the mixture:
(254.3)
Indeed, a necessary and sufficient condition for this functional dependence
is the vanishing of the following J acobians:
o(er·11· 1!1• ... , 1!5\•(1)1• ... ,w,)
0(1j1•1IB• f!l• ... , f!ft• (1)1• ... , Wf) ',','
o(e1•11• (!1, ... , (!ft,Wp ... ,wf)
0(1j1•1/s•1IJ• f!2• "., f!ft• W1, ... , WrJ '' '''
iJ(ei•1/• f!1• .. . , f!ft• W1, .. . , Wr)
0(111•112• 1!1• "., 1!5\•1/a• "., Wr) ' ... ,
(254.4)
Since the variables 171• ... , 1JR• e1, ... , eft, w1, ... , Wr are independent, the only
members of the above set of determinants that are not zero by definition are
those of the first line, which reduce to
8(e1.1j)
F(?h.1js), • .. , (254.5)
where e 1 , ••. , eR, w 1 , ••. , w, are held constant. If we set
(254.6)
for the temperature of the constituent ~. the determinants (254.5) may be expressed
in the forms
c1 c3 (01 -03), ... ,}
c2 c3 (02 - 03 ), ... .
(254.7)
Therefore we have the following local theorem1 : I f each constituent of a mixture
has its own caloric equation of state (254.1), and if the total inner energy E1 and
total entropy 1J of the mixture are defined by (254.2), then in order that there be a
caloric equation of state (254.3), it is necessary and sutficient that at each place
and time the constituents fall into two categories:
1. ~ constituents have the same temperature:
(254.8)
1 TRUESDELL [1957. 16, § 10]. The necessity of (254.8) and (254.9) is easily established
alternatively as follows: from (254.2) 2 and (254.3) we have
( --
oe1 ) -cm- (OBI) . o11m p,..,- a11 p, ."·
but by (254.2h
~) _ cm (oem) 011'11. p.w- 01jiJI p,w·
Equating these two results shows that Om = 0 if c'll =I= 0.
636 C. TRUESDELL and R. TouPIN: The Classical Field Theories.
2. The remaining ~- 5} constituents are not present:
c~.l!+l = c~.I!+S = · ··c~.l\= 0.
Sect. 255.
(254.9)
The form of the caloric equation (254.3) is thus derived; it depends not only
on the form of the caloric equatioris (254.1) for the individual constituents but
also on the particular variety defined by the conditions (254.8) and (254.9).
255. Heterogeneaus media II. Explicit form of the Gibbs equation. Although
it is only a matter of functional elimination to determine the caloric equation
of state according to the results just preceding, we cannot expect to find its
explicit form in general. However, in the practice of thermodynamics for continuous media Eq. (254-3) is never used other than through its material derivative following the mean motion:
.1\ ~
Er= 0~ -nv + L ,u~c~ + 1: O"aWa· (255.1) ~=1 a=l
Of course, this equation, often called the Gibbs equation, is a special case of the
differential relation (247.2); the differentials are taken along the mean motion
of the mixture, and the parameters Va are rendered partly explicit as v, c1, .. . , c.l\,
where the c~ satisfy (158.5). The thermodynamic pressure n, the chemical potentials1
.U'11, and the remaining tensions O"a are defined by
n-- (~~)~,c,w' ,U~t=(~;r) ~ +f(rJ,V,c,w), O"a=(:er) a· (255.2) ut 11, v, c , w Wa 11, v, c, w
The arbitrary function f in the definition of ,u~ reflects the indeterminacy
corresponding to the infinitely many possible functional forms for dependence of
8r upon the variables c~, which are related by the condition (158.5). A possible
method of resolving this indeterminacy is to regard the ~-th constituent as the
"solvent", so that 8r depends onlyupon c1 ,c2 , ••• , c.I\-J, not uponc51 ; the functionf
in (255.2h must then be taken as 0, yieldinguniquevalues for ,u1 ,,u2 , ····1-'!il.-I•
but .U!i\ is not defined.
Alternatively, (255.1) may be written in the form
Er= 0~- (n- L !i\ e'1l,u~) v+ L !i\ v,u~~ + L r O"aWa' )
~=1 ~=1 a=l
!i\ !i\ r
= 0~ + ~"'f1 V [,u~ +V (n-!B~l e!B.U!B)] e-n: + a'{;l O"aWa;
(255.3)
that is, the chemical potentials are related to the tensions corresponding to the
individual densities by the following equation:
.1\ ( :;~)~,p'l,w =V [,u'H +V (n-!8~/!B.U!B)]. (255.4)
The quantity on the left-hand side is a uniquely defined thermodynamic coefficient; the identity is valid for all possible choices of ,U'11 !8 V!ß + ~ U!Bv!B+ ~aa.wa.+ ~1J'; Pr + ( h;) + hPo~.P_ + eoq_, (257.2)
,p
by (257.1) and (156.8) 1 we have
H-~ hP~ap = f(L1 + endv, (257-3)
:7' .y
1 The criticism of TRUESDELL [1952, 21, § 33] is directed toward this assumption, which
is usually not stated in discussions of Case 2.
In evaluating the theory, it is essential to recall that its object is to relate the total stress t
to the thermodynamic tensions Ta. Werewe content, as are most writers on thermodynamics,
to accept the Ta as the actual tensions in the material, there would be no problem, since the
Ta may always be calculated from (247.8) 2 or from {251.4),. 2 Our presentation in this section and the next follows EcKART [1940, 7]. Cf. also MEIXNER [1941, 2, §§ 2-3] [1943, 3, §§ 3-4]. Earlier authors were inclined to select one or
another quantity bearing the dimension of [entropy]/[time] and on the basis of some physical
argument call it the "irreversible" production of entropy; this arbitrary procedure has been
criticized by DE GROOT [1952, 3, §§ 1, 82]. However, the criticism is not applicable to the work
of TaLMAN and FINE [ 1948, 31], who in essence follow EcKART's procedure and on the basis
of (257.3) call LI the "irreversible" rate of production of entropy, thereafter deriving other
equations in which LI occurs, e.g.
e () iJ = hf>. P - kP (log O). P + () LI + e q .
Some alternative forms involving the coefficients {248.5) are discussed by HuNT [1955, 14,
§ 1C, d].
3 There is always a question as to how the total entropy of a system should be defined
The definition {257.1), asserting that what we call the total entropy of a body is the sum
of the entropies of the several elements of mass, is unequivocal and isthat customarily introduced in continuum mechanics. We do not attempt to decide whether this definition is
consistent with others, such as those used in statistical mechanics.
Sect. 258. The entropy inequality. 643
where
(25 7.4)
Ya being given by (256.10). Eq. (257.3) for production oj total entropy is an
equation of balance of the type (157.1).
Unlike our earlier examples of equations of balance, (257.3) is not set down
by definition 1. Rather, it is derived from the assumptions expressed by previous
equations, and the quantities occurring in it have meanings in earlier associations.
From (257.3) we see that total entropy may be regarded as flowing into a body
through its boundary at the rate -hj(), where - h is the influx of non-mechanical
energy. The supply of entropy is ilfe +qjß; in particular, by (257.4), a nonzero flow of non-mechanical energy creates entropy only when it is in the presence
of a temperature gradient. Both the supply of non-mechanical energy and the
excess of total working over recoverable working contribute to the total entropy
in just the same: way as to the specific entropy. In summary, total entropy is
changed by the same factors that change:energy, with two modifications:
1. Entropy rate= ( energy rate)/ ( temperature), and
2. In addition to the sources fust stated, there is also an effective supply of energy
of amount hP (log ()), P.
The supply of entropy iJ is given by the bilinear form (257.4). The terms
occurring in this form are the building blocks of recent theories of particular
thermodynamic phenomena 2• It has grown customary to name the two terms
in each summand a thermodynamic force or affinity and a corresponding thermodynamic flux. For example, we may set up the table:
Affinity
(1/ß), k
Corresponding flux
-hk
(! Yaf()
There seems tobe no compelling reason for assigning any one term to one category
or to the other, and usage varies 3• In any case, the terms entering iJ are not
uniquely determined, since, as was remarked in Sect. 157, in an equation of
balance it is trivially possible to shift any term from the surface integral into the
volume integral at will, and conversely, though not always explicitly, to shift
any term from volume integral into the surface integral. Thus we are unable
to see any physical significance in the interpretation of any one term in (257.3)
unless accompanied by interpretation of all the other terms.
258. The entropy inequality4. It is a matter of experience that a substance
at uniform temperature and free from sources of heat may consume mechanical
work but cannot give it out. That is, whatever work is not recoverable is lost,
1 A formally analogaus treatment can be applied to the calculation of the rate of change
of J va di!R but has no interest. Indeed, as is plain from the manner in which entropy was
"'I"
introduced in Sect. 246, its properties differ from those of the parameters va only in physical
dimension.
2 Cf. footnote 1, p. 618.
a Cf. the references cited in Sect. 259. 4 The postulate (258.3), which sometimes shares with (247 .2) the name second law of
thermodynamics, for the case when h = 0 and q = 0 is due to CLAusrus [1854, 1, p. 152] [1862,
1, § 1] [1865, 1, §§ 1, 14-17]; the surface integral was added by DuHEM [1901, 7, Part. I,
Chap. 1, § 6].
41*
644 C. TRUESDELL and R. TOUPIN: The Classical Fie!d Theories. Sect. 258.
not created. Similarly, in a body at rest and subject to no sources of heat, the
flow of heat is from the hotter to the colder parts, not vice versa. The two observations, abstracted and generalized, may be put as follows:
1. PE- P1 ~ 0 when q =0 and () =const, }
2. hPO,p ~ 0 when q =0 and PE -P1 = 0. (258.1)
In both these cases, by (257.4) we conclude that
()LI~ o. (258.2)
When 0>0, by (257.3) we see that (258.2) 1s equivalent to
H-f hP~a/>. ~ J e,l dv. (258-3)
[/' '"f'"
Guided by these special cases, we might set up (258.3), or the equivalent
condition (258.2), as a general postulate oj irreversibility1. Unlike the previous
assumptions of the field theories, it is an inequality rather than an equation.
I t asserts a trend in time for various processes.
The most familiar of these is an adiabatic process, defined by the condition
that non-mechanical energy flow neither in nor out through the boundary, nor
be created or destroyed within the body: h = 0 on !/, q = 0 in "Y. Then (258.3)
yields H ~ 0: In an adiabatic process, the total entropy cannot decrease 2•
If (258.3) is to hold for all bodies, it is equivalent to the local condition (258.2)
restricting the sum of products of affinities by fluxes (cf. Sect. 257). Whether or
not this condition may be broken down into a statement that separate parts,
such as hPO,p, are to be severally non-negative depends on whether or not the
body is susceptible of independent variation of the parameters Va and () or of
sets of these parameters, and whether or not the partial sums selected are scalars
under appropriate transformations 3•
1 We are aware of the unsatisfactory nature of (258.3) in that h and q are not uniquely
defined. However, since q is to include the possibility of arbitrarily assignable sources and
sinks of heat, we cannot restriet its sign or its value.
2 Notice that the specific entropy 1] is not necessarily non-decreasing at all places and
times in an adiabatic process. Cf. MEISSNER [ 1938, 7, § 11].
3 Here we touch on a great mystery of the subject. Writers on irreversible thermodynamics appear to select partial sums at will and then demand that each one be non-negative.
Certainly, however, an arbitrary choice is not justified: E.g., it is not necessary that h1 0, 1 ~ o,
and it would make no sense to require it, since h1 0, 1 is not scalarunder co-ordinate transformations. In dealing with the more complicated situation to be considered in the next
section, PRIGOGINE and MAZUR [1951, 21,_ §3d] demand that certain terms in L1 be separately
non-negative, "tout coupJage entre quantites de caractere tensoriel different etant interdit ... "
but the meaning of this assertion is not clear to us. Cf. also KrRKWOOD and CRAWFORD
[1952, 12, p. 1050]: "We must treat scalars, vectors and tensors separately, for entities of
different tensorial character cannot interact (CuRrE's theorem)." In the publication of
CuRIE [1894, 1] sometimes cited in this connection we are unable to find anything relevant.
Interactions between quantities of different tensorial orders are well known in the kinetic
theory of gases and are illustrated in Sect. 307. Possibly what the writers on "irreversible
thermodynamics" mean to describe is the separation of effects that follows by linearization
of isotropic functions (cf. Sects. 2931] and 307). Also we stumble again over the most
serious gap in the fundamentals of thermodynamics, that the appropriate group of transformations of the thermodynamic variables and the invariance to be required arenot known.
Cf. footnote 2, p. 620.
Sect. 259. Production of entropy in a heterogeneaus medium. 645
If we raise (258.3) to the level of a general postulate of mechanics, then it
is to be applied also to regions containing surfaces of discontinuity. At such
a surface, provided a (erJ)f8t and eqf() be bounded, it is equivalent to 1
(258.4)
259. Production of entropy in a heterogeneaus medium 2• Confining attention
to the non-polar case, we eliminate i 1 between (243.9) and (255.15), obtaining
ft ft
e()'lj=h~,k+ L (e'l!.U'llu;),k- L (!'l!.U'll,ku;+D, (259.1) 'll~l ~[~1
where
ft
h~ = h1 - L (!'ll e'll ut 'll~l
f ft
D == t1m dk m + n d~ - (! L O"b Wb + L t;m U'llk, m - (259.2)
b~I 'll~l
ft
- (! L [P~ u'll k + c'l! (,U'l! + t Ufu)] + (! qr · 'll~I
Therefore
• ( sk) A f! qi erJ- 7r ,k =LJ+e· (259-3)
1 In the case when h = O, this farnaus condition was introduced into gas dynamics by
}OUGUET [1901, 9] [1904, 3, § 2] and ZEMPLEN [1905, 8] (cf. also HADAMARD [1905, 1],
RAyLEIGH [1910, 8, pp. 590- 591]), where it is used to prove that shocks of rarefaction are
impossible, since it may be shown to follow from the conditions of stability (Sect. 265) and from
205.7) that sgn [17] = sgn [e]. For proof, see Sects. 55 to 56 of the article by SERRIN, Mathematical Principles of Classical Fluid Mechanics, this Encyclopedia, Vol. VIII/1.
2 This subject has been discussed by numerous authors, e.g. EcKART [1940, 8], MEIXNER
[1943, 2, § 2] [1943, 3, § 4], PRIGOGINE and MAZUR [1951, 21, § 3] [1951, 17, § 2]. Our treatment follows TRUESDELL [1957, 16, § 12].
As mentioned in Sect. 243, detailed comparison of our results with those of thermodynamic writers is not possible, since they employ an equation for balance of energy obtained
by intuition rather than derivation. An exception is furnished by the treatment of HIRSCHFELDER, CuRTISS and BIRD [1954, 9, §§ 7.6a, 7.6b, and 11.1]; if we presume that their
phenomenological variables are to be understood as including as special cases the corresponding quantities they define explicitly for the kinetic theory of gas mixtures, then our treatment and theirs are in entire agreement up to the point where they introduce a caloric equation of state for the mixture. Instead of our (255.1) they write an equation of similar form
containing i rather than i 1 on the left-hand side. It seems to us that such an equation cannot
be exact: the total kinetic energy of diffusion cannot be a function of static parameters alone,
since any diffusion velocities, at a given place and instant, are compatible with any values
of the local thermodynamic state. From our point of view, HrRSCHFELDER, CuRTISS and
BIRD neglect the quadratic term in (243.1) when they calculate (255.1), although they do
not neglect it when deriving (243.7). This observation accounts entirely for the difference
between their equation for production of total entropy and our Eq. (259.4).
Perhaps motivated by results in the approximate kinetic theory of diffusion, numerous
writers on irreversible thermodynamics (e.g. PRIGOGINE [1949, 24, § 1]) recommend use
of a "reduced" heat flux which in our notation is given by
From (243.2) and (243.8) 2 we see that this reduced heat flux is precisely our - hr if we assume
(as in theories of diffusion) that t21 = - :7l2( 1 and if we neglect the kinetic energy of diffusion.
646 C. TRUESDELL and R. TouPrN: The Classical Field Theories.
where
5t
sk = h~ + L: e~ ,u~ u~' ~=1
5t
= M + L: e~ (.u~r- e~) ~, ~=1
5t
(.)Lf == h~ (log ll),k- ll L; e~ (,u~/ll),k ~+D- eqr, ~=1
5t f
= ( h~- L; e~e~u~) (logll),k + t1m dkm + nd~- e L; a0 w0 + ~=1 0=1
5t
+ L: [t~mu~k.m-e~e (,u~;e),k~-eP~ u~k- ec~(.u~ + t ut)J. ~=1
Sect. 259.
(259.4)
The form of L1 suggests the following possible division of the variables into affinities and fluxes:
Affinity Corresponding flux
5t
( 1/ ll) k - [ h1- L: e~ e~ u~] ~=1
dkm ~ [t~m+ngkmJ
wo - e aofll
U~k,m t~m;e
u~ - [e~ (,u~/ll),k + e P~kfliJ
ec~ 1 [ 1 2] -0 ,u~+2u~
To the doubts mentioned at the end of Sect. 257 must be added the remark that
various alternative forms and regroupings of terms in (259.4) itself are obviously
possible and lead to different selection of affinities and fluxes 1• To mention the
simplest of examples, if as in Sect. 257 we choose some of the parameters Wa
as the x7"', the second and third lines in the above table coalesce into the single
line
as has already appeared in our treatment of a simple medium. Also, part of the
flux on the first line, since it is proportional to u~, could equally well be combined
with that on the next-to-last line.
For heterogeneaus media the entropy inequality is assumed to hold in the
form (258.3), or, equivalently, (258.2). It is customary, though there appears
to be no solid reason for it, to infer that various sums occurring in L1 are separately
1 We share the view of EcKART [1940, 8] that the classification is arbitrary. However,
most thermodynamic writers disagree with this view and with each other's choices. The point
has been discussed from a physical Standpoint by MEIXNER [ 1942, 9, Zusatz bei der Korrektur] and by DE GROOT [1952, 3, §§ 2, 9-13, 18]. MEIXNER [1943. 2, § 3] has also investigated
the invariance of the classification under linear transformation, but only in the case when the
affinities and fluxes are assumed linearly related. Cf. also PRIGOGINE [1947, 12, Chap. IX, § 2],
DE GROOT [1952, 3, §§ 29, 44, 52, 78]. MEIXNER and REIK in § 5 of their article, "Thermodynamik der irreversiblen Prozesse", this Encyclopedia, Vol. 111/2, notice the infinitely
many possible ways of rearranging and regrouping terms in (259.3); they suggest that it
should be done in such a way as to render the source of entropy non-negative in all circumstances, and they assert that the only possibility of satisfying this requirement is given by
the particular form they adopt.
Sect. 260. Differential condition for homogeneity. 647
.S\
non-negative. For example, EcKART1 concluded that ~c~,u~~O; by (159A.5), ~=1
we have equivalently
.S\
t:... "" c~,u~ ~ * ~ 0. (259.5) ~=\1!+1
In particular, if only one compound ~ is being created, it is forming or dissociating
according as its potential difference ,u~ is negative or positive. Thus when but
a single compound substance can result, the reaction proceeds so as to reduce
its potential difference to zero. EcKART concluded also that from his special
cases of (259.3) and (258.3) that
c~~,U~l,kut ~ o; (259.6)
that is, the diffusion current for the substance ~ carries it toward a region of
lower chemical potential. Interesting as are such inequalities, it is not justified at present to regard them as derived from any general principle.
111. Equilibrium.
There are several different ideas of the meaning of "equilibrium"; when put
into mathematical form, they lead to conditions that are not generally equivalent
to one another. These conditions occur frequently in the Iiterature of irreversible
thermodynamics, but their interrelations have been studied only subject to the
assumption of linear constitutive equations. We confine the following sections
to the general definitions.
260. Differential condition for homogeneity. Thus far wehavenot needed to
restriet the variables va specifying the sub-state. Now, however, we presume
that they are the densities of additive set functions. The set functions themselves
are called extensive variables 2• When the thermodynamic state is homogeneous,
which here is taken to mean uniform, throughout a body, we have then
'YJ = HJIDl, e = ~JIDl, Va = TaJIDl, where Ta- f VadiDl,
.."..
and (246.1) may be written in the equivalent form
~ = m e(~ .~) = ~(H, T, IDl), (26o.1)
.S\ 1 [ 1940, 8, esp. p. 924]. The quantity A = (! ~ c~ ,uw was called the "affinity" or "chemical ~=1
affinity" by DE DoNDER and has been studied intensively in connection with chemical reactions; cf. DE DONDER [1927, 3] [1929, 1] [1932, 5], DUPONT [1932, 6]. If we use (159A.6),
we may write the chemical affinity A in the form
( .S\
A = ~ JaAa. Aa = ~ N~a.U~· a=1 ~=1
where { is the total number of chemical reactions that may occur. The quantity Aa is the
chemical affinity of the reaction a. Further developments follow by assuming that A a is
an assignable function of the thermodynamic state, etc. The Iiterature of this subject is
obfuscated by the habit of writing all rates as time derivatives of otherwise undefined quantities. 2 The distinction between extensive and intensive variables is due to MAXWELL [ 1876, 4];
the former kind, which he called magnitudes, "represents a physical quantity, the value of
which, for a material system, is the sum of its values for the parts of the system ", while the
latter "denote the intensity of certain physical properties of the substance". The defining
property of intensive variables is expressed by (260.2) and hence is an alternative statement
of the homogeneity of (260.1) when the densities of the extensive variables are uniform. In
general, then, "intensive variable" is no more than another name for "field ".
648 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 260.
where the function on the right-hand side is homogeneous of degree 1 in all
its arguments. Since by (247.1) we have
0~ - o(me) - oe - () 0~ oe
BH- o(m1J) - BrJ- ' oYa = ova =Ta, (260.2)
the temperature and thermodynamic tensions are independent of the amount
of material present. Such variables are called intensive. The Euler differential
equation expressing the homogeneity of (260.1) is1
G: = 0 H + 1fiaTa Ta +,u Wl, l 8 = O'Yj + L Ta Va + ,u, a=1
where 0~
,u == om ·
By (251.1) 4 we may write (260.3) in the form
c =,u,
(260.3)
(260.4)
(260. 5)
expressing succinctly the relation between the caloric equation of state for the
density, 8, and that for the total energy, Gl: [cf. the analogous result (255.14) 3].
While (260. 3) is the basic expression of homogeneity, other forms are more
commonly encountered. First, from (260.2) and (260.3) we have
l
d@ = 0 d H + L Ta d Ta+ ,Ud Wl (260.6) a=l
as the counterpart of (247.2). If we subtract this from the differential of (260.3)1 ,
we obtain the Gibbs-Duhem equation 2 :
l
0 = H d() + L TadTa + Wld,u. (260.7) a=1
For an alternative derivation, we need only multiply (251.3) 4 by ill( and then
take note of (260.5).
In the usual applications the Va are supposed to consist in the sr masses ill(~
of the constituents of a mixture and in a further set S~ of extensive variables Da
independent of the partial masses. These quantities, in the case of a homogeneous
system, are related to the densities used in Sect. 254 and 25 5 as follows:
9)(2! V
I c~ d ill( = c~ ill(' ~=1
.i wc~ = ill( ' I
Da= I WadWl = Wa Wl. V
(260.8)
From (260.3) we have
! ft
Gl:r = 0 H + L aaDa + L;,u~Wl~ + ,u Wl, (260.9) a=1 ~=1
where
(260.10)
1 GIBBS [1875. 1, Eq. (54)]. 2 GrBBS [1875, 1, Eq. {97)], DUHEM [1886, 2, Part Il, Chap. Il, Eq. (81)].
Sect. 260. Differential condition for homogeneity. 649
and where we have written 1 instead of ~ as areminderthat a mixture is being
considered (cf. Sects. 254 and 255). The indeterminacy represented by the occurrence of /(H, rol, .2) is the sameasthat already encountered in (255.2). So as
to eliminate this indeterminacy, we may agree to regard 1 as a function of
ml, m2, ... , mjl but not also of m; then we may set
(260.11)
It is these chemical potentials ,u~ which generally are used in works on chemical
equilibrium. They have the disadvantage of not being easily interpretable in
terms of densities. However, for any f in (260.10), by (260.4), (260.6)a, and
(260.8) we obtain
,u~ = ,U'll. + ,u '
while (260.6) and (260.7) become, respectively,
d~, ~ 9 d H +,~,"' dQ, +.~f; d!!Jl,,l
0 = H d () + I.J2adaa + L m'l(d,u~. a=1 'li.=I
The latter relation may be written in terms of densities:
f 5l
0 =1Jd() + LWadaa + L c'll.d,u~, a=1 'll=1
(260.12)
(260.13)
(260.14)
but it is important to note that the differentials of the ,u~, not generally the ,U'll.,
occur here. However, the custom among thermodynamical writers of expressing
all relations in differential form can lead to confusion, since the differentials dc'li.
arenot independent. From (158.5) it follows that
By (260.12), then,
.lt
Ldc'll. =0.
'll.=1
5l 5l
L f.l'li. d Cl]l = L f.l~ d C'l(. 'l1.=1 '11=1
(260.15)
(260.16)
Thus the Gibbs equation (260.13) 1 , for homogeneaus conditions, may be written
in either of the forms
ds1 = () d1J +a~ aa dwa +~r~l~ dem, I
f 5l (260.17)
= () d1J + L aadwa + Lf.lmdc'l(, a=1 21=1
as well as in other forms, for variations in which the total mass is kept constant.
It is customary to take one of the parameters !2a as the volume, 78. The
Gibbs-Duhem relation (260.14) then assumes the form
5l f
o = 1J d () - v d n + L c~1 d,urr + L wa daa. (260.18) ~1=1 a~1
While writers on "irreversible thermodynamics" sometimes use the relations
of this section in problems concerning deformation, we are unable to find any
650 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sects. 261, 262.
solid ground for ascribing any relevance to them except in equilibrium. Consequences of (260.13) are
ft ( o.u~) 2>~ -0- ~ =O, ~=l C~ O,n,c ,a
V~=-- ( o.u~) . o:rr: O,c,a (260.19)
Another condition of equilibrium, apparently independent of the foregoing, is the principle of detailed balancing, asserting that in thermostatic equilibrium each chemical reaction
is individually in equilibrium. In terms of the reaction rates Ja occurring in (159A.6), this
postulate asserts that 1J = const and v = const implies Ja= 0 for all a.
261. Mechanical equilibrium. By definition, a system is in mechanical equilibrium if :V= 0 in an inertial system. For simple media, no special thermodynamic
consequences result. For heterogeneaus media, however, by (215.4) and (205.2)
we have
5\
tkm,m + L (J~I ~~ == 0. (261.1) ~~1
If now we add assumptions 1 sufficient to validate the Gibbs-Duhem equation
(260.18), from its two consequences (260.19) we obtain
5\ ft [ ft o.u*) ~2;1 e~(f~,.- .U~.k) =-tk:m- (J ~2;1 c~ ];:1 ( oc: o,",c~./!!.!,k +
+ ( :o~)", c,a e,,. + ( ~~ t. o,a n, k+ bt ( ::: t. "· c, ab ab,k]' (261.2)
=- (t'l: + n b'/:),m- e .± c~ f( ~) e,k + ± ( :.u~) bab,k] · ~~1 l "· c,a b~1 ab 0, "· c, a
Further drastic assumptions are required in order to get a simple result.
If () = const and if the parameters w do not appear or if ab = const for all b,
and if further t'l: =-nb'/:, then (261.2) yields
5\
L: e~u~!k- .U~,k) = o. '2!~1
In this case, by (158.19), we may apply (158.18) to show that
5\ 5\
L (J'H {f~ k - .U~. k) ut = L !?'H (f'l! k - .U~. k) ut '!1~1 'H~l
for all definitions of the diffusion Velocity U'H.
(261.3)
(261.4)
It is also possible to reorganize the terms in (261.2) in a fashion analogaus
to that used in Sect. 256.
According to EcKART2, it should be possible to arrange the affinities and fluxes
(Sect. 259) in such a way that the vanishing of the ones is a definition of equilibrium; the vanishing of the others should then be a provable criterion of equilibrium. This interesting idea does not serve to classify the quantities uniquely.
262. Variational condition of equilibrium. Toward the end of the nineteenth
century arose a tendency to regard all gross phenomena as essentially thermodynamical, whence followed attempts to construct a system of energetics including mechanics as a subsidiary part3. The strength and the weakness of this
1 The idea was suggested by PRIGOGINE [1947, 12, Chap. IX, § 3]; we generalize the
treatment of DE GROOT [1952, 3, § 47]. 2 [1940, 8, p. 270], "D-factors" and "C-factors". 3 Cf. DuHEM [1911, 4], }AUMANN [1911, 7] [1918, 3], LoHR [1917, 5] [1924, 10] [1940,
17] [1948, 12].
Sect. 262. Variational condition of equilibrium. 651
approach are illustrated in the special case when the substate v in (246.1) consists in the 9 deformation gradients x~x from a reference state X. The objective
is to give a purely thermodynamic delinition ol stress, without reference to mechanical considerations, and to derive CAUCHY's laws (Sect. 205) as special cases
of a general criterion of equilibrium 1•
This condition is bipartite. First, the principle of virtual work in the form
(232.6) when äl =0-i.e., ~ =0-along with the postulate
~=t5fediDl, (262.1)
-r
is assumed. Second, thermal equilibrium is defined by
t5 H = o. (262.2)
All variations are assumed to respect the conservation of mass. From (262.1),
(232.5), (247.2), and (156.1) follows
r
Je [0 I5'Yj + L Ta t5 Va] dV = ~ s" t5xk da + Je I" t5xk dV,
-r a=I fl' -r (262.3)
Je I5'Yjdv= o.
-r
The second condition is equivalent to the requirement that 'YJ = const in the variations; in this case, or in the alternative case that (262.3) is used as a side
condition and the variations are executed with (J = const, the former condition
becomes r
Je L Ta t5va dV = ~ s" t5xk da + Je I" t5xk dV. (262.4)
-r a=I fl' -r
Further progress cannot be made unless we specify the relation between the
t5va and the t5x". If all variations are independent, we get
Ta = 0, sk = 0, jk = 0 (262.5)
as the conditions of equilibrium for this trivial case. However, if the Va are in
fact the ~K· as we agreed to assume, then we write (262.3) in the form
f e ( ~;) t5x~x dV = ~ s" t5xk da + f e jk t5xk dV. (262.6)
-r ' [/' -r
Since t5~x = (t5x");K• by GREEN's transformation we thus obtain
0 = J. (s" da - e ~:- dAx) t5x" + f [e jk + (e ~1----). ] t5x" dV. (262.7) 'f" ox.K OX ·K ,K [/' ' -r '
lf there are no kinematical constraints, it follows that
s"da =e~dA" ox;K
o=(-~~) +-1" (! oxk ·K (! ;K'
on .91
in "f'".
(262.8)
1 The analysis we present is due to Grass [1875. I, pp. 184-190]. The theory based on
the more special assumptions E= e(E, X) was initiated by GREEN [1839. I, pp. 248_:_255]
[1841, 2, pp. 298-300] and developed by KrRCHHOFF [1850, 2, § 1] [1852, I, pp. 770-772]
and KELVIN [1855. 4] [1856, 2, Chaps. XIII, XIV] [1863, 2, §§ 61-67] [1867, 3, § 673] and
App. C. § § (c)- (d)]. Cf. Sect. 232. Further attempts along this line have never been successful. The usual approach is to assume e depends on various variables at hand. For example,
VAN MrEGHEM [1935, 9, § 3] sets up the equation e = e(n, 0, A, ih where A is the affinity
and Eis defined by (31.6) 2 , as supposedly appropriate for a viscous fluid.
652 C. TRUESDELL and R. TouPIN: The Classical Field Theories.
By (210.5) and (210.8), the quantities defined by
F,K--~ k -(!~k uX;K
Sect. 263.
(262.9)
in virtue of (262.1) and (262.2) satisfy the same equations as those imposed on
PIOLA's double vector T,.K in virtue of the conditions of equilibrium in continuum
mechanics. Thus it is not inconsistent with mechanical principles if we set
(262.10)
and replace the laws of statics by the conditions (262.1) and (262.2) of thermodynamic
equilibrium. At bottom, this analysis rests on the same idea asthat in Sect. 256A.
If there are additional parameters Va beyond the x7K, the corresponding Ta
must vanish in equilibrium, provided there are no constraints. The details here
are akin to those in Sect. 256.
Despite the elegance of the foregoing analysis, it cannot be accepted. Indeed,
from a special thermodynamic assumption, CAUCHY's first law has been derived.
That CAUCHY's second law cannot be derived by such methods is plain from its
generalform (205.10) and from footnote 2, p. 596. In fact, in general the stresst
derived from (262.1 0) is not symmetric unless some additional assumption, such
as that e depends on the x7K only through E, is added.
But there is a deeper objection. The equations of mechanics describe a wider
range of phenomena than do the principles of thermodynamics. No one will
contest the principles of balance of mass, momentum, and energy, but the existence of a caloric equation of state is an assumption of a more special kind1.
CAUCHY's laws arevalid for all sorts of continuous media, but in essence the thermodynamic method just explained is restricted to perfectly elastic bodies 2• If there
are viscous or plastic stresses, they must be dragged in by extra assumptions having nothing to do with thermodynamic principles 3 • Finally, to obtain equations
of motion it is customary to apply the Euler-D'Alembert principle to the
equations of equilibrium, and this is at bottom no different from assuming the
balance of momentum to start with.
263. Inequalities restricting the equations of state, I. "Absolute" temperature.
Thus far in our formal structure the temperature () has been taken as defined
by (247.1) 1 with no other restriction than that the equations of state be such
as to permit any number of differentiations and functional inversions. For the
validity of one or two of the formulae it has been tacitly supposed that () =f= O,
and at one point in connection with the entropy inequality (258.3) we have assumed that () >O. This last requirement is customarily imposed throughout
the subject. When the caloric equation of state (246.1) is so restricted that
()>O,
1 This is borne out also by general statistical mechanics, whence, as shown by NoLL
[1955, 19], the field Eqs. (156.5), (205.2), (205.11), and (241.4) followin the greatest generality,
but the existence of thermodynamical equations for cases other than equilibrium remains
in doubt. In the kinetic theory of monatomic gases, there is a thermal equation of state in
all circumstances, but a caloric equation of state is valid only in conditions sufficiently near
to equilibrium.
2 Within this Iimitation, CoLEMAN and NoLL [1959, 3] have replaced the analysis in
this section by a rigorous development based upon a genuine minimum principle; they obtain
not only the classical stress-strain relations but also full conditions of· stability. 3 Cf. the method of DUHEM [1901, 7, Chap. I, §§ 1- 5] [1903, 4 and 5] [1904, 1. Chap II.
§V] [1911, 4, Chap. XIV, § 3]. who was a devotee of the thermodynamic approach. Only
the details, not the principles, of more recent allegedly thermodynamic treatments are different.
Sect. 264. Stability of equilibrium. 653
for all allowable values of the thermodynamic state 'fJ, t', then () is said to be an
absolute temperature.
It does not seem to be possible within the concepts of thermodynamics to
give any clear idea of what absolute temperature is1. We may, however, rephrase
the assumption as follows: the caloric equation of state (246.1) is such that ()
has a finite lower bound. By (247.1h this is equivalent to
00 = inf ( ;; )., >- oo. (263.2)
For a given function e ('YJ, v), put
(263-3)
Then
. f ( oe') m 87].,=0. (263.4)
While, as stated in Sect. 245, the requirements of invariance to which the caloric
equation of state is subject are not clear, it seems that e' as an internal energy
function is physically equivalent to e; by (263.4), if () is defined from e' rather
than from e, we obtain (263.1). Thus for an internal energy function satisfying
(263.2) it is possible to find a physically equivalent energy function such that
the temperature is absolute.
The requirement (263.1) has a simple physical interpretation: Addition of
heat to a body increases its internal energy. This is so because 'fJ, while itself
not a measure of heat, is to be interpreted as a quantity which increases or decreases according as heat is being supplied or drawn off. Pursuing this same
interpretation suggests also the additional restriction
lim e=oo: (263.5)
If, at a fixed substate 'L', heat is added indefinitely, the internal energy also
increases indefinitely.
264. Stability of equilibrium. At the conclusion of a paper on various forms
of the "second law of thermodynamics ", CLAUSIUS 2 asserted, "Die Energie der
Welt ist konstant. Die Entropie der Welt strebt einem Maximum zu." GIBBS 3
replaced this claim of a trend in time by a definition of thermostatic stability for
an isolated system:
(L1H) 0. (265.3)
When f is twice differentiable, (265.3) implies that f"(w) ~ 0. Conversely, the
condition f"(x) >O is sufficient, but not necessary, for convexity. Likewise the
more general condition (265.1) implies that
II ~w11 u :11 u is positive semi-definite, (265.4)
where we have put
OB
Ga- owa · (265.5)
This is the result inferred by GIBBS1 •
Suppose that for a given mixture there exists a convex set of local states w
such that the homogeneous state corresponding to each is a state of stable
equilibrium. It follows then from (265 .1) and (265 .2) that for all states ro', w" in
the set we ha ve
e (w") - e (w') - (rJ"- rj') e'IJ (w') - l
.1\-1
- (v" - v') Ev (w') - ~ (cU( - c~) Ec, (w') > 0, ~=1 ~l
(265.5 A)
1 [1875. 1, pp. 111-112]. In his earlier study of simple fluids [1873. 2, p. 29] GIBBS
had written, "The condition of stability requires that, when the pressure is constant, the
temperature shall increase with the heat received,-therefore, with the entropy .... It also
requires that, when there is no transmission of heat, the pressure should increase as the
volume diminishes ... " l.e., the "condition of stability" is
(on) ~ 0
ov 'lj
(equivalently,
""~o. O, are consequences of (265.1). DuHEM's methods did
not enable him to derive (265.8); neveriheless, he knew that this equation holds in various
applications, so he adopted it, calling it "the postulate of HELMHOL TZ".
DuHEM criticized certain inequalities of GIBBS similar to (265.7)3 [1875. 1, Ineqs. (167).
(168), (169)]: "une des rares inexactitudes"; also [1893, 2, Part I, Chap. V, §V] [1894, 2,
Chap. IV, especially §§ 5 and 7] [1898, 2] [1911, §, Chap. XVI, § 9]. Such results are very
sensitive to the choice of variables, and GrBBS is obscure in the detail of argument as weil
as reluctant to state just what is being varied. CoLEMAN and NoLL have straightened the
whole matter out; they find that DuHEM's criticism is weil taken if his interpretation of what
GrBBS meant to say is correct, but they find another interpretation in which GIBBS' inequalities are correct.
In summary, there is no conflict between the results of GrBBS and those of DuHEM, which
complement one another; all their results and many more of like nature are stated unequivocally and derived precisely in the forthcoming work of CoLEMAN and NoLL.
Sect. 265. Inequalities restricting the equations of state, II. 657
where the subscripts to c; denote partial derivatives. I t is clear that (265. 5 A),
for the special variables considered here, is a mathematical statement of the
physical notions behind (264.7). Thus the approach based upon (264.7), when
made clear, amounts to an asserting as a postulate the main result GIBBS sought
to derive.
For the case of a fluid obeying a caloric equation of state of the form (264.11)
-the case to which CoLEMAN and NoLL's analysis applies-we have 1 lJ = ((), -n,
t-t1 ,f1z, ... ,ft~- ), and the matrix occurring in (265.4) assumes the form
ao ao ao ao
OrJ ov ocl OCSt-1
on on on on
-8r/ ov
0/l)
ocl OCft-1 (265.6)
OrJ
0/lft-1
OCSt-1
where all quantities are taken as functions of 'f), v, c1 , ... , cSt_ 1 • In order that
this matrix be positive semi-definite, it is necessary that each principal minor
be non-negative. In particular,
~) ~ 0, OrJ v, c- (an) ~ O,
oe ~.c- (265.7)
The first of these, by (249.1), may be written in the form 2
. (265 .8)
that is, in order to increase the temperature of a body held at constant substate,
it is necessary to supply heat. The second of the inequalities (265.7) may be
interpreted as a statement that the Laplacian speed of sound is real [ cf. footnote 1,
p. 631 and (297.13)]. But also all principal minors of (265.6) must be nonnegative; in particular,
o(n, 0) I ~ O.
o(rJ, v) c- (265 .9)
Up to now we have identified the variables constituting the substate, conformaply to the caution stated in Sect. 246. For more general thermostatic
systems, it is clear that (265 .4) cannot generally hold. For if wb is a suitable
parameter for describing the state of a system, so also are -w0 and 1/wb; use
of - wb, leaving all other Wa unchanged, would change the sign of 8aaf8wb,
and use of 1/wb would change the sign of 8abj8wb. However, it seems likely that
for some admissible choice of the wb in any system, (265.4) will result, and its
simplest consequence,
oaa > O owa = '
has often been taken as a condition for stability.
(265.10)
1 Note that /l~ = ll'H- llSt, m = 1, 2, ... , S\'- 1, where /l~I is the chemical potential
defined by (255.2) 2 . In virtue of (158.5) we have
St St-1 ~ ll'H c'H = ~ ll~ c'1f..
'11=1 '11=1
2 GIBBS [1875. 1, Eq. (166)], HELMHOLTZ [1882, 2, § 1].
Handbuch der Physik, Bd. Ill/1. 42
658 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 265.
We now translate the condition of stability (265.1) for homogeneous states
of isolated systems into forms more a~cessible to experimental test. We have
seen that (265 .4) is a necessary but not sufficient condition for stability; if we
replace "positive semi-definite" in (265 .4) by "positive definite", we obtain a
condition which is sufficient but not necessary. We agree to consider this stronger
condition only1• Also we shall rule out the possibility of thermodynamic
degeneracy (Sect. 253), and we exclude from consideration states such that
IDa = 0 or oca = 0, for any a. Also we adopt (263 .1). A system satisfying these
conditions will be called ultrastable.
By a weil known theorem on matrices, from (252.11) we see that a necessary
and sufficient condition for ultrastability is that II abll be positive definite and
o(T, O)fo(v, 'YJ) >O. From the former statement it follows that ct> >O. By (252.22)a
we conclude that the latter statement, when the former is satisfied, is equivalent
to x .. > 0. Hence we have the
First necessary and sufficient condition for ultrastability:
llabll is positive definite, x .. > 0. (265.11)
Proceeding in just the same way from (252.11) 2 and using (242.22) 5 , yields the
S econd necessary and sutficient condition for ultrastability 2 :
u;a bll is positive definite, Xu > 0. (265.12)
In this condition, ;ab may be replaced by Vab·
From (265.11) and (252.14) or (252.16), or, alternatively, from (265.12) and
(252.15), it follows that y> 1. Conversely, if y> 1 and llc!>abll is positive definite,
from (252.14) or (252.16) it follows that x .. >O, whence by (265.11) the equilibrium is ultrastable. · Thus we have proved the
Third necessary and sutficient condition for ultrastability:
llc!>abll is positive definite, y>1.
In just the same way, from (265.12) and (252.15) we prove the
F ourth necessary and sutficient condition for ultrastability:
n;a bll is positive definite, y > 1 .
(265.13)
(265.14)
The stability of thermodynamically degenerate substances requires a separate
analysis. The particular kind of degeneracy which defines an ideal material
(Sect. 253) does not affect the arguments given above, so the conditions of
stability (265.11) to (265.14) remain applicable. For piezotropic substances, by
(253.9) and (253.10) it is easy to see directly from (252.11) that a necessary and
sufficient condition for ultrastability is
llvabll is positive definite, xu>O or x .. >O. (265.15)
We do not attempt to find conditions appropriate to other degenerate materials
or to investigate the conditions of neutral stability.
1 This was done by SAUREL [1904, 6 and 7], who phrased GrBBs' considerations more
formally in terms of differentials. Our text, in effect, replaces the Gibbs-Saurel arguments
by more efficient and precise mathematical analysis. It appears that the subtle logical
difference between stability and ultrastability was recognized by GrBBS, though he did not
emphasize it; in the work of DuHEM, it is completely obscured, and a reader with modern
standards of rigor will need to apply some Iabor if he is to disentangle DuHEM's arguments. 3 The condition (265.12h, when f = 1, is traditionally used in analysis of the stability of
a VAN DER WAALS gas, as was mentioned in Sect. 252.
Sect. 265. Inequalities restricting the equations of state, II. 659
An important consequence of (265.12) has been derived by EPSTEIN1• Consider two tensions, Ga andab; then Ga =aa(wa,Wb,wab), ab =ab(wa,wb,wab), where
wab stands for all wa's except Wa and wb. The Maxwellrelations (252.1) may all
be expressed in the form
(265.16)
Inverting the relation giving ab yields wb = wb(ab,Wa,wab) so that Ga= aa(wa,
wb (ab, Wa, wab), wab). Differentiating this relation yields
But
( oaa ) ( oaa ) ( oaa )' ( owb ) owa oo • ..,ab= owa ..,a + OWb ..,b owa oo • ..,ab'
0 = (:::Lab= (:::)ab . ..,ab+ ( :: )..,b ( ::J..,a' l = (:::)ab, ..,ab+ ( ~= )...b ( ::~ )...b'
(265.17)
(265.18)
where the second step follows by (265.16). Substituting this result into (265.17)
yields the identity
( oaa ) ( oaa ) [( oaa ) 12 ( OWb) owa ab, ..,ab = owa ..,a - owb ..,b -aab ..,b · (265.19)
By (265.12) we see that the term subtracted is non-negative; hence
(_!aa ) < (~) . owa D'b, ..,ab= &wa ..,a' (265.20)
equivalently,
~) oaa 00, ..,ab >(~) = oll'a ..,a · (265.21)
The foregoing analysis is due to EPSTEIN, who calls (265.20) and (265.21) the
restricted Le Chatelier-Braun principle. For the case of a simple fluid, the relation just derived is equivalent to y ~ 1 .
Whether all inequalities called "conditions of stability" in the Iiterature
are consequences of GIBBs' condition {265.4) is not certain.
Further inequalities to be satisfied by the equations of state have been discussed2, but not definitively.
It is to be noted that equations of state need not satisfy the conditions of
stability for all values of the thermostatic state. Rather, the conditions serve to
distinguish stable states from unstable ones. In a theory where mechanical phenomena are of primary interest, it may be natural to seek and impose a require1 [1937. 1, § 143].
2 Cf. WEYL [1949, 38, §3]. who treats only the case when f=l, and who proposes also
the postulate ( ()2~) > 0, i.e., ( :4>) < 0, and derives from it some further inequalities.
WEYL notes als:~ha1 from (252.3)rit11 follows that ( :; )
11 and (:~ ). cannot vanish simultaneously; hence (265.19) implies th.at ( :; )v and ( :~ t are of one sign. CouRANT and FRIEDRICHS [ 1948, 8, § 2] propose the stronger inequality (~::)'I;::::; o. For the effect of these inequalities in gas dynamics, see § 37 and § 56 of SERRIN's article, Mathematical Principles of Classical
Fluid Mechanics, This Encyclopedia, Vol. VIII/1.
42*
660 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sects. 266,267.
ment of universal stability, but in a theory aiming to determine criteria of stability
of equilibrium it is more natural to include the theoretical possibility of unstable
states1•
F. Charge and Magnetic Flux.
I. Introduction.
266. The scope of this chapter. Classical mechanics is founded on the principles
of conservation or balance of mass, momentum, moment of momentum, and
energy. Alongside these mechanical principles we now set up two further principles of conservation as a basis for the theory of electromagnetism. These are,
the conservation of charge and the conservation of magnetic flux. Subchapter li
formulates these principles and deduces and interpreis some of their consequences.
In Subchapter III we introduce an additional postulate, the aether relations,
and in Subchapter IV we consider the mechanics of the electromagnetic and
charge-current fields. The sequey1ce of hypotheses and the order of logical development that we adopt here depart from the traditional treatments. A principal objective is to isolate those aspects of the theory which are independent
of the assigned geometry of space-time.from those whose formulation and interpretation depend on or imply a particular space-time geometry 2• For example,
we regard the conservation of charge as a physical or intuitive concept logically
independent of the concepts of rigid rods, uniform clocks, and inertial frames,
and we have chosen to express this law in a mathematical form likewise independent of the representation of these extraneous entities.
The development of electricity and magnetism is too closely connected with
special cases and special phenomena for detailed historical references as in the
previous chapters tobe practicable here. The central importance of MAXWELL's
work is well known; in FARADAY and KELVIN he had major predecessors, and the
classical theory is in part the creation of his successors, HERTZ and LoRENTZ. An
excellent history has been written by WHITTAKER 3•
267. Antisymmetrie tensors. Since the only quantities occurring in the conservation laws of electromagnetic theory are antisymmetric tensors, we introduce
some relevant specific terminology and notation.
Let
x''=x''(:x:) (267.1)
denote an element of the group of analytic transformations of the co-ordinates
of an n-dimensional space. Let
U' = u-~u (267.2)
denote a transformation of the unit 4 U. As U runs over the real numbers, (267.2}
generates the group of unit transformations on the unit U.
1 E.g., in gas dynamics the van der Waals equation may be used for the entire density
range only so long alj the temperature exceeds the critical temperature, while in the theory
of liquefaction it is tiseful to consider it also below this temperature. 2 Our development. and ordering is similar tothat of KoTTLER [1922, 4]. Forahistory
of the mathematics and physics leading to Kot"TLER's formulation of the basic equations of
electromagnetic theory, cf. WHITTAKER [1953, 35, p. 192]. This earlier work includes that
of HARGREAVES [1908, 5], BATEMAN [1910, l], and MURNAGHAN [1921, 4]. Cf. also the remarks ofW:EvL [1921, 6, § 17]. [1950, 35, § 17]. KOTTLER's ideas were taken upbytheDutch
school of geometers and mathematical physicists and culminated in the series of papers by
VAN DANTZIG [1934, 11, '\[ 12] [1937, 10 and 11]. Cf. also ScHOUTEN and HAANTJES [1934, 8].
Shortcomings of the metric viewpoint even in strictly mechanical situations are emphasized
by KRÖNER [1960, 3, § 18]. 3 [1951. 39] [1953. 35].
' See Sect. App. 9 for a discussion of units and unit transformations.
Sect. 267. Antisymmetrie tensors. 661
A tensor field under the group of analytic co-ordinate transformations having
absolute dimension [UJ is a set of functions of the co-ordinates :JJ whose law of
transformation has the general form
' ' oxm' oxn' ox' t"f"···(:JJ' U')=I(:JJ'/:JJ)I-wsgn(:JJ'/:JJ)PU-----···------, ···fm"···(:JJ U) (267.3) '··· ' oxm ox" ox' • '··· • •
where (:JJ'/:JJ) denotes the Jacobian of the Co-ordinate transformation. If p =W =0,
I is called an absolute tensor. If p = 0, w =l= 0, I is called a relative tensor of weight w.
If p = 1, w =0, I is called an axial tensor, and if p = 1, w=j= 0, I is called an axial
relative tensor of weight w.
A covariant or contravariant completely antisymmetric tensor of rank k
will be called a k-vector. We can restriet the value of k to be less than or equal
to n since k-vectors suchthat k >n vanish identically. 0-vectors are called scalars
and 1-vectors are called vectors.
Weshall be concemed primarily with the following special types of k-vectors:
1. Absolute covariant or contravariant k-vectors.
2. Contravariant relative k-vectors of weight + 1.
3. Covariant relative k-vectors of weight - 1.
4. Covariant and contravariant axial k-vectors.
5. Contravariant axial relative k-vectors of weight + 1.
6. Covariant axial relative k-vectors of weight -1.
Tensors of types 2 and 3 will be called k-vector densities. Tensors of types 5
and 6 will be called axial k-vector densities.
If the group of co-ordinate transformations consists solely of unimodular
transformations, (:ll' /:JJ) = + 1, then the transformation laws of absolute k-vectors,
axial k-vectors, k-vector densities and axial k-vector densities coalesce. If the
group consists solely of transformations with positive Jacobian, then the trans~
formation laws of absolute k-vectors and axial k-vectors coalesce, as do the
transformation laws of k-vector densities and axial k-vector densities.
In the following definitions, Fand Y stand for k-vectors; they may be absolute,
axial, densities, or axial densities, but their variance is specified by their indices.
The dot product of a contravariant k-vector Fand a covariant m-vector Y, k-:?:. m,
is defined by (cf. Sect. App. 3)
(267.4)
(267.5)
Let e'•'•···'" and e,,,, ... , .. denote the permutation symbols such that e12···" =
e12 ..... = + 1; these symbols define axial n-vector densities. The duals, dual F
and dual Y, of covariant and contravariant k-vectors F and Y are defined by
1 (dual F)'•'•···'"-k = - e'•'•···'n-ks, ... sk F. or k! s,s, ... sk, dualF = E·F, (267.6)
dual Y=Y·E. (267.7)
These definitions imply that for any type of k-vector we have
dual dualF = F. (267.8)
662 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 268.
The cross product of a covariant (contravariant) k-vector F and a covariant
(contravariant) m-vector Y, k +m~ n, is defined by
1 (F X Y)'•'•···'"-.t-.. = --e'•'•···'"-.t-•••···•.t'•···"" F. y: k! m! ••····~ t, ... lm (267.9)
(F X Y) r,r, ... rn-.t-m =
- k! m! -F••····~Y •···"" Bs, ... s.tl1 ••• 1mr1 r1 ••• rn-.t-M . (267.10)
These definitions generalize that for the cross product of a vector and a tensor
given in Sect. App. 3. We have the useful identities relating the cross product,
dot product, and dual:
Fx Y =(dual Y) ·F, (Fand Y covariant), (267.11)
Fx Y = Y · dualF, (Fand Y contravariant). (267.12)
A constant tensor field whose components have the same values in all Coordinate systems related by a group of transformations ~ is called a constant
invariant tensor ol the group ~- The axial n-vector densities e'•'•···'" and e,,,, ... ,,.
are constant invariant tensors of the full group of general analytic transformations. The mixed Kronecker deltas .:5~:~::::~: are examples of absolute constant
invariant tensors of the group of analytic transformations. The covariant Kronecker delta d,. is a constant invariant tensor of the orthogonal group. Since
the fundamental axial n-vector densities are constant invariant tensors of the group
of general analytic transformations, the identity (267.8) establishes a one-to-one
correspondence between k-vectors and their dual (n- k)-vectors which is invariant
under the group of general analytic co-ordinate transformations. This implies
that the transformation laws of a k-vector and its (n- k)-vector dual are indistinguishable. For example, suppose that 11 , 12 , ••• , 16 is a set of six functions
of four Co-ordinates whose transformation law is suitably defined by setting
l1 = ~~· l2 = ata• Ia = al,. 1, = a2s· ls = a2,. ls = aa,.
where a is an absolute covariant 2-vector. The same transformation law for the
f's is obtained by setting
11 =(dual a) 34 , 12 =(dual a) 42 , Ia =(dual a) 23 ,
1, = (duala)14 , 15 = (duala)31 , 16 = (duala)12 •
The point we make is analogaus to the better known fact that, under the orthogonal
group, the transformation laws of covariant and contravariant tensors are indistinguishable. The similarity of the two situations can be made even more
apparent by noting that a contravariant index may be raised or lowered by the
Kronecker deltas d,. or d'" and that this process of association is invariant under
the orthogonal group because the covariant and contravariant Kroneckerdeltas
are constant invariant tensors of that group.
268. Invariant integral and differential equations independent of a metric or
connection. Let o, stand for ojox'. If F is an absolute or axial k-vector field,
the quantities defined by
(rotF),s,s, ... s.t = (k + 1) o[,.F.,s, ... s.t] (268.1)
transform as the components of an absolute or axial (k + 1 )-vector field under
general transformations of the co-ordinates. Similarly, if Y is a contravariant
k-vector density or axial k-vector density, the quantities defined by
(div Y)'•'•···'.t-• = a. Y'•'•···'~-·· (268.2)
Sect. 268. Invariant integral and differential equations. 663
transform as the components of a (k -1)-vector density or an axial (k -1)-
vector density. The (k+1)-vector field defined in (268.1) is called the natural
rotation of the field F, and the (k -1)-vector field defined in (268.2) is called the
natural divergence of the field Y. Note that the natural rotation is defined only
for covariant absolute or axial k-vectors and not for covariant k-vector densities
or contravariant k-vectors of any type. Similarly, the natural divergence is
defined only for contravariant densities or axial densities. The dual of the natural
rotation of a covariant k-vector is a contravariant (n- k -1)-vector called the
curl of F.
curl F = dual rot F. (268.3)
Wehave the following identity relating the divergence, curl, and dual of a k-vector:
curl F = div dual F. (268.4)
Consider the oriented k-dimensional surfaces in an n-dimensional space
admitting a parametric representation
(268.5)
by piecewise continuously differentiahte functions of the parameters u,., a = 1,
2, ... , k. We denote such a surface (hypersurface) by ~. If the parameters u
are transformed by continuously differentiable one-to-one parameter transformations with positive Jacobian
u"' = u"' (u), (u'ju) > 0, (268.6)
we obtain another admissible parametrization of the surface 9f given by
x' = x'(u') = x'(u(u')). (268.7)
A k-dimensional circuit is an ~ topologically equivalent to the complete
boundary of a (k + 1)-dimensional interval.
Let IDl [ ~, U] denote a quantity defined for every ~. We may think of
IDl[~. U] as a physical quantity having the dimension [U] suchthat for every
k-dimensional set ~ of events or points in space a value is assigned to it in
principle. Thus, for example, we may think of IDl as the total charge in a given
spatial region or as the total charge which has passed through a 2-dimensional
surface in space in a given interval of time. In the general case, we assume that
IDl is an additive set function. More specifically, weshall assume that IDl is expressible in the form
IDl[~. U] = J m(~(u), :::)du1 du2 ... du", (268.8)
.9'k
where the integrand is a polynomial in the vectors ~=:.
Theorem: I f IDl [ ~, U] has the transformation law
IDl' [ ~, U'] = U IDl [ ~, U] (268.9)
under independent transformations of the unit U, generat analytic transformations
of the co-ordinates ~. and transformations of the Parameters u, the set function
IDl [ ~, U] is expressible in the form
a-n = __1_ J ...,_ d Y.'•'• .. ·'k = J m · d .9.. .:u~ k! . .. ......... ~ " '" (268.10)
664 Co TRUESDELL and Ro TouPIN: The Classical Field Theorieso Secto 268.
wkere m is an absolute covariant k-vector field, of absolute dimension [UJ, independent of tke vectors ::: , and wkere d ~ is tke absolute contravariant k-vector
defined by
o x[r, o x'• · o x'k] • d oJ,,r,ooo'k = k I----. 0 • --du1 du2 du" Jk - 0 oul ou2 ouk . . . . (268.11)
Proof: First consider the invariance of !In under the subgroup of unimodular
parameter transformations: (u'fu) = +1. Under parameter transformations with
the co-ordinates held fixed, the quantities ox'fou" transform as a set of n absolute
covariant vectors in a k-dimensional space. A known theorem of classical invariant theory1 states that every invariant polynomial function of a set of
n covariant vectors in k dimensions: (n ~ k) under the group of unimodular transformations is reducible to a polynomial in the (~) determinants of the vectors
taken k at a time. In the present application of this theorem we conclude that
the integrand of (26808) must reduce to a polyn?mial in the (;) variables
(268.12)
The coefficients in the polynomial are at most functions of the co-ordinates
x'(u). Under parameter transformations with Jacobian not equal to 1, the
variables D of (268012) transform as relative scalars of weight 1, while the coefficients of these quantities transform as absolute !'!Calarso The invariance of m
under this larger group allows us to conclude . that . the integrand must reduce
to a linear homogeneous function of the variables D. Hence we can write !In
in the form (268.10). The invariance of !In under general analytic transformations
of the co-ordinates and the quotient rule of tensor algebra allow us finally to
conclude that the coefficients m,,,,o 0 o't must transform as a covariant tensor.
Only the antisymmetric part of the tensor of coefficients contributes to the
transvected sum in (268.1 0); hence, there is no loss in generality by assuming
that m is a k-vector. On transforming the unit U, we see that the absolute
dimension of m must be [U] provided we assume that the absolute dimension
of d ~ is [7]. As explained in Sect. App. 9, c;:o-ordinates and parameterswill always
be assigned the absolute dimension [7]. The physical dimension of d ~ may
differ from [7], and the physical dimension of m may differ from the absolute
dimension of !In; but this problern will be treated in some detaillater in Sect. 277.
Corollary: I I !In [ 9k, U] is an axial scalar with tke transformation law
!In[~, U'J = sgn (~'/~) U!m[~, UJ,
then it is expressible in tke form
!m[~;U] =Jm-d~,
wkere m(~(u)) is an axial k-vector.
The set function !In [ ~, UJ can be written in the dual form
!In [ ~. U] = (- 1)"(n-k> J (dual m). d,9!,
(268.13)
(268.14)
(268.15}
where d§;. =dual d9k. In an odd-dimensional space, the factor ( -1 )"(,.-k) = + 1
for every value of k; however, in even-dimensional spaces, this factor alternates
in sign as k runs over the integers.
1 WEYL [1946, 10, p. 45]0
Sect. 269. Conservative k-vector fields. 665
Let ~ be a circuit given by x' = x'(tt1, u2, ••• , uk), and Iet it form the complete
b d f ro • b r '( 1 2 k+I) 'fth t r ox' ox' ox' oun ary o Jk+.l• g1ven y x = x v , v , ... , v ; 1 e vec ors w , 7fl a--z · · · Bk
o x' o x' 8 x' . . . u . u u
and the vectors ovf, ~ • · · ~+! have the same onentanon, where w 1s a vector
directed out of ~+I• we have (Sect. App. 21)
pm · d~ = Jrotm -d~+I• (268.16)
where m is a continuously differentiable absolute or axial covariant k-vector
field in ~+I· The dual form of this fundamental integral identity is 1
(-)"+I p (dual m) · d~ = J div dual m · d~+I (268.17)
Substituting from the identity (268.4) we can write (268.17) in the alternative
form
(-)"+Ip (dualm) · d~ = J curlm · d~+I· (268.18)
269. Conservative k-vector fields. In mechanics, if we are given a force field I
such that
,~.. f dx' = 0 ':1' ', (269.1)
for every circuit, the field I is said to be conservative. Vector fields satisfying
(269.1) play an important role in many physical theories. In other contexts,
vector fields satisfying (269.1) are called lamellar (see Sect. App. 33). Hereinthis
chapter we prefer to use the name "conservative" owing to the connection we
shall establish between the conservation laws of charge and magnetic flux and
k-vector fields satisfying equations bearing a formal resemblance to (269.1).
If a vector field satisfies (269.1), there exists an infinity of scalar fields h such
that zs
h(~ )- h(~ ) = J I· d~. (269.2) illJ
where 1 and 2 are the end points of the curve ~(u). At a point where I is continuous, we have
t, = a,h. (269.3)
A scalar field h with these properties is called a potential of the conservative fieldf.
The potential h is not uniquely determined by the field I and these requirements.
If h is any one field satisfying (269.2), then so also is the field h' given by
h' = h + const. (269.4)
We shall now extend the above definitions to k-vector fields in n dimensions.
Let F denote an absolute or axial covariant k-vector field and Y an absolute
or axial covariant (k + 1 )-vector field such that
~F · d~ = f Y · d~+l (269.5)
for every k-dimensional circuit. We then call Y the source of F. A source-free
k-vector field F, i.e., one for which Y =0, is called a conservative k-vector field.
The integral theorem (268.16) states that if F is continuously differentiable, its
1 The factor (-)"+1 appears here and not in the corresponding formula of ScHOUTEN
[1954, 21, Chap. II, Eq. (8.14)] because our definitions (267.6) (267.7) of the dual tensors
differ slightly from his [ibid., Chap. I, Eq. (7.15)]. If we had followed ScHOUTEN's definitions, we should have had to insert a factor (-)10+1 before the right-hand side of the identity
(268.4).
666 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 270.
source Y is given by
Y = rotF. (269.6)
Now let F denote a conservative k-vector field so that
~F-dY',.=O. (269.7)
If F has continuous derivatives of all orders up top >O, there exists at least one
(k -1)-vector field K having continuous derivatives of all orders up to p + 1
such that
(269.8)
for every Y;., where 9";,_1 is the complete boundary of 9";. .1 A field K satisfying
(269.8) will be called a potential of the conservative field F. The field K is not
uniquely deterrnined by the field Fand (269.8). Given any potential field K of F,
we obtain any other possible potential by adding to K an appropriate conservative
(k -1)-vector field. For example, if G is any twice differentiable (k- 2)-vector
field, then K' given by
K'=K+rotG (269.9)
is an alternative potential of the field F. The gmup of transformations relating
the potentials of a given conservative field will be called the group of potential
transformations.
II. The conservation of charge and magnetic flux.
270. The electromagnetic and charge-current fields and the Maxwell-Bateman
laws. Space-time is a 4-dimensional space. A point in this space will be called
an event, whose four real co-ordinates we denote by x 0 , .Q = 1, 2, 3, 42• Quantities
transforrning as a tensor under the group of general analytic transformations
of the ~will be called world tensors (cf. Sect. 152).
W e assign an additive set function (!; [ 9;;, QJ called the charge to every 9;;
in space-time. We assume that the charge is expressible in the form
(270.1)
where lJ is the charge-current field and where Q is the unit of charge. I t follows
from (268.15) that we can also write (!;[~, QJ in the dual form
(l;[~,QJ =-J (duallJ) -d~. (270.2)
The charge (!;[~, Q] is assumed to transform as an axial scalar having absolute
dimension [0]. Therefore, it follows by the corollary (268.14) that dual lJ is
an axial 3-vector having absolute dimension [0]. Thus lJ is a vector density
having absolute dimension [0).3 Under the group of analytic transformations
of the co-ordinates ~ and the group of transformations
Q' = o-lQ (270-3)
1 Throughout this chapter we assume that the underlying space is topologically equivalent to the Euclidean space of the same dimension. For such spaces, the theorem embodied
in (269.8) follows from WHITNEY's Iemma [1957, 18, § 25]. 2 Henceforth in this chapter Greek indiceswill range over the four values 1, 2, 3, and 4. 3 Our reasons for assuming that the charge transforms as an axial scalar instead of an
absolute scalar will be explained in Sect. 283. In Sect. 285 we introduce a more general
expression for the charge.
Sect. 270. The electromagnetic and charge-current fields.
of the unit of charge Q, (J has the transformation law
oxO
aD' = 01(~'/~)1-1 oxD an.
·667
(270.4)
We now postulate the law oj conservation of charge: the world scalar invariant integral (r[ .9;, QJ vanishes for every 3-dimensional circuit,
(270.5)
In a similar fashion we assign to every 9; in space-time a world invariant
additive set function ~ [ 9;, Cl>] called the magnetic flux and assume tha t magnetic
flux: is expressible in the form
(270.6)
where p is the electromagnetic field and where Cl> is the unit of magnetic flux. The
magnetic flux and the electromagnetic field have the absolute dimension [ ].
Under the group of analytic coordinate transformations of the ~ and the transformations
Cl>'= -1 Cl> (270.7)
of the unit of magnetic flux, the electromagnetic field has the transformation
law
(270.8)
We now postulate the law oj conservation of magnetic fluz: the world scalar
invariant ~ [ 9;, Cl>] vanishes for every 2-dimensional circuit:
~p·d9;=0. (270.9)
Eqs. (270.5) and (270.9) are the Cornerstones of electromagnetic theory. The
laws of nature embodied in these postulates are perhaps the most Iasting achievements of the classical theory of electromagnetism. For the moment we regard
the charge-current and electromagnetic fields as independent. In Subchapter III
we shall see the manner in which they are related to one another by an additional
hypothesis. Also, we shall subsequently give a more general mathematical expression for the law of conservation of charge. The intended generalization, however, involves no new physical idea, and we prefer now to consider the simpler
Eqs. (270.5) and (270.9), independently of any further physical assumptions or
increased mathematical generality. When we are on more familiar ground, these
generalizations may be added with less difficulty and abstractness.
The world invariant integral equations (270.5) and (270.9) were deduced by
BATEMANl, who took as a starting point the differential equations commonly
referred to as MAXWELL's equations. Consistently with the program of Sect. 7,
we prefer, following KoTTLER2, to announce our basic premises in the stronger
form of the integral equations (270.5) and (270.9). The acceptance of these
M~well-Bateman laws may be motivated on the grounds of simplicity and
the general intuitive notion of conservation. As we shall see, the customary
field equations and boundary conditions of electromagnetic theory follow as
consequences of these integral equations if we supply appropriate assumptions
regarding the continuity of the fields and the nature of the Co-ordinates ~-
1 [1910, 1]. BATEMAN cites the earlier work of HARGREAVES [1908, 5] on invariant integral
forms.
2 [1922, 4].
668 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 271.
We trust that our rather abstract postulation of the laws of conservation
of charge and magnetic flux will not discourage the reader seeking more concrete
and familiar results. Our main reason for this approach is to emphasize the
independence of the conservation laws from any geometry of space-time. The ideas
of conservation as formulated here have, in a certain sense, topological significance, transcending both the intuition and the mathematics of length, time,
and angles.
This "metrical independence" of the conservation laws has been noted by VAN DANTZIG 1•
As he remarks, the concept of metric and the measurement of lengths, angles, and time intervals is perhaps one of the most sophisticated and complex aspects of any physical-theory.
Furthermore, is not the intuitive notion of conservation of charge, for example, quite independent of measurements of length and time? We view the conservation of charge and magnetic flux as independent of ideas like inertial frames, rigid rods, absolute or uniform time,
Lorentz transformations, Galilean transformations, etc., and hence as deserving an independent mathematical expression.
Since there is a one-to-one correspondence between k-vectors and their duals,
the same physical quantity or physical law involving only k-vectors may be
expressed in equivalent dual forms. It is a matter of taste and convenience
which representation is used. Usually the representation involving the fewer
number of indices is to be preferred. Thus the charge-current field is usually
represented by a contravariant vector density rather than by the dual covariant
axial 3-vector. As far as the electromagnetic field is concerned, both cp and dual cp
are of rank 2, and it is a matter of convention that we tend to favor the covariant
representation. However, for some purposes, a degree of formal simplification
is obtained by using one or the other representation, and in what follows we
do not restriet ourselves to any one particular choice.
271. Electromagnetic and charge-current potentials. If the electromagnetic
and charge-current fields are continuous, the Maxwell-Bateman laws (270.5)
and (270.9) are sufficient conditions for the existence of continuously differentiable fields a; and 11 such that
J (J. d,§; =- ~"' 0 d~,
fcp. d~= ~a; 0 d!l;_.
The integral theorems (268.16) and (268.17) lead to the local relatjons
(J = div 11,
In terms of components, we have
cp =rot a;.
(271.1)
(271.2)
(271.3)
(}Q=OA'YJDA, (/JuA=20[.o1XLI]· (271.4)
A contravariant 2-vector density 11 satisfying (271.1) will be called a chargecurrent potential, and a covariant absolute vector a; satisfying (271.2) will be
called an electromagnetic potential.
The existence of charge-current potentials '1/ is a consequence of the law of conservation
of charge. In Subchapter 111 we shall subsequently introduce the aether relations fixing a
particular charge-current potential in terms of the electromagnetic field. Our procedure
here is not unlike a procedure sometimes adopted in mechanics, where the stress t's may
be introduced as a solution of the equations
fef,dv=~t ,da ,
J (! Z[r /s] dv = p Z[r tPs] dap.
where the force field (! f is regarded as prescribed (cf. Sect. 203). These equations do not uniquely determine the stress field t since addition of a null stress, i.e., any symmetric tensor
I [1934, lJj.
Sect. 272. Field equations and boundary conditions. 669
satisfying o, t's = 0, to a given Solution of these equations yields another solution (cf. the
remarks at the end of Sect. 205). Nevertheless, theories of continuum mechanics generally
e_mploy constitutive equations for the stress which arenot invariant under the process of adding
a divergence-free symmetric tensor. For the moment we emphasize that the existence of an
infinity of distinct charge-current potentials follows from the law of conservation of charge.
Similarly, conserva:tion of magnetic flux is a sufficient condition for the existence of aninfinity
of distinct electromagnetic potentials provided we assume rather weak continuity properties
for the electromagnetic field. The group of transformations of the electromagnetic potentials
has been called the group of gauge transformations 1 . It is customary to render the electromagnetic potential unique by imposing upon it additional restrictions in the form of boundary
conditions, continuity requirements, and algebraic or differential relations amongst its
components. Same remarks on these conditions are given in Sect. 276.
272. Field equations and boundary conditions. If the electromagnetic and
charge-current fields are continuously differentiable in a region, by applying the
classical argument based on (268.16) to (268.17) (cf. Sect. 157), from the MaxwellBateman laws (270.5) and (270.9) we derive the jield equations
div 6 = 0,
curl fJ! = 0 =dual rot fJ!.
In terms of components, we have
Let
ou an= 0,
c;n.J'l'€1 O,j CfJ'l'B = 0.
(272.1)
(272.2)
(272-3)
(272.4)
(272.5)
be the equation of a 3-dimensional surface in space-time dividing a region f!A
into two regions !JI+ and &~-. W e suppose the electromagnetic and chargecurrent fields continuous in the closure of !JI+ and &~- but possibly discontinuous
at L'=O. We may think of the surface L'=O as the set of events representing
the history of a 2-dimensional surface in space across which the electromagnetic
and charge-current fields have finite jumps. Applying the basic integral laws
(270. 5) and (270.9) to circuits divided by the surface .E = 0, we conclude that
the discontinuities [dual fJ!] and [ 6] are such as to satisfy the restrictions
[dual 'f!] · grad .E = 0, } (272.6) [ 6] · grad .E = 0.
In terms of components, these equations read
(dual 'f!)nc. od .E = 0, }
an On .E = 0. (272.7)
These are the electromagnetic and charge-current boundary conditions. They imply
that the most general discontinuities in the electromagnetic and charge-current
fields allowed by the conservation laws are expressible in the form
[dual ffJ]= ßxgrad.E,
[6] = w X grad .E,
or, in terms of components,
[(dualffJ)n.1]=en.J'l'eß'P8e.E, }
[an] =-lc:D.J'l'ew.J'P8e.E,
where ß and w are arbitrary fields defined on .E = 0.
1 BERGMANN [1942, J, p. 115].
(272.8)
(272.9)
(272.10)
670 C. TRUESDELL and R. TouPrN: The Classical Field Theories. Sect. 273.
273. The geometry of space-time, reduction of k-vectors, and the units of length
and time. So as to obtain equations and results familiar from conventional
treatments of electromagnetic theory, we now introduce a space-time geometry.
In Sect. 152 we studied two such geometries based on the idea of the underlying
Euclidean and Galilean groups of transformations. We showed that the dass
of Eudidean frames could be defined as the dass of Co-ordinate systems in spacetime for which the space metric g (x) and the covariant space normalt (x) assume
the canonical forms
= [b'• o] g 0 0' t = (0, 0, 0, 1). (273.1)
The subdass of these frames for which the Galilean connection r(x) = 0 were
identified as the inertial frames of dassical mechanics. The question of inertial
frames in electromagnetic theory will be taken up in Subchapter III, where we
shall discuss the Lorentz invariance of the aether relations, Lorentz frames,
and the relation between these entities and their Galilean counterparts. For the
remainder of this Subchapter it suffices that we consider the entire dass of reetangular Cartesian co-ordinate systems in space-time characterized by the canonical forms (273.1). In addition to the co-ordinate transformations considered in
Sect. 152, we shall here consider transformations of the units oflength and time
L' = L -1 L, T' = r-1 r. (273.2)
The space metric g will be assigned the absolute dimension [L -2], and the covariant space normalt will be assigned the absolute dimension [T]. Thus, under
transformations of the co-ordinates ;r and the units of length and time, the
world tensors g and t have the transformation laws
g Q'A'( ;r' ' L') = L-2 ox.Q' OXQ ~xA' oxA g QA( X, L) ' l
oxQ (273.3)
ttJ' (x', T) = T -
0 xQ' t!J (x, T).
Recall that ta=o!Jt, where t(x) is the time. The absolute dimensionoft is [T].
If g and t are to retain their canonical form (273.1) when the units of length
and time are changed, the co-ordinates belanging to the two systems of units
must be related by a transformation of the form
z'' (L') = L[A'',(t) z'(L) + d'' (t)], }
z4' (T') = t' = T z4(T) + const = T t + const, (273.4)
where Ais an orthogonal matrix. Weshall denote the co-ordinates ;r by z and t
when (273.1) holds. That is, z denotes reetangular Cartesian spatial co-ordinates
and t = x4 denotes the time. Recall that in Sect. 152 we called a co-ordinate
system for which we have the canonical forms (273.1), a Euclidean frame.
Let F denote a covariant world k-vector. Let us denote the components of F
referred to a Euclidean frame by
(273.5)
All the components of F are determined from these particular components. With
reference to the decomposition (273.5) weshall use the notation
F = (a, b). (273.6)
Sect. 273. The geometry of space-time, reduction of k-vectors.
For a contravariant world k-vector Y we write
Y = (c, d),
where
Note the different position of the index 4 in (273.5) and (273.8).
671
(273.7)
(273.8)
If Fis an absolute covariant k-vector of absolute dimension [UJ, the transformation laws of its components a and b can be put in the form
r;r; ... r_t= r; r;··· r_t s,s, ... sk (
273 .
9 a U L ) -kAs• As• Ask a l
b ,, , -uL-k+lT-lAs~As, Ask-l(b +a ut) '•'•···'k-l- 'i r;. •. 'k-1 s,s, ... sk-1 s, ... sk-lt '
where u1==oijoz4 '. Thus the quantities a transform as a 3-dimensional tensor;
but unless a = 0, the quantities b do not.
For contravariant absolute k-vectors, the components c and d in (273.8) have
the transformation laws
c';r; ... rk =ULk A~: A:: ... A~: (cs, ... sk _ kurs,ds,s, ... skl), l
d';r; ... r;,_1 = uLk-1 TA'; ... A'; A'.t-1 ds,s, ... sk-1. S1 s1 Sk-1
(273.10)
Similar considerations apply in the case of covariant and contravariant axial
k-vectors, k-vector densities, and axial k-vector densities. For example, if Y is
a contravariant k-vector density we have
c';,; ... rk = U u-a r-1 A'; sl A'; Sz . .. A'k sk (cs, ... sk- k urs,d s, ... skl), l
d';r; ... •A:-1 = ULk-4Ar; A'; ... A'L1 ds,s, ... sk-1. 51 Sa S.t-1
(273.11)
For axial k-vectors and axial k-vector densities a factor, det A =± 1, will occur
in the transformation laws of the Euclidean components.
The reetangular Cartesian co-ordinates z (L) of a Euclidean frame based on
the unit of length L are assigned the physical dimension [L J (cf. Sect. App. 9) and,
consistent with (273.4), z4(T) is assigned the physical dimension [T]. The physical
dimensions of the components a, b, c, etc., of world tensors referred to a Euclidean
frame are determined by the absolute dimension of the corresponding world
tensor, its variance, and its weight by writing the transformation laws for its
components in the forms (273.9), (273.10), (273.11), etc. Thus if Fis an axial
or absolute covariant k-vector of absolute dimension [UJ, then
phys. dim. a = [UL-k], } (273.12) phys. dim. b = [U L -k+I PJ.
If Y is an absolute or axial contravariant k-vector, then
phys. dim. c = [ULk], }
phys. dim. d =[ULk-t T].
If Fis a covariant density or axial density, then
phys. dim. a = [U L -k+a T], }
phys. dim. b = [U L -kHJ.
Finally, if Y is a contravariant density or axial density, then
phys. dim. c = [U Lk-a r-1], }
phys. dim. d = [U U-4] •
(273.13)
(273.14)
(273.15)
672 C. TRUESDELL and R. ToUPIN: The Classical Field Theories. Sect. 274
Consider a world covariant k-vector or axial k-vector F and a world covariant
(k -1)-vector or axial (k -1)-vector G such that
F =rotG.
Let
F = (a, b), G = (g,h).
We shall then have
a =rotg, b =roth+ (-1)k-1 ~~,
or, equivalently,
duala = curl g, dual b = curlh + (-1)k-1 o(d~~Ig).
(273.16)
(273.17)
(273.18)
(273.19)
274. The charge density, current density, electric field, and magnetic flux
density. If we assume the geometry of space-time to be Euclidean in the sense
of Sect. 152, then we can apply the considerations of Sect. 273 to a reduction
of the charge-current and electromagnetic fields into four distinct fields. Refer
the electromagnetic field p and the charge-current field to a Euclidean frame, and
set
p = (dualB,E), lJ = (J, Q), (274.1)
where B is called the density of magnetic flux, E the electric field, J the current
density, and Q the charge density. From the rules for determining the physical
dimensions of these quantities given in Sect. 273 we get immediately,
phys. dim. B = [ L-2], l
phys. di~. E = [ L -1 PJ,
phys. d1m. J = [0 L -2 r-1],
phys. dim. Q = [0 L-3].
(274.2)
Let vn be the world velocity field of a motion as defined in Sect. 152. In
terms of v, the electromagnetic field, and the charge-current field, we can define
the space tensors
(274-3)
Recall that a space tensor was defined in Sect. 152 as a contravariant world
tensor for which pnA · · · tn = pAn · · · tn = · · · = 0. A k-vector which is a space
tensor has the representation
F = (c, 0), (274.4)
so that, according to (273.10) and (273.11), c transforms as a 3-dimensional
Cartesian tensor. If we refer the components of lf and il to a Euclidean frame,
we get
lf =~+~X B, 0) _ (lf, 0), }
il-(J Q V, 0) - (il, 0), (274.5)
where i = E +v X Bis called the electromotive intensity at a point moving with
the particles of the motion, and il = J- Qv is called the conduction current
relative to the particles of the motion. The electromotive intensity and the conduction current of a motion transform as vectors under time-dependent transformations of the spatial coordinates. However, one should note that the electric
field and current density do not transform as vectors under transformations
between co-ordinate frames in relative motion. Thus, if the electric field and
current density vanish in one Euclidean frame, they need not necessarily vanish
Sect. 275. The 3-dimensional integral form of the laws of conservation. 673
in every Euclidean frame. A distribution of charge in one frame will constitute
a current density in a frame in relative motion. Similarly, a density of magnetic
flux in one frame will be interpreted as an electric field in a frame in relative
motion. However, if the density of magnetic flux and the density of charge
vanish in one Euclidean frame, they vanish in every Euclidean frame.
275. The 3-dimensional integral form of the laws of conservation of
charge and magnetic flux. Consider first the 3-dimensional circuits in spacetime formed by 3-dimensional tubes
(Sect. App. 29) of the world velocity
field v defined by a given motion
closed on either end by surfaces lying
in the instantaneous spaces t (x) = t1
and t(x) =t2 • (See Fig. 44.) If we
apply the law of conservation of
charge (270.5) to such a circuit and
refer all quantities to a Euclidean
frame, we get
t, t,
J Qdv] + J dt~!J·da = 0, (275.1)
" t, t, !:I' Fig. 44. Tube of integration in Eq. (275.1).
where v is a spatial region moving with the particles of the motion and 9" is its
complete boundary. Set LI t = t2- t1 and consider the limit
(275 .2)
If the limits of the two terms exist separately, we get
:tJQdv+~!J·da=O, (275.3)
which has the traditional form of an equation of balance when sources are
excluded (cf. Sect. 157) and puts the law of conservation of charge in terms
easy to understand. The conduction current relative to the particles 1 of the
motion is the efflux of chargeout of the moving region through its boundary.
Consider next the 2-dimensional circuits in space-time formed by segments
of 2-dimensional tubes of the world velocity field closed on either end by surfaces
lying in the instantaneous spaces t (x) = t1 and t (x) = t2 • If we apply the law
of conservation of magnetic flux (270.9) to such a circuit and refer all quantities
to a Euclidean frame, we get
f 2 ta
fB·d(l] + fdt~fi·dX =0, (275.4)
!:I' t, t, '{/
where 9" is a 2-dimensional surface in space moving with the particles of the
motion and ~ is its complete boundary. Applying a limit argument as in (275.2),
we then get
(275.5)
1 These particles are defined mathematically by integration of the given velocity field
and are a convenient device for visualizing it; they need not be mass-bearing.
Handbuch der Physik, Bd. lll/1. 43
674 C. TRUESDELL and R. TOUPIN:The Classical Field Theories. Sect. 276
which is the traditional form of Faraday's law of induction for moving circuits.
If we apply (270.9) to a 2-dimensional circuit which lies in the surface t (~) = const
(Fig. 45) we get the condition
pB-da =0. (275.6)
A visualization of FARADAY's law of induction is obtained by introducing
the notion of lines of magnetic flux or lines of induction. The number of lines of
magnetic flux threading a closed curve in space is measured by the integral J B ·da, where [/ is any surface having
the curve for its boundary. Eq. (275.6) states that this measure is independent of the choice of [/ and depends only on
the curve. FARADAY'S law of induction states that the time
rate of change of the total number of lines of magnetic flux
threading a moving circuit is measurde by the negative line
integral of the electromotive intensity around the moving circuit.
Since we have (275.)), (275.5), and (275.6) for any motion,
Fig.45. surtaceofinte· the corresponding laws for circuits which are at rest in some
gration in Eq. (275·
6l· Euclidean frame follow as a special case by setting v' = 0.
276. The 3-dimensional integral form of the potential equations. Let the components of the electromagnetic potential and the charge-current potential in a
Euclidean frame be denoted by
Cl= (A,- V), 1'1 =(dual H, D), (276.1)
where A is called the magnetic pote11.tial, V the electric potential, H the current
potential, and D the charge potentiall.
If we apply the world invariant equations (271.1) and (271.2) to circuits in
space-time constructed in a manner similar to those used to obtain (275.)),
(275.5), and (275.6), we obtain the equations
J~·da=~(H+Dxv)·da- :tJD·da, (276.2)
JQdv=pD·da, (276.)}
JB·da=pA·da, (276.4)
~ ~
Ji·da=- :tJ A·da+(A·v-V)], (276.5)
••
where z1 and z2 denote the end points of the moving curve c.
Eq. (276.2) will be called the current equation, and (276.)) will be called the
charge equation 2•
The potential equations for stationary circuits follow simply by setting v' = 0. For the
classical theory, (276.4) and (276.5) are less used than are the current and charge equations,
although the introduction of magnetic and electric potentials and their relative importance
1 The axial vector H is generally called the "magnetic field intensity" and the vector
density D the "electric displacement". Since the charge-current field and the electromagnetic field have not been related to one another in any way, this customary terminology
would be inappropriate for our purposes here. The conventiooal names for the magnetic
potential A and the electric potential V of the electromagnetic field are appropriate and
suggestive, and we have assigned names to H and D an the basis of their similar relation
to the charge-current field. The terms "magnetic field intensity" and "electric displacement"
will take an their proper connotation after we introduce the aether relations in Subchapter III.
2 Eq. (276.2) is HERTz's form of the current equation for moving circuits. HERTZ [1900,
4, Chap. XIV], WHITTAKER [1951, 39, pp. 329-331].
Sect. 277. Time derivatives of scalar integrals over moving curves, surfaces, and regions. 675
is a matter of debate. The charge and current potentials Hand D play a more fundamental
role in the classical theory than do the magnetic and electric potentials A and V. MIE and
DIRAC1 have proposed theories of electromagnetism in which the four potentials H, D, A,
and V enter the theoretical structure on a more equal footing. However, in most treatments
of the classical theory, the magnetic and electric potentials are introduced as auxiliary fields,
determined only to within a gauge transformation. Additional restrictions are often imposed
in the form of boundary conditions, continuity requirements or algebraic and differential
relations between the four fields A1 , A2 , A3 , and V. For example. many authors introduce
the "gauge condition" V= 0, while others impose the "Lorentz gauge condition" div A +
c-2 ~ = o. MAXWELL 2 .imposed the condition div A = 0 and called A the electromagnetic ot
momentum. Perhaps a suitably "gauged" electromagnetic potential and the ideas of MIE
and DIRAC will serve as the basis of a better theory of electromagnetism. We do not take
up these questions, resting content with our choice of charge and magnetic flux as the fundamental quantities entering the equations of conservation.
277. Time derivatives of scalar integrals over moving curves, surfaces, and
regions. The electromagnetic equations for moving circuits involve the time
derivatives of scalar integrals having the form
~ = k\ J F,,,, ... ,kdY{•'•···'k = J F · dY" = J (dualF) ·dfi{, (277.1)
where Fis a 3-dimensional k-vector (k =0, 1, 2, or 3) and .9k is a spatial region,
surface, or curve moving with the particles of a motion with velocity field v'
in a Euclidean frame. Now the motion can be presented in the form z'=z'(ZK, t)
where the zK are material Co-ordinates. In general, the k-vector field F depends
explicitly on the time t as weil as the spatial co-ordinates z. The surface .9k
is given by parametric equations
z' = z' (u1, u2, ... , uk, t) = z' (ZK (u), t). (277.2)
The limits of integration (277.1) correspond to fixed values of the Z and u independent of the time t. Thus we can write
!!Sl_ = ~1 ()FK,K, ... Kk df/'.K,K, ... Kk
at k! ot k (277.3)
provided that the field F and the motion z (Z, t) are continuously differentiable.
In (277.3), the FK,K, ... Kk are defined by
oz'• oz'• oz'k P,KK K =----•••---F. ' •··· k- ezK, ezK, ezKk '•'•···'k' (277.4)
and d.9J.K,K, ... Kk is the corresponding transform of dSJ.•'•···'k. lf follows from
(277.3) and the results in Sect. 150 that d~jdt can be written in the form
d;J - ~j~F. df/'.'•'•···'k dt - k! dt '•'•···'k k ' (277.5)
where dßfdt denotes the convected time-flux of the k-vector field F.
We can also write d~fdt in the dual form
d;J }. d ~ dt = d~ (dualF) · dY". (277.6)
Now the convected time-flux of an absolute or axial k-vector field has the particularly simple form
d 8F 8F -ftF = Tt + v · rotF + rot(v · F) = Tt + curlFxv + rot(v · F). (277.7)
1 [1912, 6]; [1951, 5].
2 [1881, 4, § 618].
43*
C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 278.
Similarly, the convected time-flux of a contravariant k-vector density or axial
density has the simple form
:; (dual F) =dual (div dual F x v) + curl (dual F X v). (277.8)
One can easily verify that
dual a~; = :~ dual F. (277.9)
Let us set F- dual F. Eqs. (277.7) and (277.8) written in terms of components
are
The foregoing integral formulae and vector indentities are useful for the study
and interpretation of the electromagnetic equations for moving circuits. In what
follows we shall also need to consider a moving surface given in the form of an
equation
E(z,t) =0, (277.12)
where z denotes the reetangular Cartesian spatial co-ordinates of a Euclidean
frame, and t denotes the time. The vector n defined by
8,E ( ) n, = 277.13
V85E85 E
is called the instantaneotts unit normal to the moving surface E = 0, and the
quantity s defined by
(277.14)
is called the speed ( cf. ( 177.5)) of the surface E = 0 relative to the reference frame
(z, t). The quantities n and un may be regarded as the Euclidean components of
the world vector
8aE
va = --::Vrcg=;cLl~e~aLl~Ec===a=ec=:E (277.15)
That is, in the notation of Sect. 273
v = (n, -un). (277.16)
278. Field equations and boundary conditions in 3-dimensional form. The
3-dimensional form of the field equations and boundary conditions may be obtained
in two ways. One method is to work with the 3-dimensional integral equations
(275-3), (275.5), (275.6), (276.2) to (276.5). By assuming the fields in these equations continuously differentiable, one can use the results of Sect. 277 to express
the time derivatives of integrals over moving surfaces, curves and regions in
terms of the convected time-flux. Then, by appropriate application of the fundamental integral theorem, line integrals can be transformed into surface integrals,
surface integrals into volume integrals, etc. Various terms cancel and, by the
usual type of limit arguments, certain local conditions in the form of partial
Sect. 279. The 3-dimensional form of the aether relations. 677
differential equations follow from the integral equations. Boundary conditions
in 3-dimensional form follow by applying these same integral equations to
appropriate circuits divided by a surface of possible discontinuity. A simpler
and more direct method applies the reduction formulae derived in Sect. 273 to
the world tensor field equations and boundary conditions already obtained. By
this latter method we obtain as an immediate consequence of the field equations
(272.1) and (272.2):
d. J oQ
lV +Bt =0,
aB
curlE + aT = o,
divB =0.
The field equations (271.3} yield the relations
Q =divD,
an J=curlH-at,
B =curlA,
aA E = ---grad V. ae
(278.1}
(278.2}
(278.3)
(278.4}
(278.5)
(278.6)
(278.7)
Fig. 46. Geometry of an electromagnetic discontinuity.
Eq. (278.1} is the differential or local form of the law of conservation of charge,
and (278.2} is the differential form of FARADAY's law of magnetic induction.
The 3-dimensional forms of the boundary conditions follow easily from the
world tensor equations (272.6) and the definitions (277.13) and (277.14) of the unit
normal n and the speed un of a moving surface of discontinuity1.
nx [E]- un[B]= 0,
[B]·n =0,
[J]·n- un[Q] = 0.
(278.8}
(278.9}
(278.10)
Eq. (278.8) implies that the most general discontinuities in the magnetic
flux density and the electric field allowed by the law of conservation of magnetic
flux are expressible in the form
[E] =ln-unk,
[B] =kxn,
(278.11}
(278.12}
where the fields I and k are arbitrary fields defined on the surface of discontinuity.
We see that conservation of magnetic flux restricts considerably the geometry
of any electromagnetic discontinuity. Eq. (278.9) demands that any discontinuity
in the magnetic flux density be transversal. For a stationary surface of discontinuity, we have un =0, and from Eq. (278.8} it follows that the discontinuity in
the electric field is longitudinal. If the surface of discontinuity is moving, the
discontinuity in the magnetic flux density must be normal to the plane determined
by the discontinuity in the electric field and the normal n (see Fig. 46).
111. The Maxwell-Lorentz aether relations.
279. The 3-dimensional form of the aether relations. In all that precedes,
we have treated the electromagnetic and charge-current fields as independent.
We now introduce a fundamental relation between these fields by postulating the
Mazwell-Lorentz aether relations.
1 LuNEBERG [1944, 9, p. 21] has obtained these boundary conditions holding at a moving
surface of discontinuity in the e!ectromagnetic field by other means.
678 C. TRUESDELL and R. TouPrN: The Classical Field Theories. Sect. 279.
So as to introduce the aether relations in their simplest and most farniliar
form, we shall· begin with a statement of thein in terms of the vector fields E,
B, D, and H. First of all, we assume the existence of at least one Euclidean
frame in which the Maxwell-Lorentz aether relations
(279.1)
are valid. The quantities e0 and f1o are fundamental constants depending
only on the units of length, time, charge, and magnetic flux. That is, we
assume the existence of at least one Eudidean frame in which a charge potential
D is proportional to the electric field E and in which a current potential H is
proportional to the magnetic flux density B. Now it is easy to show that if the
Maxwell-Lorentz aether relations (279.1) hold in one Euclidean frame, they cannot hold generally in every Eudidean frame. To see this, simply consider the
transformation laws of the four fields E, B, H, and Dashave been determined
in Sect. 273. These can be written in the form
D'=D E'=E,+uxB, l
B'=B,
H'=H-·uXD,
(279.2)
from which it is immediately plain that the aether relations cannot hold
generally and simultaneously in Euclidean frames in relative motion. Thus the
M axwell-Lorentz aether relations are not Euclidean nor even Galilean invariants.
If we adopt the aether relations we see that the equations of electromagnetic
theory will not have the same form in every Euclidean frame, or more importantly,
they will not havc the same form in every Galilean frame, as do the equations of
dassical mechanics. Thus it appeared, from the dassical point of view, that if
the aether relations could be verified in some selected and preferred inertial
frame, then by a simultaneaus application of the laws of mechanics and electromagnetic theory one should be able to detect the relative motion of inertial
frames. To put it otherwise, the aether relations determine a preferred dass
of Euclidean frames all at rest relative to one another. It may also be assumed
for our purposes here that this preferred dass of frames are inertial or Galilean.
In Galilean frames which are in motion relative to this preferred dass, the aether
relations do not generally hold. This gives rise in a natural way to the notion of
an" aether". That is, we may think of the preferred inertial frames for which the
aether relations are valid as the dass of inertial frames in which the "aether"
is at rest. In all other inertial frames, there will be an "aether wind", controverting the validity of the simple relations (279.1).
Matter resides in an aether characterized by the relations (279.1). These relations are not to be regarded as constitutive equations for matter. We may
think of them as constitutive relations for the aether. Allow us to remind the
reader already familiar with the dassical theory of dielectric and magnetic
materials that the aether relations apply only to the charge and current potentials
of the resultant charge and current distribution of all kinds of charge and current.
Charge and current distributions arising from the polarization and magnetization
of a material medium are to be induded in the charge-current field 6, and (H, D)
is a resultant potential.
Sect. 280. The world tensorform of the Maxwell-Lorentz aether relations. 679
We shall enlarge on this point in Sect. 283. Here it suffices that our intuitive motivation
for adopting the aether relations {279.1) as valid both inside and outside of "matter" is
based on the guiding principle used by LoRENTz [1915, 3]. In LoRENTz's theory, matter is
regarded as a collection of charged point particles which exist and move through an aether
or vacuum whose properties are unaffected by the presence or motion of the particles. Here
we treat matter as a continuous medium but carry over the Lorentz hypothesis that the
aether relations are unaffected by the presence of matter. The effects of polarization and
magnetization in the Lorentz electron theory were determined by arguments based on a
particle model of a dielectric and magnetic material medium. A treatment of these effects
regarding matter as a continuous medium is given in Sect. 283.
At the same time, we give warning that the aether relations represent an
assumption not adopted in every existing theory of electromagnetism1, whereas
the conservation laws of charge and magnetic flux are, to our knowledge, common
to all. Thus the considerations of this subchapter are of a rather special nature,
and we have attempted to present but one point of view. The principal objective
of the chapter as a whole is to formulate and develop the conservation laws of
electromagnetic theory. These laws hold independently of the aether relations 2,
and the contents of this subchapter do not in any way restriet the considerations
in Subchapter II.
280. The world tensor form of the Maxwell-Lorentz aether relations. Let the
constant c2 be defined by
(280.1)
The fundamental nature of this constant will become apparent as we proceed;
it is called the square of the speed oflight in vacuum. Consider a space-time coordinate system (z, t) for which the aether relations are valid. As we have seen,
from the classical view of space-time geometry the frame (z, t) will be one of
a restricted subdass of all Galilean and Euclidean space-time frames. Consider
a world contravariant, absolute, symmetric tensor of rank two whose components
in the frame (z, t) have the particular values
(280.2)
Now the components of a tensor in a general system of co-ordinates are determined
uniquely by its transformation law and its components in any one co-ordinate
system, so that (280.2) determines the components of y in every space-time Coordinate system. The determinant of the inverse y of y is given by
det y = - c2 (280.3)
in the frame (z, t) and transforms as a scalar density of weight 2 under general
transformations of the co-ordinates. Now consider the world tensor equation
r;!M =V8 o (- dety)fyD'l'yLieiP'l'e·
flo
(280.4)
By the quotient rule of tensor algebra, we easily verify that (280.4) is indeed a
tensor equation; i.e., the rank and weight of both sides of the equation agree.
1 ABRAHAM [1909, J] reviews several of the points of view regarding the aether relations
in material media. 2 A similar distinction was made in Chap. E, where developments resting on an equation
of state were separated from the general theory of energy. Cf. the remarks at the end of
footnote 5. pp. 617-618.
680 C. TRUESDELL and R. TouPIN: The Ciassical Field Theories. Sect. 281.
Since it is a tensor equation, it will be satisfied in every Co-ordinate system if
it is satisfied in one Co-ordinate system. In the frame (z, t) where the r have the
values (280.2), we can verify that the tensor equation (280.4) is satisfied if the
Maxwell-Lorentz aether relations (279.1) are satisfied. Thus we call (280.4) the
world tensorform of the Maxwell-Lorentz aether relations.
For some purposes it proves convenient to write (280.4) in a somewhat different form. Consider the world contravariant tensor of weight l defined by
-1 - 1DLI
@!M =='= y .
V-detf (280.5)
-1
The determinant of ~ is an absolute world scalar having the value - 1 in every
co-ordinate system. We find that the aether relations can be put in the alternative invariant form
V
--1 -1
'YJ!M == ~ @D'P @LIS !f''PS.
flo
-1 -1
In the frames where r has the specialform (280.2), ~will have the form
~ = Vc [ d~ • - o ;2]·
(280.6)
(280.7)
281. Dimensional transformations and the aether relations. The absolute
dimension of the world tensor fields a, p, 'rj, and cx are given by
abs. dim. a = [0],
abs. dim. 'r/ = [0],
abs. dim. p = [], }
abs. dim. cx =[]. (281.1)
From these dimensions and the rules of Sect. 273 follow the physical dimensions
of the fields B, E, J, Q, D, H, A, and V. Some of these have been stated elsewhere already, but we shall list all of them now for easy reference:
phys. dim. B =[ L -2],
phys. dim. J = [OL-2T-1J,
phys. dim. D = [OL-2],
phys. dim. A =[ L-1],
phys.dim.E= [L-1T-1], l
phys. dim. Q = [OL -s],
phys.dim.H= [OL-IJ-1],
phys. dim. V = [ r-1] .
(281.2)
Thus, in order that the aether relations be invariant under independent transformations of the units of length, time, charge, and magnetic flux, we must have
phys. dim. e0 = [ -1 0 L -1 T], phys. dim. #o = [ <1> o-1 L -1 T].
The constant c defined in (280.1) has, therefore, the dimension
phys. dim. c = [L r-1].
(281.3)
(281.4)
The dimension of the constant l [i; occurring in the world invariant forms of v;.; the aether relations (280.4) and (280.6) is given by
phys. dim. V eo = [0 -1].
flo
(281. 5)
The absolute dimension and physical dimension of absolute world scalars and
constants are equal. From (281.5) and (280.6) we see that dimensional invariance
Sect. 282. The Lorentz invariance of the Maxwell-Faraday aether relations. 681
of (280.6) requires that
abs. dim. 6; = 1 , (281.6)
which is consistent with the fact that the determinant of this tensor has the
value - 1 in all co-ordinate systems and unit systems. Dimensional invariance
of the canonical form (280.2) requires that
abs. dim. y = [L - 2]. (281.7)
Of course, this requires that the absolute dimension of the inverse y be [L 2].
When magnetic flux, charge, length, and time are measured in units of
Q = 1 Coulomb, = 1 Weber, L = 1 Meter, T = 1 Second, (281.8)
the fundamental constants in the aether relations have the values1 :
e = 8_854 X 10_12 Coulomb-Second
O Weber-Meter '
= 1.257 X 10_6 Weber-Second
flo Coulomb-Meter '
C = 2.998 X 108 _M~ter_ Second '
(281.9)
V e0 = 2_654 X 10_3 Cou!omb .
P.o Weber
282. The Lorentz invariance of the Maxwell-Faraday aether relations, Lorentz
transformations, and Lorentz frames. In Sect. 152 we showed that the Galilean
(inertial) frames of classical mechanics could be characterized as the preferred
co-ordinates in space-time for which the world tensors g!M, tA, and the Galilean
connection r assumed the canonical forms
g =[do" oo]' t = (0, 0, 0, 1), (282.1)
The Co-ordinates of any two such frames must be related by a Galilean transformation having the form
z'' = A'~ z' +ur' z"' + const, }
z"'' = z4 + const,
(282.2)
where A is an orthogonal matrix and the u'' are constants representing the
relative velocity of the two Galilean frames. If we also consider transformations
of the units of length and time and assign g the absolute dimension [L - 2] and
t the absolute dimension [T], the co-ordinates of a Galilean frame based on the
units of length L and T are related to the co-ordinates of a Galilean frame based
on the units L' and T' by a transformation of the form
z''(L') =L(A•;z'(L) +u''z4 (T) +const),} (282.3) z"'' (T') = T (z4' (T) + const).
So as to distinguish these more general transformations from those having the
special form (282.2), we shall call (282.3) a generalized Galilean transformation.
Following the same general procedure outlined above for the case of Galilean
space-time, let us consider the co-ordinate transformations and space-time frames
1 These values are quoted from STRATTON [1941, 8, p. 601].
682 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 282.
defined by the canonical form (280.2) of the Lorentz tensor r. A Co-ordinate transformation which leaves the Lorentz tensor invariant must satisfy the equations
" .Q' ., <1' -l.Q'<1'(L' T') = L_2 _u_x _ _ v_x_-l.Q<1 (L T) y ' oxD ox<1 y ' ' (282.4)
where
-1
y (L,T) = [!5" 0 - 0
~2 l , [ -y, (L', T') = - !5' s 0 __E._ l •
0 c2 L2
(282.5)
Since the ·:./ and y' are constants, it follows that the co-ordinate transformation
must be linear. A coordinate transformation satisfying the (282.4) will be called
a generalizecl Lorentz transformation. One can deduce the general form of these
transformations by multiplying the well known special Lorentz transformation 1,
Z~= (282.6)
by an orthogonal transformation of the spatial Co-ordinates, simple extensions
of the co-ordina tes corresponding to transformations of L and T, and by time
inversions. The generalized Lorentz transformations have the form
z'' (L') = LA'~{Ws + (C- 1) u-2 u' u.] z• (L) -Cu' z4 (T)}, )
z4 '(T') =±TC {z4 (T)- -}u,z'(L)}, (
282·7)
where Ais an orthogonal matrix, C = ( 1 - :: r~. and u2 == u, u' < c2 is the squarecl
relative speecl of the two Lorentz frames.
Various subgroups of the group of generalized Lorentz transformations have received
special attention and special names. If we keep the units of length and time fixed, so that
L = T = 1 in (282.7), we get what has been called the extended Lorentz group 2 • If we transform the units of length and time by the same factor so that L = T and c is invariant, we
get what BATEMAN3 has called the group of spherical wave transformations. This subgroup
is also called the conformal Lorentz group. BATEMAN noted that the aether relations are
invariant under the group of spherical wave transformations. If we require that det A = + 1
and disallow time inversions corresponding to the minus sign in (282.7) 2 , we get the proper
Lorentz transformations. On writing the aether relations in the form (280.6), we see that
they are invariant under the conformal Lorentz group.
A Lorentz transformation with ufc <{::: 1 approximates a Galilean transformation. In
special relativity theory, the notion of a stationary aether is abandoned. The inertial frames
of relativistic mechanics are identified with the Lorentz frames. The equationsand definitions
of mechanics are revised so as to be Lorentz invariant rather than Galilean invariant as in
classical mechanics.
The world invariant form of the aether relations (280.6) provides us with our principal
motivation for assuming that the charge ([ [( .9"3 , O)] transforms as an axial scalar rather
than as an absolute scalar under general transformations of the space-time coordinates.
With magnetic flux an absolute scalar and charge an axial scalar as we have assumed, the -1
transformation law of the tensor density (f) is
(282.8)
Had we assumed that both charge and magnetic flux were absolute scalars, the charge-current -1
potential 11 would have been an axial density. The transformation law (280.8) for the (f)
1 See, for example, BERGMANN [ 1942, J]. 2 CORSON [1953, 6, Chap. 1].
3 BATEMAN [1910, J].
Sect. 283. Polarization and magnetization. 683
would then have involved the square root of the Jacobian rather than the square root of
-1
the absolute value of the J acobian. Thus, in some coordinate systems, the coefficients 6)
would have been imaginary or complex-valued. We have considered this undesirable.
Another motivation for assuming charge to be an axial scalar is that the current density J
and the charge density Q, being tensor densities under time-independent transformations
of the spatial co-ordinates when this assumption is adopted, transform as absolute tensors
under the group of time independent orthogonal transformations of the spatial co-ordinates.
This is the transformation usually assumed for these quantities in traditional treatments of
electromagnetic theory. The distinction between axial tensors and absolute tensors and the
distinction between axial densities and densities is made necessary only because we find the
consideration of improper co-ordinate transformations to be of some major concern in many
physical theories. It is for this reason that we did not restriet ourselves at the outset to
the group of proper co-ordinate transformations.
283. Polarization and magnetization. a.) The principle of Ampere and Lorentz.
Polarization and magnetization are auxiliary fields introduced into the general
theory so as to serve in formulating constitutive relations for special types of
materials called dielectrics and magnets or magnetic materials. In the classical
theory of dielectrics and magnets, surfaces across which the electromagnetic
properties of the medium change abruptly are important for the description
of many electromagnetic phenomena. These surfaces are most often associated
with surfaces of discontinuity in the polarization and magnetization. Sudaces
of discontinuity in the polarization and magnetization are in turn associated
with surface distributions of charge and current. The point of view we adopt
here is that charge and current are the fundamental entities while polarization
and magnetization are simply auxiliary fields introduced as mathematical devices
providing a convenient description of special distributions of charge and current
in special types of materials. This may be called the principle of Ampere and
Lorentz1. Thus, if we anticipate a treatment of surfaces of discontinuity in the
polarization and magnetization, we must first introduce surface distributions
of charge and current into the conservation law of charge. For mathematical
simplicity, at the outset we did not burden the reader with this additional complication. The fundamental physical idea remains the same, conservation of
charge, but now we give a slightly more general mathematical expression for
the measure of charge. The mathematician will readily infer a general mathematical statement for the physical idea of conservation in terms of additive
set functions.
ß) Surface distributions of charge and current. Let E (a:) = 0 be the equation
of a 3-dimensional surface in space-time representing the history of a 2-dimensional
surface in space bearing a surface distribution of charge and current. We now
replace (270.1) by the more general assumption that the charge is expressiblein
the form
a: [ Ya. QJ = J (J. dYa + J w . d~*' (283.1)
.9',n 2'
where (J is the volume density of charge-current and the world contravariant 2-vector density w is the surface density of charge-current. The field w is defined
only at events corresponding to the surface E (~) = 0. In (283 .1), ~ n E denotes
the intersection of the arbitrary 3-dimensional surface ~ and the special3-dimensional surface E(~) =0. Here we have assumed that, if the intersection is not
empty, Ya n E is a 2-dimensional surface and d~* is its differential element.
Generalizing (270.5). we now write the law of conservation of charge in the
form
~ (J • d~ + J w . d~* = 0. (283.2)
1 WHITTAKER [1951. 39, p. 88, PP· 393-400]
684 C. TRUESDELL and R. TOUPIN: The Classical Field Theories. Sect. 283.
We know that, for all surfaces .9':; not intersecting E=O, as a consequence
of (283 .2) there exist potential fields 117'" such that
f6·d.9':i=-~11'f'""d~. (283-3)
W e can also construct fields 11.'7' such that
~ 11.'/'" d~ = 0
for every circuit not intersecting E(a:) =0, and suchthat
~ 11.'/'. d~ = - J w . d~*
(283.4}
(283.5)
for every circuit intersecting E = 0. Therefore, we can write the charge-current
potential equation in the form
J (J. d.9':; + J w. d~* =- ~ 11" d~. (283.6)
Y',n.z:
where 11 =w-+11.'7'· Eq. (283.6) generalizes (271.1). If the fields are continuously
differentiable except perhaps at E (:~:) = 0, at every event not on E (:~:) = 0 we
have
6 = div 11· (283.7)
Howcver, at E(a:) =0, the potential11 must have a discontinuity consistent with
the condition
or, in component form
[11] ·gradE= -w·gradE,
[1JDLI] oLlE=- wDLI oLlE.
(283.8)
(283.9)
The aether relations (280.4) or (280.6) are postulated to hold between the
electromagnetic field p and the generalized potential 11 of (283.6). Thus, the
electromagnetic field is discontinuous at a surface bearing a distribution of charge
and current.
The 3-dimensional form of the law of conservation of charge and the charge
and current equations taking into account a surface distribution of charge and
current can be obtained from the world invariant equations (283.2) and (283.6)
by introducing a Euclidean frame. The resulting equations have an interesting
but complicated structure. Since we do not make use of these equations here,
we leave to the reader the task of deriving them.
Having established these preliminary results on surface distributions of
charge and current, we are in a position to give a fairly general treatment of
polarization and magnetization in moving and deforming material media, allowing that these fields may be discontinuous at special surfaces.
y) Polarization charge and current and magnetization current. As an introduction to the more difficult dynamical case, we note first some simpler and better
known results of the static theory of dielectrics and magnets. These latter
theories are concerned with the determination of the electric and magnetic fields
in regions of space containing dielectric and magnetic materials at rest.
It is customary in the electrostatic theory of dielectrics to divide the charge
distribution into two types: ( 1) the bound charge or polarization charge; (2) the
free charge1• The polarization charge contained in a 3-dimensional spatial region v
is given by
[[v, QJ =-p P · dd (283.10)
1 The polarization charge is also called the "induced" charge, and the free charge is
also called the "real" or "true" charge. See, for example, STRATTON [1941, 8, Sec.t. 3.13,
p. 183].
Sect. 283. Polarization and magnetization. 685
where 6 is the complete boundary of v and P is the density of polarization. One
generally considers problems where P is a continuously differentiable field
except at special surfaces where it suffers a jump or discontinuity. Excepting
points on these surfaces, we can use the integral theorem to transform the surface
integral (283.10) to a volume integral. Thus, if v is a region in which the polarization field is everywhere continuously differentiable, we have
C\:[v,QJ =- JdivPdv. (283.11)
The scalar field - div P may be called the volume density of polarization charge.
If the region v is divided into two regions v+ and v- by a surface 6* across which P
suffers a jump [P], and if P is continuously differentiable in the closure of the
regions v+ and v-, we can write (283.10) in the form
. (290.7)
Thus, for example, transformations of the unit of magnetic flux are generally
regarded as fixed in terms of the transformations of M, L, T, and Q by the relation
(290.8)
This is consistent with our having called DxB a momentum density and
Ex H a flux of energy without introducing any additional dimensional constants
of proportionality. Thus we have
phys. dim. D X B = [M L - 2 y-1],
phys. dim. EX H = [M T- 3],
(290.9)
(290.10)
in agreement with the customary dimensions of momentum per unit volume,
and energy per unit area per unit time.
291. Conclusions. The principal result contained in this chapter is a world
tensor invariant formulation of the laws of conservation of charge, magnetic
flux, energy, and momentum, independent of special assumptions as to the geometry of space-time.
In Subchapter IV, we motivated the law of conservation of energy and momentum by considering the classical energy and momentum equations of a continuous
medium interacting with an electromagnetic field. We assumed that the stressenergy-momentum tensor T/} was a 2-event world tensor field. While the chargecurrent potential and the electromagnetic field are connected through the aether
relations, no general relation between the stress-energy-momentum tensor and
the other fields has been proposed here. The basic content of Subchapter IV
is thus independent of the two preceding. Rather, the conservation law of energy
and momentum as formulated here is to be regarded as encompassing and
generalizing the purely mechanical considerations of Subchapters D II and E I;
it also includes and extends the various purely electromagnetic proposals for
conservation of energy and momentum.
The conservation laws of charge, magnetic flux, momentum, energy and
angular momentum constitute an underdetermined system of equations. They
must be supplemented by relations between the conservative fields. The aether
relations are an example of such relations. The many determinate special theories
700 C. TRUESDELL and R. TouPIN: The C!assical Field Theories. Sects. 292, 293·
encompassed by the conservation laws is evidence of the different special ideas
of material behavior consistent with them. The conservation laws provide a
general framework common to all electrodynamic theories, classical and relativistic, with which we are familiar. Special and general relativity theory entail
modification of our concepts of time, mass, energy, momentum, the aether relations, and inertial frames, but have yet to alter in principle· either the formal or
the intuitive aspects of the laws of conservation of charge, magnetic flux, energy,
and momentum.
G. Constitutive Equations.
I. Generalities.
292. The nature of constitutive equations. In Sect. 7 we explained that the
field equations and jump conditions express the general principles of mechanics,
thermodynamics, and electromagnetism, while constitutive equations define
ideal materials, which are mathematical models of particular classes of materials
encountered in nature. The preceding treatise has developed in detail the properties of motions and of the fundamental physical principles of balance, which
imply both field equations and jump conditions.
In this final chapter we illuminate the concept of constitutive equation by
stating general principles and adducing common examples.
293. Principles to be used in forming constitutive equations. It should be
needless to remark that while from the mathematical standpoint a constitutive
equation is a postulate or a definition, the first guide is physical experience,
perhaps fortified by experimental data. However, it is rarely if ever possible
to determine all the basic equations of a theory by physical experience alone.
Every theory abstracts and simplifies the natural phenomena it is intended to
describe (cf. Sect. 4). Supposing that the theorist has assembled the facts of
experience he wishes to use in defining an ideal material, we now list the mathematical principles he may call to his aid when he attempts to formulate definite
constitutive equations.
We know of no ideal material for which all these principles have been demonstrated to hold, although for the simpler classical theories it is generally believed
that they do.
rx) Consistency. Any constitutive equation must be consistent with the general
principles of balance of mass, momentum, energy, charge, and magnetic flux.
This is obvious and easy to say, but to test it in a special case may be difficult.
ß) Co-ordinate invariance. Constitutive equations must be stated by a rule
which holds equally in all inertial Co-ordinate systems, at any fixed time. Otherwise, a mere change of description would imply a differentresponsein the material.
Such a rule may be achieved, usually trivially, by stating the equations either
in tensorial form or by the aid of direct notations not employing co-ordinates
at all1 .
y) I sotropy or aeolotropy. Materials exhibiting no preferred directions of
response are said to be isotropic. There are differences of opinion as to how
this somewhat vague concept should be rendered mathematically precise. In
the common phrase "isotropic material" we are unable to discem any meaning.
Only after a kind of material has been defined by a particular constitutive equa1 However, the Iiterature of hydraulics abounds in "power laws" which are not invariant,
as was pointed out by KLEITZ [1873, 4, § 22] in a work dating from 1856; the same may be
said of rheology.
Sect. 293. Principles to be used in forroing constitutive equations. 701
tion does it become possible to state an unequivocal concept of isotropy. Pursuant
to this idea, NoLL1 defines the isotropy group of a material as the group of transformations of the material coordinates which leave the constitutive equations
invariant. The nature of this group specifies the symmetries of the material.
A material is then isotropic if its isotropy group is the full orthogonal group.
Various kinds of aeolotropy, which is a symmetry with respect to certain preferred
directions, are shown by materials whose isotropy groups are proper subgroups
of the orthogonal group, or are other groups.
While these definitions reflect broader ideas, up to this time they have been
rendered concrete only in pure mechanics, since the general constitutive equations
expressing energetic and electromagnetic response are not yet known. In theories
not limited to purely mechanical phenomena it is usual to define isotropy in
respect to the variables A, B, C, ... entering a constitutive equation by requiring
the functional dependence of A, B, C, ... to be isotropic in the mathematical
sense. A material which is isotropic in respect to one property, e.g. stress
and strain, need not be so in respect to another, e.g., electric displacement
and electric field. There is a large current Iiterature 2 concerning means of expressing isotropy and aeolotropy, but there is no adequate survey of the fieldas yet.
(J) Just setting. Constitutive equations connecting a given set of variables
should be such that, when combined with all the principles of balance affecting
these same variables, there should result a unique solution corresponding to
appropriate initial and boundary data, and a solution that depends continuously
on that data. This principle can rarely be used. With great mathematicallabor,
it has been proved to hold only in the simplest classes of boundary-value problems
in the simplest classical theories 3• At best, it shifts the problern to another domain,
that of finding adequate initial and boundary conditions 4•
e) Dimensional invariance. It is essential that included in each constitutive
equation should be a full list of all the dimensionally independent moduli or
material constants upon which the response of the material may depend. This
requirement has often been neglected in recent work, although it is a commonly
accepted principle of physics, known at least since GALILEo's day, that any fully
stated physical result is dimensionally invariant. (It is tacitly understood that
dimensionless moduli need not be listed, since they cannot be specified until
a particular functional form is stated.)
The use of dimensional invariance to classify constitutive equations and in
some cases to exclude inappropriate terms was initiated by TRUESDELL 5• The
tool used is the classical n-theorem, a precise statement and simple rigorous
proof of which was recently given by BRAND 6•
1 [1948, 8, §§ 19-20]. 2 The basic principles derive froro work of CAUCHY. Nuroerous articles on the subject
have appeared in the Journal and Archive for Rational Mechanics and Analysis, 1952 to date. 3 The honored custoro of verifying that the nurober of equations equals the nurober of
unknown functions seeros to bring corofort.
' In roany cases the traditional setting is known to fail to lead to a unique solution. E.g.,
in buckling probleros the specification of the loads on the boundary is not sufficient to insure
a unique solution. Such probleros roay nevertheless be regarded as weil set if we aroplify
either the boundary conditions or the constitutive relations. E.g., in the case of buckling we
roay include as a part of the constitutive equations the requireroent that the elastic energy
shall be a roinirouro with respect to coropatible deforroations; a unique solution results.
Still better, the solution for static buckling roay be regarded as the liroit of a uniquely soluble
dynaroic problero. 5 [1947, 17] [1948, 32 and 33] [1949, 33 and 34] [1950, 32, §§ 7-8] [1951, 27, §§ 22, 25]
[1952. 22] [1952, 21, 1 § 47, 62-65. 67-69. 74-75] [1955. 28, § 1]. 6 [1957. 1].
702 C. TRUESDELL and R. TouPIN : The Classical Field Theories. Sect. 293.
C) Material indiflerence. The most important and the most frequently used
correct idea for formulating constitutive equations is that the response of a material
is independent of the observer. In proposing the first of all constitutive equations
for deformable materials, namely, the law of linear elasticity, HoOKE1 gave the
first dim hint toward this principle in the suggestion that by carrying a spring
scale to the bottom of a deep mine or to the top of a mountain, the change of
gravity could be measured. A more striking example is furnished by use of a
spring, attached to the center of a horizontal, uniformly rotating table, to measure
the centrifugal force acting upon a terminal mass. The assumption implicit in
these measurements is that the force exerted by the spring in response to a given
elongation is independent of the observer, being the same to an observer moving
with the table as to one standing upon the floor.
Another example is furnished by FoURIER's law of heat conduction, to be
discussed in Sect. 296 below. According to this law, the flow of heat is proportional to the temperature gradient. Since both of these quantities are independent
of the motion of the observer, in order that FOURIER's theory satisfy the principle
of material indifference the constant of proportionality, or heat conductivity,
must also be so invariant and hence is an absolute scalar under general transformations of space-time.
This invariance or material indiffere:nce has nothing to do with the coordinate invariance described in Subsection ß, above. Indeed, if we conceive forces
and motions directly, without the intermediary of co-ordinates and compon~nts,
the requirement of material indifference is not thereby satisfied automatically 2•
Neither can it be achieved by a thoughtless tour-dimensional formulation, for
the precise meaning of "observer" in classical mechanics, where the time is a
preferred co-ordinate, must 'be stated in terms of time-dependent orthogonal
transformations, not of more general ones.
That this requirement has a fundamental physical meaning is shown by the
fact that the laws of motion themselves do not enfoy invariance with respect to the
observer. "Apparent" forces and torques, in general, are needed to reconcile
the descriptions of mechanical phenomena given by two observers in relative
motion (Sect. 197). The principle of material indifference states that these are
the only mechanical effects of the motion. For example, the deflection of the
spring on the rotating table is supposed to be proportional to the entire force
acting parallel to the spring, and thus, since the end is free, to measure precisely
the force measurable by an observer at rest upon the table, this force being the
"apparent" force that accompanies the rotation of the observer's frame with
respect to one in which the spring would suffer no extension 3•
All the classical linear theories, and some of the non-linear ones, satisfy the
principle of material indifference trivially and automatically. For example, since
mutual distances are invariant under change of observer, any theory in which
the stress is determined only by mutual distances and differences or derivatives
of mutual distances or relative velocities of particles exhibits material indifference. This is the case with most theories of elasticity and fluid motion 4 •
1 [1678, 1]. 2 Cf. THIRRING [1929, 10, § 6]: "Es sei ausdrücklich darauf hingewiesen, daß Unabhängigkeit von der Koordinatenwahl durchaus nicht mit Unabhängigkeit vom Bezugssystem zu
verwechseln ist." 3 If the spring is not idealized as massless, a simple correction is made for the centrifugal
force acting upon the mass of the spring itself, but otherwise the response of the spring is
unchanged.
' It is important not to confuse material indifference with any kind of tensorial invariance.
Some constitutive equations relate quantities which transform as tensors under change of
Sect. 293. Principles to be used in forming constitutive equations. 703
But as soon as time rates of stressing come into consideration, this invariance
holds no longer. The problern was first faced and solved correctly, in a special
case, by CAUCHY1 ; the idea was expressed more clearly and given a somewhat
different form by ZAREMBA 2• While several recent studies have attacked the
problern in one way or another, a general mathematical statement, for purely
mechanical theories, was first achieved by NoLL3• Since this principle has not
yet been formulated in a scope broad enough to cover all situations envisaged in
this treatise, and since in the purely mechanical case it will be presented in the
article, "The Non-linearField Theories of Mechanics" (Vol. VIII, Part2), we do not
discuss it further here, though we give an example of its use in Sect. 298, below.
rJ) Equipresence. In the most general physical situations, mass, motion, energy,
and electromagnetism are simultaneously present. In the classical theories, the
variables describing these phenomena are divided more or less arbitrarily into
classes, the members of each of which are supposed to influence only each other,
not the members of the other classes. Stress is coupled with strain or stretching,
flux of energy with temperature gradient, electric displacement with electric field
strength, magnetic intensity with magnetic induction.
Such simple divisions of phenomena, perhaps lingering remnants of old views
on "causes" and "effects ", suffice to describe a great range of physical experience,
but, as is well known, the members of the different classes often react upon
each other, and there are also various special theories of energetico-mechanical,
electro-energetic, and electro-mechanical interactions. Here the process of the
theorists has been conservative. They have maintained as much of the old
separation of variables as possible, relinquishing just enough of it to include
the bare existence of a particular phenomenon of interaction.
In truth, the separation in itself is unnatural and unjustified by physical
principle. Resulting only from the gradual discovery of individual phenomena,
it reflects the old opinions that break physics up into compartments. The present
treatise is conceived in the belief that the classical field theories, if rightly understood, describe most of the gross physical phenomena of nature. Constitutive
equations, then, should not artificially divert these theories into disjoint channels.
This same view, for energetico-mechanical effects alone, results also from
statistical models of continuous matter. There, stress and flux of energy appear
as gross or mean expressions of a purely mechanical process. The separation
frame; e.g., the theories of fluids discussed in Sects. 298-299 relate t to d, both of which,
by the results in Sects. 144 and 211, are indeed tensors under change of frame. The principle
of material indifference requires that any relation between such tensors be a tensor relation
under change of frame. But some variables occurring in constitutive relations are not tensors
under change of frame, the vorticity w and the deformation tensor C being examples. The
classical theory of finite elastic strain (Sect. 302) relates t to C; its constitutive equations
satisfy the principle of material indifference, but the elastic coefficients do not generally
transform as tensors under change of frame. 1 [1829, 4].
2 [1903, 20 and 21] [1937, 12, Chap. I, § 2]. 3 [1955. 18, § 4], "isotropy of space"; [1957. 11, § 9] [1959, 9, § 7] [1958, 8, § 11],
"principle of objectivity". A noteworthy earlier attempt is given by ÜLDROYD [1950, 22]:
"The form of the completely general equations must be restricted by the requirement that
the equations describe properties independent of the frame of reference ... Moreover, only
those tensor quantities need be considered which have a significance for the material element
independent of its motion as a whole in space." ÜLDROYD's process is to write all constitutive
equations in convected co-ordinates, extension to spatial systems then being achieved by
imposing tensorial invariance. Cf. LonGE [1951, 16]. While this procedure leads to correct
results, it is not the only one possible, nor is ÜLDROYD's formulation of the problern unequivocal. TheprinciplesofRIVLINandERICKSEN [1955, 21, § 11 ff], ofTHOMAS [1955, 24and25],
and of CoTTER and RrvLIN [1955. 4], which likewise rest upon use of special co-ordinate
systems, are special cases of NoLL's.
704 C. TRUESDELL and R. TouPrN: The Classical Field Theories. Sect. 294.
of stress as arising from deformation and flux of energy from changes of temperature emerges, not without reason, as a first approximation; but on a finer
scale, it is illusory. Any attempt to draw precise conclusions from statistical
theories show that the classical separation of phenomena is artificial and unphysical1.
For energetico-mechanical phenomena, a principle denying any fundamental
separation was formulated by TRUESDELL 2 : The stress and the flux of energy
depend upon the same variables. Here we propose the following more general
principle oj equipresence: A variable present as an independent variable in one
constitutive equation should be so present in alt.
Let it not be thought that this principle would invalidate the classical separate
theories in the cases for which they are intended, or that no separation of effects
remains possible. Quite the reverse: The various principles of invariance, stated
above, when brought to bear upon a general constitutive equation have the
effect of restricting the manner in which a particular variable, such as the spin
tensor or the temperature gradient, may occur. The classical separations may
always be expected, in one form or another, for small changes-not as assumptions, but as proven consequences of invariance requirements. The principle of
equipresence states, in effect, that no restrictions beyond those of invariance are
to be imposed in constitutive equations. It may be regarded as a natural extension of ÜCKHAM's razor as restated by NEWTON 3 : "We are to admit no more
causes of natural things than such as are both true and sufficient to explain
their appearances, for nature is simple and affects not the pomp of superfluous
causes." This more general approach has the added value of showing in what
way the classical Separations fail to hold when interactions actually occur. We
illustrate these remarks in Sect. 307, below.
II. Examples of kinematical constitutive equations.
294. Rigid bodies. A body is rigid if it is susceptible of rigid motions only:
The distance between any two particles of the body is constant in time. The
motion of the body is then completely determined when the position of one particle and the orientation of an appropriate rigid frame with origin at that particle
are given as functions of time. For any body, rigid or not, the motion of the
center of mass c is determined from the total assigned force by integration of
(196.6). This fact is of little or no use when deformation occurs, since the center
of mass generally moves about and fails to reside in any one particle, but the
center of mass of a rigid body is stationary within it; more precisely, in any frame
with respect to which the body is at rest, the center of mass c is also at rest.
Thus we may say that the motion of the center of mass determines, or, more
properly, constitutes the translatory motion of the rigid body.
The kinematical statement that there exists a frame with respect to which
the body is permanently at rest may be expressed in terms of the tensor of
inertia and the moment of momentum as follows: In a suitably selected frame,
denoted by a prime, we have
il'[O'l = 0, ~'[0'] = 0 (294.1)
for all t. These are the constitutive equations of a rigid body. Substituting them
into the general equations of balance of moment of momentum (197.3) 4 and
1 M. BRILLOUIN [1900, 1, § 37]. supported by many later results. 2 [1949, 34, § 19] [1951, 27, § 19], "BRILLOUIN's Principle". 3 [1687, 1, Lib. III, Hypoth. I].
Sect. 295. Diffusion of mass in a mixture. 705
(197.4) yields Euler's equations forarigid body1 :
w -~'[ '1 +wx~'[ 'l. w = ~[O'J_ c'xWlb. (294.2)
Here ~[O'l is the applied torque with respect to 0'. The last term on the righthand side vanishes if 0' is the center of mass: c' = 0; it vanishes also if there is
some one particle of the body which moves at uniform velocity in an inertial
frame and if 0' is chosen at that particle: b = 0. In these cases and in any others
when c' X b is known, the angular velocity w of any frame rigidly attached to the
body is determined to within an arbitrary initial value by the applied torque, the
mass and the translatory motion of the body, and the tensor of inertia ~'[O'J with
respect to the same frame. This conclusion follows from the existence and uniqueness of
solution to (294.2), which is a non-linear differential equation of first order for w.
Thus the purely kinematical assumption that the motion is persistently rigid
renders it simply determinable. This is no wonder, for the infinite nurober of
degrees of freedom generally present in a body of finite mass and volume have
been reduced to six. While the laws of motion (196-3) are linear in an inertial
frame, EULER's equations (294.2) are non-linear, since they state the balance of
moment of momentum in a non-inertial frame, the coupling between these frames
being that expressed by the principles of apparent torques in Sect. 197.
In a rigid body, we are at liberty to imagine internal stress and flux of energy
and their respective balances, but since the motion is determinate without these
considerations, we invoke ÜCKHAM's razor 2 to excise them.
A theorem of balance of mechanical energy follows from (294.2). Choosing
the origin of the primed frame at c, by (168.8) and (168.9) we obtain
2~ = Wlc 2 + w. ~[o'J. w.
By differentiating this result and using (196.6) and (294.2) we obtain
Sr= iJ. c + ~'[O'J. w.
(294.3)
(294.4)
This equation asserts that the entire rate of working of the applied force iJ in
producing the translatory motion and of the applied torque ~'[O'J in producing
the rotation is converted into kinetic energy. Assumptions analogaus to those
discussed in Sect. 218 suffice to derive from (294.4) a law of conservation of total
energy.
The theory of rigid bodies is presented in detail in the article by SYNGE in
this volume.
295. Diffusion of mass in a mixture. A simple theory of diffusion of mass in
a non-uniform binary mixture was proposed by FrcK3 : The rate of diffusion of
1 The theory is usually and justly attributed to EULER [ 1765, 1]; it has a lang prior and
a short but important subsequent history, which has never been adequately traced. 2 Adopted byNEWTON [1687, 1] as the first ofthe"Hypotheses", calledinlatereditions
Rules of Philosophizing", set at the head of Book III. On p. 704, we have quoted in full the
statement in the edition of 1687.
Here we mention that the usual derivation of EuLER's equations given in textbooks
must also be excised by ÜCKHAM's razor. That derivation rests on an argument, summarized
in Sect. 196 A, showing that if all the mutual forces between the particles of a body are central
and pairwise cancelling, balance of momentum implies balance of moment of momentum.
In a rigid body, by definition, the mutual forces never manifest themselves in any way except
to maintain rigidity. Rigidity has already been assumed, and it suffices; to hypothecate
mutual forces is to multiply causes. In any case, the derivation is absurd from the Standpoint
of modern physics, which does not represent the smallest portians of any kind of body as
stationary centers of force. 3 FicK [1855, 1, pp. 65-66] gave as his only physical basis an assertion that diffusion
of matter in a binary mixture at uniform total density and pressure seems analogaus to flow
of heat according to FouRIER's theory (Sect. 296).
Handbuch der Physik, Bd. Ill/1. 45
706 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 295.
mass of the constituent m: is proportional to the gradient of the density of that
constituent and is such as to tend to restore uniformity. In equations,
D>O, m:=1,2, (295.1)
where the coefficient D, the diffusivity, has the physical dimensions [L 2 r-1].
FICK assumed that f!,k = 0, as indeed is necessary in orderthat (295.1) be compatible with (158.8) 1 • Substituting (295.1) into (159.2) 2 , an expression of the conservation of mass, we have the diffusion equation:
(295.2)
where (159.4) is assumed to hold. This equation is of the same form as the
isotropic case of FouRIER's equation (296.10).
FrcK's law was proposed and has been applied under various specializing
hypotheses, typically that c~1 =0, that there is no mean motion, and that the
pressure is constant. Also, only a binary mixture was envisioned, and while formally (295.1) is meaningful for an arbitrary nurober of constituents, it does not
allow for their possible mutual actions. While there have been numerous studies1
1 The main features of a general theory of diffusion might have been found in the kinetic
theory of gases; indeed, MAXWELL [1860, 2, Prop. XVIII] [1867, 2, Eq. (76)] emphasized
the fact that diffusion arises from equal and opposite forces proportional to the differences of
velocities, so that (295-5) is virtually suggested by him. However the later development of
the formal theory (e.g. [1939, 6, § 8.4]), emphasizing special features of the binary case,
tended to obscure the simple mechanical idea. The effort of HELLUND [1940, 13] to calculate
the basic equations for multi-constituent mixtures according to the kinetic theory did not
lead to clear results.
The general continuum theory based upon (295.5) and (295.8) may fairly be attributed
to STEFAN [1871, 6, Eqs. (3)], who gave the special case of (295.6) appropriate to a ternary
mixture of perfect fluids, but his work seems to have attracted no attention, and many
inferior attempts were published later. DuHEM [1893, 2, Chap. VI, §I] in effect noted that
a constitutive equation forafluid mixture should specify the functional form of piJ!, but
instead of perceiving a connection with Frcx's law, he proposed to determine the piJ! by
assuming the mixture to be such that all constituents may participate in a common infinitely
small isentropic motion. We do not understand the meanings of all the terms in the theories
of diffusion proposed by }AUMANN [1911, 7, §XXXVI] [1918, 3, §§ 136-141, 150-154]
and LoHR [1917, 5, §§ 5, 7, 12, 15].
A thermodynamic theory of diffusion, including thermal diffusion and the diffusionthermal effect, seems first to have been proposed by EcKART [1940, 8, Eq. (49)]. Cf. the
earlier and more special theory of ÜNSAGER and Fuoss [1932, 10, § 4.12]. A more fully
elaborated theory of the same kind, adopting the "Onsager relations ", was proposed independently by MEIXNER [1941, 2, § 5] [1943, 2, § 2] [1943, 3, § 4]. The later Iiterature
in this field does not always take pains to recognize all the conditions laid down by MEIXNER
but otherwise seems to diverge from his work only in minor points; cf. LAMM [1944, 7], LEAF
[1946, 7, Eq.(48)), PRIGOGINE [1947, 12, Chap. 10, §§ 1 and 5, Chap. 11, §3], DE GROOT [1952,
3, §§ 45-46], KIRKWOOD and CRAWFORD [1952, 12], HIRSCHFELDER, CURTISS and BIRD [1954,
9, § 11.2d]. There is also a more primitive theory of ÜNSAGER [1945, 4, pp. 242-247],
resting upon an inverse of (295.4); apparently only special circumstances were envisioned,
but it is not clear what they are. In the thermodynamic theories, little if any use is made
of mechanical concepts and principles.
Meanwhile, clear results from the kinetic theory approximations for a multi-constituent
mixture had been obtained byCowLING [1945, 2, Eq.(34)]; since his dl}! is approximately
(though not exactly) proportional to our piJ!, his result is equivalent to the inverse of a special
caseof (295-5). Hisanalysis was generalized by CuRTISS and HIRSCHFELDER [1949, 4, Eqs. (19),
(20), (21)] so as to include thermal diffusion; after allowance is made for the approximations
of the kinetic theory, their result is seentobe a special case of (295-5). In the kinetic theory
the relation (295-8) is valid in first approximation even if st >2 but has not been shown
to hold in a more accurate treatment.
The theory of STEFAN was proposed anew by ScHLÜTER [1950, 25, Eq. (3)], [1951, 23,
Eqs. (1) to (3)] and by JoHNSON [1951, 14]; the latter gave an argument indicating that
F!BIJ!/(!!B(!IJ! is roughly independent of the densities.
Sect. 295. Diffusion of mass in a mixture. 707
of more general cases, we prefer to follow an independent argument 1 toward
a linear theory of diffusion which reduces to FicK's law under the conditions in
which it is usually applied.
First, we note that by means of (158.8), we may express (295.1) alternatively in
theform
1 em e!B e'2l e!B , , e'll,k= 2n (elßu!Bk-e'Hu'Hk)=-eD(u!Bk-u'Hk) =en(x!Bk- xu), (295-3)
still for a binary mixture. This symmetrical expression indicates that it is an
excess of the mass flow of one constituent over another's that increases the density
gradient of the latter. Thus for a general mixture we might have
~ ~
em,k = L Y!B'H (e'H u!B k - e'll u!Bk) = L .F!B ~~ (x!Bk - Xmk). (295 .4) !B~l !B~l
But since the right-hand side of this relation is equal to a linear combination
of the excess of the momenta of all constituents above the momentum of the
constituent lll:, a still more natural idea of diffusion is expressed by the more
general constitutive equation ~
eP'H = L .Flß'll (x!B- x'll), (295.5) !B~l
where P'H is the supply of momentum, defined by (215.2) and restricted by (215.5).
In this relation the purely kinematical view expressed in FICK's law (295.1) is
replaced by a dynamical one: Diffusion gives rise to a force tending to restore
uniformity. The dynamical equations corresponding to (295.5) follow at once
from (215 .2) : ~
("k jk) km " ;= ( 'k 'k) e~( x'H - 'll - t'll, m = L.. \ll'll x!B - x'll . (295.6) !B~l
Since c'll=O, in order that (295.5) be consistent with the balance of total
momentum in the mixture, expressed by (215.5), it is necessary and sufficient
that ~
L (.F!B'll- .F'llm) = o. 'H~l
When ~ =2, this result is equivalent to 2
.F'llm= .F!B'll·
1 The ideas given here are presented more fully by TRUESDELL [1960, 6].
(295.7)
(295 .8)
2 This condition, for the binary case, originated in the work of MAXWELL [ 1860, 2,
Prop. XVIII], to whom is due the appeal to "NEWTON's third law" which JoHNSON [1951, 14]
phrased more generally: "The force exerted by species ~ on species 18 is equal in magnitude
and opposite in direction to that exerted by 18 on ~ .... " The argument in the text above,
basically different in that it rests upon the principle of linear momentum, was first suggested
by STEFAN [1871, 6, p. 74]: "Da die ... Kräfte aus Wechselwirkungen zwischen den im
Element dx dy dz zur selben Zeit befindlichen Teilchen des ersten und zweiten Gases entspringen, so ändern sie die Bewegung des Schwerpunktes dieses Elementes nicht." Indeed,
this establishes (295.8) in the binary case, but if .lt >2, only (295-7) follows. For the ternary
case, STEFAN assumed (295.8) without analysis or comment. On the other hand, there is
twofold insufficiency in the argument of MAXWELL and J OHNSON, for in order to invoke
"NEWTON's third law" we have to know that the diffusive term represents the entire force
arising from the mutual forces of the two species, which it generally does not, and, secondly,
the influence of species GI: upon the mutual action of species ~ and 18 is left out of account.
We feel the true content of the Maxwell-Johnson argument does not yield a proof but merely
suggests the supplementary assumption (1) in the text above, stating that species GI: has
no effect upon the relative diffusion of species ~ and 18.
The relations (295.8) for arbitrary .lt may weil be named "STEFAN's relations ". The
"Onsager reciprocity relations" for pure diffusion refer to a different set of coefficients.
LAMM [1954, llA], who considered STEFAN's relations to be "self-evident for physical
reasons ", showed that they are equivalent to ÜNSAGER's relations when .lt = 3; for general .lt,
this is established by TRUESDELL [1960, 6] ..
45*
708 C. TRUESDELL and R. TouPrN: The Classical Field Theories. Sect. 295·
but if ~ >2, such a relation of syrnmetry does not hold without further assumptions. It may be shown tobe sufficient that (1) r~~ shall not depend upon 1?e=O, no wave can travel at the speed given by (297.16) except in the very special
circumstances when (J = const. This difference ceases to seem paradoxical when we recall
that the conduction of heat is a dissipative mechanism, and that when the second main dissipative mechanism, that of viscosity, is included in the mathematical model, both the kinds of
waves considered here become impossible, as is shown in Sect. 298. The Iimit behavior is
discussed by DuHEM [1901, 7, Part III, §§ 2- 3] and by SERRIN, § 57 of Mathematical Principles of Classical Fluid Mechanics, this Encyclopedia, Vol. VIII. Part 1.
3 The first analysis of this kindisthat of STOKES [1848, 4, p. 355], faulty from failure to
take account of the balance of energy. Fora perfect gas, the full results are due to HuGONIOT
[1887, 2, §§ 154, 161-163]; cf. the summary of VIEILLE [1900, 10, p. 185]. The general theory
of weak shocks had been given earlier by CHRISTOFFEL [1877. 2, §§ 2-5]. An exposition for
general tri-variate fluids is given by SERRIN, Sect. 56 of the article just cited.
Sect. 298. Linear!y viscous fluids. 715
a jump of 17 which is not uniform, thus rendering the flow baroclinic and destroying the circulation preserving property, as weil as inducing a jump of vorticity1.
Further general theorems of gas dynamics follow from furth~r assumptions;
typically, that the flow is steady, that f = 0, etc. 2•
298. Linearly viscous fluids. To consider a substance which in equilibrium has
the same behavior as that predicted by EuLER's equation (297.1), viz.
grad p = ef, (298.1)
yet when in motion can support appropriate shearing stresses, assume that the
stress tensor t be a linear function of the velocity x and the velocity gradient.
By (90.1) we may write
t =g(x,w,d), (298.2)
where g is a linear function. The constitutive equations (298.2) define a linearly
viscous fluid, it being supposed also that there is no couple stress: m = 0.
We now apply the principle of material indifference (Sect. 293 0). The constitutive equation (298.2) is to have the same form for all observers. To an
observer in a co-rotational frame, x = 0 and w = 0; for him, (298.2) reduces to
a relation giving t as a function of d alone, and therefore, since both t and d
transform independently of x and w [cf. (144.3) and (211.1)], it must reduce to
such a relation for all observers. l.e. 3,
t = f(d). (298.3)
N ow consider frames whose axes coincide with the principal directions of d
(Sects. 82 and 83), so that (298-3) becomes
(298.4)
For a definite assignment of the undirected axes different assignments of the
positive senses yield four different positively oriented frames, any one of which
may be obtained from any other by a rotation through a straight angle about
one axis. Under such rotations, it follows from (82.6) that da is invariant. By
the principle of material indifference, then, fkm in (298.4) is invariant under
these rotations: ftm=fkm• say. But t transforms according to the tensor law;
in particular, under rotation through a straight angle about the k-axis we have
ttm = - tk m for k =I= m . Comparing these two results yields tk m = 0 if k =I= m. Thus
the principal axes of stretching are also principal axes of stress. Since further
orthogonal transformations may permute the da in any way, the principal stress tb
is a symmetric function of them. Hence the relation (298.3) reduces to one giving
1 Indicated by HADAMARD [1903, 10, Notelll]. A modern exposition is given by SERRIN,
§ 54 of the article, just cited.
Despite the above-stated facts, the speed of sound behind the shock is still given by
(297.13). since, as remarked by TRUESDELL [1951, 36], this follows from (297.12) and the
more general theorem proved in footnote 4 on p. 712.
2 A development of these principles is given by TRUESDELL [1952, 23].
3 This equation, in an inertial frame, along with f(O) = - p 1, was taken as the definition
of a fluid by STOKES [1845, 4, § 1].
716 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 298.
t as an isotropic function of d. In other words, all fluids included in the definition
(298.2) are necessarily isotropic1•
The mostgenerallinear isotropic function t of a symmetric second-order tensor
d may be written in the form 2 of the Na vier-Poisson law:
t=-P1+Ud1+2p,d, }
t" m m = - p (Jk + Ä. dq lJk + 2 IL d" q m r m•
(298.5)
where we have used the requirement that t= -p1 when d =0. From (298.5)
it follows that t is symmetric. Therefore CAUCHY's second law in the form
(205.11) is satisfied automatically. In other words, it is a consequence of the
principle of material indifference that a fluid cannot support extrinsic couples;
when there are no extrinsic couples, balance of moment of momentum follows from
balance of momentum 3• Substitution of (298.5) into CAUCHY's first law (205.2)
yields a system of three differential equations known, at least when subjected
to further simplifying assumptions, as "the Navier-Stokes equations ". If the
pressure p and the three components i" of the velocity are regarded as unknowns
and the coefficients Ä., p, are regarded as given, the system is still underdetermined,
but since the means of rendering it determinate are essentially the same as those
used in the theory of perfect fluids (Sect. 297), they will not be discussed here.
Since the physical dimensions of the coefficients Ä. and p, in (298.5) are
[M L -1 r-1], it is plain that our definition (298.2) violates the principle stated
in Sect. 293 e. In rectification, we replace (298.2) by the complete definition
of a linearly viscous fluid:
t = /(~. w, d, p, 0, 00 , fto}, (298.6)
where, as before, f isalinear function of ~. w, and d and a continuous function
of its remaining arguments, and where 00 and p,0 are material constants such
that
phys. dim ()0 = [8], phys. dim. fto = [M L -1 y-1]. (298.7)
The presence of the first of these constants makes it possible for the fluid's response
to deformation to vary with the temperature. Motivation for introducing the
secon:i constant may be found in simple experiments on the resistance of fluids
in viscometers of various kinds. Since it would be possible, still within the framework of pure mechanics, to lay down a relation like (298.6) but involving four
1 Sometimes encountered are "anisotropic fluids" defined by constitutive relations of
the type t~ = C~ d$ + D~, where 0 and D are general tensors of the orders indicated. Such
an equation does not generally satisfy the principle of material indifference. Given such a
relation in an inertial frame, for example, transformation of t and d by the appropriate laws
to a non-inertial frame yields a relation of the same form except that the components of 0
and D in the non-inertial frame depend upon the timet, violating the original postulate (298.2).
In order to obtain a properly invariant theory of anisotropic fluids, it is necessary to modify
(298.2) by introduction of some vector or vectors specifying preferred directions. Cf. the
oriented bodies studied in Sects. 60-64 and the theory recently proposed by ERICKSEN
[1960, 1]. 2 The simplest case of this law was proposed by NEWTON [1687, 1, Lib. II, Chap. IX].
For incompressible fluids, equivalent dynamical equations were obtained from a molecular
model by NAVIER [1821, 1] [1822, 2] [1825. 1] [1827, 6]. The continuum theory of CAUCHY
[1823, 1] [1828, 2, § 111, Eqs. (95). (96)] lacks the term -pt5:_. The general formula was
obtained by PorssoN [1831, 2, ~~ 60-63] from a molecular model. The continuum theory,
subject to an unjustified but easily removed specialization, is due to ST. VENANT [1843. 4]
and STOKES [1845, 4, §§ 3- 5]. 3 The reader is to recall that in Sect. 205 symmetry of t was proved equivalent to balance
of moment of momentum only under the assumptions that l = 0, that m= 0, and that linear
momentum was balanced.
Sect. 298. Linearly viscous fluids. 717
rather than two dimensionally independent material constants, our definition
(298.6) implies not only the kinematical restrictions we have already demonstrated
but also dimensional ones, which we now proceed to determine.
First, the same reasoning as given above suffices to reduce (298.6) to an
equation of the form1
t = f(d, p, (), ()o, f.lo) · (298.8)
The dimensional matrix of the 11 quantities appearing in any one component
of (298.8) is
L T M e
tkm -1 -2 0 (k and m fixed)
p -1 -2 0
d 0 -1 0 0 (6 rows)
flo -1 -1 0
() 0 0 0
Oo 0 0 0
The first 2 rows are alike: so are the next 6 and the last 2; the ninth may be
obtained by subtracting the third from the first; thus the rank is at most 3;
in fact, it is exactly 3· By the n-theorem, (298.8) is equivalent to a relation
among 11 - 3 = 8 dimensionless ratios formed from the quantities entering it.
A possible set of such ratios is tkm/P, f.lo dfp, fJj()0 • Hence (298.8) is equivalent
to a relation of the form 2
tjp = g (f.lo dfp, ()f()o), (298.9)
where the function g is a dimensionless function depending linearly upon f.lodfp
and continuously upon fJjfJ0 • Thus the requirement of dimensional invariance
implies that p may enter the dynamical equations only in a strikingly restricted
way. Re-examining the argument used to reduce (298.3), we see that the presence
of additional scalar arguments does not affect it. Hence we again obtain (298.5),
except that now the coefficients A. and f-l are shown to have restricted functional forms:
(298.10)
the functions f1 and /2 being dimensionless. While, as shown by (298.9), it has
been tacitly assumed that p =f= 0 in the region considered, the assumed continuity
of f as a function of p in (298.6) allows this restriction to be removed by inspection
of the final result.
The material coefficients A. and f.l are the viscosities of the fluid. The dimensional constant flo is a parameter which, in any system of units selected, may
be assigned a numerical value representing the amount of shearing stress or other
resistance to a given stretching that a particular physical fluid may offer. The
dimensional constant ()0 and the dimensionles functions / 1 and /2 are parameters
which enable representation of a viscous response varying with temperature.
From their definitionvia (298.6), the viscosities are independent 3 of all kinematical
1 In the following argument, we tacitly employ reetangular Cartesian co-ordinates throughout, so that all tensor components bear the physical dimensions shown in the table. The final
results are in tensorial form and hence valid in all co-ordinates. 2 TRUESDELL [1949, 33, §§ 3, 6] [1950, 32, §§ 4, 7] [1952, 2], §§ 63-65] [1952, 22, p. 90]. 3 To speak of a "frequency-dependent viscosity" or a "non-linear viscosity", as is not
uncommon in the Iiterature of ultrasonics and rheology, is a misleading way of saying that
the linear law (298.5) does not hold but some particular (and usually imperfectly specified)
non-linear law does.
718 C. TRUESDELL and R. TouPrN: The Classical Field Theories. Sect. 298.
variables; according to the results (298.1 0), the viscosities are also independent
of the pressurel, being functions of the temperature only.
When .?. = fJ, = 0, the dynamical equations for viscous fluids reduce to EuLER's
equation (297.1) for perfect fluids. Thus perfect fluids are often called inviscid.
That, irrespective of the values of .?. and f-t, the equations of the theory of
viscous fluids reduce to the same statical equation (298.1) when d = 0, is a standard example to show that the commonest statement of "D'ALEMBERT's principle"
is false 2 ; for this reason, the portion .?. Id 1 + 2ft d of the stress is regarded
as arising from internal friction.
Since
t~=2f-td~ when k=j=m, (298.11) . the quantity fJ, is the ratio of shear stress to the corresponding shearing (Sect. 82)
of any two orthogonal elements; hence it is called the shear viscosity.
The mean pressure (204.7) is given by .
p - p = - (.?. + i fh) d~ = (.?. + i fh) log e . ( 298.12)
Thus.?. + if-t, the bulk viscosity, is the ratio of the excess of the mean of the three
normal pressures over the static pressure to the rate of condensation. To render
this last interpretation definite, it is best to distinguish two cases. First, for an
incompressible fluid we have e =0, and from (298.12) follows p = p in all circumstance5. Thus, for an incompressible fluid, the pressure p occurring in the constitutive relation (298.5) is always the mean pressure, and .?. drops out of all
equations. Second, for a compressible fluid, in accord with the agreement that
(298.1) holds in equilibrium, we take p in (298.5) tobe the same pressure as would
hold in equilibrium under the same thermodynamic conditions, i.e., p = n =
n (e, 0) as given by the thermal equation of state. This pressure is then determined,
independently of the motion, as soon as e and (} are known. (298.12) relates this
pressure to the mean of the pressures actually exerted upon three perpendicular
planes at the point in question. (A pressure-measuring device, in general, measures
some component of the stress tensor; in a viscous fluid, it is not justifiable to
identify either p or p with the results of measurement, a theory of the flow near
the instrument being required in order to interpret the experimental values in
terms of the variables occurring in the theory.)
1 I.e., to obtain viscosities which depend also upon the pressure, it is necessary to start
with a definition more generat than (298.6): A constant or scalar bearing the dimension of
time or stress must be included, as is done in the following section. Cf. the discussion of
this point by TRUESDELL [1950, 30, §§ 4, 7, 11] [1952, 21, §§ 62-63]. 2 I.e., to obtain "the" equations of motion from the statical equations, for a given system,
add the "inertia force" -x to the assigned forcelper unit mass. lf (298.1) are the statical
equations, then this form of "D' ALEMBERT's principle" is to be supplemented by adding
to the "inertial force" any "frictional forces" that may be present. "Frictional forces"
are then defined as any forces arising from motion other than inertial force. Putting all these
definitionstagether Ieads to the conclusion that D' ALEMBERT's principle asserts that equations
of motion follow from statical equations by supplying inertial force plus such other forces
as may arise in conjunction with motion. Thus D' ALEMBERT's principle appears to have no
content at all.
To the reader who finds this confusing we remark that it was not by oversight that from
the !ist of guiding principles in Sect. 293 we omitted "D' ALEMBERT's principle ", for we consider it to be either trivial or false in the usual statements in this context. (This does not
affect the validity of the different "D'ALEMBERT-LAGRANGE principle" given in Sect. 232.)
Acorrect statement, revealing the limited but useful validity ofthisform of the principle,
is as follows: Given the statical equations for a material, constitutive equations for a dynamically
possible material resuZt if I is replaced by I- x. This special kind of material, being only one
of the infinitely many that share the same statical properties, is called "perfect ". This definition
applies to several classical theories.
Sect. 298. Linearly viscous fluids.
The stress power (217.4) assumes the form
P =PE= tkmdkm =-Pd~+ J.(d~} 2 + 2fl d!, d'/:,
while by the above agreement regarding p we have from (256.4) 1
P1 =- pd~.
Hence Eq. (256.6) for production of total entropy becomes
e o ~ = cJ> + h~p + e q.
where
719
(298.13)
(298.14)
(298.15)
(298.16)
[It is plain that t (/> is a dissipation function in the sense of (241 A.2).] If we adopt
the corollary (258.1)1 of the entropy inequality, we conclude that (/> as given
by (298.16) must be a positive semi-definite quadratic form. An easy analysis
of (298.16) shows that this is the case if and only if1
ft-;;;,o, 3J.+2fl-;;;,o. (298.17)
These results have immediate mechanical interpretations. By (298.11), (298.17)1
asserts that the shear stress always opposes the shearing. By (298.12), (298.17) 2
asserts that in order to produce condensation ( expansion), a mean Pressure not
less (not greater) than that required to maintain equilibrium at the same density
and temperature must be applied. These interpretations, showing that the effect
of the viscous stresst +P 1 as given by (298.5) is always toresist change of shape
or bulk, reinforce the view that the viscous stress is of the nature of frictional
resistance.
The quantity C/> in (298.15) is the viscous dissipation of energy per unit
volume. As (298.15) shows, this energy goes into the increase of entropy, or is
carried off by the flux - h, or is drawn away by sinks - q. It is customary
torender the theory moredefinite by setting q =0 and adopting FoURIER's law
(296.4). The resulting equations furnish an example, somewhat more typical
than the perfect non-conducting gas (Sect. 297), of a fully thermomechanical
theory, in which the basic principles of mechanics, energetics, and thermodynamics are employed.
The presence of viscosity has the effect of rendering propagation of most
kinds of waves impossible. We consider here only the simplest case, that of an
acceleration wave, across which p, x, and e are continuous. By substituting
(298.5) into the dynamical conditions (205.5) we obtain
Ank[x;q] +!lnm[im,k] +!lnm[.ik,m] =0, (298.18}
where it is assumed that },, fl· and ef are continuous. Writing sk=- U s~ in
(190.5)1 and substituting the result into (298.18) yields
(J. + fl) nms,. nk + flSk = 0.
Taking the scalar product of this equation by n, we have
(A + 2fl) nmsm = 0.
(298.19)
(298.20)
It is a consequence of (298.17) that in order for }. + 2fl = 0 to hold, it is necessary and sufficient that 2 = 0 and fl = 0. Hence by (298.20) we conclude that
in a viscous fluid nm sm = 0. Putting this result back into (298.19) yields fl sk = 0,
and hence sk = 0. What has been proved 2 is that the instantaneous existence of
1 DuHEM [1901, 7, Part I, Chap. 1, § 3], STOKES [Note, pp. 136-137 of the 1901 reprint
of [1851, 2]]. 2 KorcHINE [1926, 3, § 3]. but the result is really included in an earlier one of DUHEM
[1901, 7, Part II, Chap. III], who uses a different terminology.
720 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 299.
a surface upon which x and p are continuous but ik,m suffers a jump discontinuity
is incompatible with the law oflinear viscosity (298.5).
We recall that kinematical analysis alone (Sect. 190) shows that a discontinuity in ik m in order to persist must be propagated as an acceleration wave.
In Sect. 29l we showed that acceleration waves are propagated in a barotropic
flow of a perfect fluid at the speed c, determined by the barotropic relation p = f (e).
Any thermal conductivity, however small, generally renders the flow baroclinic
and results in the Newtonian speed of propagation (297.16) instead of the Laplacian speed (297.13) that holds when there is a non-degenerate equation of
state p =P (e, 'YJ) and when there is neither flux nor supply of energy. Now we
have seen that in fluid with non-vanishing viscosity, however small, such a surface
cannot exist at even at a single instant. This last result is included in a general
theorem of DuHEM1 : In a linearly viscous fluid, no waves of order greater than 1
are possible.
Since viscous and thermally conducting fluids are regarded as a refined model,
superior to the perfect fluid, for the same physical materials, a device must be
found whereby wave propagation in some sort may occur. Moreover, perfect
fluids are formally a special case of viscous fluids; thus, properties of perfect fluids
must be reflected in corresponding properties of viscous fluids. The appropriate
device is the quasi-wave of DuHEM, a thin layer in which some of the variables
suffer rapid but nevertheless continuous changes. The solutions containing waves
that occur in perfect fluids are to be regarded as limit cases of an appropriate
solution of "the same" problern for viscous fluids, in the limit as ;., f-l, and "
approach 0, and hence the layer becomes arbitrarily thin. The rate at which
x-+0, relative to those at which A-+0 and p-+0, influences the results. A full
analysis of the plane quasi-wave has been given by GILBARG (1951) 2•
A similar difficulty arises in connection with the boundary conditions. For
viscous fluids it is customary to impose the condition (69.3) representing adherence
of the fluid to the boundary. For perfect fluids, solutions satisfying this condition
generally fail to exist, and only the weaker condition (69.1) is employed. In many
cases the solutions afforded by the two theories are sensibly the same throughout
most of the region occupied by the fluid but differ only in a thin boundary layer
near solid objects. The existence ofthissmall region of difference, while perhaps
effecting little alteration of the gross appearance of the flow, can yield results
which are dynamically of a different kind; e.g., the force exerted by the fluid
on an obstacle will generally be very different, as is plain from the results given
in Sect. 202.
The purpose of the foregoing remarks is to point up a major instance of the
ideas sketched in the second paragraph of Sect. 5.
299. Non-linearly viscous fluids. A simple and now familiar theory of nonlinear viscosity is obtained by taking (298.2), but without the restriction that f
be linear, as the definition of a fluid. The analysis at the beginning of Sect. 298
made no use of the linearity of f; thus it follows, in full generality, that (298.2)
must reduce to a form giving t as an isotropic function of d. By the representation
theorem for such functions we thus ha ve
(299.1)
1 [1901, 6] [1901, 7, Part II, Chap. 111]. DuHEM asserted also that shock waves arenot
possible in a viscous fluid, but this result holds only subject to some qualification. Cf. Sect. 54
of the article by SERRIN, Mathematical Principles of Classical Fluid Mechanics, this Encyclopedia, Vol. VIII, Part 1.
2 For an exposition, see § 57 of SERRIN's article, just cited.
Sect. 299. Non-linearly viscous fluids. 721
where
N0 (0,0,0) =0. (299.2)
This reduction was first given by REINER1. In an incompressible fluid, there is
no loss in generality in taking N0 = 0. The scalar coefficient functions generalize
the classical viscosities occuring in the linear law (298.5). If we set 2p, = N1 (0, 0, 0)
and ). = 8N0j(:Hd when d =0, then (298.5) results from (299.1) by linearization.
The coefficient N2 , the cross-viscosity, gives rise to effects of a type not present
in the linear theory.
For the dimensional analysis, we may proceed just as in Sect. 298, but instead
we prefer to replace (298.6) by the more general defining equation
t = f(x, w, d, p, fJ, fJ0 , p,0 , to)
where t0 , the natural time, is a material constant such that
(299.3)
phys. dim. t0 = [T] . (299.4)
Use of the principle of material indifference and the n:-theorem reduces (299.3)
to the form2
(299.5)
where, as shown already, the dimensionless function g is an isotropic function of
its first argument. Thus in (299.1) the coefficients have the more explicit forms
Nr = f!:'!_ • t[- 1 !:lr, (F unsummed) } lo
!:lr = fr(to Id, t~ Ild, tg IIId, to Plflo, fJ!fJo),
(299.6)
where the functions f r are dimensionless functions of their five dimensionless
scalar arguments.
When the definition of a fluid is narrowed, as it was in Sect. 298, so as to
exclude the time constant t0 , the forms (299.6) must be replaced by others, as
follows 3 :
Nr=P ( p, ; )r-1 !Ir, l
( f.lo ,u~ p,~ ) !Ir= Ir p Id, -:p2 Ild, pa IIId, ()J()o ·
(299.7)
The theory based upon (299.7) is easily seentobe a special case ofthat based
upon (299.6). The specialization, however, is one of consequence. The theory
devoid of a time constant is nearer to the classicallinear theory in that it possesses
no more dimensionally independent material constants. A possible parameter
governing dynamical similarity is the truncation number ], given by 4
(299.8)
1 [1945, 5, § 4]; cf. the treatment of the incompressible case by RrVLIN [1947, 13] [1948,
24]. A simple rigorous proof is given in Sect. 59 of the article by SERRIN, just cited. 2 TRUESDELL [1950, 32, § 11] [1952, 21, §§ 68-69] [1952, 22, PP· 89-90]. 3 TRUESDELL [1949, 33, § 4] [1950, 32, § 5], "Stokesian fluid". 4 TRUESDELL [1949, 33, § 7] [1950, 32, § 8]. lt was stated erroneously by TRUESDELL
that ] must be controlled independently of the classical scaling parameters. In fact, in two
geometrically similar flows having the same Euler and Reynolds numbers the two truncation
numbers (299.8) are necessarily also the same. The only further scaling parameters required
for Stokesian non-linear viscosity are such dimensionless material constants as are used to
specify the functions fr in (299.7). No such remark applies to (299.9).
Handbuch der Physik, Bd. III/1. 46
722 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 300.
A condition for approximating the constitutive relation (299.1), supposed differentiable with respect to d, by the linear relation (298.5), is 3J< 1; that is, the intensity of stretching (Sect. 83) shall be small in respect to the ratio of pressure
to linear viscosity. While the parameter (299.8) may be introduced in the more
general theory defined by (299.6), it does not suffice; for dynamical similarity
it is necessary to control also a parametersuch as
(299-9)
The criterion for truncation is now that the product of the intensity of stretching
by the natural time of the fluid shall be small. The difference between the two
theories is reflected even in the linear case, for in the theory based upon (299.7)
the linear viscosities Ä. and ll• as we have seen in Sect. 298, are independent of the
pressure, while in the theory based upon (299.6) this need not be so, for they may
depend upon to Pl!lo. Further, it can be shown that the stress of the "Stokesian"
theory is of a type that occurs also in results following from the kinetic theory
of monatomic gases, while the more general theory based upon (299.6) is generally
in contradiction with the kinetic theory, in which there is no time constantl.
A summary of existing knowledge of the theories of non-linear viscosity is
given in "The Non-linear Field Theories of Mechanics ", Vol. VIII, Part 2 of this
Encyclopedia.
300. Perfectly plastic bodies. The theory of perfectly plastic bodies is intended
to describe an elastic rather than viscous flow in response to stretching and hence,
while adopting the constitutive relation (298.3), relinquishes the requirement of
consistency with (298.1). Thus we have at once in the linear case 2
t!.=J.d~<5~+2!ld~. (300.1)
but Ä. and !l are not material constants, being rather functions of d to be determined by additional conditions.
Writing (300.1) in terms of the deviators oft and d (cf. App. 38.12), we have
\
-3P=IXIc~. 1X=3.1.+21l·}
0t = 2/.l 0d. (300.2)
The additional conditions imposed as an essential part of the theory are (1) the
mean pressure is a function of the dilatation only, and (2) some scalar invariant
of 0t vanishes. These conditions are said to represent plastic flow or yield. The
first amounts to replacing (300.2)1 by a general relation f (p, Ic~) = 0 and is a law
of compressibility. Often it is replaced by the condition of incompressibility,
Id = 0; in this case p becomes an additional unknown function. The second,
characteristic of the theory, may be put in the form
y ( II,!_ III,!) = O az ' a3 ' (300.3)
where a is an elastic modulus called the yield stress of the material and where
the yield function Y is dimensionless. This condition represents a material which
responds, or at least responds in the manner indicated by (300.2), only when
stresses of a certain kind reach an appropriately !arge value, while the state of
plastic flow is assumed such as to maintain this value unaltered.
1 The "relaxation time" in a Maxwellian gas is notamaterial constant such as t0 , being
in fact p.fp, a function of temperature and pressure. 2 CAUCHY [1823, 1] [1828, 2, § III, Eqs. (95), (96)] gave these formulae as appropriate
for a "soft" body but stated that Ä and p. are constants.
Sect. 301. Linearly elastic bodies. 723
A more general theory for incompressible substances1 is obtained by replacing
(300.2) by a relation of the form
2u dk = Y~ (300.4) r m ot'k''
where the dimensionless plastic potential P(tfa) is subject to the condition
~k aP _ ) Um ot'k' -0, (300.5
so that (300.4) is consistent with the condition of incompressibility, Id = 0. For
isotropic materials, P is taken as a function of II,1/a2 and III,1/a3 only.
While (300.4) might seem to determine d uniquely when t is known, this is
not so. Indeed, t-tdfa is so determined; conversely, if Pis a sufficiently smooth
function, tja is determined as a function of f-t dfa. Substituting this function into
the yield condition (300.3), in the isotropic case we obtain a functional relation
of the form
t(~: nd, ~: IIId) = o. (300.6)
Assuming this can be solved for t-tfa, we see that
= (J t ( yliid ) .
t-t Vnd Vnd (300.7)
Thus the factor f-t occuring in (300.4) is determined, not assignable.
Comparing a formula of the type (300.4), f-t being eliminated by means of
(300.7), with the relation (299.1) for a non-linearly viscous fluid, we see that
a still more general theory of perfectly soft bodies, including both viscous fluids
and perfectly plastic bodJ.es, results by allowing the coefficients N0 , N1 , and N2
to be discontinuous at d = 0 and by leaving the physical dimensions of the moduli
unrestricted. The principal difference between the theories of viscosity and plasticity arises from the different physical dimensions of the material constants.
While A. and f-t are viscosities, having the dimension [M L -1 r-1], a is a stress
or elasticity, so that phys. dim a = [M L -1 r-2].
The theory of perfectly plastic bodies, as defined by (300.2h and (300.3),
is due to ST. VENANT, LEVY, and v. MISES 2• The most commonly employed
yield condition is that of MAXWELL and v. MISES 3, viz., II,t =K a2, where K
is a constant (cf. the alternative forms of II,t given in Sect. App. 38); this amounts
to taking f = const in (300.7). In this theory P = Y.
As appears at once by confronting (300.1) with CAUCHY's first law (205.2),
the theory of perfectly plastic bodies is a dynamical theory. Nevertheless, almost
all of its large Iiterature treats it as if it were statical. Thus, as far as the exact
theory is concerned, very little is known regarding it. A survey of the mathematical developments as practised by current specialists in the field is given in
the article by FREUDENTHAL and GEIRINGER in Vol. VI of this Encyclopedia.
301. Linearly elastic bodies. The simplest kind of elastic or springy body
is one such that the stress arises solely in response to such change of shape as the
body has undergone from its "natural" or unstressed state. Considering the strain
tobe very small, by the results in Sect. 57 we may take the tensor e as a measure
of it, and we assume the stress depends linearly upon it. Moreover, t = o if e = o.
1 GEIRINGER [1953, 11, § 3]. 2 [1870, 7]; [1870, J]; [1913, 6]. 3 [1937. 5, pp. 32-33] (written in 1856), [1913. 6, §§ 2-3].
46*
724 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 301.
There results the constitutive equation
t"'k = CkmPq epq• CkmPq = Cmkpq = Ckmqp, (301.1)
where for simpler writing we have dropped the tilde from e. A linearly elastic
body is isotropic if the relation (301.1) is an isotropic one. For an isotropic body,
(301.1) reduces to Cauchy's lawl:
(301.2)
To the extent that e is an approximate measure of the mutual distances of
particles, the principle of material indifference is satisfied by (301.1) and (301.2).
If (57.1) is interpreted strictly, however, the principle of material indifference is
violated, although (301.1) and (301.2) areinvariant under infinitesimal rigid timeindependent displacements. Linear elasticity theory does not represent exactly the
kind of behavior possible in any real material. Rather, it is to be regarded as a
mathematical approximation to the properly invariant theories described in the
two following sections.
The elasticities Ä., f.l, and CkmPq in (301.1) and (301.2) arematerial constants
or functions of the temperature or entropy; their physical dimensions are those
of stress, [M L -l r-2], and they bear no physical connection with the mathematically analogous viscosities appearing in (298.5). The static theory isalinear one:
Uniformly doubled displacements always result from uniformly doubled loads,
and, more generally, from displacements u', u" corresponding to stresses t', t",
assigned forces j', f", and assigned surface loads t{n)• t(~) we may construct a
displacement u=u'-u" answering to the stress t=t'-t", forcef=f'-f",
and surface load t(nl =t'. Consider two solutions u' and u" corresponding to the same assigned loads and boundary values. Form a solution
u = u'- u" as indicated above; for this solution we have f = 0 in z>, t(n) = 0
on j 1 , and u =0 on j 2 • Substitution of (301.6) 2 into CAUCHY's first law (205.2)
yields
(;~L=o. (301.8)
Hence
0 = J um (8 -:) dv = J [(um~-:) -um k _o-:] dv, öek ,k öek ,k • öek U V
(301.9)
V
=-2JI:dv,
V
where we have used EuLER's theorem on homogeneaus functions as well as the
fact that on <1 either u or t(n) vanishes. Now if J:(e) is of one sign for all values
of e, it follows from (301.9) that L'=O in z>. Looking back at (301.4), we see that
if 1: is a definite quadratic form ( whether positive or negative), the generat boundaryvalue Problem of static linear hyperelasticity cannot have two distinct solutions.
What has been proved is that e = 0; the strains is thus unique, and from the
results in Sect. 57 it follows that the displacement u is determined uniquely to
within an infinitesimal rigid displacement. This degree of indeterminacy is inherent in the linear theory of elasticity and is to be understood in all statements concerning it, except in cases where this indeterminacy is removed by
specification of the displacement on the boundary.
There is physical reason to require that }; be a positive definite form, for then
in any given small strain from an unstressed state, the stress must do positive
work. This idea seems tobe related to, but is not identical with, the requirement
(258.1) 1 following from the entropy inequality, which is expressed rigorously
in terms of time rates rather than displacements. In the isotropic case, 1: is
positive definite if and only if
f-l > 0, 3 Ä + 2J.l > 0. (301.10)
There is a remarkable principle enabling us, in the case of equilibrium subject
to given surface displacements and vanishing assigned force in the interior, to
select among all kinematically possible deformations that one which is consistent
with the theory of hyperelasticity, a positive definite stored energy function
1 KIRCHHOFF [1859, 2, § 1] [1876, 2, Vor!. 27, § 2].
726 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 301.
being assigned 1 : The displacement that satisfies the equations of equilibrium
as well as the conditions at the bounding surface yields a smaller value for the total
stored energy than does any other displacement satisfying the same conditions at the
bounding surface.
To prove this theorem, we writl~ e for the strain leading to a stressthat satisfies the conditions of equilibrium, and e + e' for some other strain, where it is
assumed that u' =0 on the boundary 11. Then
J .E(e + e') dv = J .E(e) dv + J (e'ZO a~~ej__ + .E(e')) dv, (301.11)
since .E(e) isahomogeneaus quadratic form in the components e!,. Now
J •m aE(e) d J , tkmd
e k aezo V= U(k,m) V,
" "
(301.12)
= ~ u_; tf .. ) da = 0, 4
where we have used (301.6) and (205.2) and the fact that u' =0 on 11. Substitution
of (301.12) into (301.11) yields
J .E(e + e') dv = J .E(e) dv + J .E(e') dv. (301.13)
This identity shows that the energy stored by a deformation corresponding
to the vector sum of two displacements, one of which leads to an equilibrated
elastic stress and both of which have the same values on the boundary, is the sum
of the energies of the constituent displacements. Since the stored energy is
assumed to be a positive definite form, the above-stated theorem of minimum
energy follows immediately.
The two theorems just proved are representative of the many that are known 2
in this classical subject, the theory of which has been brought to a state of analytical completeness second only to that of the theory of the potential 3•
Having given some consideration to static theorems, we turn now to the pro·
pagation of waves.
For a body of continuous constant elasticity C, putting (301.1h into (205.2)
yields
(301.14)
where we have supposed ef tobe continuous. In linear· elasticity theory we have
[up,q,J =g~ g~ [xp,apl· By applying the general identities (190.1) and (190.2)
for an acceleration wave, when the present configuration is taken as the initia1
one, from (301.14) we thus obtain
or
(301.15)
(301.16)
1 KELVIN [1863, 2, § 62] took the assertion as "the elementary condition of stable equilibrium"; in this sense, that of a postulated variational principle, its history may be traced
back to an idea of DANIEL BERNOULLI in respect to elastic bands (1738). As a proved theorem
of linear three-dimensional elasticity, it seems first to have been given by LovE [1906, 5,
§ 119].
2 A masterly exposition of some of them is given by LovE [1927, 6, Chap. VII]. 3 Despite this fact, there exists no general exposition of the theory from a rigorous mathematical standpoint.
Sect. 302. The rotationally elastic aether. 727
Thus in order for an acceleration wave with normal n to exist and propagate,
the jump s which it carries must be a proper vector of Ck mpqnqnm corresponding
to the proper number e U2• For a body such that the work of the stress in any
deformation is positive, as is the case for a hyperelastic body with positive definite stored energy, the tensor C k mpq nq nm is positive definite, its quadric being called
Fresnel's ellipsoid for the direction n; therefore allproper numbers e U2 are positive, and therefore all possible speeds U are real. In the general case, then,
in any linearly elastic body such that the work of the stress is positive for arbitrary
deformations, a wave with given normal n may carry a discontinuity of the acceleration parallel to any one of three uniquely determined, mutually orthogonal directions,
and corresponding to each of these directions there is a speed of Propagation determined uniquely by the elasticities of the material and by n.
When the proper numbers e U2 are not distinct, the above conclusion must
be modified, as is seen most easily by considering the isotropic case, for then
(301.14) assumes the more special form
[exk] = (J. +,u)[u~pk] +,u[ui.~p]. (301.17)
so that for an acceleration wave we have
{301.18)
specializing (301.15). Taking the scalar and vector products of this equation
by n yields
{e U2 - (J. + 2,u)} s · n = o, }
{e U2 - ,u} s x n = o. (301.19)
If s . n =f= 0, the first equation yields e U2 = J. + 2,u, and the second, if we exclude
the case when J. + ,u = 0, yields s X n = 0. If s · n = 0 but s X n =f= 0, the second
equation yields e U2 =,u. Summarizing these results, we see that in an isotropic
linearly elastic body for which A + ,u =f= 0, a necessary and sufficient condition that
acceleration waves be propagated at positive speeds is A +2,u >0, ,u >O. This
condition is satisfied when the stored energy is positive definite. Two kinds of
acceleration waves are possible: longitudinal waves, whose speed of propagation is
given by
and transverse waves, for which
U2 = Jl-..
e
(301.20)
(301.21)
In view of the kinematical interpretation furnished by HADAMARD's theorem in
Sect.190, the longitudinal waves are called expansion waves or irrotational waves,
while the transverse waves are called equivoluminal waves or shear waves. The
foregoing results, which are due to CHRISTOFFEL and HUGONIOT1, illustrate the
far-reaching effect of isotropy: instead of three speeds of propagation, for an
isotropic body there are only two, but instead of there being only three possible
directions for the discontinuity, there are infinitely many, though the possible
directions are still far from arbitrary.
302. The rotationally elastic aether. The quest for a mechanical theory of light
as a vibration of an elastic medium attracted the attention of many of the
1 CHRISTOFFEL [1877, 3] obtained really all of the above results and more, but he did
not present them very clearly, nor did he recognize as such the isotropic case, for which
HuGONIOT [1886, 3] gave a very simple treatment. Our proof is essentially that of HADAMARD [ 1903, 11, ~~ 260, 267- 268] and DuHEM [ 1904, 1, Part IV, Chap. I, § V].
728 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 302.
illustrious physicists of the nineteenth century1• The most successful of these
mechanical theories as applied to the deduction of the laws of transmission,
reflection, and refraction of light in transparent media was devised by MAcCuLLAGH in 1839 2• Some thirty years after the publication of MAcCULLAGH's theory,
FITZGERALD 3 applied the new electromagnetic theory of MAXWELL to derive
the laws of reflection and refraction of light at the interface between two dissimilar
dielectric media. In this memoir, FITZGERALD called attention to the formal
analogy between the equations of the new theory and the old equations of
MAcCuLLAGH's theory. This seems remarkable, for the physical ideas underlying
the two theories are totally different. The constitutive equations of MAcCuLLAGH'S "rotationally elastic aether" do not conform to the principle of material
indifference known to be satisfied by ordinary elastic materials. In all other
regards, however, the theory is based on the usual equations of mechanics,
specialized somewhat by linearization.
a.) MacCullagh's equations and boundary conditions. As a starting point we
consider the equations of motion (205.2) and boundary conditions (205.5) at a
stationary surface of discontinuity which is not a shock:
(302.1)
Now contrary to the theory of ordinary elastic materials, where the stress arises
solely in response to changes in shape, MAcCULLAGH supposed the aether to be
a medium in which the stress, while completely insensitive to changes in shape, arises
only in response to rotations about its relaxed state. This implies the existence
of a preferred dass of reference frames. We may regard these preferred frames
as the inertial frames of classical mechanics. The aether in its relaxed state will
be at rest or in uniform translatory motion with respect to one of these frames.
Let uk(~. t) denote the displacement vector of the medium taken in this sense.
Then the quantities
w .. = U[, s] = -l (u, s- U5 ,) ' ' ' (302.2)
measure an infinitesimal rotation of the medium (cf. Sect. 57). Thus MAcCULLAGH
assumed that for small rotations of the aether medium the stress is given by
(302.3)
Although MAcCULLAGH considered the more difficult case of crystalline media,
we shall here treat only the isotropic case, where A••Pq is an isotropic tensor.
It follows that, in this case, the relations (302.3) reducc to
t, 5 = Kw, 5 • (302.4)
The constant K is the gyrostatic rigidity. The aether is assumed to pervade all
ordinary material media and to have the same density ein all materials. However,
the gyrostatic rigidity of the aether is assumed to have a different value in materials
with different indices of refraction. Thus at the interface between two dissimilar
isotropic media ofthissimple type we have [e] =0, [K] =1=0. Moreover, at such
an interface, the displacement u is assumed continuous, [u] =0. Thus it follows
from the results of Sect. 175 that if u is differentiable in each of the adjoining
media with differentiable limit values for aujat and grad u on each side of the
1 WHITTAKER [1951, 39] has given us a detailed and fascinating account of the evolution
of these theories and their interrelations. 2 [1848, 1]. 3 [1880, 8].
Sect. 302. The rotationally elastic aether. 729
interface, then
[ ~7] = 0, [curl u]. n = 0. (302.5)
Substituting the constitutive relation (302.4) into (302.1), linearizing the acceleration with respect to u and its derivatives, and collecting our assumptions thus
far, we have
K curlcurl u + e a;;- = o, I
[K curl u] xn = o, (302.6)
[e ~] =0, [curlu] ·n =0.
These are the basic equations of MAcCULLAGH's theory of light.
ß) Fitzgerald's analogy. MAXWELL's constitutive equations for a non-magnetic,
linear, rigid, homogeneous, isotropic stationary dielectric were discussed briefly
in Sect. 283 and are treated in detail in Sect. 308. For such media we have
~=eE, l "= ~B (302.7)
'!I' /1-o '
where the dielectric constant e is a measure of the ease of electric polarization
of the medium. If we substitute these relations into the general electromagnetic
equations (278.2), (278.3), (278.8), (278.9), and (283.42) to (283.45), we obtain
the system aB
curlE + -ät = 0, div B = 0, ) (
302_8)
[E]Xn=O, [B] =0, [eE]·n=O,
1 aE - - curl B - e- = 0 div E = 0.
/1-o ot ' (302.9)
These equations, generalized to the case of crystalline media, formed the basis
of FrTZGERALD's derivation of the laws of reflection and refraction of light from
the Maxwell theory. We see that, leaving aside the question of physical interpretation of the symbols, if we make the substitutions
Kcurlu~oc.E, e ~7 ~oc.B, ) (302.10)
K ~ßfe, (! ~ßfto•
in (302.6), we obtain (302.8) and (302.9). The quantities cx and ß in (302.10) are
arbitrary non-vanishing constants. The last two Maxwell equations (302.9) are
a consequence of the identities div curl u = 0, curl ~ - -fft curl u = o. This
mathematical equivalence between the electromagnetic equations in ideal media
characterized by the constitutive equations (302.7) and MAcCuLLAGH's equations
was first perceived by FrTZGERALD1•
As remarked by WHITTAKER2, " ••• there can be no doubt that MAcCuLLAGH
really solved the problern of devising a medium whose vibrations, calculated
1 [1880, 8]. The analogy between MAcCuLLAGH's equations and MAXWELL's equations
based on the replacements (302.10) is discussed briefly by SoMMERFELD in his lectures [1947.
14]. He also considers a second set of replacements which renders the equations equivalent.
HEAVISIDE [1893. 4, Chap. 111] gives an elaborate account of electro-mechanical analogies
based on the Maxwell equations of dielectrics, conductors, etc., and the mechanical equations
of elastic solids, viscous fluids, etc. 2 [1951. 39, p. 144].
730 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 303.
in accordance with the correct laws of dynamics, should have the same properties
as the vibrations of light".
303. Finitely elastic bodies. The classical theory of finite elastic strain generalizes the concepts of linear elasticity by considering a body in which the
stress arises solely in response to the difference of the present shape from that
in an unstressed "natural state". An apparently still more general notion is
included in the constitutive equation
tkm = fkm(X'L,a.), {303.1)
where X'l,a.=oxqjoXa. and where ~=~(X) is the deformation from the natural
state to the present configuration. The prominence of the natural state is extreme:
neither any intermediate state nor any event in the stress and strain history
need be considered in ascertaining the stress. We may say that the body exhibits
a perfect memory of its natural state and is entirely oblivious to every other
except the present one.
The principle of material indifference (Sect. 293 0) makes it possible to reduce
(303.1) to the form1
T=f(C), (303.2)
where C is the deformation tensor defined by (26.2) and where T is ProLA's
stress tensor, connected with t through {210.9).
Although both C and T transform as double tensors under time-independent
changes of material and spatial co-ordinates, they do not transform as tensors
under change of frame. Thus the principle of material indifference does not
force the relation {303.2) tobe an isotropic one; indeed, any relation of the form
(303.2) satisfies that principle. Herein lies the explanation of why the classical
theory of elasticity includes anisotropic as weil as isotropic behavior, in contrast
to the theories of fluids described in Sects. 298 and 299.
An elastic body is isotropic if (303.2) reduces to an isotropic relation. The
principles given in Sect. 32 suffice to show that in this case, alternatively, t
may be regarded as an isotropic function of c, allowing a statement of the law
of elasticity in terms of spatial tensors only2 :
t = N0 1 + N1 c + N2 c2
(303-3)
where the coefficients Nr and !:l.r are scalar functions of c, and hence may be
taken as functions of Ic, Ilc, IIIc or as functions of Ic, Ic-1 and IIIc or IIIc-1, etc.
Thus far we have followed CAUCHY's concept of elasticity. GREEN's concept,
which may be stated exactly as in Sect. 301, Ieads to a more restricted theory,
as we shall see now. The basic assumption of finite hyperelasticity is that there
exists a stored energy E, a scalar function of the material co-ordinates X and
of the deformation gradients x". a., such that all the work of the stress is recoverable. We have then, by hypothesis,
JLi=t"md !!o km• (303.4)
where the inessential factor eleo is added for conformity with standard usage.
The principle of material indifference requires that in fact E = E (X, C). A classical argument, which has been given in three different forms in Sects. 218, 232A,
1 NoLL [1955. 18, § 15a]. A weaker theorem had been proved under stronger hypotheses
by earlier writers.
2 REINER [1948, 23, §§ 1-2].
Sect. 304. Hypo-elasticity.
and 256A, enables us to conclude thatl
T,cx- ol:
"'- ax"' • ,cx
etc. These stress-strain relations furnish a special case of (303.2).
731
(303.5)
Within hyperelasticity, a body is isotropic if its stored energy is an isotropic
function of C. From the fundamentallemma in Sect. 30, it follows that, alternatively, 1: =l:(c) =l:(Ic-'• Ilc-'• Illc-,)· It is easy to show by using (App. 38.16)
that for isotropic bodies any one of Eqs. (303.5) reduces to the form given by
FINGER2 :
t!, = ~: [ (nc-, a:~_, + Illc-' 81~~-~) <5!, + )
a.1: -1,. al: ,. ] + 8Ic-' Cm- Illc-, olle-' cm •
(303.6)
Equivalently, the principal axes of stress coincide with the principal axes of
strain at z, and the principal stresses ta are related to the principal stretches Äa
as follows 3 :
a =1,2,3. (303.7)
The definition (303.1) is incomplete in that the dimensional moduli are not
specified; it should be amplified to read
where
phys. dim. p. = [M L -1 r-z].
(303.8)
(303.9)
Thus, as in the theory of perfectly plastic solids, there is but a single independent
dimensional material constant, and it has the dimensions of stress. By the
n-theorem we see at once that (303.8) must reduce to t"m/p. =h"m(xq,cx), and
hence (303.2) may be replaced by
T /i =g(C), (303.10)
where g is dimensionless. Corresponding modifications are easily made in all
equations of the theory.
After a quiescence of half a century, the theory of finite elastic strain has
undergone remarkable development in the past decade. See "The Non-linear
Field Theories of Mechanics" in Vol. VIII, Part 2 of this Encyclopedia.
304. Hypo-elasticity. A different generalization of the classicallinear theory
of elasticity is obtained by regarding it as describing approximately a material
not necessarily having any natural state but rather experiencing a stress increment
arising in response to the rate of strain from the immediately preceding state. Thus
no finite memory is ascribed to the material. From the results at the beginning
of Sect. 95 it is plain that the appropriate tensor measuring the rate of strain is
the stretching, d. From the principle of material indifference in Sect. 293 (} we
see that one of the various time fluxes introduced in Sects. 149 to 151 may be used
1 The argument is due in principle to GREEN [1839, 1, p. 249] [1841, 2, pp. 295-296]
whose analysiswas corrected by KELVIN [1863, 2, §§ 51- 57]. All essential arguments occur
in the treatment of a special case by KIRCHHOFF [1852, 1, pp. 770-772]. The many known
equivalent forms are summarized by TRUESDELL [1952, 21, §§ 39-40]. 2 [1894, 4, Eq. (35)]. 3 KöTTER [1910, 6, § 1)], ALMANSI [1911, 1, § 7].
732 C. TRUESDELL and R. TOUPIN: The Classical Field Theories. Sect. 304.
to measure the rate of stress in a properly invariant way, and that the constitutive equations must be sufficiently general as to render immaterial the choice
of a particular flux. We select the co-rotational flux drtfdt as defined by (148.7);
allowing the magnitude of the present stress t to moderate the response of the
material to stretching, we write the constitutive equation in the form
dtt = j(t, d, p,), (304.1)
where f-l = const and
phys. dim. f-l = [M L -1 7-2]. (304.2)
This last requirement asserts that all material moduli shall be elasticities, just
as in the theories of linear or finite elasticity or of perfectly plastic materials.
Fluid behavior, to the extent that it is accompanied by effects of viscosity or
relaxation, is thus explicitly and intentionally excluded. (Effects of temperature
differences are easily taken into account if desired but are here omitted for
simplicity.)
Application of dimensional analysis shows that t may enter (304.1) only in
the ratio lff-l and that f, assumed continuous at d =0, must be linear in d:
drskm =KkmPqd skm:==tkmj2u, dt Pq• r (304.3)
where K is a function of s. Since drsfdt and d are both tensors under change of
frame (Sects. 144 and 148), K must alsobesuch a tensor; consequently K is an
isotropic tensor function of s. By a representation theorem due to RIVLIN and
ERICKSEN1 it follows that the most general constitutive equation satisfying the
hypotheses (304.1) and (304.2) is
where
dt = N0 I.t 1 + N 1 d + N2 I.t s + N3 M 1 + l
t + t N4 (d s + s d) + N5 I.t s 2 + N6 M s +
+ N7N1+!N8 (ds2 +s2 d) +N9 Ms2 +
+ N10 Ns + N11 Ns 2,
M - s~ dk', N s~ srp dt
Nr = Nr (18 , II8 , III8 ) F= 1, 2, ... 11.
(304.4)
(304.5)
(304.6)
The theory based upon these constitutive equations is called hypo-elasticity 2•
In keeping with the expressed aim of elastic rather than viscous response,
Eqs. (304.4) areinvariant under a change of the time scale. While the stress at a
given time generally depends upon the manner in which the load has been applied
during previous instants, it is thus independent of the actual speed of deformation.
The exact theory is a fully dynamical one. However, Eqs. (304.4) reduce to those
of the linear theory of elasticity (Sect. 301) under the assumptions usually made
in formulating that theory. Moreover, every isotropic elastic body is also hypoelastic 3. This result, however, must not be regarded as reducing hypo-elasticity
to elasticity, for the converse is generally false, and in particular the simpler
special cases of (304.4) are not elastic cases. Finally, it has been shown that
most of the common "incremental" theories of plasticity other than that of
1 [1955, 21, §§ 40 and 33]. A result of the sameform had been deduced from a too restrictive definition by TRUESDELL [1951, 27, § 26].
2 TRUESDELL [1953, 32, §56 (revised)J [1955, 27 and 28]. 3 NoLL [1955, 18, § 15b].
Sect. 305. Visco-elastic and accumulative theories. 73.3
perfectly plastic materials are included as special cases of hypo-elasticity, providing they be first corrected so as to satisfy the principle of material indifference1 .
In much of the foregoing discussion, it has been assumed that the body is
initially unstressed, but this assumption is not at all necessary in hypo-elasticity.
Suppose that in an interval dt a body subject to initial stress s 0 undergoes a
displacement having small gradients in the sense defined in Sect. 57. Then we
have wk, dt R:j Rk, and dpq dt f'::j epq• where il and e are the tensors of infinitesimal
rotation and strain for the displacement considered. Also skmdt R:j skm_ s~m.
From (.304._3) and (148.7) we have then
(.304.7)
The essential content of these equations, which define a theory of small deformation of an initially stressed body, was given by CAUCHY 2• It reflects the fundamentally greater generality of hypo-elasticity in comparison to elasticity. While,
as is plain from (303.6), the most general theory of finite elasticity is not capable
of representing an elastically isotropic body which in its natural state suffers
any but hydrostatic stress, in hypo-elasticity there is, in general, no natural
state, and the stress at any given instant may be arbitrary. It was CAUCHY's
theory of initially stressed bodies, a special case ofthat defined by (.304.7), that
suggested the theory of hypo-elasticity. Hypo-elasticity may be regarded as a
theory in which relations of CAUCHY's type are applied in each time interval dt.
305. Visco-elastic and accumulative theories. Studies of elasticity and viscosity, according to the classical theories presented in Sects. 298 and .301, made
it natural to attempt to combine the two kinds of phenomena within a single
theory. Two ideas immediately present themselves:
1. The total stress is the sum of an elastic stress arising from the strain and a
viscous stress arising from the stretching, and
2. The total rate of strain is the sum of an elastic rate arising from the rate of
stressing and a viscous rate arising from the stress.
Schematically, the two alternatives may be written
t=f(e) +g(e) and e =h(t) +k(t). (305.1)
Both imply necessarily the existence of material constants having the physical
dimensions of elasticity and of viscosity; hence there is a modulus having the
physical dimensions of time, and relaxation effects may be expected. The former
alternative was proposed by 0.-E. MEYER, VoiGT, and DUHEM 3; the latter, by
MAXWELL, NATANSON, and ZAREMBA 4. Theoriesofthis type, including generalizations obtained by allowing time derivatives up to arbitrarily high order to occur
on the each side of (.3 0 5 .1 h or (3 0 5 .1) 2 , are called visco-elastic. If e is the infinitesimal strain tensor, the constitutive relations are
n (o)k m ( 0 )k ~ Ak 8t t = J:o Bk 8t e, (305.2)
1 GREEN [1956, 9 and 10]. Perfectly plastic·materials are included.formally by a limit
process. 2 [1829, 4, Eqs. (36), (37)]. 3 [1874. 3 and 5] [1875, 4]; [1889, 10] [1892, 12 and 13] [1910, 10, §§ 395-396];
[1903, 6-9] [1904, 1, Part I, Chap. II, and Part IV, Chaps. II-III]. 4 [1867, 2, pp. 30-31]; [1901, 11-13] [1902, 7-9] [1903. 12-14]; [1903, 17-20]
[1937. 12].
734 C. TRUESDELL and R. TOUPIN: The Classical Field Theories. Sect. 306.
where the coefficients Ak and Bk are usually assumed constant. These theories
possess a large recent literat~re, mostly in one-dimensional andjor linearized
contexts.
Such linear theories generally violate the principle of material indifference,
as was first noticed by ZAREMBA, who was led to his form of the principle in
this context. His is the only early attempt in viscoelasticity to yield a theory
that is a mechanically possible one for unrestricted motions.
A generaland properly invariant theory of the MEYER-VOIGT type, allowing
the stress to depend upon the first spatial derivatives of the displacement and
of the accelerations of all orders, has been achieved by RrvLIN and ERICKSEN1 .
The Maxwell-Zaremba theory is generalized by NoLL's theory of hygrosteric
materials 2, which are defined by (304.1) without requiring f to be linear in d,
and by a more general theory sketched by CoTTER and RIVLIN 3•
BoLTZMANN and VoLTERRA4 proposed a theory in which the stress is determined by the entire sequence of strains undergone by the body in the past.
The material is represented as having a weaker memory for older experiences;
the stress is obtained by integrating the strains from - oo to t, with a suitable
damping or "memory" function. The linear theory, sometimes called "hereditary" but more fitly named accumulative, has a considerable literature.
It is still more general to allow the stresstobe determined in any way by the
strain history, i.e., 1
t = F [x~cx], (305.3) -oo
where F denotes a functional. This extremely general and natural concept of
material behavior has been put into properly invariant from by NoLL5 and by
GREEN, RrVLIN, and SPENCER6 (1956). The latter authors determined conditions
under which the functional in (305.3) may be replaced by a sum of integrals of
VüLTERRA's type or by a finite combination of time derivatives as in the theory of
RrvLIN and ERICKSEN. NoLL's method, which rests directly upon properties
of invariance as the definitions of particular materials, is presented in "The
Non-linear Field Theories of Mechanics ", This Encyclopedia, Vol. VIII, Part 2.
V. Examples of thermo-mechanical constitutive equations.
306. Irreversible thermodynamics. While even the classical theory of isotropic viscous and thermally conducting fluids furnishes an example where both
mechanical and energetic principles must be used in order to solve any definite
problem, the constitutive equations themselves, namely, (296.1) and (298.5),
embody separate principles, one being purely mechanical and the other, purely
energetic.
The various theories of "irreversible thermodynamics" attempt to describe
the numerous physical phenomena which cannot be separated as belonging to
one or the other category to the exclusion of the other. Such phenomena occur
especially in heterogeneaus substances. Current practice concentrates upon the
production of entropy owing to such interactions and supposes that the "affinities" and "fluxes" which enter into the production of entropy depend functionally, and usually linearly, upon one another.
1 [1955. 21]. 2 [1955. 18, §§ 6-9]. 3 [1955. 4]. 4 [1874, 1]; [1909, 10 and 11] [1930, 9]. 5 [1957. 11] [1958, 8].
6 [1957. 6] [1959. 6].
Sect. 307. The "Maxwellian" fluid. 735
In Sects. 257 and 259 we have explained the concepts in terms of which such
theories are constructed. Since we do not consider that the problern of dynamical
and energetic invariance is yet sufficiently understood, we rest content with
citing expositians of the subject as it is currently received 1.
307. The "Maxwellian" fluid. As our sole example of the use of the principle
of equipresence (Sect. 29317), we mention a fully thermomechanical theory of
fluids based on two constitutive assumptions:
1. Both the stress and the flux of energy depend upon the spatial and temporal
derivatives of the thermodynamic state and of the velocity, of all orders.
2. The constitutive relations involve material coefficients having the physical
dimensions of viscosity, thermal conductivity, and temperature, but no others.
These constitutive relations define the M axwellian fluid of TRUESDELL 2•
While this theory is known 3 to stand in need of revision so as to be rendered
properly invariant, the general forms of the leading terms in the expansions of
the stress and flux of energy in powers of the viscosity give an idea of what is tobe
expected. From the first of the defining conditions, expressing an application
of the principle of equipresence, it might be thought that the expansions for
the stress t and flux of energy h would be very similar. However, the different
tensorial characters of t and h, combined with requirements of dimensional
invariance, force the the counterparts of terms present in the expansion of t to
be absent from the expansion of h, and conversely.
To avoid long formulae we summarize the forms of the terms of orders 0, 1, 2
in words. The stress t is the sum of
TL The terms in the linear law of fluid viscosity (298.5).
T2. The quadratic viscous terms according to the theory of RIVLIN and
ERICKSEN mentioned in Sect. 305.
T}. The mostgenerallinear isotropic function of P,k P,m·
T4. The mostgenerallinear isotropic tensor function of P,(k(),ml·
T 5. The most generallinear isotropic tensor function of (), k (), m.
T6. The mostgenerallinear isotropic tensor function of P,km·
T7. The mostgenerallinear isotropic tensor function of O,km·
Similarly, the flux of energy h is the sum of
H 1. A linear term,
where IX is dimensionless, and
H 2. The most general linear isotropic vector function of
P,k' e,m' and Xp,q.
H3. The mostgenerallinear isotropic vector function of ip,q,.
(307.1)
It is most striking that in these results the separation of effects follows from
considerations of invariance alone. For example, despite the much more general
definition of a fluid used here, the linear terms T 1 in the stress are exactly the
same as occur both in the classical linear theory and in the simpler non-linear
1 PRIGOGINE [1947, 12], DE GROOT [1952, 3], J. MEIXNER and H. REIK in Vol. III, Part 2
of this Encyclopedia.
2 [1949, 34, §§ 2-4) [1951, 27, §§ 19-21). We have altered slightly TRUESDELL'S
definition and results. 3 Cf. the remarks of TRUESDELL [1956, 13, § 16).
736 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 308.
"Stokesian" theory (Sect. 299). While FouRIER's law (296.4) is slightly generalized
by (307.1), the added term is of a thermodynamic type, parallel to the classical
thermal one, and in the linear approximation the effects of deformation cannot
enter the constitutive equation for h. Similar, though more elaborate, separations
of effects are seen in the quadratic terms T2 to T7 and H2 to H3. These quadratic terms allow for interactions of a definite, not arbitrary kind.
Restrietions such as these, which follow from principles of invariance alone,
are only now coming to be studied.
VI. Electromagnetic constitutive equations.
308. The Maxwellian dielectric. The problern of formulating constitutive
equations for moving and deforming material media is one of the most difficult
and controversial in electromagnetic theory. We shall illustrate some of the
relevant ideas by treating a simple example.
rx.) The Euclidean invariant constitutive equations of a Maxwellian dielectric.
In classical electromagnetic theory as refined by LoRENTZ we have the aether
relations (cf. Sect. 279), which can be written in the 4-dimensional tensorform
r;nd = 1 ~ ( _ det y)~ yll'Fyde CfJ'I'e V~
or the 3-dimensional vector form
H = _!_B.
f.lo
D =e0 E,
(308.1)
(308.2)
According to the Lorentz point of view, these relations hold in all material media,
moving or stationary, but it must be recalled that D and H are the charge and
current potentials for the total charge including that due to polarization and
magnetization of the material medium.
We define the ideal material called a Maxwellian dielectric by the relations
P=e0 xE,
M=O
X= const } (308.3)
for the polarization P and the magnetization M, provided the medium is at
rest in a Euclidean frame which is simultaneously a Lorentz frame. Recall that
such a frame is one for which we have the canonical forms (cf. Sects. 280 and 152)
_ 1= [<5rs 0 l = [<5rs 01 1 Y 0 _ ~ , g 0 0 , t = (0, 0, 0, 1), c2 =-. (308.4) ~ ~~
If we introduce the potential T= (~, ~) of free charge and current (cf. Sect. 283)
in a polarizable and magnetizable medium, the constitutive relations (308.3)
can be put in the 4-dimensional form
(308.5)
with
-)(1 = [ 15,5 0 1'
0 - Bflo
e = e0 K, K = 1 +X (308.6)
or in the 3-dimensional vector form
~=eE, 1 ~=-B. f.lo (308.7)
Sect. 308. The Maxwellian dielectric. 737
To see how these relations are generalized to the case of moving media, it is
easiest to employ the world tensor formalism of Sect. 152 as applied to the problern
of constructing invariants of fields under the Euclidean group of transformations
(rigid motions). It was noted in Sect.152 that if lJI"was a world tensor satisfying
the conditions
(308.8)
then the non-vanishing components prs. · · in a Euclidean frame transform as a
3-dimensional tensor under the group of Euclidean transformations. This special
kind of world tensor was called a space tensor. Consider then the electromagnetic
field cp and the world velocity vector v of a motion. In terms of these quantities
we can define two associated space tensors
(i;!.l = gD-1 f/J-1s vS'
58!.1-1 =g!.I'Pg-1Sf/J'l'S·
In a Euclidean frame we have (274.1), so that, in such a frame,
i = (E + vxB, 0), !8 = (dualB, 0).
(308.9}
(308.10}
(308.11)
Consider next the polarization-magnetization world tensor :rt defined in (283.23).
In terms of this tensor we defined the space tensors 1.)3!.1 and 9Jl!.l -1 called the world
polarization and magnetization densities in (283.24) and (183.25). In a Euclidean
frame we have
!l3 = (P, 0), IJJl =(dual M, 0). (308.12)
Therefore, the world tensor equations
1.)3D = eo Vg X (i;!.l, 9JlDLI = 0, (308.13)
reduce to (308.3) when the medium is at rest in a Euclidean frame. Moreover,
since these equations involve only space tensors, they are rotationally invariant.
Stated more explicitly, the proportionality of polarization and electromotive
intensity and the vanishing of the magnetization are invariant conditions under
the group of Euclidean transformations relating the dass of reference frames
moving rigidly with respect to one another. Now from (283.33), (283.24), (283.25),
(279.13), and the aether relations (308.1) we have
yn.1 = "1!.1.1 _ ;n;D.1,
= 'V*(-dety)![y!.l'~'y-1 q;'l'fJ- ~g-1'Pv!.lvSq;'Ps-J (308.14)
- _K_gnSv-1v'P m ] c2 r'PS •
or, in a frame for which we have the canonical forms (308.4),
~ = e0 [E +X (E + v X B], l ~ = - 1 [B + ~ v X (E + v X B)]. 1-'o c
(308.15)
We have arrived at these constitutive equations for the potentials ~ and ~
by demanding that the relations between polarization and magnetization be
invariant under the group of rigid (Euclidean) transformations. We are led
to this idea of invariance by thinking of polarization and magnetization as
quantities characteristic of the moving material medium and carried with it.
Handbuch der Physik, Bd. 111/t. 4 7
738 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 308.
Thus, on the basis of classical thought, any relation betwcen them and the electromagnetic field should be invariant under the group of rigid motions. This is an
application of the principle of material indifference.
We wish now to cantrast with the results (308.14) and (308.16) obtained by
this classical procedure the corresponding results obtained by MINKOWSKI 1 , whose
reasoning was guided by the idea of Lorentz invariance of the constitutive relations in moving media.
ß) The Lorentz-invariant constitutive equations for a moving M axwellian
dielectric. Suppose a Maxwellian dielectric is at rest in some Lorentz frame
which, for our present purpose, we may imagine as coinciding with some
Euclidean frame. Recall that the dass of Lorentz frames are those for which the
tensor y has the canonical form (308.4)1 . Let us assume that, when the medium
is at rest, it is characterized by the relations (308.5). Now suppose that such a
medium is set into motion. Any selected particle of the medium undergoing
the motion will be instantaneously at rest in some other Lorentz frame. MINKOWSKI reasoned that the relations (308.5) should hold in that Lorentz frame as
a consequence of the correct constitutive equations for the moving medium.
To put this idea into effect, let LD .1 be the coefficients of a Lorentz transformation
( cf. Sect. 282), so that by definition we have
-1 -1 LD.-1 L'~'e yL1e = yD'P. (308.16)
Let wD be the relativistic velocity vector of the motion defined by
c vD
w0 == ~· , Yn.-1 wDw.-1 =- c2 , (308.17) - Y~e v~ve
where, as before, v is the classical world velocity vector of thc motion. In the
reference framein which the medium is moving we have
w-( v 1 ) (308.18) - v1- ~ v1 - v2 • c2 c2
We can choose the coefficients LD.-1 of a Lorentz transformation so that
has the form
w = (0, 0, 0, 1)
(308.19)
(308.20)
at some selected event. If the medium is in uniform translatory motion, so that
v =const, then w will have the form (308.20) throughout a space-time region,
but, for general motions, we shall have (308.20) only at a single event. Let if'DL1
and (jjDL1 denote the transformed components of r and Cf!· That is,
if'DL1 =LD'PLL1e T'~'e, fPnA = L'Pn Le.-1 CfJ'Pe· (308.21)
Thus, in the new coordinate system, MINKOWSKI assumes that
-!J.-j_,;~( d t-)1-!JljF-Ltf')-
- V f.lo - e """ " CfJ'P&• (308.22)
where x has rest values consistent with (308.6). Applying the inverse Lorentz -1 -1
transformation LeA (LD.-1 LL1'P= bfl,) to (308.22) we see that the relation between
I [1910, 7).
Sect. 308. The Maxwellian dielectric.
the unbarred components of r and gJ must be
TQLJ = v-~- (- det x)~ XQ 'P xL1 e IP'Pe'
flo
where, for the moving medium,
x!JLl = L~'PLLJ e x_'Pe.
739
(308.23)
(308.24)
To compute the values of the unbarred components x!JLl it is convenient to decompose x as follows:
X,!JLl = yQLl- f-lo (e- eo) w!J wLl' l
=yQLl _ Lw!JwLl.
c2
It then follows easily from (308.16) and (308.19) that
(308.25)
(308.26)
Substituting this result into (308.23), we get the Minkowski relations in the form
T!JLJ =V;:(- dety)~ x I
X [JJQ'PyLJe IP'Pe -1dYQ'P wLl we IP'Pe- "J}Ll'P w!2we IP'Pe)]
or, in terms of ~ and ~.
~= e0 [E + x 2 (E + vxB)- ~-x- -vv ·E], 1 - !____ c2 (1 - !____) c2 c2
(308.27)
(308.28)
Eqs. (308.28) are to be compared with the classical expressions (308.15). We
see that if one neglects all terms 0 (v2jc2) in the Minkowski constitutive equations,
they reduce to the classical ones. MINKOWSKI's method of deriving Lorentzinvariant electromagnetic constitutive relations was extended to the case of
moving crystals by EINSTEIN and LAUB1 and by BATEMAN 2•
y) M oving surfaces of discontinuity, the velocity of light, and Fresnel' s dragging
formula. If we substitute the constitutive relations (308.15) into the electromagnetic boundary conditions (278.8), (278.9) and (283.44), (283.45) at a moving surface of discontinuity, we obtain the following system of equations:
nx[EJ-un[B]=O, [B]·n=O, [E+x(E+vxB)]·n=O,l
(308.29)
nx [B + 4-vx(E+vxB)] + 4-[E + x(E + vxB)]=O. c c
The second and third of these conditions are satisfied as a consequence of the
first and the fourth provided we assume un =f= 0. Let us suppose that the velocity v
and the polarizability X of the dielectric medium are continuous. In this case,
the latter two vector equations constitute a system of six homogeneaus linear
1 [1908, 3]. 2 [1922, 1].
47*
740 C. TRUESDELL and R. TOUPIN: The Classical Field Theories. Sect. 309.
equations in the six quantities E, and B'. Therefore, the system admits nonzero solutions for the jumps [E] and [B] if and only if the determinant of the
coefficients vanishes. To simplify matters, let us consider the case where the normal
n to the surface of discontinuity is parallel to the velocity of the medium v. In
this special case, the vanishing of the determinant requires that the speed of
propagation un satisfy the quadratic equation
K (u")2- ~ (11..!!.)' + ~~ -1 =0. c c c c2 (308.30)
The solution of this equation is
x-± V v--V2 K-x-- c c2
--------
K (308.31)
N ow when the polarizability of the medium vanishes (x = 0, K = 1) the speed
of propagation of the surface of discontinuity is ± c. Thus the fundamental
constant c == V1/e~.U~ is the speed of propagation of an electromagnetic discontinuity
in vacuum or any medium in which the polarization and magnetization vanish
and there is no free charge or current. We see that the motion of such a medium
devoid of polarization has no effect on this result. Now set n = VK, where n
is the index of refraction. We can then write the solution (308.31) in the form
Un = _1_ [± 1 + ~_11_ + 0 (v2jc2)J]' c n n c {308.32)
which, when the terms 0 (v2fc2) are neglected, reduces to FRESNEL's dassie formula 1
for the dragging of light by a moving polarizable medium.
If one substitutes MINKOWSKI's constitutive relations into the jump conditions, the equation analogaus to (308.30) is
(308.33)
Again we see that neglecting the terms 0 (v2fc 2) leads to FRESNEL's result. However, the two equations for the determination of the speed of an electromagnetic
discontinuity in a moving Maxwellian dielectric based on the classical and Minkowski constitutive equations of the moving medium differ by terms 0 (v2jc2).
309. VoLTERRA's electromagnetic constitutive equations. We next mention
VoLTERRA's generalization of MAXWELL's constitutive relations for ~ and ~
in stationary dielectric and magnets 2• In mostreal materials, the magnetization
is not a single-valued function of the magnetic flux B, much less a linear function.
The magnetization at any instant, however, can be considered as a functional
of the history of the field B (cf. Sect. 305). If the "heredity" is linear, VaLTERRA
writes 00
B(~.t) =ft~(x,t) + Jif>(-r)~(~.t--r)d-r, (309.1) 0
where if>(-r) is the coefficient of heredity. When constitutive equations of this
type are substituted into the differential field equations obtained from the
conservation laws, one obtains a system of integro-differential equations which
govern the evolution of the physical system rather than a system of partial differential equations as in the simpler theories.
I WHITTAKER [1951, 39, p. 403]. 2 [1912, 7], [1930, 9, p. 195].
Sects. 310,311. ÜHM's law for moving conductors. 741
310. MIE's theory. The role of the potential fields fj and a in the classical
theory is rather odd and asymmetrical. We see that the electromagnetic potential
a is but a subsidiary field which is generally introduced in order to facilitate the
solution of problems. The whole system of electromagnetic equations is invariant
under potential transformations of the electromagnetic potential (gauge transformations). On the other hand, the charge-current potential fj plays a more
fundamental role in the classical theory because of the Maxwell-Lorentz aether
relations, which are not invariant under potential transformations of f/· In
MIE's theory1, the potentials a and fj are made to enter the theory in a more
symmetrical manner. In addition to the conservation laws of charge and magnetic
flux, MIE assumes the existence of a "universal" function A such that
a oA
a = aoc!i. (310.1)
The function A is supposed to be a Lorentz-invariant function of the electromagrtetic field (/! and the electromagnetic potential a. MIE's theory is concerned
with the fundamental question of the electromagnetic constitution of matter.
VAN DANTZIG 2 proposed a theory somewhat similar to MIE's in which the potentials
fj and a were assumed tobe linear fundionals of the fields (/! and lJ. We cite these
examples of constitutive relations to illustrate the variety of viewpoints which
have been expressed as regards the appropriate constitutive relations to accompany the conservation laws of electromagnetic theory.
311. OnM's law for moving conductors. We consider the dass of ideal materials
such that, when they are at rest in a Euclidean frame, the current J is a linear
isotropic function of the electric field E:
J=CE, C = const. (311.1)
This relation is called Ohm's law. We generalize ÜHM's law to the case of a moving
medium in much the same way as the constitutive equations of a Maxwellian
dielectric were generalized in Sect. 308. First we consider the classical or rotationally invariant generalization. In terms of the charge-current vector (J and
the velocity vector v of a motion, we define two space tensors
:O==anta, }
~n =an- vn aLl tLl. (311.2)
In a Euclidean frame we have
:0 = Q, ~ = (~, 0)' (311.3)
where Q is the charge density and ~ is the conduction current. MAXWELL's
generalization 3 of ÜHM's law to the case of a moving medium can then be stated
in the form
(311.4)
Since ~ and ~ are space tensors, the 4-dimensional formalism insures that the
constitutive equation (311.4) is rotationally invariant. In a Euclidean frame it
assumes the form
~ =C~, }
J- Q v = C (E + V X B), (311. 5)
1 [1912, 6]. A summary of Mm's theory is given by WEYL [1921, 6, § 26] [1950, 3.5, § 28]. 2 [1934, 10]. 3 [1873. 5, §609].
742 C. TRUESDELL and R. TOUPIN: The Classical Field Theories. Sect. 312.
where (.! is the electromotive intensity at a point moving with the medium.
When the medium is at rest, (311.5) reduces to (311.1).
MINKOWSKI's generalization of ÜHM's law is obtained by requiring (311.1)
to hold in the Lorentz frame in which the particle of the moving medium is
instantaneously at rest. Thus we set
(j!i = LQLJ afl' l
cpgLJ = L'~' D LeLJ T'Pe• -1
wD = (o, o, o, 1), wD = LD Ll wfl,
(311.6)
so that the relation (311.1) in the moving Lorentz frame can be put in the form
(311.7)
Applying the inverse Lorentz transformation, we then get
Q + 1 'P e D- C D'P . e a (JY'Pea w w - y T'Pflw . (311.8)
The spatial component of this last equation is
J + _1_ ('J · v- c2 Q) V = C (E + v X B) c2 1 - v2(c2 1 - v2(c2 . (311.9)
Taking the scalar product of (311.9) by v, we obtain an equation which can be
solved for J · v:
J · v = C V1= v2jc2 v . E + v2 Q. (311.10)
We then eliminate J · t' from (311.9) and obtain finally
J-Qv=. C (E+vxB-~-VV·E), Vt-v2fc2 c (311.11)
which is the relativistic form of ÜHM's law for moving media. Again we see
that the relativistic or Lorentz-invariant constitutive equation (311.11) differs
from the classical or Euclidean-invariant constitutive equation (311.5) only by
terms 0 (v2jc2).
VII. Electromechanical constitutive equations.
312. Elastic dielectrics. The phenomena of piezoelectricity, photoelasticity,
and electrostriction in elastic solids are closely related. Because of their importance in engineering applications, the classical theories for these effects have become highly specialized disciplines 1. Any such theory must be based on simultaneaus application of the principles of mechanics and of electromagnetism. The
laws of conservation of energy and momentum, agumented so as to include the
effects of the electromagnetic field, have been formulated in Chap. F. The
relevant equations of mechanics, aside from the boundary conditions, are conveniently summarized in the set of Eqs. (288.11) to (288.13) and (286.14). For
dielectric media in the absence of free charge and current and of magnetization,
the charge and current densities occurring in (288.11) and (288.12) are expressed
in terms of the polarization Pas in (283.22). If these expressions for the charge
1 As sources of experimental and theoretical results and references to original and contemporary Iiterature in this field we may cite the following works: VorGT [1910, 10], MASON
[1950, 17], CADY [1946, 1], CoKER and FrLON [1931, 3], STRATTON [1941, 8].
Sect. 312. Elastic dielectrics. 743
and currcnt are substituted into (288.11) and (288.12) we obtain the basic field
equations of mechanics as applied to the case of dielectric media:
(! x' = trs - ~ div P (§;' + -1- (~!'I>_ X B)' ,s yg yg dt '
• • 1 dc P a:. d" h (!e =t'sxs,, + Vfat. ~ + IV , (312.1)
0(! ( "r) - Tt + QX ,r- 0,
tfrs]=Q.
We have written these equations as they appear in a general curvilinear inertial
co-ordinate system. Under general time-independent transformations of the spatial
Co-ordinates xk, pr and ßr transform as vector densities of weight 1. The quantities (!, e, t' 5 , h' and E, transform as absolute tensors. A comma denotes, as
usual, covariant differentiation based on the ChristoHel symbols of the metric
tensor g,,. Eqs. (312.1) are supplemented by the conservation laws of charge
and magnetic flux and by the aether relations of electromagnetic theory. The
classical theory of piezoelectricity is based on the linearized version of this system
of equations corresponding to infinitesimal deformations and weak fields and the
linear piezoelectric constitutive equations of VorGT, which can be expressed in
the form
t's = crsmn e + _1 mn yg yrs m pm ' l 1 -1
E, = yg Xrs ps +rmn,emn•
(312.2)
where e,s =u(r,s> (u =displacement vector) is the classical measure of infinitesimal
strain. Because of the relation D = e0E +P, the Voigt relations (312.2) can be
written in various forms corresponding to different choices of independent variables.
A thermodynamic treatment of the Voigt relations can be based on the energy
equation (312.1) 2 by making the assumptions necessary to yield the equation
e ()ij = div h, (312.3)
where r; is the entropy density and () is the absolute temperature. Then by
assuming that the internal energy s, the stresst, the electric field E, and the temperature () are functions of the infinitesimal strain measure e, the polarization P,
and the entropy r;, we obtain VorGT's relations in the linear approximation if
the energy equation is assumed to be satisfied identically in the independent
variables e, P, and r;. The coefficients in the Voigt relations are then given by
mn - lfa oe r s - e v g a;--a:ps . mn
-1 oe
X - ng--- (312.4) rs- o: oP•(!ps .
The classical linear theory of piezoelectricity outlined above has been generalized to the case of finite deformation and large field strengths by ToUPIN1.
The non-linear theory stands in the same relation to the Voigt theory as the theory
of finite elastic deformations stands in relation to linear elasticity theory. In
the non-linear theory, the internal energy e is assumed initially to be a general
polynomial function of the displacement gradients xi;.l of a deformation, the
1 [ 19 56, 20].
744 C. TRUESDELL and R. TouPIN: The Classical Field Theories. Sect. 313.
polarization P, and the entropy density 'YJ· One then shows that if the energy
is invariant under the group of rigid motions, it must reduce to a function of the
variables
1]. (312. 5)
Assuming that (312.3) holds for this ideal medium, that the electromotive intensity, the stress, and the absolute temperature are functions of the displacement
gradients, polarization, and entropy, and that the energy equation is satisfied
identically yields the constitutive relations of the general theory:
OE trs = 2 (! --X' X5 = tsr
()CIJV ;IJ ;v '
(!)r- "'-~X"' i}I/IJ ;r• (312.6)
0=~. OTJ
Theserelations generalize the Voigt relations (312.2) to the case of finite deformations, large field strengths, and moving media. They reduce to the Voigt relations
in the linear approximation and for stationary media.
313. Magnetohydrodynamics. Electrodynamics of continuous media is currently enjoying a new birth of interest. This contemporary work is generally
classified under the title, magnetohydrodynamics. The abundance of theoretical
and experimental labor now directed toward the novel behavior of conducting
fluids moving in a magnetic field should yield progress toward an understanding
of the general theory of electrodynamics of continuous media. In the present
theories of magnetohydrodynamics, the constitutive relations for the stress
generally take the form
(313.1)
as in the classical theory of viscous fluids. ÜHM's law for moving media
(311.5) is generally assumed and a simple type of constitutive relation such
as M = kB is introduced to account for the effects of magnetization. The
general equations of magnetohydrodynamics are obtained by substituting these
rather special constitutive relations into (288.11) and augmenting this set of equations by the equations of conservation of charge and magnetic flux. The linear
magnetohydrodynamic equations can then be obtained by casting away the
non-linear terms in the general equations. Qualitative features of magnetohydrodynamic solutions are discussed by ELSASSER and ALFVEN 1.
List of Works Cited.
(Italic numbers in parentheses following the reference indicate the sections
where the work is mentioned.)
1678 1. HooKE, R.: Lectures de Potentia Restitutiva, or of Spring Explaining the Power
of Springing Bodies. London = R. T. GtrNTHER: Early Science in Oxford 8,
331-356 (1931). (293)
1687 1. NEWTON, I.: Philosophiae Naturalis Principia Mathematica. London. 3rd ed.,
ed. H. PEMBERTON, London 1726; trans. A. MoTTE, Sir Isaac Newton's Mathematical Principles of Natural Philosophy and his System of the World. London:
1729. There are many later editions, reprints, and translations. Our references
are to the first edition. (3, 77, 89, 196, 196A, 293, 294, 298)
---
1 [1956, 6] [1950, 1].
List of Works Cited. 745
1736 1. EuLER, L.: Mechanica sive Motus Seientia Analytice Exposita 1, Petropoli =
Opera omnia (2) 1. (4, 99, 155)
1738 1. BERNOULLI, D.: Hydrodynamica sive de Viribus et Motibus Fluidarum Commentarii. Argentorati. (66A, 77, 89)
1743 1. BERNOULLI, J.: Hydraulica nunc primum detecta ac demonstrata directe ex fundamentis pure mechanicis, Opera 4, 387-493. Virtually the same work appears
under different titles in Comm. Acad. Sei. Petrop. 9 (1737), 3-49 (1744); 10 (1738).
207-260 (1747). (200)
2. D'ALEMBERT, J. L.: Traite de Dynamique ... Paris. (196)
1744 1. D' ALEMBERT, J. L.: Traite de l'Equilibre et du Mouvement des Fluides pour servir
de Suite au Traite de Dynamique. Paris; 2nd ed., 1770. (89) ·
1745 1. CLAIRAUT, A.: Sur quelques prineipes qui donnent la solution d'un grand nombre
de problemes de dynamique. Mem. Acad. Sei. Paris (1742), 1-52. (143, 197)
2. EuLER, L.: Neue Grundsätze der Artillerie, aus dem Englischen des Herrn Benjamin Rohins übersetzt und mit vielen Anmerkungen versehen. Berlin = Opera
omnia (2) 14. (66A, 70, 202, 248)
1746 1. EuLER, L.: De motu corporum in superficiebus mobilibus. Opusc. var. arg. 1,
1-136 = Opera omnia (2) 6, 75-174. (168)
17 51 1. KoENIG, S.: De universaH prineipio aequilibrii et motus, in vi viva reperto, deque
nexu inter vim vivam et actionem, utriusque minimo, dissertatio. Nova acta
erudit., 125-135, 162-176 = L. Euleri Opera omnia (2) 5, 303-324. (167)
1752 1. D'ALEMBERT, J. L.: Essai d'une Nouvelle Theorie de la Resistance des Fluides.
Paris. (66A, 68, 70, 77, 86A, 88, 98, 108, 130, 140, 156, 202, 206)
2. EuLER, L.: Decouverte d'un nouveau principe de mecanique. Mem. Acad. Sei.
Berlin [6] (1750), 185-217 = Opera omnia (2) 5, 81-108. (1, 7, 168, 196)
3. EuLER, L.: Recherehes sur l'effet d'une machine hydraulique proposee par Mr. Segner Professeur a Göttingue. Mem. Acad. Sei. Berlin [6] (1750), 311-354 =Opera
omnia (2) 15, 1-39. (143, 197)
1753 1. EuLER, L.: Recherche sur une nouvelle maniere d'elever de l'eau proposee par
Mr. DeMour. Mem. Acad. Sei. Berlin [7] (1751), 305-330 =Opera omnia (2) 15,
134-156. (163)
1755 1. SEGNER, A.: ... Specimen Theoriae Turbinum, Halae Typis Gebaverianis, 27 April
17 55. 40 pp. (168)
17 56 1. EuLER, L.: Theorie plus complette des machines qui sont mises en mouvement
par la reaction de l'eau. Mem. Acad. Sei. Berlin [10] (1754), 227-295 = Opera
omnia (2) 15, 157-218. (163, 197)
17 57 1. EuLER, L.: Prineipes generaux de l'etat d'equilibre des fluides. Mem. Acad. Sei.
Berlin [11] (1755). 217-273 = Opera omnia (2) 12, 2-53. (200, 207, 248)
2. EuLER, L.: Prineipes generaux du mouvement des fluides. Mem. Acad. Sei. Berlin
[11] (1755). 274-315 = Opera omnia (2) 12, 54-91. (66A, 70, 76, 88, 89, 98,
108, 109, 110, 156, 205, 248, 297)
3. EuLER, L.: Continuation des recherches sur la theorie du mouvement des fluides.
Mem. Acad. Sei. Berlin [11] (1755), 316-361 = Opera omnia (2) 12, 92-132.
(89, 99, 109, 110, 125, 130, 137, 143, 161, 197, 248)
1761 1. D' ALEMBERT, J. L.: Remarques sur les lois du mouvement des fluides. Opusc.
1, 137-168. (99, 108, 133, 140, 161)
2. EuLER, L.: Prineipia motus fluidarum (1752-1755). Novi Comm. Acad. Sei.
Petrop. 6 (1756-1757). 271-311 = Opera omnia (2) 12, 133-168. (66A, 76,
77, 84, 86A, 88, 98, 108, 109, 130, 141)
1762 1. EULER, L.: Lettre de M. Euler a M. de La Grange, Recherehes sur la propagation
des ebranlemens dans un milieu elastique. Mise. Taur. 22 (1760-1761), 1-10 =
Opera omnia (2) 10, 255-263 = Oeuvres de Lagrange 14, 178-188. (15, 17, 20,
66A, 156)
2. LAGRANGE, J. L.: Nouvelles recherches sur la nature et la propagation du son.
Mise. Taur. 22 (1760-1761). 11-172 = Oeuvres 1, 151-316. (20, 108)
3. LAG RANGE, J. L.: Application de la methode exposee dans le memoire precedent
a la solution de differens problemes de dynamique. Mise. Taur. 22 (1760-1761).
196-298 = Oeuvres 1, 365-468. (86A, 140, 160, 233)
1764 1. EULER, L.: De motu fluidarum a diverse caloris gradu oriundo. Novi Comm.
Acad. Sei. Petrop. 11 (1765), 232-267 = Opera omnia (2) 12, 244-271. (248)
1765 1. EuLER, L.: Theoria Motus Corporum Solidarum seu Rigiderum ex Primis nostrae
Cognitionis Principiis Stabilita et ad Omnis Motus, qui in hujusmodi Corpora Cadere
Possunt, Accomodata. Rosteck = Opera omnia (2) 3 and 4, 3-293. (165, 168, 196,
294)
746
2.
3.
1766 1.
1767 1.
1768 1.
2.
1769 1.
1770 1.
1771 1.
2.
1776 1.
2.
3.
4.
1783 1.
1786 1.
1788 1.
1806 1.
1821 1.
1822 1.
2.
1823 1.
2.
3.
1824 1.
1825 1.
2.
1827 1.
C. TRUESDELL and R. TouPIN: The Classical Field Theories.
EuLER, L.: Recherehes sur Ia connaissance mecanique des corps. Mem. Acad.
Sei. Berlin [14] (1758), 131-153. (165, 168)
EuLER, L.: Du mouvement de rotation des corps solides autour d'un axe variable.
Mem. Acad. Sei. Berlin [14] (1758), 154-193. (143, 170)
EuLER, L.: Supplement aux recherches sur Ia propagation du son. Mem. Acad.
Sei. Berlin [15] (1759), 210-240 = Opera omnia (3) 1, 452-483. (17, 20, 66A,
156)
EuLER, L.: Recherehes sur le mouvement des rivieres (1751). Mem. Acad. Sei.
Berlin [16] (1760), 101-118 = Opera omnia (2) 12, 212-288. (66A, 125)
D' ALEMBERT, J. L.: Sur l'equation qui exprime Ia loi du mouvement des fluides.
Opusc. math. 5, No. 33, 95-131. (161)
D' ALEMBERT, J. L.: Suite des recherches sur le mouvement des fluides. Opusc.
math. 5, No. 34, 132-170. (202)
EuLER, L.: Sectio prima de statu aequilibrii fluidorum. Novi Comm. Acad. Sei.
Petrop. 13 (1768), 305-416 = Opera omnia (2) 13, 1-72. (7, 248)
EuLER, L.: Sectio secunda de principiis motus fluidorum. Novi Comm. Acad. Sei.
Petrop. 14 (1769), 270-386 = Opera omnia (2) 13, 73-153. (15, 17, 20, 66A,
72, 76, 82A, 84, 88, 89, 109, 156, 160, 164, 210, 297)
EuLER, L.: Sectio tertia de motu fluidarum lineari potissimum aquae. Novi Comm.
Acad. Sei. Petrop. 15 (1770), 219-360 = Opera omnia (2) 13, 154-261. (248)
EuLER, L.: Genuina principia doctrinae de statu aequilibrii et motu corporum
tarn perfecte flexibilium quam elasticorum. Novi Comm. Acad. Sei. Petrop. 15
(1770), 381-413 = Opera omnia (2) 11, 37-61. (200, 214)
CouLOMB, C. A. : Essai sur une application des regles de maximis et minimis a
quelques problemes de statique, relatifs a l'architecture. Mem. divers savants 7
(1773), 343-382. (200, 204)
EULER, L.: Formulae generales pro translatione quacunque corporum rigidorum.
Novi Comm. Acad. Sei. Petrop. 20 (1775), 189-207. (168, 196)
EuLER, L.: Nova methodus motum corporum rigidorum determinandi. Novi Comm.
Acad. Sei. Petrop. 20 (1775), 208-238. (170)
EuLER, L.: De gemina methodo tarn aequilibrium quam motus corporum flexibilium determinandi et utriusque egregio consensu. Novi Comm. Acad. Sei. Petrop.
20 (1775), 286-303 = Opera omnia (2) 11, 180-193. (214)
LAGRANGE, J. L.: Memoire sur Ia theorie du mouvement des fluides. Nouv. mem.
Acad. Sei. Berlin (1781), 151-198 =Oeuvres 4, 695-748. (72, 74, 99, 134, 161,
163, 184, 206)
CousiN, J. A. J.: Memoire contenarrt quelques remarques sur Ia theorie mathematique du mouvement des fluides. Mem. Acad. Sei. Paris (1783), 665-692. (72)
LAGRANGE, J. L.: Mechanique Analitique. Paris. Oeuvres 11, 12, are the 5th ed.
Our references are to the first edititon. (6, 15, 20, 72, 74, 99, 134, 196, 232)
EULER, L. : Die Gesetze des Gleichgewichts und der Bewegung flüssiger Körper.
Leipzig (being a transl., with some changes and additions, by H. W. BRANDES,
of [1769, 1], [1770, 1], [1771, 1], and another paper). (108)
NA VIER, C.-L.-M.-H.: Sur !es lois des mouvements des fluides, en ayant egard a
l'adhesion des moh~cules. Ann. chimie 19, 244-260. (298)
FouRIER, J.: Theorie Analytique de Ia Chaleur. Paris = Oeuvres 1. (241, 296)
NA VIER, C.-L.-M.-H.: Sur ]es lois du mouvement des fluides, en ayant egard a
l'adhesion de leurs molecules. Bull. Soc. Philomath. 7 5-79. (298)
CAUCHY, A.-L.: Recherehes sur l'equilibre et le mouvement interieur des corps
solides ou fluides, elastiques ou non elastiques. Bull. Soc. Philomath. 9-13 =
Oeuvres (2) 2, 300-304. (7, 21, 26, 27, 82A, 200, 203, 298, 300, 301)
CAUCHY, A.-L.: Memoire sur une espece particuliere de mouvement des fluides.
J. Ecole Polytech. 12, cahier 19, 204-214 = Oeuvres (2) 1, 264-274. (77, 99)
NA VIER, C.-L.-M.-H.: Sur !es lois de l'equilibre et du mouvement des corpssolides
elastiques (1821). Bull. Soc. Philomath., 177-181. (Abstract of [1827, 7].) (301)
CARNOT, S.: Reflexions sur Ia Puissance Motrice du Feu et sur !es Machirres Propres
a Developper cette Puissance. Paris= Ann. Ecole Norm. (2) 1, 393-457 (1872).
See [1878, 1]. (240)
NA VIER, C.-L.-M.-H.: Memoire sur !es lois du mouvement des fluides, en ayant
egard a l'adhesion des mo!ecules (1822). Bull. Soc. Philomath. 49-52. (Abstract
of [1827, 6].) (298)
PIOLA, G.: Sull' applicazione de' principj della meccanica analitica del Lagrange
ai principali problemi. Milano: Regia Stamparia, VII + 252 pp. (17, 82, 156)
CAUCHY, A.-L.: De Ia pression ou tension dans un corps solide. Ex. de math. 2,
42-56 = Oeuvres (2) 7, 60-78. (21, 200, 203, 204, 205)
List of Works Cited. 747
2. CAUCHY, A.-L.: Sur la eondensation et la dilatation des eorps solides. Ex. de math.
2, 60-69 = Oeuvres (2) 7, 82-83. (21, 22, 26, 27, 30, 54, 57, 82A)
3. CAUCHY, A.-L.: Sur les moments d'inertie. Ex. de math. 2, 93-103 = Oeuvres
(2) 7, 124-136. (168)
4. CAUCHY, A.-L.: Sur les relations qui existent dans l'etat d'equilibre d'un eorps
solide ou fluide, entre les pressions ou tensions et les forees aeeeleratriees. Ex.
de math. 2, 108-111 = Oeuvres (2) 7, 141-145. (205)
5. CAUCHY, A.-L.: Theorie de la propagation des ondes a la surfaee d'un fluide pesant
d'une profondeur indefinie (1815). Mem. divers savants (2) 1 (1816), 3-312 =
Oeuvres (1) 1, 5-318. (20, 86A, 134, 206)
6. NAVIER, C.-L.-M.-H.: Memoire sur les lois du mouvement des fluides (1822). Mem.
Aead. Sei. lnst. Franee (2) 6, 389-440. (289)
7. NA VIER, C.-L.-M.-H.: Memoire sur les lois de l'equilibre et du mouvement des
eorps solides elastiques (1821). Mem. Aead. Sei. lnst. Franee (2) 7, 375-393- (301)
1828 1. CAUCHY, A.-L.: Sur les eentres, les plans prineipaux et les axes prineipaux des
surfaees du seeond degre. Ex. de math. 3, 1-22 = Oeuvres (2) 8, 9-35. (28)
2. CAUCHY, A.-L.: Sur les equations qui expriment !es eonditions d'equilibre, ou !es
lois du mouvement interieur d'un eorps solide, elastique, ou non elastique. Ex. de
math. 3, 160-187 = Oeuvres (2) 8, 195-226. (205, 298, 300, 301)
3. CAUCHY, A.-L.: Sur quelques theoremes relatifs a Ia eondensation Oll a Ia dilatation
des eorps. Ex. de math. 3, 237-244 = Oeuvres (2) 8, 278-287. (27, 82A)
1829 1. CAUCHY, A.-L.: Sur !es pressions ou tensions supportees en un point donne d'un
eorps solide par trois plans perpendieulaires entre eux. Ex. de math. 4, 30-40 =
Oeuvres (2) 9, 41- 52. (203)
2. CAUCHY, A.-L.: Sur Ia relation qui existe entre !es pressions ou tensions supportees
par deux plans quelconques en un point donne d'un eorps solide. Ex. de math. 4,
41-46 =Oeuvres (2) 9, 53-55. (205)
3. CAUCHY, A.-L.: Sur !es eorps solides ou fluides dans lesquels Ia eondensation ou
dilatation lineaire est la meme en tous sens autour de ehaque point. Ex. de math.
4, 214-216 = Oeuvres (2) 9, 254-258. (142)
4. CAUCHY, A.-L.: Sur l'equilibre et le mouvement interieur des eorps eonsideres
eomme des masses eontinues. Ex. de math. 4, 293-319 =Oeuvres (2) 9, 243-369.
(150, 293, 301, 304)
5. PorssoN, S.-D.: Memoire sur l'equilibre et le mouvement des eorps elastiques
(1828). Mem. Aead. Sei. Inst. Franee (2) 8, 357-570. (205, 301)
1830 1. CAUCHY, A.-L.: Sur !es diverses methodes a l'aide desquellesOll peut etablir !es
equations qui representent !es lois d'equilibre, ou le mouvement interieurdes eorps
solides ou fluides. Bull. sei. math. soe. prop. eonn. 13, 169-176. (301)
1831 1. FARADAY, M.: On a peeuliar class of optieal deeeptions. J. Roy. Inst. Gt. Brit. 1,
205-223 =Res. Chem. Phys. 291-309 (1859). (70)
2. PorssoN, S.-D.: Memoire sur !es equations generales de l'equilibre et du mouvement
des eorps solides elastiques et des fluides ( 1829). J. Ecole Polytech. 13, eahier 20,
1-174. (134, 184, 205, 298, 301)
1832 1. DuHAMEL, J .-M.-C.: Memoire sur !es equations generales de Ia propagation de Ia
ehaleur dans !es eorps solides dont Ia eonduetibilite n'est pas Ia meme dans tous
!es sens (1828). J. Eeole Polyteeh. 13, eahier 21, 356-399- (296)
1833 1. FouRIER, J.: Sur le mouvement de Ia ehaleur dans !es fluides. Mem. Aead. Sei.
Inst. Franee (2) 12, 507-530 = Oeuvres 2, 595-614. (241, 296)
2. LAME, G., et E. CLAPEYRON: Memoire sur l'equilibre des eorps solides homogenes.
Mem. divers savants (2) 4, 465-562. (21)
3. PIOLA, G.: La meeeaniea de' eorpi naturalmente estesi trattata eol ealcolo delle
variazioni. Opuse. mat. fis. di diversi autori. Milano: Giusti 1, 201-236. (29,
9.5, 210, 232)
4. PorssoN, S.-D.: Traite de Meeanique, 2nd ed. Paris. (134, 184)
1834 1. CLAPEYRON, E.: Memoire sur Ia puissanee motriee de Ia chaleur. J. Eeole Polyteeh.
14, cahier 23, 153-190. (245, 252)
1835 1. CoRIOLIS, G.: Memoire sur la maniere d'etablir !es differents principes de Ia mecanique pour des systemes de eorps, en !es eonsiderant eomme des assemblages de
moleeules. J. Eeole Polyteeh. 15, eahier 24, 93-132. (86)
2. CoRTOLIS,G.: Memoire sur !es equations du mouvement relatif des systemes de eorps.
J. Eeole Polytech. 15, eahier 24, 142-154. (143)
1836 1. ProLA, G.: Nuova analisi per tutte Je questioni della meeeaniea moleeolare. Mem.
Mat. Fis. Soe. Ital. Modena 21 (1835), 155-321. (17, 26, 72, 82A, 95, 96, 101,
109, 156, 210)
748 C. TRUESDELL and R. TouPIN: The Classical Field Theories.
1837 1. EARNSHAW, S.: On fluid motion, so far it is expressed by the equation of continuity.
Trans. Cambridge Phil. Soc. 6 (1836-1838), 203-233. (110)
2. jACOBI, C. J. J.: Über die Reduktion der Integration der partiellen Differentialgleichungen erster Ordnung zwischen irgend einer Zahl Variabeln auf die Integration eines einzigen Systems gewöhnlicher Differentialgleichungen. J. reine angew.
Math. 17, 97-162 = Werke 4, 57-127. (219)
3. MöBIUS, A. F.: Lehrbuch der Statik, 2 vols., xx + 356 + 314 pp. Leipzig:
Göschen. (219)
4. MoHR, F.: Ansichten über die Natur der Wärme. Ann. der Pharm. 24,141-147.
Trans!., P.G.TAIT, Views of the nature of heat. Phil. Mag. (5) 2 (1876), 110-114.
(240)
1839 1. GREEN, G.: On the laws of reflection and refraction of light at the common surface
of two non-crystallized media (1837). Trans. Cambridge Phil. Soc. 7 (1838-1842),
1-24 = Papers, 245-269. (33A, 218, 232A, 262, 301, 303)
2. SEGUIN, M.: De l'Influence des Chemins de Fer et de l'Art de les Tracer et de les
Construire. Paris. (240)
1841 1. CAUCHY, A.-L.: Memoire sur !es dilatations, les condensations et les rotations
produits par un changement de forme dans un systeme de points materiels. Ex.
d'an. phys. math. 2, 302-330 = Oeuvres (2) 12, 343-377. (26, 2'1, 28, 33A,
36, 56, 5'1, 85, 86, 86A)
2. GREEN, G.: On the propagation of light in crystallized media (1839). Trans. Cambridge Phil. Soc. 7 (1839-1842), 121-140 = Papers 293-311. (26, 31, 262,
301, 303)
3. J ACOBI, C. J. J.: De determinantibus functionalibus. J. reine angew. Math. 22,
319-359 = Werke 3, 393-438. (1'1)
4. LAME, G.: Memoire sur les surfaces isostatiques dans les corps solides homogenes
en equilibre d'elasticite. J. Math. pures appl. 6, 37-60. (205A, 209)
5. SVANBERG, A. F.: On fluides rörelse. K. Vetenskapsacad. handlingar (1839),
139-154. Trans!., Sur le mouvement des fluides. J. reine angew. Math. 24 ( 1842),
153-163. (99, 133)
1842 1. CHALLIS, J.: A general investigation of the differential equations applicable to the
motion of fluids. Trans. Cambridge Phil. Soc. 7 (1839-1842), 371-396. (136)
2. MAYER, J. R.: Bemerkungen über die Kräfte der unbelebten Natur. Liebigs Ann.
Chem. 42, 233-240 = pp. 3-12 of J. R. MAYER, Die Mechanik der Wärme.
Stuttgart 1867 = [1929, 8, pp. 35-42]. Trans!., G. C. FosTER, Remarks on the
forces of inorganic nature. Phil. Mag. (4) 24, 371-377 (1862); transl., G. SARTON
[1929, 8, pp.27-33]; transl., W. F. MAGIE, A Source Bookin Physics pp.197-203.
New York and London 1934. (240, 253)
3. PowER, ]. : On the truth of the hydrodynamical theorem, that if u dx + v dy + w dz be a complete differential with respect to x, y, z at any one instant, it is always
so. Trans. Cambridge Phi!. Soc. 7 (1839-1842), 455-464. (134)
4. STOKES, G. G.: On the steady motion of incompressible fluids. Trans. Cambridge
Phil. Soc. 7 (1839-1842), 439-453 = Papers 1, 1-16. (125, 126, 133, 161, 162)
1843 1. CAUCHY, A.-L.: Note sur les pressions supportees, dans un corps solide ou fluide,
par deux portians de surface tres voisine, l'une exterieure, l'autre interieure a ce
meme corps. C.R. Acad. Sei., Paris 16, 151-156 = Oeuvres (1) 1, 246-251. (205)
2. ] ouLE, ] . P.: On the caloric effects of magneto-electricity, and on the mechanical
value of heat. Phil. Mag. (3) 23, 263-276, 347-355. 435-443 = Papers 1,
123- 159. (Abstract in Rep. Brit. Assn. (1843). 33 = [1929, 8, p. 43].) (240)
3. ST. VENANT, A.-] .-C. B. DE: Memoire sur Je calcul de la resistance et de la flexion
des pieces solides a simple ou a double courbure, en prenant simultanement en
consideration !es divers efforts auxquels elles peuvent etre soumises dans tous les
sens. C. R. Acad. Sei., Paris 17, 942-954, 1020-1031. (60, 63A, 214)
4. ST. VENANT, A.-J.-C. B. DE: Note a joindre au memoire sur la dynamique des
fluides, presente Je 14 Avril 1834. C. R. Acad. Sei., Paris 17, 1240-1243. (298)
5. WATERSTON, J. J.: Note on the physical constitution of gaseaus fluids and a theory
of heat, appendix to Thoughts on the Mental Functions. Edinburgh = Papers,
183-206. (240)
1844 1. BINET, J.: Memoire sur l'integration des equations de la courbe elastique a double
courbure. C. R. Acad. Sei., Paris 18, 1115-1119. (63A)
2. J ACOBI, C. J. J.: Theoria novi multiplicatoris systemati aequationum differentialium vulgarium applicandi. J. reine angew. Math. 27, 199-268 = Werke 4,
317-394. (17)
3. ST. VENANT, A.-J .-C. B. DE: Sur les pressions qui se developpent a l'interieur des
corps solides lorsque les deplacements de leurs points, sans alterer l'elastieite, ne
4.
1845 1.
2.
3.
4.
1846 1.
1847 1.
2.
3.
1848 1.
2.
3.
4.
5.
1849 1.
2.
3.
4.
List of Works Cited. 749
peuvent cependant pas etre consideres comme tres petits. Bull. Soc. Philomath.
5, 26-28. (31, 55)
STOKES, G. G.: On some cases of fluid motion. Trans. Cambridge Phi!. Soc. 8
(1844-1849). 105-137 = Papers 1, 17-68. (114)
JoULE, J. P.: On the changes of temperature produced by rarefaction and condensation of air (1844). Phi!. Mag. (3) 26, 369-383 =Papers 1, 171-189. (240)
JOULE, J. P.: On the existence of an equivalent relation between heat and the
ordinary forms of mechanical power. Phi!. Mag. (3) 27, 205-207 = Papers 1,
202-205. (240)
ST. VENANT, A.-J.-C. B. DE: Note sur l'etat d'equilibre d'une verge elastique ä.
double courhure lorsque !es deplacements eprouves par ses points, par suite de
l'action des forces qui Ia sollicitent, ne sont pas tres petits. C. R. Acad. Sei., Paris
19, 36-44, 181-187. (63A)
STOKES, G. G.: On the theories of the internal friction of fluids in motion, and
of the equilibrium and motion of elastic solids. Trans. Cambridge Phi!. Soc. 8
(1844-1849). 287-319 =Papers 1, 75-129. (72, 82A, 86, 90, 96, 134, 142, 298)
STOKES, G. G.: Report on recent researches in hydrodynamics. Rep. Brit. Assn.
Pt. I, 1-20 = Papers 1, 151-187. (134)
HoPKINS, W.: On the internal pressure to which rock masses may be subjected,
and its possible influence in the production of the Iaminated structure. Trans.
Cambridge Phi!. Soc. 8 (1844-1849). 456-470. (203, 204)
JOULE, J. P.: On matter, living forces, and heat. Manchester Courier, May 5 and
12 = Papers 1, 265-276 = pp. 385-390 of E. C. WATSON: Joule's only general
exposition of the principle of conservation of energy. Am er. J. Phys. 15, 383-390
(1947). (240)
ST. VENANT, A.-J .-C. B. DE: Memoire sur l'equilibre des corps solides, dans !es
limites de leur elasticite, et sur !es conditions de leur resistance, quand !es deplacements ne sont pas tres petits. C. R. Acad. Sei., Paris 24, 260-263. (31, 55)
MAcCuLLAGH, J.: An essay towards a dynamical theory of crystalline reflection
and refraction (1839). Trans. Roy. Irish Acad. Sei. 21, 17-50 = Works, 145-184.
(86A, 302)
PIOLA, G.: Intorno alle equazioni fondamentali del movimento di corpi qualsivogliono, considerati secondo Ia naturale loro forma e costituzione (1845). Mem.
Mat. Fis. Soc. Ital. Modena 241, 1-186. (17, 20, 26, 72, 82, 82A, 95, 96, 156, 204,
205, 210, 232)
STOKES, G. G.: Notes on hydrodynamics (4), Demonstration of a fundamental
thcorem. Cambr. Dubl. Math. J. 3, 209-219 = (with added notes) Papers 2,
36-so. (129, 134)
STOKES, G. G.: On a difficulty in the theory of sound. Phi!. Mag. 23, 349-356.
A drastically condensed version, with an added note, appears in Papers 2, 51-55.
(7, 177, 185, 189, 205, 241, 297)
THOMSON, W. (Lord KELVIN): Notes on hydrodynamics (2), On the equation of
the bounding surface. Cambr. Dubl. Math. J. 3, 89-93 = Papers 1, 83-87.
(74, 177, 184)
HAUGHTON, S.: On the equilibrium and motion of solid and fluid bodies (1846).
Trans. Roy. Irish Acad. 21, 151-198. (205, 232A)
ScHWEINS, F.: Fliehmomente, oder die Summe L (x X+ yY) bei Kräften in der
Ebene, und L (x X+ y Y + zZ) bei Kräften im Raume. J. reine angew. Math.
38, 77-88. (219)
THOMSON, W. (Lord KELVIN): Notes on hydrodynamics (5), On the vis-viva of
a liquid in motion. Cambr. Dubl. Math. J. 4, 90-94 =Papers 1, 107-112. (94,
113, 169)
THOMSON, W. (Lord KELVIN): An account of Carnot's theory of the motive power
of heat; with numerical results deduced from Regnault's experiments on steam.
Trans. Roy. Soc. Edinb. 16, 541-574 = Ann. Chimie 35, 248-255 (1852) =Papers
1, 113-155· (245)
1850 1. CLAUSIUS, R.: Über die bewegende Kraft der Wärme und die Gesetze, welche sich
daraus für die Wärmelehre selbst ableiten lassen. Ann. Physik (3) 19, 368-398,
500-524 = Abh. 1, 16-78. Trans!., On the moving force of heat, and the laws
regarding the nature of heat itself which are dedueible therefrom. Phi!. Mag. (4)
2 (1851). 1-21, 102-119; transl., W. F. MAGIE, On the motivepower of heat,
and on the laws which can be deduced from it for the theory of heat, pp. 65-107
of The Second Law of Thermodynamics. New York and London 1899. (245, 248,
252, 253)
750 C. TRUESDELL and R. TouPIN: The Classical Field Theories.
2. KIRCHHOFF, G.: Über das Gleichgewicht und die Bewegung einer elastischen Scheibe.
J. reine angew. Math. 40, 51-88 = Ges. Abh. 237-279. (232A, 262)
3. THOMSON, W. (Lord KELVIN): Remarks on the forces experienced by inductively
magnetized ferromagnetic or diamagnetic non-crystalline substances. Phi!. Mag.
(3) 37, 241-253 = Papers Electr. Magn., §§ 647-668. (102)
1851 1. RANKINE, W. J. M.: Laws of the elasticity of solid bodies. Cambr. Dubl. Math. J.
6, 41-80, 178-181, 185-186 = Papers 67-101. (25)
2. STOKES, G. G.: On the effect of the internal friction of fluids on the motion of pendulums (1850). Trans. Cambridge Phi!. Soc. 92, 8-106 =Papers 3, 1-141. (72,
170, 217, 298)
3. STOKES, G. G.: On the conduction of heat in crystals. Cambr. Dubl. Math. J. 6,
215-238=Papers 3, 203-227. (241)
1852 1. KIRcHHOFF, G.: Über die Gleichungen des Gleichgewichts eines elastischen Kör·
pers bei nicht unendlich kleinen Verschiebungen seiner Theile. Sitzgsber. Akad.
Wiss. Wien 9, 762-773. (Not repr. in Abh.) (19, 210, 217, 236A, 256A, 262, 303)
2. LAME, G.: Le~ons sur Ia Theorie Mathematique de l'Elasticite. Also 2nd ed.,
Paris 1866. (21)
1853 1. RANKINE, W. J. M.: On the mechanical action of heat I-V (1850-1851). Trans.
Roy. Soc. Edinb. 20, 147-210. (245)
2. REECH, M. F.: Theorie generale des effets dynamiques de Ia chaleur. J. Math.
pures appl. 18, 357-568. (245, 252)
3. THOMSON, W. (Lord KELVIN): On the dynamical theory of heat I-V (1851).
Trans. Roy. Soc. Edinb. 20, 261-293, 475-483 =Phi!. Mag. (4) 4, (1852) 8-20,
105-117, 168-176, 424-434, 9 (1855), 523-531 = Papers 1, 174-232. Pts.
I-III = pp. 111-146 of The Second Law of Thermodynamics, ed. W. F. MAGIE.
New York and London 1899. (245)
1854 1. CLAUSIUS, R.: Über eine veränderte Form des zweiten Hauptsatzes der mechanischen Wärmetheorie. Ann. Physik 93, 481-506 = Abh. 1, 126-154. (245, 248,
249, 252, 258)
2. ScHWEINS, F.: Theorie der Dreh- und Fliehmomente der parallelen Seitenkräfte,
in welche Kräfte im Raume zerlegt werden können. J. reine angew. Math. 47,
238-245. (219)
1855 1. FICK, A.: Über Diffusion. Ann. der Phys. 94, 59-86. (158, 295)
2. HAUGHTON, S.: On a classification of elastic media, and the laws of plane waves
propagated through them ( 1849). Trans. Roy. Irish Acad. Pt. I-Science 22, 97-138.
(205, 232)
3. MEISSEL, E.: Über einen speciellen Fall des Ausflusses von Wasser in einer verticalen Ebene. Ann. Physik (4) 5, 276-283. (161)
4. THOMSON, W. (Lord KELVIN): On the thermo-elastic and thermo-magnetic properties of matter. Quart. J. Math. 1 (1855-1857) 55-77 = (with notes and
additions) Phi!. Mag. (5) 5 (1878), 4-27 = Pt. VII of On the dynamical theory
of heat. Papers 1, 291-316. (218, 262)
1856 1. RANKINE, W. J. M.: On axes of elasticity and crystalline forms. Phi!. Trans. Roy.
Soc. Lond. (146) 46, 261-285 = Papers 119-149. (200)
2. THOMSON, W. (Lord KELVIN): Elements of a mathematical theory of elasticity.
Phi!. Trans. Roy. Soc. Lond. 146, 481-498. (See [1877, 5].) (262)
1857 1. CLEBSCH, A.: Über eine allgemeine Transformation der hydrodynamischen Gleichungen. J. reine angew. Math. 54, 293-312. (125, 136, 137, 162, 210)
1858 1. HELMHOLTZ, H.: Über Integrale der hydrodynamischen Gleichungen, welche den
Wirbelbewegungen entsprechen. J. reine angew. Math. 55, 25-55 = Wiss. Abh.
1, 101-134. Trans!., P. G. TAIT, On integrals of the hydrodynamical equations,
which express vortex-motion. Phi!. Mag. (4) 33, (1867) 485-512. (8, 75, 88, 93,
106, 108, 113, 133, 175, 185, 186, 218)
1859 1. CLEBSCH, A.: Über die Integration der hydrodynamischen Gleichungen. J. reine
angew. Math. 56, 1-10. (136, 136A)
2. KIRCHHOFF, G.: Über das Gleichgewicht und die Bewegung eines unendlich dünnen elastischen Stabes. J. reine angew. Math. 56,285-313 = Ges. Abh., 285-316.
(57, 63A, 214, 236A, 301)
3. LAME, G.: Le~ons sur !es Coordonnees Curvilignes et leurs divers Applications.
Paris: Mallet-Bachelier. xxvii + 368 pp. (205 A, 209)
1860 1. DIRICHLET, G. LEJEUNE: Untersuchungen über ein Problem der Hydrodynamik.
Gött. Abh., math. Cl. 8 (1858-1859), 3-42 = J. reine angew. Math. 58, 181-216
(1861) = Werke 2, 263-301. (66)
2. MAXWELL, J. C.: Illustrations of the dynamical theory of gases. Phi!. Mag. (4) 19,
19-32 and 20,21-37 =Papers 1, 377-409. (295)
3.
4.
1861 1.
2.
1862 1.
2.
3.
4.
5.
1863 1.
2.
1864 1.
2.
3.
4.
5.
1865 1.
2.
1866 1.
2.
3.
4.
1867 1.
2.
3.
1868 1.
2.
3.
4.
List of Works Cited. 751
NEUMANN, C.: Zur Theorie der Elasticität. J. reine angew. Math. 57, 281-318
(210)
RIEMANN, B.: Ueber die Fortpflanzung ebener Luftwellen von endlicher Schwingungsweite. Gött. Abh., math. Cl. 8 (1858-1859), 43-65 = Werke, 145-164.
(185, 189, 205)
HANKEL, H.: Zur allgemeinen Theorie der Bewegung der Flüssigkeiten. Göttingen.
(66A, 88, 108, 128, 136)
WARREN, J. W.: On invariant points, lines, and surfaces in space, and their physical
significance. Amer. J. Math. 4, 306-310. (38)
CLAUSIUS, R.: Ueber die Anwendung des Satzes von der Aequivalenz der Verwandlungen auf die innere Arbeit. Vjschr. nat. Ges. Zürich 7, 48-95 = Ann.
Physik 116, 73-112 = Abh. 1, 242-279. (245, 258)
CLEBSCH, A.: Theorie der Elasticität fester Körper. Leipzig. (63A, 214)
EULER, L.: Fragmentum 97, Opera postuma 1, 494-496 = Opera omnia (1) 29,
437-440. (33A)
EULER, L.: Anleitung zur Natur-Lehre, worin die Gründe zu Erklärung aller in
der Natur sich ereignenden Begebenheiten und Veränderungen festgesetzt werden
(probably written between 1755 and 1759). Opera postuma 2, 449-560 = Opera
omnia (3) 1, 16-178. (76)
W ARREN, J. W.: On the internal pressure within an elastic solid. Quart. J. Math.
5, 109-117. (21)
AIRY, G. B.: On the strains in the interior of beams. Phi!. Trans. Roy. Soc. Lond.
153, 49-80. Abstract in Rep. Brit. Assn. 1862, 82-86 (1863). (224)
THOMSON, W. (Lord KELVIN}: Dynamical problems regarding elastic spheroidal
shells and spheroids of incompressible liquid. Phi!. Trans. Roy. Soc. Lond. 153,
583-616 =Papers 3, 351-394. (262, 301, 303}
JouLE, J. P.: Note on the history of the dynamical theory of heat. Phi!. Mag.
(4) 28, 150-151. (240)
RANKINE, W. J. M.: On plane water-lines in two dimensions. Phi!. Trans. Roy.
Soc. Lond. 154, 369-391 = Papers 495-521. (71, 161)
ST. VENANT, A.-J .-C. B. DE: Etablissement elementaire des formules et equations
generales de Ia theorie de l'elasticite des corps solides. Appendix in: Resurne des
Lec;:ons donnees a l'Ecole des Ponts et Chaussees sur l'Application de Ia Mecanique,
premiere partie, premiere section, De Ia Resistance des Corps Solides, par C.-L.-
M.-H. NA VIER, 3rd. ed. Paris. (57)
ST. VENANT, A.-J .-C. B. DE: Theorie de l'elasticite des solides, ou cinematique de
leurs deformations. L'Institut 321, 389-390. (27)
WARREN, J. W.: Note on a transformation of the generat equation of wave propagation, due to internal force. Quart. J. Math. 6, 137-139. (209)
CLAUSIUS, R.: Ueber verschiedene für die Anwendung bequeme Formen der Hauptgleichungen der mechanischen Wärmetheorie. Vjschr. nat. Ges. Zürich 10, 1-59
= Ann. der Phys. 125. 353-400 = Abh. 2, 1-44. (245, 246, 258, 264)
THOMSON, W. (Lord KELVIN}: On the elasticity and viscosity of metals. Proc.
Roy. Soc. Lond. 14, 289-297. (See [1877, 5].)
DARBOUX, G.: Sur les surfaces orthogonales. Ann. Ecole Norm. (1) 3, 97-141.
(122)
]ACOBI, C. J. J.: Vorlesungen über Dynamik (1842-1843). Werke, Suppl. (219).
LIPSCHITZ, R.: Ueber einen algebraischen Typus der Bedingungen eines bewegten
Massensystems. J. reine angew. Math. 66, 363-374. (219)
V ERDET, E.: Introduction, Oeuvres de Fresnel 1, IX-XCIX. (301)
BERTRAND, J.: Theoreme relatif au mouvement le plus generat d'un fluide. C. R.
Acad. Sei., Paris 66, 1127-1230. (78, 86, 86A, 89)
MAXWELL, J. C.: On the dynamical theory of gases ( 1866). Phi!. Trans. Roy.
Soc. Lond. 157, 49-88 = Phi!. Mag. (4} 35 (1868), 129-145, 185-217 =Papers,
2, 26-78. (215, 243, 295, 305)
THOMSON, W. (Lord KELVIN}, and P. G. TAIT: Treatise on Natural Philosophy.
Part I. Cambridge. (8, 12, 21, 27, 35, 36, 38, 42, 45, 46, 167, 205, 262)
Anonymous: Les Mondes 17, 620-623. (89)
BERTRAND, J.: Note relative a Ia theorie des fluides. Reponse a Ia communication
de M. Helmholtz. C. R. Acad. Sei., Paris 67, 267-269. (86A)
BERTRAND, J.: Observations nouvelles sur un memoire de M. Helmholtz. C. R.
Acad. Sei., Paris 67, 469-472. (86A)
BERTRAND, J.: Reponse a Ia note de M. Helmholtz. c. R. Acad. Sei., Paris 67,
773-775- (86A)
752 C. TRUESDELL and R. TouPrN: The Classical Field Theories.
5. FRESNEL, A.: Premiere memoire sur Ia double refraction (1821). Oeuvres 2,
261-308. (208, 209)
6. FRESNEL, A.: Supplement au memoire sur la double refraction (1822). Oeuvres
2, 343-367. (21, 208, 209)
7. FRESNEL, A.: Second supplement au memoire sur Ia double refraction (1822).
Oeuvres 2, 369-442. (21, 203, 205, 208, 209)
8. HELMHOLTZ, H.: Sur le mouvement le plus general d'un fluide. Reponse a une
communication precedente de M. J. Bertrand. C. R. Acad. Sei., Paris 67, 221-225
= Wiss. Abh. 1, 135-139. (86A)
9. HELMHOLTZ, H.: Sur le mouvement des fluides. Deuxieme reponse a M. J. Bertrand.
C. R. Acad. Sei., Paris 67, 754-757 = Wiss. Abh. 1, 140-144. (86A)
10. HELMHOL TZ, H.: Reponse a Ia note de M. J. Bertrand, du 19 Octobre. c. R. Acad.
Sei., Paris 67, 1034-1035 = Wiss. Abh. 1, 145. (86A)
11. KIRCHHOFF, G.: Ueber den Einfluß der Wärmeleitung in einem Gase auf die Schallbewegung. Ann. Physik 134, 177-193 = Ges. Abh. 1, 540-556. (241, 248)
12. MAXWELL, J. C.: On reciprocal diagrams in space, and their relation to Airy's
function of stress. Proc. Lond. Math. Soc. (1) 2 (1865-1969), 58-60 = Papers
2, 102-104. (227)
13. RIEMANN, B.: Über die Hypothesen, welche der Geometrie zu Grunde liegen (1854).
Abh. Ges. Wiss. Göttingen 13, (1866-1867) 133-150 =Werke, 2nd. ed. (1892),
272-287. (34)
14. TAIT, P. G.: Sketch of Thermodynamics. London. (240)
15. WALTER, A.: Anwendung der Methode Hamiltons auf die Grundgleichungen der
mathematischen Theorie der Elasticität. Diss. Berlin. (We have not been able
to see this work.) (236)
16. WARREN, J.: Theorem with regard to the three axes of invariable direction in a
strained elastic body. Quart. J. Math. 9, 171-172. (38)
17. WEBER, H.: Über eine Transformation der hydrodynamischen Gleichungen. J.
reine angew. Math. 68, 286-292. (135)
1869 1. CHRlSTOFFEL, E. B.: Über die Transformation der homogenen Differentialausdrücke
2ten Grades. J. reine angew. Math. 70, 46-70 = Abh. 1, 352-377 (34)
2. CHRISTOFFEL, E. B.: Über ein die Transformation homogener Differentialausdrücke
zweiten Grades betreffendes Theorem. J. reine angew. Math. 70, 241 -24 5 =
Abh. 1, 378-382 (34)
3. LIPSCHITZ, R.: Untersuchungen in Betreff der ganzen homogenen Functionen von
n Variablen. J. reine angew. Math. 70, 71-102. (34)
4. MASSIEU, F.: Sur les fonctions caracteristiques des divers fluides. C. R. Acad.
Sei., Paris 69, 858-862. (251)
5. MAssrEu, F.: Addition au precedent memoire sur les fonctions caracteristiques.
C. R. Acad. Sei., Paris 69, 1057-1061. (251)
6. ST. VENANT, A.-J .-C. B. DE: Probleme des mouvements que peuvent prendre les
divers points d'une liquide, ou solide ductile, contenue dans un vase a parois verticales, pendant son ecoulement par un orifice horizontal interieur. C. R. Acad. Sei.,
Paris 68, 221-237. (89, 134)
7. THOMSON, W. (Lord KELVIN): On vortex motion. Trans. Roy. Soc. Edinb. 25,
217-260 = Papers 4, 13-66. (79, 88, 105, 106, 108, 128, 129, 297)
1870 1. CLAUSIUS, R.: Über einen auf die Wärme anwendbaren mechanischen Satz. Ann.
Physik u. Chem. (5) 21 = 141, 124-130. (219)
2. LIPSCHITZ, R.: Fortgesetzte Untersuchung in Betreff der ganzen homogenen Funktionen von n Differentialen. J. reine angew. Math. 72, 1-65. (34)
3. L:Evv, M.: Memoire sur les equations generales des mouvements interieurs des
corps solides ductiles au dela des limites ou l'elasticite pourrait les ramener a leur
premier etat. C. R. Acad. Sei., Paris 70, 1323-1325. (300)
4. MAXWELL, J. C.: On reciprocal figures, frames, and diagrams of forces. Trans.
Roy. Soc. Edinb. 26 (1869-1872), 1-40 = Papers 2, 161-207. (216, 224, 227)
5. MAXWELL, J. C.: On the displacement in a case of fluid motion. Proc. Lond.
Math. Soc. 3 (1869-1871), 82-87 = Papers 2, 208-214. (71)
6. RANKINE, W. J. M.: On the thermodynamic theory of waves of finite longitudinal
disturbance. Phil. Trans. Roy. Soc. Lond. 160, 277-288 = Paptrs, 530-543.
(183, 189, 205, 241, 297))
7. ST. VENANT, A.-J.-C. B. DE: Memoire sur l'etablissement des equations differentielles des mouvements interieurs operes dans les corps solides ductiles au dela
des limites ou l'elasticite pourrait les ramener a leur premier etat. C. R. Acad.
Sei., Paris 70, 473-480. (300)
List of Works Cited. 753
8. WARREN, J.: Note on a fundamental theorem in hydrodynamics. Quart. J. Math.
10, 128-129. (101)
1871 1. BELTRAMI, E.: Sui principi fondamentali della idrodinamica. Mem. Acad. Sei.
Bologna (3) 1, 431-476; 2 (1872). 381-437; 3 (1873), 349-407; 5 (1874). 443-484
= Richerche sulla cinematica dei fluidi. Opere 2, 202-379. (78, 82A, 83, 84.
86, 101, 105, 129, 136)
2. BoussiNESQ, J.: Etude nouvelle sur l'equilibre et le mouvement des corps solides
elastiques dont certaines dimensions sont tres petites par rapport a d'autres.
Premier Memoire. J. Math. pures appl. (2) 16, 125-240. (57)
3. CHERBULIEZ, E.: Geschichtliche Mittheilungen aus dem Gebiete der mechanischen
Wärmetheorie. Mittheil. Naturf. Ges. Bern 1870, 291-324. (240)
4. DuRRANDE, H.: Extrait d'une theorie du deplacement d'une figure qui se deforme.
C. R. Acad. Sei., Paris 73, 736-738. (82A)
5. MAXWELL, J. C.: Theory of Heat. London. (252)
6. STEFAN, J.: Über das Gleichgewicht und die Bewegung, insbesondere die Diffusion
von Gasmengen. Sitzgsber. Akad. Wiss. Wien 632, 63-124. (158, 295)
1872 1. BoussiNESQ, J.: Theorie des ondes liquides periodiques (1869). Mem. divers sav.
20, 509-615. (210)
2. Rapport sur un memoire deM. Kleitz intitule: .,Etudes sur !es forces moleculaires
dans !es liquides en mouvement, et application a l'hydrodynamique" (ST. VENANT).
C. R. Acad. Sei., Paris 74, 426-438. (82A, 208)
3. LIPSCHITZ, R.: Extrait d'une lettre. Bull. sei. math. astr. (1) 3, 349-352. (219)
4. VILLARCEAU, Y.: Sur un nouveau theoreme de mecanique generale. C. R. Acad.
Sei., Paris 75, 232-240. (219)
1873 1. BoBYLEW, D.: Einige Betrachtungen über die Gleichungen der Hydrodynamik.
Math. Ann. 6, 72-84. (211)
2. GIBBS, J. W.: Graphical methods in the thermodynamics of fluids. Trans. Connecticut Acad. 2, 309-342 = Works 1, 1-32. (245, 247, 248, 265)
3. GIBBS, J. W.: A method of geometrical representation of the thermodynamic
properties of substances by means of surfaces. Trans. Connecticut Acad. 2,
382-404 = Works 1, 33-54. (245, 246, 247)
4. KLEITZ, C.: Etudes sur !es forces moleculaires dans !es liquides en mouvement et
application al'hydrodynamique. Paris. (Cf. [1872, 2].) (82A, 110, 204, 208, 293).
5. MAXWELL, J. C.: A Treatise on Electricity and Magnetism. 2 vols., Oxford. Cf.
[1881, 4]. (175, 187, 200, 276, 311)
6. STRUTT, J. W. (RAYLEIGH): Some general theorems relating to vibrations. Proc.
Lond. Math. Soc. 4 (1871-1873). 357-368 = Papers 1, 170-181. (241A)
1874 1. BoLTZMANN, L.: Zur Theorie der elastischen Nachwirkung. Sitzungsber. Akad.
Wiss. Wien 702, 275-306 = Wiss. Abh. 1, 616-639. (305)
2. MAXWELL, J. C.: V an d·~r Waals on the continuity of the gaseous and liquid states.
Nature 10, 477-480 = Papers 2, 407-415. (219)
3. MEYER, 0.-E.: Zur Theorie der inneren Reibung. J. reine angew. Math. 78, 130
to 135. (305)
4. MEYER, 0.-E.: Theorie der elastischen Nachwirkung. Ann. Physik (6) 1, 108-119.
(305)
5. NANSON, E. J.: Note on hydrodynamics. Mess. math. 3, 120-121. (75, 108)
6. UMov, N.: Ableitung der Bewegungsgleichungen der Energie in continuirlichen
Körpern. Z. Math. Phys. 19, 418-431. (217)
187 5 1. GIBBS, J. W.: On the equilibrium of heterogeneaus substances. Trans. Connecticut
Acad. 3 (1875-1878), 108-248, 343-524 = Works 1, 55-353. (14, 245, 246,
247, 251, 260, 262, 264, 265)
2. GREENHILL, A. G.: Hydromechanics. Encycl. Britt., 9th ed. (carried through
14th ed., 1926). (99, 207)
3. LIPSCHITZ, R.: Determinazione della pressione nell' intorno d'un fluido incompressibile soggetto ad attrazioni interne ed esterne. Ann. mat. (2) 6 (1873-1875).
226-231. (102, 141)
4. MAXWELL, J. C.: 4th. ed. of [1871, 5]. (245, 246)
5. MEYER, 0.-E.: Zusatz zu der Abhandlung zur Theorie der inneren Reibung. J. reine
angew. Math. 80, 315-316. (305)
1876 1. CoTTERILL, J.: On the distribution of energy in a mass of liquid in a state of steady
motion. Phil. Mag. 1, 108-111. (125)
2. KIRCHHOFF, G.: Vorlesungen über mathematische Physik: Mechanik. Leipzig.
2nd ed., 1877; 3rd ed., 1883. (57, 63A, 113, 129, 196, 200, 205, 214, 232, 236,
301)
Handbuch der Physik, Bd. 111/1. 48
754 C. TRUESDELL and R. ToUPIN: The Classical Field Theories.
3. MASSIEU, F.: Memoire sur les fonctions caracteristiques de divers fluides et sur Ia
theorie des vapeurs. Mem. divers savants 22, No. 2, 92 pp. (251)
4. MAXWELL, J. C.: On the equilibrium of heterogeneaus substances. Repts. South
Kensington conf. spec. loan coll. sei. app. 144-150 = (with a note added by
LARMOR) Phil. Mag. (6) 16, 818-824 (1908). Condensed version, Proc. Cambridge
Phil. Soc. 2, 427-430 (1876) =Papers 2, 498-500. (260)
5. RIEMANN, B.: Commentatio mathematica, qua respondere tentatur quaestioni ab
mma Academia Parisiensi propositae. Werke, 370-399 = Werke, 2nd ed. (1892),
391-404. (34)
6. TAIT, P. G.: Recent Advances in Physical Science. London. 1st and 2nd eds.
(240)
7. ZHUKOVSKI, N. E.: Kinematics of fluid bodies [in Russian]. Mat. Sborn. 8, 1-79.
163-238 = repr. separ., Moscow 1876 = Sobr. Soch. 2, 7-144. (8, 12, 70, 102,
108)
1877 1. BoussiNESQ, J.: Sur la construction geometrique des pressions que supportent
les divers elements plans se croisant en un meme point d'un corps, et sur celle
des deformations qui se produisent autour d'un tel point. J. Math. pures appl.
(3) 3, 147-152. (23, 35, 208)
2. CHRISTOFFEL, E. B.: Untersuchungen über die mit dem Fortbestehen linearer partieller Differentialgleichungen verträglichen Unstetigkeiten. Ann. Mat. (2) 8,
81-113 = Abh. 2, 51-80. (173, 177, 180, 183, 189, 205, 297)
3. CHRISTOFFEL, E. B.: Über die Fortpflanzung von Stößen durch elastische feste
Körper. Ann. Mat. (2) 8, 193-244 = Abh. 1, 81-126. (301)
4. LAMB, H.: Note on a theorem in hydrodynamics. Mess. math. 7 (1877-1878),
41-42. (76) - 5. THOMSON, W. (Lord KELVIN): Elasticity, Encycl. Britt., 9th ed. = (with reVisions)
Papers 3, 1-112. The first part is a revision of [ 1865, 2]; the second part, a revision of [1856, 2]. (33A)
1878 1. Reprint of [1824, 1], with the addition of MS notes of CARNOT (1824-1832) Paris.
Trans!., R. H. THURSTON, Reflections on the Motive Power of Heat. New York
1890; newed., NewYork 1943. Trans!., W. F. MAGIE, pp. 2-60 of The Second Law
of Thermodynamics. N ew Y ork and London 1899. The most important portians
of the MS notes = [1929, 8, pp. 33-34, 44]. (240)
2. CLIFFORD, W. K.: Elements of Dynamic, an Introduction to the Study of Motion
and Rest in Solid and Fluid Bodies, I. Kinematic. London. (86A)
3. GILBERT, P.: Sur le probl~me de la composition des accelerations d'ordre quelconque.
C. R. Acad. Sei., Paris 86, 1390-1391. (143)
4. LAISANT, C.-A.: Note sur un theor~me sur les mouvements relatifs. C. R. Acad.
Sei., Paris 87, 204-206. Cf. remarks on pp. 259 and 377. (143)
5. LAMB, H.: On the conditions for steady motion of a fluid. Proc. Lond. Math.
Soc. 9, (1877-1878). 91-92. (122, 125, 133)
6. L:Evv, M.: Sur la composition des accelerations d'ordre quelconque et sur un problerne plus generale que celui de la composition des mouvements. C. R. Acad.
Sei., Paris 86, 1068-1071. (143)
7. MüLLER, J.: Einleitung in die Hydrodynamik. Vjschr. naturf. Ges. Zürich 23,
129-159. 242-265. (107)
8. NANSON, E. J.: Note on hydrodynamics. Mess. Math. 7 (for 1877-1878), 182-185.
(20)
1879 1. FoRSYTH, A. R.: On the motion of a viscous incompressible fluid. Mess. Math.
9 (1879-1880), 134-139. (121)
2. LAMB, H.: A Treatise on the Mathematical Theory of the Motion of Fluids. Cambridge. (See [1895, 2], [1906, 4] [1916, 3], [1924, 9], [1932, 8].) (117, 122, 125)
3. OBERDECK, A.: Über die Wärmeleitung der Flüssigkeiten bei Berücksichtigung
der Strömungen infolge von Temperaturdifferenzen. Ann. Physik (2) 7, 271-292.
(241)
4. SHILLER, N.: Certain applications of the mechanical theory of heat to the changes
of state of elastic bodies [in Russian]. Zh. Rus. Fiz.-Chim. Obsch. 11, Fiz. Otd.
55-77. (246)
1880 1. BoussiNESQ, J.: Sur la mani~re dont les frottements entrent en jeu dans un fluide
qui sort de l'etat de repos, et sur leureffet pour empecher l'existence d'une fonction
des vitesses. C. R. Acad. Sei., Paris 90, 736-739. (134)
2. BousSINESQ, J.: Quelques considerations a l'appui d'une note du 29 mars, sur
l'impossibilite d'admettre, en general, une fonction des vitesses danstaute question
d'hydraulique ou les frottements ont un röle notable. C. R. Acad. Sei., Paris 9,
967-969. (134)
List of Works Cited. 755
3. BRESSE, J.: Fonction des vitesses; extension des theoremes de Lagrange au cas
d'un fluide imparfait. C. R. Acad. Sei., Paris 90, 501- 504. (134}
4. BRESSE, J.: Reponse a une note de M.J.Boussinesq. C. R. Acad. Sei., Paris 90,
857-858. (134)
5. CRAIG, T.: Motion of viscous fluids. J. Franklin Inst. 110, 217-227. (121, 125)
6. CRAIG, T.: On steady motion in an incompressible viscous fluid. Phi!. Mag. (5)
10, 342-357- (125)
7. CRAIG, T.: On certain possible cases of steady motion in a viscous fluid. Amer.
J. Math. 3, 269-293. (121)
8. FITZGERALD, G. F.: On the electromagnetic theory of the reflection and refraction
of light. Phi!. Trans. Roy. Soc. Lond. 171, 691 -711. (302)
9. RoWLAND, H. A.: On the motion of a perfect incompressible fluid when no solid
bodies are present. Amer. J. Math. 3, 226-268. (121)
10. Sr. VENANT, A.-J .-C. B. DE: Geometrie cinematique. - Sur celle des deformations
des corps soit elastiques, soit plastiques, soit fluides. C. R. Acad. Sei., Paris 90,
53-56. 209. (8, 27)
1881 1. BELTRAMI, E.: Sulle equazioni generali dell' elasticita. Ann. Mat. (2) 10 (1880
-1882}, 188-211 = Opere 3, 383-407. (232, 238)
2. CRAIG, T.: Methods and Results, General Properties of the Equations of Steady
Motion. U.S. Treasury Dept. (Coast and Geod. Survey) Document 71, Washington.
(122, 136)
3. HILL, M. J. M.: Some properties of the equations of hydrodynamics. Quart. J.
Math. 17, 1-20, 168-174. (66, 136 136A, 137)
4. MAXWELL, J. C.: 2nd ed. of [1873, 5]. (175, 187, 276)
1882 •: 1. HELMHOLTZ, H.: Zur Geschichte der Entdeckung des Gesetzes von der Erhaltung
der Kraft. Abh. 1, 71-74. (240)
2. HELMHOL TZ, H.: Die Thermodynamik chemischer Vorgänge. Sitzgsber. Akad.
Wiss. Berlin 1882, 22-39 = Abh. 2, 958-978. (246, 249, 251, 263, 265)
3. MoHR, 0.: Über die Darstellung des Spannungszustandes und des Deformationszustandes eines Körperelementes und über die Anwendung derselben in der Festigkeitslehre. Civilingenieur 28, 112-155. (24)
4. VorGT, W.: Allgemeine Formeln für die Bestimmung der Elasticitätsconstanten
von Krystallen durch die Beobachtung der Biegung und Drillung von Prismen.
Ann. Physik {2) 16, 273-321, 398-416. (224)
1884 J. GrLBERT, P.: Abstract. Ann. Soc. Sei. Bruxelles 8 {1883-1884}, A53-A56. (143}
1885 1. BoBYLEW, D.: Über die relative Bewegung eines Punktes in einem in continuirlicher Deformation begriffenen Medium. Z. Math. Phys. 30, 336-344. (144)
2. HrLL, M. J. M.: The differential equations of cylindrical annular vortices. Proc.
Lond. Math. Soc. 16 {1884-1885). 171-183. (161, 162)
3. HuGONIOT, H.: Sur Ia propagation du mouvement dans un fluide indefini (Premiere
Partie). C. R. Acad. Sei., Paris 101, 1118-1120. (177, 180, 181, 183, 190, 297)
4. HuGONIOT, H.: Sur Ia propagation du mouvement dans un fluide indefini (Deuxieme
Partie). C. R. Acad. Sei., Paris 101, 1229-1232. (177, 181, 182, 183, 190, 246, 297)
5. LAMB, H.: Proof of a hydrodynamical theorem. Mess. Math. 14, 87-92. (138)
6. MoRERA, G.: Sulle equazioni generali per l'equilibrio dei sistemi continui a tre
dimensioni. Atti Accad. Sei. Torino 20 (1884-1885), 43-53. (205A) 209, 232)
7. SoMOFF, P.: Über einen Satz von Burmester. Z. Math. Phys. 30, 248-250. (74)
8. THOMSON, W. (Lord KELVIN}: On the motion of a liquid within an ellipsoidal
hollow. Proc. Roy. Soc. Edinb. 13, 114, 370-378 = Papers 4, 193-201. (70)
1886 1. BELTRAMI, E.: Sull' interpretazione meccanica delle formale di Maxwell. Mem.
Accad. Sei. Bologna {4} 7, 1-38 = Opere 4, 190-223. (57)
2. DUHEM, P.: Le Potentiel Thermodynamique et ses Applications a Ia Mecanique
Chimique et a l'etude des phenomenes electriques. Paris: Hermann. (246, 252,
260)
3. HuGONIOT, H.: Sur un theoreme general relatif a la propagation du mouvement.
C.R. Acad. Sei., Paris 102, 858-860. (194A, 301)
4. ToDHUNTER, 1., and K. PEARSON: A History of the Theory of Elasticity and of
the Strength of Materials from Galilei to the Present Time, 1. Cambridge. (21,
33A, 45, 204, 205A, 232)
188 7 1. HuGoNroT, H.: Memoire sur Ia propagation du mouvement dans !es corps et specialerneut dans !es gaz parfaits. J. Ecole Polytech. 57, 3-97; 58, 1-125 (1889). The
date of the memoir is 26 October 1885. (180, 205, 241, 297)
2. HuGONIOT, H.: Memoire sur Ia propagation du mouvement dans un fluide indefini.
J. Math. pures appl. (4} 3, 477-492; {4) 4, 153-167. (183, 190, 194A, 297)
48*
756
3.
4.
5.
1888 1.
2.
3.
4.
5.
6.
7.
8.
C. TRUESDELL and R. TouPrN: The Classical Field Theories.
PLANCK, M.: Über das Princip der Vermehrung der Entropie. Dritte Abhandlung.
Ann. Physik (2) 32, 162-503. (264)
REIFF, R.: Zur Kinematik der Potentialbewegung. Math. naturw. Mitt. (Tübingen)
13 (1884-1886). 41-48. (88)
VorGT, W.: Theoretische Studien über die Elasticitätsverhältnisse der Krystalle.
Abh. Ges. Wiss. Göttingen 34, 100 pp. (200, 217)
BASSET, A. B.: A Treatise on Hydrodynamics, 2 vols. Cambridge. (99, 125, 136,
136A, 137, 162)
BRILLOUIN, M.: Chaleur specifique pour une transformation quelconque et thermodynamique. J. phys. theor. appl. (2) 7, 148-152. (250)
BRILLOUIN, M.: Deformations permanentes et thermodynamique. J. phys. theor.
appl. (2) 7, 327-347; 8, 169-179 (1889). Much of the first part appears in C. R.
Acad. Sei., Paris 106, 416-418, 482-485, 537-540, 589-592. (250)
DoNATI, L.: Sul lavoro di deformazione dei sistemi elastici. Mem. Accad. Sei.
Bologna (4) 9, 345-367. (205)
GrLBERT, P.: Sur !es composantes des accelerations d'ordre quelconque suivant
trois directions rectangulaires variables. J. Math. pures appl. (4) 4, 465-473. (143)
LovE, A. E. H.: The small free vibrations and deformation of a thin elastic shell.
Phil. Trans. Roy. Soc. Lond. A 179, 491-546. (212)
Rrccr, G.: Delle derivazioni covarianti e controvarianti e delloro uso nella analisi
applicata. Studi Univ. Fadova comm. ottavo centenar. Univ. Bologna 3, No. 12,
23 pp. (84)
RrECKE, E.: Beiträge zur Hydrodynamik. Nachr. Ges. Wiss. Göttingen, 347-357
= Ann. Physik (2) 36, 322-334 (1889). (71)
1889 1. BELTRAMI, E.: Considerazione idrodinamiche. Rend. Ist. Lombardo (2) 22,
121-130 = Opere 4, 300-309. (99)
1890
1891
2. BELTRAMI, E.: Sur Ia theorie de Ia deformation infiniment petite d'un milieu.
C. R. Acad. Sei., Paris 108, 502-504 = Opere 4, 344-347. (57)
3. GrLBERT, P.: Recherehes sur les accelerations en general. Ann. Soc. Sei. Bruxelles
13,261-315. (143)
4. LEONARDO DA V IN er: Les ... manuscrits F and I de Ia bibliotheque de l'institut
publies en fac-similes phototypiques ... , ed. RAVAISSON-MOLLIEN. Paris. (70)
5. LIPPMANN, M.: Cours de Thermodynamique. Paris. (240)
6. MoRERA, G.: Sui moti elicoidali dei fluidi. Rend. Lincei (4) 5, 511-617. (99, 112)
7. PADOVA, E.: Sulle deformazioni infinitesimi. Rend. Lincei (4) 51, 174-178. (34,
57)
8. PorNCARE, H.: Let;:ons sur la Theorie Mathematique de la Lumiere, redigees par
J. BLONDIN. Paris. (233)
9. VOIGT, W.: Ueber adiabatische Elasticitätsconstanten. Ann. Physik 36, 743-7 59.
(256A, 296)
10. VorGT, W.: Über die innere Reibung der festen Körper, insbesondere der Krystalle.
1.
2.
3.
4.
5.
6.
7.
8.
1.
2.
3.
Abh. Ges. Wiss. Göttingen 36, No. 1. (305)
BASSET, A. B.: On the extension and flexure of cylindrical and spherical thin
elastic shells. Phil. Trans. Roy. Soc. Lond. A 181, 433-480. (213)
GrLBERT, PH.: Sur quelques formules d'un usage general dans la physique mathematique. Ann. Soc. Sei. Bruxelles 14, 1-18. (99)
GosrEWSKI, W.: 0 eisnieniu kinetycznem w plynie nieseisliwym i jednorodnym.
Pami~tnik Akad. Krakow, mat.-przyr. 17, 128-134. (121)
GosiEWSKI, W.: 0 naturze ruchu wewna,trz elementu plynnego. Pami~tnik Akad.
Krakow, mat.-przyr. 17, 135-142. (86, 88)
Is:E, E.: Sulla deformazione elastica di un corpo isotropo. Atti Accad. Pontaniana
20, 241-260. (42)
LAMB, H.: On the deformation of an elastic shell. Proc. Lond. Math. Soc. 21,
119-146. (212, 213)
LEVY, M.: L'hydrodynamique moderne et l'hypothese des actions a distance. Rev.
Gen. sei. pures appl. 1, 721-728. (86, 107, JIJ8, 109, 110)
PADOVA, E.: 11 potenzialedelleforze elastiche di mezzi isotropi. Atti Ist. Veneto
sei. lett. arte 48 = (7) 1, 445-451. (61)
BRILLOUIN, M.: Deformations homogenes finies. Energie d'un corps isotrope.
C. R. Acad. Sei., Paris 112, 1500-1502. (36)
DUHEM, P.: Sur les equations gem\rales de la thermodynamique (1888). Ann.
Ecole Norm. (3) 8, 231-266. (246)
DuHEM, P.: Hydrodynamique, Elasticite, Acoustique, Paris: Hermann, 2 vols.
(194A, 241)
List of Works Cited. 757
4. NATANSON, L.: Thermodynamische Bemerkungen. Ann. Phys. 278 = (2) 42,
178-185. (251, 252)
5. SAMPSON, R. A.: On Stokes' current function. Phi!. Trans. Roy. Soc. Lond. A 182,
449-518. (162)
1892 1. BASSET, A. B.: On the theory of elastic wires. Proc. Land. Math. Soc. 23
(1891-1892), 105-127. (214)
2. BELTRAMI, E.: Osservazioni sulla nota precendente. Rend. Lincei (5) 11 , 141-142
= Opere 4, 510-512. (227)
3. DUHEM, P.: Commentaire aux principes de Ia thermodynamique, Premiere partie.
]. math. pures appl. (4) 8, 269-330. (240)
4. FINGER, J.: Über die gegenseitigen Beziehungen gewisser in der Mechanik mit
Vortheil anwendbaren Flächen zweiter Ordnung nebst Anwendung auf Probleme
der Astatik. Sitzgsber. Akad. Wiss. Wien (IIa) 101, 1105-1142. (14, 21, 22,
29, 30, 37)
5. HELMHOLTZ, H.: Das Princip der kleinsten Wirkung in der Elektrodynamik.
Sitzgsber. Akad. Wiss. Berlin 1892,459-475 = Ann. Phys. (2) 47, 1-26 = Abh.
3, 476-504. (150)
6. KILLING, W.: Über die Grundlagen der Geometrie. J. reine angew. Math. 109,
121-186. (84)
7. LARMOR, J.: The equations of propagation of disturbances in gyrostatically loaded
media, and of the circular polarization of light. Proc. Land. Math. Soc. 23
(1891-1892), 127-135 = Papers 2, 248-255· (200)
8. LovE, A. E. H.: A Treatise on the Mathematical Theory of Elasticity, vol. 1.
Cambridge. (See [1906, 5] [1927, 6].) (31, 35, 36, 39, 44, 45, 46, 205)
9. MORERA, G.: Soluzione generale delle equazioni indefinite dell' equilibrio di un
corpo continuo. Rend. Lincei (5) 11 , 137-141. (227)
10. MoRERA, G.: Appendice alla Nota: Sulla soluzione piu generale delle equazioni
indefinite dell' equilibrio di un corpo continuo. Rend. Lincei (5) 11 , 233-234. (227)
11. POINCARE, H.: Lec,:ons sur la Theorie de l'Elasticite. Paris. (55, 205, 210, 233)
12. VoiGT, W.: Ueber innere Reibung fester Körper, insbesondere der Metalle. Ann.
Phys. (2) 47, 671-693. (305)
13. VoiGT, W.: Bestimmung der Constanten der Elasticität und Untersuchung der
inneren Reibung für einige Metalle. Gött. Abh. 38, No. 2. (305)
14. WIEN, W.: Über den Begriff der Localisirung der Energie. Ann. Phys. (2) 47,
1-26 = Abh. 3, 476-504. (157, 217)
1893 1. DuHEM, P.: Le potentiel thermodynamique et Ia pression hydrostatique. Ann.
Ecole Norm. (3) 10, 187-230. (60, 215, 241, 245, 247)
2. DUHEM, P.: Dissolutionset melanges. Trav. Mem. Fac. Lilie 3, Nos. 2 and 12. (215,
255, 265, 295)
3. DUHEM, P.: Commentaire aux principes de la thermodynamique, Deuxieme partie.
]. Math. pures appl. (4) 9, 293-359. (245)
4. HEAVISIDE, 0.: Electromagnetic Theory, vol. 1. London. (302)
5. LovE, A. E. H.: A Treatise on the Mathematical Theory of Elasticity, vol. 2.
Cambridge. (See [1906, 5] [1927, 6].) (63A, 212, 213)
6. PoiNCARE, H.: Lec,:ons sur Ia Theorie des Tourbillons. Paris. (106, 107, 109, 117,
122, 125, 134, 145)
7. TaucHE, P.: Transformation des equations generales du mouvement des fluides.
Bull. Soc. Math. France 21, 72-7 5. (99)
1894 1. CuRIE, P.: Sur la symetrie dans !es phenomenes physiques, symetrie d'un champ
electrique et d'un champ magnetique. J. Physique (3) 3, 393-415 =Oeuvres
118-141. (258)
2. DuHEM, P.: Commentaire aux principes de Ia thermodynamiques, Troisieme partie.
]. Math. pures appl. (4) 10, 207-285. (246, 249, 265)
3. FABRI, C.: I moti vorticosi di ordine superioreal prima in relazione alle equazioni
pel movimento dei fluidi viscosi. Mem. Accad. Sei. Bologna (5) 4, 383-392. (123)
4. FINGER, J.: Über die allgemeinsten Beziehungen zwischen Deformationen und den
zugehörigen Spannungen in aeolotropen und isotropen Substanzen. Sitzgsber. Akad.
Wiss. Wien (IIa) 103, 1073-1100. (29, 30, 303)
5. KIRCHHOFF, G.: Vorlesungen über die Theorie der Wärme. Leipzig: Teubner.
X + 210 pp. (241)
6. METZLER, W. H.: Homogeneaus strains. Ann. of Math. 8, 148-156. (42)
7. NEUMANN, C.: Über die Bewegung der Wärme in compressiblen oder auch incompressiblen Flüssigkeiten. Ber. Verh. Ges. Wiss. Leipzig 46, 1-24. (241)
8. SHILLER, N.: On the problern of the thermodynamic potential [in Russian]. Isv.
imp. Obschestva Moskov. Univ. 7, No. 1, 22-31. (246)
758 C. TRUESDELL and R. TouPIN: The Classical Field Theories.
1895 1. BASSETT, A. B.: On the deformationofthin elastic wires. Amer. ~- Math. 17,
281-317. (214)
2. LAMB, H.: Hydrodynamics. Cambridge. (2nd ed. of [1879. 2], see [1906, 4], [1916,
3], [1924, 9], [1932, 8].) (94, 106, 108)
3. OuMOFF, N.: Une expression gem\rale du potentiel thermodynamique. Bull. Soc.
Imper. Naturalistes de Moscou (2) 8 (1894), 138-145. (246)
4. PAINLEVE, P.: Le~ons sur l'Integration des Equations Differentielles de la mecanique et applications. 291 pp. Paris: Hermann. (Lithographed.) (167, 170)
5. ScHÜTZ, ] . R.: Über die Herstellung von Wirbelbewegungen in idealen Flüssigkeiten durch conservative Kräfte. Ann. Physik (N. F.) 56, 144-147. (108)
6. ToucHE, P.: Equation d'une trajectoire fluide. Bull. Soc. Math. France 23,
111-113. (108)
7. VOIGT, W.: Über Medien ohne innere Kräfte und eine durch sie gelieferte mechanische Deutung der Maxwell-Hertz'schen Gleichungen. Gött. Abh. 1894, 72-79-
(200)
8. VoiGT, W.: Kompendium der theoretischen Physik 1. Leipzig. (256A, 296)
1896 1. CossERAT, E. et F.: Sur la theorie de l'elasticite. Ann. Toulouse 10, 1-116.
(12, 14, 26, -31, 46, 67, 69, 96, 210, 217)
2. DuHEM, P.: Sur les deformations permanentes et l'hysteresis, Parts I-V. Mem.
Couronn. Mem. Sav. Etrangers Acad. Belg. 54, 56 ( 1898); 62 ( 1902). (250)
3. MAcH, E.: Die Prinzipien der Wärmelehre historisch-kritisch entwickelt. Leipzig:
Johann Ambrosius Barth. (240, 246, 253)
1897 1. APPELL, P.: Sur les equations de l'hydrodynamique et la theoriedes tourbillons.
]. Math. pures appl. (5) 3, 5-16. (108, 129, 135)
2. FERRARIS, G.: Teoria geometrica dei campi vettoriali come introduzione allo studio
della elettricita, del magnetismo, ecc. Mem. Accad. Sei. Torino (2) 47, 259-338.
(175)
3. FINGER, ] . : Über das innere Virial eines elastischen Körpers. Sitzgsber. Akad.
Wiss. Wien (IIa) 106, 722-738. (219)
4. FöPPL, A.: Die Geometrie der Wirbelfelder. Leipzig. (93, 118, 129)
6. MEHMKE, R.: Zum Gesetz der elastischen Dehnungen. Z. Math. Phys. 42, 327-338.
(26, 33A)
6. MILLER, W. L.: On the second differential coefficients of Gibbs' function C. The
vapour tensions, freezing and boiling points of ternary mixtures. ] . Phys. Chem.
1, 633-642. (252)
7. NEUMANN, C.: Die Anwendung des Hamilton'schen Princips in der Hydrodynamik
und Aerodynamik. Ber. Verh. Ges. Wiss. Leipzig 94, 611-615. (160)
8. ToucHE, P.: Equations d'une trajectoire fluide dans le cas general. Bull. Soc.
Math. France 27, 5-8. (108)
9. TREVOR, J. E.: Inner thermodynamic equilibria. J. Phys. Chem. 1, 205-220.
(2/j])
1898 1. BJERKNES, V.: Über die Bildung von Circulationsbewegungen und Wirbeln in
reibungslosen Flüssigkeiten. Vidensk. Skrift. No. 5. (170)
2. DUHEM, P.: On the general problern of chemical statics. ] . Phys. Chem. 2, 1-42,
91-115. (266)
1899 1. APPELL, P.: Lignes correspondentes dans la deformation d'un milieu; extension
des theoremes sur les tourbillons. ]. Math. pures appl. (5) 5, 137-153. Abstract
in Bull. Soc. Math. France 26, 135-136 (1898). (132)
2. BROCA, A.: Sur le principe de l'egalite de l'action et de la reaction. c. R. Acad.
Sei., Paris 129, 1016-1019. (176)
3. VOLTERRA, V.: Sul flusso di energia meccanica. Atti Accad. Sei. Torino 34
(1898-1899). 366-375- (217)
4. VoLTERRA, V.: Sul flusso di energia meccanica. Nuovo Cim. (4) 10, 337-359·
(209)
1900 1. BRILLOUIN, M.: Theorie moleculaire des gaz. Diffusion du mouvement et de l'energie.
Ann. chimie (7) 20, 440-485. (293)
2. BRocA, A.: Sur les masses vectorielles de discontinuite. C. R. Acad. Sei., Paris
130, 317-319. (175)
3. DuHEM, P.: Sur le theoreme d'Hugoniot et quelques theoremes analogues. C. R.
Acad. Sei., Paris 131, 1171-1173. (180, 181, 187)
4. HERTZ, H.: Electric Waves. London. (276)
5. HILBERT, D.: Mathematische Probleme. Nachr. Ges. Wiss. Göttingen 1900, 253
-297. Repr. with addns., Arch. Math. Phys. 1, 44-63, 213-217 (1901). Transl.
with addns., Sur les problemes futurs des mathematiques. C. R. 2me Congr. Int.
6.
7.
8.
9.
10.
11.
12.
1901 1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
1902 1.
2.
3.
4.
5.
List of Works Cited. 759
:Math. Paris (1900). 58-114 (1902). Trans!., Mathematical problems. Bull. Amer.
Math. Soc. (2) 8, 437-479 (1902). (3)
MICHELL, J. H.: On the direct determination of stress in an elastic solid, with
applications to the theory of plates. Proc. Lond. Math. Soc. 31 (1899), 100-124.
(224)
MICHELL, J. H.: The uniform torsion and flexure of incomplete tores, with application to helical springs. Proc. Lond. Math. Soc. 31, (1899) 130-146. (228)
MoHR, 0.: Welche Umstände bedingen die Elastizitätsgrenze und den Bruch eines
Materials? Z. VDI 44, 1524-1530. (24)
STRUTT, J. W. (Lord RAYLEIGH): On a theorem analogaus to the virial theorem.
Phi!. Mag. (5) 50, 210-213 =Papers 4, 491-493. (216)
VIEILLE, P.: Etude sur le röle des discontinuites dans !es phenomenes de propagation. Mem. poudres salpetres 10 (1899-1900), 177-260. (297)
ZoRAWSKI, K.: Über die Erhaltung der Wirbelbewegung. C. R. Acad. Sei. Cracovie,
335-341. (7·5, 80, 107, 153)
ZoRAWSKI, K.: Über gewisse Änderungsgeschwindigkeiten von Linienelementen
bei der Bewegung eines continuirlichen materiellen Systems. C. R. Acad. Sei.
Cracovie, 367-374. (78, 153)
ABRAHAM, M.: Geometrische Grundbegriffe. Enz. math. Wiss. 43, 3-47. (163)
APPELL, P.: Deformation speciale d'un milieu continu; tourbillans de divers ordres.
Bull. Soc. Math. France 29, 16-17. (20)
DARBOUX, G.: Sur !es deformations finies et sur !es systemes triples de surfaces
orthogonales. Proc. Lond. Math. Soc. 32 (1900), 377-383. (38)
DUHEM, P.: Sur !es equations de l'hydrodynamique. Ann. Toulouse (2) 3, 253
to 279. (99, 136)
DuHEM, P.: Sur Ia condition supplementaire en hydrodynamique. C. R. Acad.
Sei., Paris 132, 117-120. (241)
DuHEM, P.: Sur !es theoremes d'Hugoniot, !es lemmes de M. Hadamard et Ia propagation des ondes dans !es fluides visqueux. C. R. Acad. Sei., Paris 132, 1163-1167.
(181, 298)
DuHEM, P.: Recherehes sur l'hydrodynamique. Ann. Taulause (2) 3, 315-377,
379-431; 4 (1902), 101-169; 5 (1903). 5-61, 197-255. 353-404 = repr. separ.
Paris, 2 vols., 1903, 1904. (113, 134, 181, 241, 246, 256, 258, 262, 297, 298)
HADAMARD, J.: Sur Ia propagationdes ondes. Bull. Soc. Math. France 29, 50-60.
(134, 173, 176, 177, 181, 183, 187, 190)
J ouGUET, E.: Sur Ia propagation des discontinuites dans !es fluides. C. R. Acad.
Sei., Paris 132, 673-676. (189, 205, 241, 258)
MUIRHEAD, R. F.: Stress - its definition. Nature, Land. 64, 207. (200)
NATANSON, L.: Sur !es lois de Ia viscosite. Bull. Int. Acad. Sei. Cracovie 95-111.
(305)
NATANSON, L.: Sur Ia double refraction accidentelle dans !es liquides. Bull. Int.
Acad. Sei. Cracovie 161-171. (305)
NATANSON, L.: Über die Gesetze der inneren Reibung. Z. physik. Chem. 38, 690
-704. (305)
RICCI, G., et T. LEVI-CIVITA: Methades de calcul differentiel absolu et leurs applications. Math. Ann. 54, 125-137. (205A)
WEINGARTEN, G.: Sulle superficie di discontinuita nella teoria della elasticita dei
corpi solidi. Rend. Accad. Lincei (5) 101, 57-60. (49, 175)
WEINGARTEN, J.: Über die geometrischen Bedingungen, denen die Unstetigkeiten
der Derivierten eines Systems dreier stetigen Funktionen des Ortes unterworfen
sind, und ihre Bedeutung in der Theorie der Wirbelbewegung. Arch. Math. Phys.
(3) 1, 27-33. (106, 175)
ZoRAWSKI, K.: Über gewisse Änderungsgeschwindigkeiten von Linienelementen
bei der Bewegung eines continuierlichen materiellen Systems, 2. Mitth. Bull. Int.
Acad. Sei. Cracovie 486-499· (86)
BJERKNES, V.: Cirkulation relativ zu der Erde. Meteor. Z. 37, 97-108. (145)
CoMBEBIAC, G.: Sur !es equations generales de l'elasticite. Bull. Soc. Math. France
30, 108-110, 242-247. (200, 217)
CouLoN, J.: Sur !'Integration des Equations aux Derivees Partielles du Second
Ordre per la Methode des Caracteristiques. Paris: Hermann. 122 pp. (175, 176,
180, 181)
HELMHOLTZ, H.: Vorlesungen über theoretische Physik, 2, Dynamik continuirlich
verbreiteter Massen. Leipzig (lectures of 1894). (4, 6)
THOMSON, W. (Lord KELVIN): A new specifying method for stress and strain in
an elastic solid. Proc. Roy. Soc. Edinb. 24 (1901-1903), 97-101 = Phil. Mag.
760
6.
7.
8.
9.
10.
1903 1.
2.
3.
4.
6.
6.
7.
8.
9.
10.
11.
12.
13.
14.
16.
16.
17.
18.
19.
20.
21.
1904 1.
C. TRUESDELL and R. TouPIN: The Classical Field Theories.
(6) 3, 444-448 = Math. Phys. Papers 4, 556-560. Prelim. note, Phi!. Mag. (6}
3, 95-97. (33A)
NATANSON, L.: Sur Ia propagation d'un petit mouvement dans un fluide visqueux.
Bull. Int. Acad. Sei. Cracovie 19-35. (306)
NATANSON, L.: Sur Ia fonction dissipative d'un fluide visqueux. Bull. Int. Acad.
Sei. Cracovie, 488-494. (306)
NATANSON, L.: Sur Ia deformation d'un disque plastico-visqueux. Bull. Int. Acad.
Sei. Cracovie, 494-512. (306)
NATANSON, L.: Sur Ia conductibilite calorifique d'un gaz en mouvement. Bull.
Int. Acad. Sei. Cracovie 137-146. (306}
ZERMELO, E.: Hydrodynamische Untersuchungen über die Wirbelbewegungen in
einer Kugelfläche. Z. Math. Phys. 47, 201-237. (162)
APPELL, P.: Sur quelques fonctions de point dans Je mouvement d'un fluide. J.
Math. pures appl. (5) 9, 5-19. An abstract is given in C. R. Acad. Sei., Paris
136, 186-187 (1903). (101, 102)
APPELL, P.: Sur quelques fonctions et vecteurs de point contenarrt uniquement
!es derivees premieres des composantes de Ia vitesse. Bull. Soc. Math. France 31,
68-73. (90, 101)
DuHEM, P.: Sur quelques formules de cinematique utiles dans Ia theorie generale
de l'elasticite. C. R. Acad. Sei., Paris 136, 139-141. (82A}
DuHEM, P.: Sur Ia viscosite en un milieu vitreux. C. R. Acad. Sei., Paris 136,
281-283. (262, 306)
DuHEM, P.: Sur !es equations du mouvement et Ia relation supplementaire au
sein d'un milieu vitreux. C. R. Acad. Sei., Paris 136, 343-345. (241, 262)
DuHEM, P.: Sur Je mouvement des milieux vitreux, affectes de viscosite, et tres
peu deformes. C. R. Acad. Sei., Paris 136, 592-595. (306)
DuHEM, P.: Sur !es ondes au sein d'un milieu vitreux, affecte de viscosite et tres
peu deforme. C. R. Acad. Sei., Paris 136, 733-735- (306}
DuHEM, P.: Des ondes du premierordre par rapport a Ia vitesse au sein d'un milieu
vitreux doue de viscosite, et affecte de mouvements finis. C. R. Acad. Sei., Paris
136, 858-860. (306}
DuHEM, P.: Des ondes du second ordre par rapport a Ia vitesse au sein des milieux
vitreux, doues de viscosite, et affectes de mouvements finis. C. R. Acad. Sei.,
Paris 136, 1032-1034. (306)
DuHEM, P.: Sur une generalisation du theoreme de Reech. Proces-Yerbaux Soc.
Sei. Bordeaux 1902-1903, 65-73. (262)
HADAMARD, J.: Le~ons sur Ja Propagation des Ondes et !es Equations de l'Hydrodynamique (lectures of 1898-1900). Paris. (20, 66, 86, 88, 172, 173, 174, 175,
176, 177, 180, 181, 182, 183, 184, 186, 186, 187, 188, 189, 190, 191, 218, 241, 246,
297, 301)
NATANSON, L.: Über einige von Herrn B. Weinstein zu meiner Theorie der inneren
Reibung gemachten Bemerkungen. Physik. Z. 4, 541-543. (306)
NATANSON, L.: Sur l'application des equations de Lagrange dans Ia theorie de Ia
viscosite. Bull. Int. Acad. Sei. Cracovie, 268-283. (306)
NATANSON, L.: Sur l'approximation de certaines equations de Ia theorie de Ia
viscosite. Bull. Int. Acad. Sei. Cracovie, 283-311. (306)
REYNOLDS, 0.: The Sub-Mechanics of the Universe. Papers 3. (81, 168, 169,
216, 243)
STEPHANSEN, E.: Von der Bewegung eines Continuums mit einem Ruhepunkte.
Arch. math. og naturv. 25, No. 6, 29 pp. (70)
ZAREMBA, s.: Remarques sur !es travaux de M. Natanson relatifs a Ia theorie de Ia
viscosite. Bull. Int. Acad. Sei. Cracovie 85-93. (306)
ZAREMBA, S.: Sur une generalisation de Ia theorie classique de Ia viscosite. Bull.
Int. Acad. Sei. Cracovie 380-403. (266, 306}
ZAREMBA, S.: Sur un problerne d'hydrodynamique lie a un cas double refraction
accidentale dans !es liquides et sur !es considerations theoriques de M. Natanson
relatives a ce phenomene. Bull. Int. Acad. Sei. Cracovie 403-423. (306)
ZAREMBA, S.: Sur une forme perfectionee de Ia theorie de Ia relaxation. Bull.
Int. Acad. Sei. Cracovie 594-614. (148, 293, 306)
ZAREMBA, S.: Le principe des mouvements relatifs et !es equations de Ia mecanique
physique. Reponse a M. Natanson. Bull. Int. Acad. Sei. Cracovie 614-621. (148,
160, 293)
DuHEM, P.: Recherehes sur l'elasticite. Ann. Ecole Norm. (3) 21, 99-139. 375-414;
22 (1905), 143-217; 23 (1906}, 169-223 = repr. separ., Paris 1906. (96, 200,
262, 301, 306)
List of Works Cited. 761
2. DuHEM, P.: Sur Ia stabilite de l'equilibre en thermodynamique et !es reeherehes de
J. W. Gibbs au sujet de ee probleme. Proees-Verbaux Soe. Sei. Bordeaux 1903-
1904, 112-121. (245, 246, 265)
3. J OUGUET, E.: Remarques sur la propagation des pereussions dans les gaz. C. R.
Aead. Sei., Paris 138, 1685-1688. (258)
4. LovE, A. E. H.: Wave-motions with diseontinuities at wave-fronts. Proe. Lond.
Math. Soe. (2) 1, 37-62. (180)
5. MANVILLE, 0.: Sur Ia deformation finie d'un milieu eontinu. Mem. soe. sei. Bordeaux
(6) 2, 83-162. According to SrGNORINI [1943, 6; Chap. I, ~ 1], this paper is not
exempt from errors. (34)
6. SAUREL, P.: On the stability of the equilibrium of a homogeneous phase. J. Phys.
Chem. 8, 325-334. (265)
7. SAUREL, P.: On the stability of equilibrium of bivariant systems. J. Phys. Chem.
8, 436-439. (265)
8. WHITTAKER, E. T.: A Treatise on the Analytical Dynamics of Particles and Rigid
Bodies; with an Introducbon to the Problem of Three Bodies. Cambridge: Cambridge Univ. Press. xiii + 414 pp. (99)
1905 1. HADAMARD, J.: Remarque sur Ia note deM. Gyczö Zemph~n. C. R. Acad. Sei.,
Paris 141, 713. (258)
2. J AUMANN, G.: Die Grundlagen der Bewegungslehre von einem modernen Standpunkte aus. Leipzig. (8, 12, 66 B, 79, 80, 81, 107)
3. KLEIN, F., u. K. WIEGHARDT: Über Spannungsflächen und reziproke Diagramme,
mit besonderer Berücksichtigung der Maxwell'schen Arbeiten. Arch. Math. Phys.
(3) 8, 1-10,95-119. (227)
4. MARCOLONGO, R.: Le formule del Saint-Venant per le deformazioni finite. Rend.
Circ. Nat. Palermo 19, 151-155. (34)
5. RrgurER, C.: Sur l'integration d'un systeme d'equations aux derivees partielles
auquel conduit l'etude des deformations finies d'un milieu continu. Ann. Ecole
Norm. (3) 22, 475-538. (34)
6. VoLTERRA, V.: Sull'equilibrio dei corpi elastici piu volte connessi. Rend. Accad.
Lineei 141, 193-202. (49, 185)
7. ZEMPLEN, G.: Kriterien für die physikalische Bedeutung der unstetigen Lösungen
der hydrodynamischen Bewegungsgleichungen. Math. Ann. 61, 437-449. (7, 9, 193,
205, 236A)
8. ZEMPLEN, G.: Sur l'impossibilite des ondes de choc negatives dans les gaz. C. R.
Acad. Sei., Paris 141, 710-712. (258)
9. ZEMPLEN, G.: Besondere Ausführungen über unstetige Bewegungen in Flüssigkeiten. Enz. math. Wiss. IV 3, art. 19. (172, 205, 236A, 241)
1906 1. CAFIERO, D.: La deformazione finita di un mezzo continuo. Atti Accad. Peloritana
21, 179-228. (12, 34, 42, 44, 45)
2. CESARO, E.: Sulle formale del Volterra, fondamentali nella teoria delle distorsioni
elastiehe. Rend. Aeead. Napoli (3 a) 12, 311-321. (57, 175)
3. CrsoTTr, U.: Sul paradosso di d'Alembert. Atti Accad. Veneto 65, 1291-1295.
(202)
4. LAMB, H.: 3rd ed. of [1879. 2]. (See [1895. 2], [1916, 3], [1924, 8], [1932, 9].) (99,
136)
5. LovE, A. E. H.: 2nd ed. of [1892, 8] and [1893, 5]. (12, 27, 41, 45, 195, 214, 218,
228, 301)
6. MüLLER, C. H., u. A. TrMPE: Die Grundgleichungen der mathematischen Elastizitätstheorie. Enz. math. Wiss. IV4 , art. 23, 1-54. (231, 256A)
1907 1. CossERAT, E. et F.: Sur la mecanique generale. C. R. Acad. Sei., Paris 145,
1139-1142. (60)
2. CossERAT, E. et F.: Sur Ia statique de Ia ligne deformable. C. R. Aead. Sei.,
Paris 145, 1409-1412. (214)
3. HrLBERT, D.: Mechanik der Continua, lectures of 1906-1907; MS notes by A. R.
CRATHORNE in Univ. Illinois Library. (9, 158, 256)
4. HrLBERT, D.: ibid., more complete MS notes by W. MARSHALL in Purdue Univ.
Library. (9, 158, 245, 256)
5. LoRENTZ, H. A.: Über die Entstehung turbulenter Flüssigkeitsbewegungen und
über den Einfluß dieser Bewegung bei der Strömung durch Röhren. Abh. theor.
Phys. 43-71. (207)
6. NEUMANN, E. R.: Über eine neue Reduktion bei hydrodynamischen Problemen.
]. reine angew. Math. 132, 189-215. (224, 225)
7. VoLTERRA, V.: Sur l'equilibre des corps elastiques multiplement connexes. Ann.
Ecole Norm. (3) 24, 401-517. (57)
762 C. TRUESDELL and R. TouPIN: The Classical Field Theories.
8. WILSON, E. B.: Oblique reflections and unimodular strains. Trans. Amer. Math.
Soc. 8, 270-298. (46)
1908 1. CossERAT, E. et F.: Sur la statique de la surface deformable et la dynamique
de la ligne deformable. C. R. Acad. Sei., Paris 146, 68-71. (63A, 64)
2. CossERAT, E. et F.: Sur la theorie des corps minces. C. R. Acad. Sei., Paris
146, 169-172. (60)
3. EINSTEIN, A., u. ]. LAUB: Über die elektromagnetischen Grundgleichungen für
bewegte Körper. Ann. Physik, 26, 532-540. (308)
4. HAMEL, G.: Über die Grundlagen der Mechanik. Math. Ann. 66, 350-397. (2,
231)
5. HARGREAVES, R.: On integral forms and their connection with physical equations.
Cambr. Phil. Trans. 21, 107. (266, 270)
6. PLANCK, M.: Bemerkungen zum Prinzip der Aktion und Reaktion in der allgemeinen
Dynamik. Verh. dtsch. phys. Ges. 10, 728-732. (289)
1909 1. ABRAHAM, M.: Zur Elektrodynamik bewegter Körper. Rend. Palermo 28, 1-28.
(80, 279)
1910
1911
2. BRILL, A.: Vorlesungen zur Einführung in die Mechanik raumerfüllender Massen.
Leipzig u. Berlin. x + 236 pp. (8, 237)
3. CARATHEODORY, C.: Untersuchungen über die Grundlagen der Thermodynamik.
Math. Annalen 67, 355-386. (245)
4. CISOTTI, U.: Sul moto di un solido in un canale. Rend. Palermo 28, 307-3 51.
(202)
5. CossERAT, E. et F.: Theorie des Corps DMormables. Paris= pp. 953-1173 of
0. D. CHWOLSON, Traite de Physique, transl. E. DAVAUX, 2nd ed., 2. Paris 1909.
(12, 60, 61, 63A, 64, 200, 210, 212, 214, 217, 238, 256)
6. GIBBS, J. W., and E. B. WILSON: Vector Analysis. New Haven. (66B)
7. LUDWIK, P.: Elemente der Technologischen Mechanik. Berlin. (25, 33)
8. MisEs, R. v.: Theorie der Wasserräder. Z. Math. Phys. 57, 1-120. (99, 137,
138, 145, 162, 170, 202, 217)
9. UsAI, G.: Sulle deformazioni di 2° ordine di una superficie flessibile ed estendibile.
Rend. Ist. Lombardo (2} 42, 416-426. (64)
10. VoLTERRA, V.: Sulle equazioni integro-differenziali della teoria dell' elasticita.
Rend. Lincei (5) 181 , 295-301 = Opere 3, 288-293. (305)
11. VoLTERRA, V.: Equazioni integro-differenziali della elasticita nel caso della iso1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
1.
2.
3.
4.
5.
tropia. Rend. Lincei (5) 181 , 577-586 = Opere 3, 294-303. (305)
BATEMAN, H.: The transformation of the electrodynamical equations. Proc. Land.
Math. Soc. (2} 8, 223-264. (266, 270, 282)
BoGGIO, T.: Sul moto permanente di un solido in un fluido indefinito. Atti Accad.
Veneto 692, 883-891. (202)
CALLENDER, H. L.: art. Heat. Encycl. Britt., 1 tth ed. to present ed. (240, 245)
CISOTTI, U.: Sul moto permanente di un solido in un fluido indefinito. Atti Accad.
Veneto 692, 427-445. (202)
DARBOUX, G.: Le<;:ons sur les Systemes Orthogonaux et les Coordonnees Curvilignes,
2nd ed. Paris. 567 pp. (38)
KöTTER, F.: Über die Spannungen in einem ursprünglich geraden, durch Einzelkräfte in stark gekrümmter Gleichgewichtslage gehaltenen Stab. Sitzgsber. Preuss.
Akad. Wiss., part 2, 895-922. (303}
MINKOWSKI, H.: Die Grundgleichungen für die elektromagnetischen Vorgänge in
bewegten Körpern. Math. Annalen 68, 472-525. (308}
STRUTT, J. W. (Lord RAYLEIGH): Aerial plane waves of finite amplitude. Proc.
Roy. Soc. Land. A 84, 247-284 = Papers 5, 573-610. (258)
VEBLEN, 0., and J. W. YouNG: Projective Geometry. Boston. (21)
V OIGT, W.: Lehrbuch der Kristallphysik Leipzig (repr. 1928}. (256 A, 296, 305,
312)
ALMANSI, E.: Sulle deformazioni finite dei solidi elastici isotropi, I. Rend. Lincei
(SA) 201, 705-714. (31, 303)
ALMANSI, E.: Sulle deformazioni finite dei solidi elastici isotropi, I li. Rend. Lincei
(SA) 202, 287-296. (30)
CRUDELI, V.: Sopra le deformazioni finite. Le equazioni del De Saint-Venant.
Rend. Lincei (5 A) 202, 306-308, 470. (34)
DuHEM, P.: Traite d'Energetique. Paris. 2 vols. (201, 241, 247, 262, 265)
GwYTHER, R. F.: The conditions that the stresses in a heavy elastic body should
be purely elastic stresses. Mem. Manchester Lit. Phil. Soc. 55, No. 20 (12 pp.).
(227)
List of Werks Cited.
6. HAMEL, G.: Zum Turbulenzproblem. Nachr. Ges. Wiss. Göttingen, math.-phys.
Kl., 261-270. (116)
7. J AUMANN, G.: Geschlossenes System physikalischer und chemischer Differentialgesetze. Sitzgsber. Akad. Wiss. Wien (IIa) 120, 385-530. (6, 159, 215, 243,
245, 262, 295)
8. KARMAN, T.v.: Festigkeitsversuche unter allseitigem Druck. Z. VDI 55, 1749-1757-
(24)
9. LE Roux, J.: Sur les covariants fondamenteaux du secend ordre dans la deformation finie d'un milieu continu. C. R. Acad. Sei., Paris 152, 1002-1005. (14, 20,
26)
10. LE Roux, J.: Sur l'incurvation et la flexion dans les deformations finies. C. R.
Acad. Sei., Paris 152, 1655-1657- (20)
11. LE Roux, J.: Etude geometrique de la torsion et de la flexion dans la deformation
infinitesimale d'un milieu continu. Ann. Ecole Norm. (3) 28, 523-579. (20)
12. VESSIOT, E.: Sur les transformations infinitesimales et la cinematique des milieux
continus. Bull. Sei. Math. (2) 351, 233-244. (107, 109)
13. ZoRAWSKI, K.: Über stationäre Bewegungen kontinuierlicher Medien. Bull. Int.
Acad. Sei. Cracovie A, 1-17. (143, 146, 148, 153)
14. ZoRAWSKI, K.: Invariantentheoretische Untersuchung gewisser Eigenschaften der
Bewegungen kontinuierlicher Medien. Bull. Int. Acad. Sei. Cracovie A, 175-218.
(84, 90, 104, 143, 144, 145, 153)
1912 1. BoussrNESQ, J.: Theorie geometrique, pour un corps non rigide, des deplacements
bien continus, ainsi que les deformations et rotations de ses particules. J. Math.
pures appl. (6) 8, 211-227. Abstract, C. R. Acad. Sei., Paris 154, 949-954. (30,
57)
2. DE DoNDER, T.: Sur la einematique des milieux continus. Bull. Acad. Belg.,
Cl. Sei., 243-251. (134, 146)
3. GwvTHER, R. F.: The formal speeification of the elements of stress in cartesian,
and in cylindrical and spherical polar Coordinates. Mem. Manchester Lit. Phil.
Soc. 56, No. 10 (13 pp.). (227)
4. HAMEL, G.: Elementare Mechanik. Leipzig u. Berlin. (30, 31)
5. KARMAN, T. v.: Festigkeitsversuche unter allseitigen Druck. Forsch.-Arb. Ing.-Wes.
1912, 37-68. (24)
6. MrE, G.: Grundlagen einer Theorie der Materie. Ann. Physik 37, 511-534; 39,
1-40; 40, 1-66. (276, 310)
7. VOLTERRA, V.: Sur les equations integre-differentielles et leurs applications. Acta
Math. 35, 295-356. (309)
8. ZoRAWSKI, K.: Über gewisse Eigenschaften der Bewegungen kontinuierlicher
Medien. Bull. Int. Acad. Sei. Cracovie A 269-292. (104, 153)
1913 1. FERRARI, M.: Flusse di energia e veloeitä. di gruppo. Rend. Lincei (5) 22, 761 -766.
(157)
2. GwvTHER, R. F.: The specification of the elements of stress. Part II. A simplification of the speeification given in Part I. Mem. Manchester Lit. Phil. Soc. 57,
No. 5 (4 pp.) (227)
3. HAVELOCK, T. H.: The displacement of the particles in a case of fluid motion.
Proc. Univ. Durharn Phil. Soc. 4 (1911-1912), 62-74. (71)
4. HEUN, K.: Ansätze und allgemeine Methoden der Systemmechanik. Enz. math.
Wiss. 42 (1904-1935), art. 11. (12, 66B, 159, 200, 212, 214, 236, 243)
5. LE Roux, J.: Recherehes sur la geometrie des deformations finies. Ann. Ecole
Norm. (3) 30, 193-235. (20)
6. MrsEs, R. v.: Mechanik der festen Körper im plastisch-deformablen Zustand.
Gött. Nachr. 582-592. (300)
7. MoRTON, W. B.: On the displacements of the particles and their paths in some
cases of two-dimensional motion of a frictionless liquid. Proc. Roy. Soc. Land.
A 89, 106-124. (71)
1914 1. BuRGATTI, P.: Sulle deformazioni finite dei corpi continui. Mem. Accad. Sei.
Bologna (7) 1, 237-244. (34)
2. DUHEM, P.: Sur le paradoxe hydrodynamique de d' Alembert. C. R. Acad. Sei.,
Paris 159, 592-595. (202)
3. DUHEM, P.: Remarque sur le paradoxe hydrodynamique de d' Alembert. C. R.
Acad. Sei., Paris 159, 638-640. (202)
4. HELLINGER, E.: Die allgemeinen Ansätze der Mechanik der Kontinua. Enz. math.
Wiss. 44, 602-694. (7, 9, 195, 231, 232, 232A, 233, 236, 237, 238, 256)
5. HESSELBERG, TH., u. H. U. SVERDRUP: Das Beschleunigungsfeld bei einfachen
Luftbewegungen. Veröff. geophys. Inst. Univ. Leipzig (2), No. 5. 117-146. (70)
764 C. TRUESDELL and R. TouPrN: The Classical Field Theories.
HuNTINGTON, E. V.: The theorem of rotation in elementary mechanics. Amer.
math. monthly 21, 315-320. (170)
MATTIOLI, G. D.: Sulle equazioni fondamentali della dinamica dei sistemi continui.
Rend. Lincei (5) 232, 328-334. (157, 207, 217, 240)
8. MoHR, 0.: Abhandlungen aus dem Gebiete der technischen Mechanik, 2nd ed.
Berlin. (24)
9. PrcARD, E.: A propos du paradoxe hydrodynamique de d' Alembert. C. R. Acad.
Sei., Paris 159, 638. (202)
10. RoEVER, W. H.: The design and theory of a mechanism for illustrating certain
systems of lines of force and stream lines. Z. Math. Phys. 62 (1913-1914),
376-384. Abstract in Bull. Amer. Soc. 18,435-536 (1912) and 19, 220 (1913). (70)
11. SoMIGLIANA, C.: Sulla teoria delle distorsioni elastiche. Rend. Lincei (5) 231 ,
463-472. (175)
1915 1. BöKER, R.: Die Mechanik der bleibenden Formänderung in kristallinisch aufgebauten Körpern. Forsch.-Arb. Ing.-Wes. 1915, 1-51. (24)
2. EMDE, F.: Zur Vektorrechnung. Archiv Math. (3) 24, 1-11. (173)
3. LoRENTZ, H. A.: The Theory of Electrons. Leipzig. (279)
4. MoRTON, W. B., and J. VrNT: On the paths of the particles in some cases of motion
of frictionless fluid in a rotating enclosure. Phil. Mag. (6) 30, 284-287. (71)
1916 1. BrRKELAND, R.: Developpements sur le mouvement d'un fluide parallele a un
plan fixe. C. R. Acad. Sei., Paris 163, 200-202. (161)
2. FRIEDMANN, A.: Sur les tourbillans dans un liquide a temperature variable. c. R.
Acad. Sei., Paris 163, 219-222. (110)
3. LAMB, H.: 4th ed. of [1879. 2]. (See [1895. 2], [1906, 4], [1924, 9], [1932, 8].) (86A)
4. PALATIN!, A.: Sulle quadriche di deformazione per gli spazi S3 . Atti Ist. Veneto
76, 125-148. (84)
5. SPIELREIN, J.: Lehrbuch der Vektorrechnung nach den Bedürfnissen in der Technischen Mechanik und Elektrizitätslehre. Stuttgart. (66B, 79, 80, 81, 173)
6. TAYLOR, G. I.: Motion of solids in fluids when the flow is not irrotational. Proc.
Roy. Soc. Lond., Ser. A 93 (1916-1917), 99-113. (143, 207)
1917 1. ALMANSI, E.: L'ordinaria teoria dell' elasticita e Ia teoria delle deformazioni finite.
Rend. Lincei (5) 262, 3-8. (55)
2. APPELL, P.: Sur un theoreme de J oseph Bertrand relatif a Ia cinematique des
milieux continus. Bull. Sei. Math. (2) 41, 23-28. (101)
3. APPELL, P.: Sur une extension des equations de Ia theorie des tourbillans et des
equations de Weber. C. R. Acad. Sei., Paris 164, 71-74. (101, 136)
4. CrsoTTI, U.: Sulle azioni dinamiche di masse fluide continue. Rend. Lombardo
(2) 50, 502-515. (170, 202)
5. LOHR, E.: Entropieprinzip und geschlossenes Gleichungssystem. Denkschr. Akad.
Wiss. Wien 93, 339-421. (6, 159, 243, 245, 262, 295)
1918 1. CRUDELI, U.: Le formule del Cauchy e i fluidi viscosi. Rend. Lincei (5) 272, 49-52.
(134)
2. DIETZIUS, R.: Die Gestalt der Stromlinien in der Nähe der singulären Punkte.
Beitr. Phys. fr. Atmosph. 8, 29- 52. (70)
3. J AUMANN, G.: Physik der kontinuierlichen Medien. Denkschr. Akad. Wiss. Wien
95, 461-562. (6, 245, 262, 295)
1920 1. ErNSTEIN, A.: Schallausbreitung in teilweise dissoziierten Gasen. Sitzgsber. Akad.
Wiss. Berlin, math.-phys. Kl., 380-385. (159A)
2. PASCAL, M.: Circuitazione superficiale. Rend. Lincei (5) 292, 3 53-356; 301, 117-119,
249-251 (1921). (105)
3. PAuLI, W.: Relativitätstheorie. Enz. math. Wiss. 5, No. 2. (284)
1921 1. APPELL, P.: Traite de Mecanique Rationnelle, 3, Equilibre et Mouvement des Milieux
Continus, 3rd. ed. Paris. (86, 101, 109, 122, 136, 138)
2. BoRN, M.: Kritische Betrachtungen zur traditionellen Darstellung der Thermodynamik. Physik. Z. 22, 218-224, 249-254, 282-286. (245)
3. ]AFFE, G.: Über den Transport von Vektorgrößen, mit Anwendung auf Wirbelbewegung in reibenden Flüssigkeiten. Physik. Z. 22, 180-183. (101, 131)
4. MURNAGHAN, F. D.: The absolute significance of Maxwell's equations. Phys. Rev.
(2) 17, 73-88. (266)
5. PAscAL, M.: Circuitazione superficiale. Giorn. mat. 59, 215-234. (105)
6. WEYL, H.: Raum-Zeit-Materie. 4th ed. Berlin: Springer. (266, 310)
1922 1. BATEMAN, H.: Equations for the description of electromagnetic phenomena. Bull.
Nat. Res. Council4, 96-191. (308)
2. CARTAN, E.: Le.yons sur !es invariants integraux. Paris: Hermann. x + 210 pp.
(129)
3.
4.
.).
6.
1923 1.
2.
3.
4.
5.
List of Works Cited. 765
KoTTLER, F.: Newton'sches Gesetz und Metrik. Sitzgsber. Akad. Wiss. Wien (IIa)
131, 1-14. (7)
KoTTLER, F.: Maxwell' sehe Gleichungen und Metrik. Sitzgsber. Akad. Wiss. Wien
(II a) 131, 119-146. (7, 266)
LAMPE, E.: Review of [1914, 10]. Jb. Fortschr. Math. 1914-1915, 963. (70)
MUNK, M.: Notes on aerodynamic forces, I. Rectilinear motion. Nat. Adv. Comm.
Aero. T. N. 104. (76)
CARTAN, E.: Sur les varit\tes a connexion affine et la theorie de la relativite glneralisee. Ann. Ecole Norm. (3) 40, 325-412; 41, 1-25 (1924); 42, 17-88 (1925).
(7, 152. 211)
CISOTTI, U.: Sull' energia einetica di masse fluide continua: viriale degli sforzi.
Rend. Lincei (5) 322, 464-467. (219)
CISOTTI, U.: Considerazione sulla nota formula idrodinamica di Daniele Bernoulli.
Boll. Un. Mat. Ital. 2, 125-128. (125)
CISOTTI, U.: Sul carattere necessariamente vorticoso dei moti regolari permanenti
di un fluido qualsiasi in ambienti limitati, eppure in quiete all' infinito. Boll. Un.
Mat. Ital. 2, 170-172. (125)
LEONARDO DA VINCI: Del moto e misura dell' acqua. Ed. CARUSI e FAVARO:
Bologna. Parts of this work were printed in various publications from 1765 onward.
(77)
1924 1. ARIANO, R.: Deformazioni finite dei sistemi continui. Rend. Palermo 48, 97-120.
(12)
2. CALDONAzzo, B.: Sulla geometria differenziale di superfieie aventi interesse idrodinamico. Rend. Lincei (5A) 332, 396-400. (122)
3. COLLINET, E.: Sur l'energie interne d'un corps elastique. c. R. Acad. Sei., Paris
178, 373-375- (210)
4. HEYMANS, P.: Note on a property of rectilinear lines of principal stress. J. Math.
Phys. 3, 182-185. (223)
5. lNGRAM, W. H.: Note on the curl. J. Math. Phys. 3, 186-187. (86)
6. JouGUET, E.: Le potential interne des corps elastiques. C. R. Acad. Sei., Paris
178, 840-842. (210)
7. JouGUET, E., et M. Rov: Le paradoxe de d'Alembert dans le cas des fluides compressibles. C. R. Acad. Sei., Paris 178, 1470-1472. Cf. remarks by CISOTTI, C. R.
Acad. Sei., Paris 178, 1792-1793 and by the authors, C. R. Acad. Sei., Paris 179,
142-143. (202)
8. KELLOGG, 0. D.: The theorem of the moment of momentum. Amer. Math. Monthly
31, 429-432. (170)
9. LAMB, H.: 5th ed. of [1879, 2]. (See [1895. 2], [1906, 4], [1916, 3], [1932, 8].) (136)
10. LoHR, E.: Das Entropieprinzip der Kontinuitätstheorie. Festschr. Techn. Hochschule Brünn 176-187. (243, 245, 262)
11. NoAILLON, P.: Reponse aux observations de M. Pascal sur la circulation superficielle. C. R. Acad. Sei., Paris 178, 311-314. (101, 105)
1925 1. ARIANO, R.: Deformazioni finite di sistemi continui, I. Ann. Mat. (4) 2, 217-261.
(12)
2. BRILLOUIN, L.: Les lois de l'elasticite sous forme tensorielle valable pour des coordonnees quelconques. Ann. de Phys. (10) 3, 251-298. (12, 31, 210)
3. CALDONAZZO, B.: Un' osservazione a proposito del teorema di Bernoulli. Boll.
Un. Mat. Ital. 4, 1-3. (122)
4. CARATHEODORY, C.: Über die Bestimmung der Energie und der absoluten Temperatur mit Hilfe von reversiblen Prozessen. Ber. Akad. Wiss. Berlin, 39-47. (245)
5. EHRENFEST-AFANASSJEWA, T.: Zur Axiomatisierung des zweiten Hauptsatzes der
Thermodynamik. Z. Physik 33, 933-945; 34, 638. (245)
6. FRIEDMANN, A.: Über Wirbelbewegung in einer kompressiblen Flüssigkeit. Z. angew.
Math. Mech. 4, 102-107. (110)
7. HENCKY, H.: Die Bewegungsgleichungen beim nichtstationären Fliessen elastischer
Massen. Z. angew. Math. Mech. 5, 144-146. (66B, 150)
8. LELLI, M.: Una forma piu generale del principio di Archimede. Boll. Un. Mat.
Ital. 4, 63-64. (202)
9. LELLI, M.: Sui teoremi delle quantita di moto per i sistemi continui. Boll. Un.
Mat. Ital. 4, 204-206. (202)
10. LICHTENSTEIN, L.: Über einige Existenzprobleme der Hydrodynamik homogener,
unzusammendrückbarer, reibungsloser Flüssigkeiten und die Helmholtzschen Wirbeisätze. Math. Z. 23, 89-154. (101)
766 C. TRUESDELL and R. TOUPIN: The Classical Field Theories.
1926 1. EISENHART, L. P.: Riemannian Geometry. Princeton. (34)
2. FINZI, B.: Constatation energetique du paradoxe de d' Alembert dans !es liquides
visqueux. C. R. Acad. Sei., Paris 182, 1077-1079. (202)
3. KoTCHINE, N. E.: Sur la theorie des ondes de choc dans un fluide. Rend. Circ.
Mat. Palermo 50, 305-344. (7, 172, 175, 180, 185, 186, 189, 191, 193, 194, 205,
241, 297, 298)
4. LANDE, A.: Axiomatische Begründung der Thermodynamik durch Caratheodory.
Handbuch Physik 9, Chap. 4. (245)
1927 1. BouLIGAND, G.: Sur le signe de la pression dans un liquide pesant en mouvement
irrotationnel. Verh. 2. Int. Congr. Techn. Mech. (1926) 460-461. (121)
2. BouLIGAND, G.: Un theoreme relatif a la pression au sein d'un liquide parfait
en mouvement irrotationnel. J. Math. pures appl. (9) 427-433. (121)
3. DE DoNDER, T.: L'affinite. Mem. acad. Sei. Belg. in 8° (2) 9, No. 7. 94 pp. Also
issued separately, Paris 1927. 2nd ed., revised by P. VAN RYSSELBERGHE. Paris:
Gauthier-Villars 1936. xv + 142pp. (159A, 259)
4. EISENHART, L. P.: Non-Riemannian Geometry. New York: Amer. Math. Soc.
(152)
5. LICHTENSTEIN, L.: Über einige Existenzprobleme der Hydrodynamik II: Nichthomogene, unzusammenrückbare, reibungslose Flüssigkeiten. Math. Z. 26, 196-323.
(101)
6. LovE, A. E. H.: 4th ed. of [1892, 8] and [1893, 5]. (38, 256A, 301)
7. MASOTTI, A.: Osservazioni sui moti di un fluido nei quali e stazionaria Ja distribuzione del vortice. Rend. Lincei (6) 6, 224-228. (111, 125)
8. VEBLEN, 0.: Invariants of quadratic differential forms. Cambr. tracts math. and
math. phys. No. 24, Cambridge. (34, 147)
1928 1. ARIANO, R.: Deformazioni finite di sistemi continui I-III. Ann. di mat. (4) 5,
55-71; 6, 265-282 (1929). (12)
2. BRILLOUIN, L.: Les Jois de l'elastieite en coordonnees quelconques. Proc. Int.
Congr. Math. Toronto (1924) 2, 73-97 ( a preliminary version of [1925, 2]). (31,
210)
3. HALDANE, J. S.: Gases and Liquids. Edinburgh. (245)
4. HENCKY, H.: Über die Form des Elastizitätsgesetzes bei ideal elastischen Stoffen.
Z. techn. Phys. 9, 214-223, 457. (33)
5. LEVI-CIVITA, T.: Fondamenti di meccanica relativistica. Bologna: Zanichelli.
vii + 185 pp. (211)
1929 1. DE DoNDER, T.: L'affinite II. Bull. Sei. Acad. Belg. (5) 15, 615-625, 900-912;
17, 298-314, 507-515, 653-663, 780-787, 874-887, 1001-1007 (1931). Repr., with 2 suppl. papers, Paris 1931. (259)
2. HENCKY, H.: Welche Umstände bedingen die Verfestigung bei der bildsamen Verformung von festen isotropen Körpern? Z. Physik 55, 145-155. (33)
3. HENCKY, H.: Das Superpositionsgesetz eines endlich deformierten relaxationsfähigen elastischen Kontinuums und seine Bedeutung für eine exakte Ableitung
der Gleichungen für die zähe Flüssigkeit in der Eulerschen Form. Ann. Physik
(5) 2, 617-630. (33, 95, 148)
4. LICHTENSTEIN, L.: Grundlagen der Hydromechanik. Berlin: Springer. (88, 101,
105, 121, 172, 174, 182, 187, 189)
5. LICHTENSTEIN, L.: Bemerkung über einen Verzerrungssatz bei topologischer Abbildung in der Hydromechanik. Math. Z. 30, 321-324. (65)
6. PoLUBARINOVA, P. YA.: Singularpoints of stream lines of affine motion in space
[in Russian]. Isv. Glavn. Geofiz. Observ. No. 1, 1-16. (70)
7. PoMPEIU, D.: Sur la condition des vitesses dans un fluide incompressible. Bull.
math. phys. Ecole Polyt. Bucarest 1, 42-43. (160)
8. SARTON, G.: The discovery of the law of conservation of energy. Isis 13 ( 1929-1930),
18-34. (240)
9. SouTHWELL, R. V.: Mechanics. Encycl. Britt. 14th ed. In the 1944 printing,
15, 156-168. (3)
10. THIRRING, H.: Begriffssystem und Grundgesetze der Feldphysik. Handbuch
Physik 4, 81-177. (293)
1930 1. FINZI, B.: Sopra i1 tensore di deformazione di un velo. Rend. Ist. Lombardo (2)
63, 975-982. (34, 84)
2. KIRCHHOFF, W.: Reduktion simultaner partieller Differentialgleichungen bei hydrodynamischen Problemen. J. reine angew. Math. 164, 183-195. (161, 225)
3. MISES, R. v.: Über die bisherigen Ansätze in der klassischen Mechanik der Kontinua. Proc. Jrd Int. Congr. appl. Mech. Stockholm 2, 1-9. (7)
List of Works Cited. 767
4. SrGNORINI, A.: Sulle deformazioni finite dei sistemi continui. Rend. Lincei (6)
12, 312-316. (34)
5. SrGNORINI, A.: Sulla meccanica dei sistemi continui. Rend. Lincei (6) 12, 411-416.
(210)
6. SrGNORINI, A.: Sulle deformazioni termoelastiche finite. Proc. 3rd Int. Congr.
appl. Mech. Stockholm 2, 80-89. (33A, 43, 44, 210)
7. ToNoLo, A.: Forma intrinseca delle equazioni dell' equilibrio dei mezzi elastici.
Rend. Lincei (6) 12, 247-250, 347-351. (205A)
8. VrLLAT, H.: Leyons sur Ia Theorie des Tourbillons. Paris. (101)
9. VoLTERRA, V.: Theory of Functionals and of Integral and Integro-differential
Equations. London and Glasgow: Blackie. (305, 309)
1931 1. ANDRUETTO, G.: Sulle equazioni intrinseche dell' equilibrio elastico. Rend. Lincei
(6) 13, 489-494. (205A)
2. BATEMAN, H.: On dissipative systems and related variational principles. Phys.
Rev. (2) 38, 815-819. (129)
3. CoKER, E. G., and L. N. G. FrLoN: Photoelasticity. London: Cambridge Univ.
Press. (31, 312)
4. DE DoNDER, T.: Theorie invariantive de l'elasticite a deformations finies. Bull.
Sei. Acad. Belg. (5) 17, 1152-1157. (152, 245)
5. DUPONT, Y.: Quelques contributions a Ia theorie invariantive de l'elasticite. Bull.
Sei. Acad. Belg. (5) 17, 441-459. (19, 31, 104)
6. DUPONT, Y.: Sur la theorie invariantive de l'elasticite a deformations finies. c. R.
Acad. Sei., Paris 192, 873-875. (31)
7. MrMURA, Y.: On the foundation of the second law of thermodynamics. J. Hiroshima Univ. 1 (1930-1931), 43-53. (245)
8. MIMURA, Y.: On the equations of motion in thermodynamics. J. Hiroshima Univ.
1 (1930-1931), 117-123. (245)
9. SBRANA, F.: Sulla validita del teorema di Bernoulli per un fluido reale. Boll. Un.
Mat. Ital. 10, 77-78. (124)
10. SLEBODZINSKI, w.: Sur les equations canoniques de Hamilton. Bull. Sei. Acad.
Belg. (5) 17, 864-870. (150)
11. WEISSENBERG, K.: Die Mechanik deformierbarer Körper. Abh. Akad. Wiss. Berlin
No. 2. (241)
1932 1. ANDRUETTO, G.: Le formnie di Saint-Venant per gli spazi curvi a tre dimensioni.
Rend. Lincei (6) 15, 214-218. (84)
2. ANDRUETTO, G.: Le formnie di Saint-Venant perle varieta V,, a curvatura costante.
Rend. Lincei (6) 15, 792-797- (84)
3. BoNVICINI, D.: Sulle deformazioni non infinitesimi. Rend. Lincei (6) 16, 607-612.
(29, 86, 95, 104)
4. CrsoTTr, U.: Spostamenti rigidi finiti. Rend. Lincei (6) 16, 381-386. (55)
5. DE DoNDER, T.: L'affinite (III). Bull. Sei. Acad. Belg. (5) 18, 578-595, 888-898,
1124-1137; 19, 881-892, 1140-1152, 1364-1376 (1933); 20, 268-281 (1934).
Reprinted, with 8 supplementary papers, Paris 1934. (259)
6. DuPONT, Y.: Applications physico-chimiques de la thermodynamique des systemes
en mouvement. Bull. Sei. Acad. Belg. (5) 18, 83-94 = reprint of [1932, 5, suppl.,
23-34]. (259)
7. IWATSUKI, T., and MIMURA: On the adiabatic process of the thermodynamical
system in which the entropy cannot be defined. J. Hiroshima Univ. 2 (1931-1932),
127-138. (245)
8. LAMB, H.: 6th ed. of [1879, 2]. (See [1895, 2], [1906, 4], [1916, 3], [1924, 9].)
(256)
9. MASOTTI, A.: Sul teorema di König. Atti Pontif. Accad. Sei. Nuovi Lincei 85
(1931-1932), 37-42. (167)
10. ONSAGER, L., and R. M. Fuoss: Irreversible properties in electrolytes. Diffusion,
conductance, and viscous flow in arbitrary mixtures of strong electrolytes. J. Physic.
Chem. 36, 2689-2978. (295)
11. SrGNORINI, A.: Aleune proprieta di media nella elastostatica ordinaria. Rend.
Lincei (6) 15, 151-156. (220)
12. SrGNORINI, A.: Sollecitazioni iperastatiche. Rend. Ist. Lombardo (2) 65, 1-7. (221)
13. VAN DANTZIG, D.: Zur allgemeinen projektiven Differentialgeometrie, II. Xn+l
mit eingliedriger Gruppe. Proc. Akad. Wet. Amsterd. 35, 535-542. (150)
14. WuNDHEILER, A.: Kovariante Ableitung und die Cesarosche Unbeweglichkeitsbedingungen. Math. Z. 36, 104-109. (16, 150)
15. WUNDHEILER, A. W.: Rheonome Geometrie. Absolute Mechanik. Proc. matem.-
fiz. (Warsaw) 49, 97-142. (153)
768 C. TRUESDELL and R. TouPIN: The Classical Field Theories.
1933 1. AGOSTINELLI, C.: Le condizioni di Saint-Venant perle deformazioni di una varietä.
riemanniana generica. Rend. Lincei (6) 18, 529-533; (6) 19, 22-26 (1934). (84)
2. ARRIGHI, G.: Una generalizzazione dell' equazione di continuita. Rend. Lincei
(6) 18, 302-307. (157)
3. BJERKNES, V., J. BJERKNES, H. SoLBERG u. T. BERGERON: Physikalische Hydrodynamik. Berlin: Springer. xviii + 797 pp. (3, 185, 256A, 297)
4. FROMM, H.: Stoffgesetze des isotropen Kontinuums, insbesondere bei zähplastischem Verhalten. lng.-Arch. 4, 432-466. (148)
5. GALLI, A.: Sulle deformazioni pure infinitesime. Rend. Napoli (4) 3, 55-58.
6. HLAVATY, V.: Über eine Art der Punktkonnexion. Math. Z. 38, 135-145. (16)
7. KLoTTER, K.: Über die graphische Darstellung zugeordneter Spannungs- und Verzerrungszustand. Z. angew. Math. Mech. 13, 433-434. (24)
8. MüLLER, W.: Über den Impulssatz der Hydrodynamik für bewegte Gefäßwände
und die Berechnung der Reaktionskräfte der Flüssigkeit. Ann. Physik (5) 16,
489-512. (202)
9. ÜDONE, F.: Deformazioni finite e deformazioni infinitesime. Boll. Un. Mat. Ital.
11, 238-241. (57)
10. SIGNORINI, A.: Sopra alcune questioni di statica dei sistemi continui. Ann. Scuola
Norm. Pisa (2) 2, 231-257- (216, 221, 222, 223)
11. TESAR, V.: A graphical representation of two-dimensional stress distributions.
]. Franklin Inst. 216, 217-224. (209)
12. VALCOVICI, V.: Sur le mouvement d'un solide rigide. Bull. math. phys. Ecole
Polyt. Bucarest 3 (1932-1933). 171-179. (160)
13. VAN LEHRBERGHE, G.: La conservation de l'energie dans les milieux continus.
Publ. Assoc. lngen. Mons 47, 601-618. (241)
1934 1. BRAHTZ, J.: Notes on the Airy stress function. Bull. Amer. Math. Soc. 40, 427-430.
(224)
2. FINZI, B.: Integrazione delle equazioni indefinite della meccanica dei sistemi continui. Rend. Lincei (6) 19, 578-584, 620-623. (84, 211, 224, 227, 229, 230)
3. PALATIN!, A.: Sulle condizioni di Saint-Venant in una V,. qualsivoglia. Rend.
Lincei (6) 19, 466-469. (84)
4. PHILLIPS, H. B.: Stress functions. J. Math. Phys. 13, 421-425. (224, 228)
5. PucHER, A.: Über den Spannungszustand in gekrümmten Flächen. Beton und
Eisen 33, 298-304. (229)
6. SBRANA, F.: Moti fluidi dipendenti da due sole coordinate. Atti Pontif. Accad.
Sei. Nuovi Lincei 87 (1933-1934), 256-262. (162)
7. SCHOUTEN, J., u. E. VAN KAMPEN: Beiträge zur Theorie der Deformation. Proc.
mat.-fiz. 41, 1-19. (150)
8. ScHOUTEN, J., u. J. HAANTJES: Über die konforminvariante Gestalt der Maxwellsehen Gleichungen und der elektromagnetischen Impulsenergiegleichungen.
Physica, Haag 1, 869-872. (266)
9. TRICOMI, F.: Un'interpretazione intuitiva del rotore e della condizione d'irrotazionalita. Rend. Lincei (6) 19, 399-401. (87)
10. VAN DANTZIG, D.: Electromagnetism, independent of metrical geometry, I. The
foundations; II. Variational principles and a further generalization of the theory;
III. Mass and Motion; IV. Momentum and energy; waves. Proc. Akad. Wet.
Amsterd. 37, 521-525, 526-531, 643-652, 825-836. (166, 238, 270, 310)
11. VAN DANTZIG, D.: The fundamental equations of electromagnetism, independent
of metrical geometry. Proc. Cambridge Phil. Soc. 30, 421-427. (266)
1935 1. BOGGIO, T.: Sull' integrazione delle equazioni idrodinamiche di Helmholtz. Rend.
Lincei (6) 21, 415-419. (101)
2. BONVICINI, D.: Sulla deformazione pura nel caso di spostamenti finiti e sulla
relazione di essa colla tensione nei corpi anisotropi. Ann. di Mat. (4) 13, 113-117.
(29, 95)
3. LEVI-CIVITA, T.: Movimenti di un sistema continuo ehe rispettano l'invariabilita
sostanziale del baricentro. Atti Pontif. Accad. Sei. Nuovi Lincei 88 (1934-1935),
102-106. Abstract in Pontif. Accad. Sei. Nuncius No. 37, 4 (1935). (165)
4. ScHoUTEN,J.A., u. D.J.STRUIK: Einführung indieneueren Methoden der Differentialgeometrie I, 2nd ed. Groningen: Noordhoff. xii + 202 pp. (150)
5. ScHWERDTFEGER, A.: Über schwerpunktbehaltende Bewegungen materieller Systeme.
Atti Pontif. Accad. Sei. Nuovi Lincei 88 (1934-1935), 302-307. (165)
6. SHAW, A. N.: The derivation of thermodynamical relations for a simple system.
Phil. Trans. Roy. Soc. Lond. A 234, 299-328. (251)
7. SOBRERO, L.: Dei significato meccanico della funzione di Airy. Ric. di Ingegneria
3, 77-80 = Rend. Lincei (6) 21, 264-269. (224)
8.
9.
10.
1936 1.
2.
3.
4.
5.
6.
7.
8.
9.
1937 1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
1938 1.
2.
3.
4.
5.
List of Works Cited. 769
SuDRIA, J.: L'action euclidienne de detormation et de mouvement. Mem. Sei.
Phys. Paris. No. 29, 56 pp. (60)
VAN MIEGHEN, J.: Thermodynamique des systemes non-uniformes en vue des
applications a Ia meteorologie. Geofys. Pub!. Norske Vid.-Akad. 10, No. 14, 18 pp.
(243, 262)
WEISSENBERG, K.: La mecanique des corps detormables. Arch. Sei. phys. nat.
Geneve (5) 17, 44-106, 130-171. (33, 42, 241)
BATEMAN, H.: Progressive waves of finite amplitude and some steady motions
of an elastic fluid. Proc. Nat. Acad. Sei. U.S.A. 22, 607-619. (224)
BERKER, A. R.: Sur quelques cas d'integration des equations du mouvement d'un
fluide visqueux incompressible. Paris and Lilie. 161 pp. (68)
CRocco, L.: Una nuova funzione di corrente per lo studio del moto rotazionale
dei gas. Rend. Lincei ( 6A) 23, 11 5- 124; transl. Eine neue Stromfunktion für die
Erforschung der Bewegung der Gase mit Rotation. Z. a~gew. Math. Mech. 17
(1937), 1-7. (123, 133, 161, 162)
HAMEL, G.: Ein allgemeiner Satz über den Druck bei der Bewegung volumbeständiger Flüssigkeiten. Mh. Math. Phys. 43, 345-363. (115, 121)
KILCHEVSKI, N.: Foundation and generalization of Hooke's law on the basis of a
law of economy of mass-energy and the problern of two non-linear measures in the
theory of elasticity [in Ukrainian]. Zh. Irrst. Mat. Akad. Nauk USRR. 1936-1937,
No. 2, 77-89. (211)
KosTYUK, A.: Equation of continuity [in Russian]. Uchenie Zapiski Kazanskogo
Univ. 964- 5, 44-59. (141)
MUNK, M.: On the common space integrals of aerodynamics. J. Aeronaut. Sei.
3, 243-247. (76)
PLATRIER, C.: Cinematique des milieux continus. Actual. sei. industr. No. 327,
34 pp. (12)
SYNGE, J. L.: Conditions satisfied by the velocity and the streamfunction in a
viscous liquid moving in two dimensions between fixed parallel planes. Proc.
Lond. Math. Soc. (2) 40, 23-36. (116)
EPSTEIN, S.: Textbook of Thermodynamics. New York. (240, 265)
HAMEL, G.: Potentialströmungen mit konstanter Geschwindigkeit. Sitzgsber.
preuss. Akad. Wiss., phys.-math. Kl. 5-20. (111)
KARMAN, T.v.: On the statistical theory of turbulence. Proc. Nat. Acad. Sei.
U.S.A. 23, 98-105. (101)
LAGALLY, M.: Die zweite Invariante des Verzerrungstensors. Z. angew. Math.
Mech. 17, 80-84. (121)
The Origin of CLERK MAXWELL's Electric Ideas, as Described in Familiar Letters
to William Thomson. Ed. J. LARMOR. Cambridge. (300)
MERLIN, E.: Un theoreme sur Je mouvement des fluides depourvus d'acceleration.
C. R. Acad. Sei., Paris 205, 1128-1130. (141)
MURNAGHAN, F. D.: Finite deformations of an elastic solid. Amer. J. Math. 59,
235-260. (14, 30, 31, 95, 232)
ODQVIST, F.: Equations de compatibilite pour un systeme de coordonnees triples
orthogonaux quelconques. C. R. Acad. Sei., Paris 205, 202-204. (57)
THEODOREsco, N.: Sur l'equilibre des milieux continus. Bull. math. phys. Ecole
Polyt. Bucarest 8 (1936-1937), 179-185. (223)
VAN DANTZIG, D.: Über das Verhältnis von Geometrie und Physik. C. r. Congr.
Int. Math. Oslo ( 1936) 2, 225-227. (266)
VAN DANTZIG, D.: Some possibilities of the future development of the notions
of space and time. Erkenntnis 7, 142-146. (7, 266)
ZAREMBA, S.: Sur une conceptjon nouvelle des forces interieures dans un fluide
en mouvement. Mem. sei. math. No. 82. (148, 293, 305)
BATEMAN, H.: The Iift and drag functions for an elastic fluid in two-dimensional
irrotational flow. Proc. Nat. Acad. Sei. U.S.A. 24, 246-251. (224)
BRILLOUIN, L.: Les Tenseurs en Mecanique et en Elasticite. Paris. (12, 31, 210,
246, 256A)
DE DONDER, T.: Sur Ia vitesse reactionelle. Bull. Acad. Belg., Cl. Sei. (5) 24, 15-18
(159A, 245)
EcKART, C.: The electrodynamics of material media. Phys. Rev. (2) 54, 920-923.
(236A)
KILCHEVSKI, N.: A new theory of the mechanics of continuous media [in Ukrainian]. Zbirnik Inst. Mat. Akad. Nauk URSR No. 1, 17-114. (33A, 34, 152, 211,
212, 21.3)
Handbuch der Physik, Bd. Ill/1. 49
770 C. TRUESDELL and R. TouPIN: The Classical Field Theories.
6. LAMPARIELLO, G.: Varieta sostanziali nel moto di un sistema continuo. Atti 1°
Congr. Un. Mat. Ital. (1937). 391-393. (75)
7. MEISSNER, W.: Thermodynamische Behandlung stationärer Vorgänge in Mehrphasensystemen. Ann. Physik (5) 32, 115-127. (159, 243, 258)
8. MERLIN, E.: Etude du mouvement d'un fluide parfait depourvu d'acceleration.
Ann. Ecole Norm. (3) 55, 223-255. (141)
9. MILNE-THoMsoN, L. M.: Theoretical Hydrodynamics. London: Macmillan. (71,
202)
10. PAILLOUX, H.: Mouvements fluides fournissant une suite de surfaces applicables
C. R. Acad. Sei., Paris 207, 319-321. (83)
11. RIABOUCHINSKI, D.: Comparaison de la methodedes variables (tp, tp1 , tp2 , t) a celles
des variables d'Euler et de Lagrange. C. R. Acad. Sei., Paris 206, 295-297. (70)
1939 1. BATEMAN, H.: The aerodynamics of reacting substances. Proc. Nat. Acad. Sei.
U.S.A. 25, 388-391. (159)
2. BIEZENO, C. B., u. R. GRAMMEL: Technische Dynamik. Berlin: Springer. (33)
3. BroT, M. A.: Non-linear theory of elasticity and the linearized case for a body
under initial stress. Phil. Mag. (7) 27, 468-489. (33A)
4. BroT, M. A.: Theory of elasticity with large displacements and rotations. Proc.
5th Int. Congr. appl. Mech., New York and London, 117-122. (33A)
5. BroT, M. A.: Theorie de l'elasticite du secend ordre avec application a la theorie
du flambage. Ann. Soc. Sei. Bruxelles (1) 59, 104--112. (33A)
6. CHAPMAN, S., and T. G. CowLING: The Mathematical Theory of Non-uniform
Gases. Cambridge. (215, 243, 295)
7. EICHENWALD, A.: La funzione hamiltoniana nei mezzi continui. Rend. Sem. Mat.
Fis. Milano 13, 15-34. (231)
8. FRITSCHE, B.: Die Spannungsellipse. Ing.-Arch. 10, 427-428. (21)
9. GuEST, J. J.: A graphical construction for stress. Phil. Mag. (7) 27, 445-448.
(23)
10. }ARDETZKY, W.: Zur Frage der Axiomatik des zweiten Hauptsatzes der Thermodynamik. Bull. Acad. Serbe Sei. (A) 5, 33-47. (245)
11. KAPPUS, R.: Zur Elasticitätstheorie endlicher Verschiebungen. Z. angew. Math.
Mech. 19, 271-285. 344-361. (55)
12. PucHER, A.: Über die Spannungsfunktion beliebig gekrümmter dünner Schalen.
Proc. 5th lnt. Congr. appl. Mech. Cambridge (1938), 134-139. (229)
13. SIGNORINI, A.: V alori medi delle caratteristiche dello stress in stereodinamica.
Rend. Lincei (6) 29, 536-541. (220)
1940 1. BALLABH, R.: Superposahle fluid motions. Proc. Benares Math. Soc. (2) 2, 69-79.
(112, 114) '
2. BALLABH, R.: Self-superposable fluid motions of the type ~ = ilu, etc. Proc. Benares
Math. Soc. (2) 2, 85-89. (114)
3. BroT, M. A.: Elastizitätstheorie zweiter Ordnung mit Anwendungen. Z. angew.
Math. Mech. 20, 89-99. (33A)
4. BR0NSTED, J. N.: The fundamental principles of energetics. Phil. Mag. (7) 29,
449-470. (245)
5. BR0NSTED, J. N.: The derivation of the equilibrium conditions in physical chemistry
on the basis of the work principle. J. Phys. Chem. 44, 699-712. (245)
6. CrsoTTI, U.: Elementi di media nella meccanica dei sistemi continui. Rend. Sem.
Mat. Fis. Milane 14, 128-138. (216, 219)
7. EcKART, C.: The thermodynamics of irreversible processes, I. The simple fluid.
Phys. Rev. (2) 58, 267-269. (257)
8. EcKART, C.: The thermodynamics of irreversible processes, li. Fluid mixtures.
Phys. Rev. (2) 58, 269-275. 924. (159, 159A, 215, 243, 245, 255, 259, 261, 295)
9. EISENHART, L. P.: An lntroduction to Differential Geometry with the Use of the
Tensor Calculus. Princeton: Univ. Press. x + 304 pp. (64)
10. FADLE, J.: Über eine zweckmäßige Darstellung des Mohrsehen Spannungskreises
zur unmittelbaren Auftindung sämtlicher Spannungs- Richtungspfeile. Ing.-Arch.
11, 319-322. (24)
11. FrNZI, B.: Principio variazionale nella meccanica dei continui. Rend. Accad.
Ital. (7) 1, 412-417. (232A)
12. GRIOLI, G.: Una proprieta di minimo nella cinematica delle deformazioni finite.
Boll. Un. Mat. Ital. (2) 2, 452-455. (42)
13. HELLUND, E. J.: Generalized theory of diffusion. Phys. Rev. (2) 57, 319-333.
(295)
14. LEVI-CIVITA, T.: Nozione adimensionale di vortice e sua applicazione alle onde
trocoidale di Gerstner. Acta Pontif. Acad. Sei. 4, 23-30. (91)
15.
16.
17.
18.
1941 1.
2.
3.
4.
5.
6.
7.
8.
9.
1942 1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
1943 1.
2.
3.
4.
5.
6.
List of Works Cited. 771
LocATELLI, P.: Sulla congruenza delle deformazioni. Rend. Ist. Lombardo 73 =
(3) 4 (1939-1940), 457-464. (234, 235)
LocATELLI, P.: Sul principio di Menabrea. Bol!. Un. Nat. Ital. (2) 2, 342-347.
(234)
LoHR, E.: Ein thermodynamischer Weg zum Planckschen Strahlungsgesetz. Z.
Physik 116, 454-468. (262)
WrsE, J. A.: Circles of strain. J. Aeronaut. Sei. 7, 438-440. (24)
DEUKER, E.-A.: Beitrag zur Theorie endlicher Verformungen und zur Stabilitätstheorie des elastischen Körpers. Deutsche Math. 5 (1940-1941), 546-562. (65,
66, 95, 210)
MEIXNER, J.: Zur Thermodynamik der Thermodiffusion. Ann. Physik (5) 39,
333-356. (159, 241, 243, 245, 255, 257, 295)
MuRNAGHAN, F. D.: The compressibility of solids under extreme pressures. Karman
Anniv. Vol., 121-136. (33, 33A)
MuNK, M.: On some vortex theorems of hydrodynamics. J. Aeronaut. Sei. 5,
90-96. (118)
RErSSNER, E.: A new derivation of the equations for the deformation of elastic
shells. Amer. J. Math. 63, 177-184. (212)
SERINI, R.: Deduzione de!le equazioni della dinamica dei sistemi continui senza
far uso del principio di d'Alembert. Bol!. Un. Mat. Ital. (2) 3, 281-283. (170}
SrGNORINI, A.: Sulle proprieta di media communi a tutti i sistemi continui. Rend.
Accad. Ital. (7) 2, 728-734. (222)
STRATTON, J. A.: Electromagnetic Theory. New York. (281, 283, 312)
SYNGE, J. L., and W. z. CHIEN: The intrinsic theory of elastic shells and plates.
Karman Anniv. Vol., 103-120. (64, 212, 213)
BERGMANN, P. G.: Introduction to the Theory of Re!ativity. New York. (271, 282)
CARSTOIU, I.: Nouveaux points de vue sur quelques theoremes fondamentaux
de Ia mecanique des fluides. Bull. Politech. Bucarest 13, 42-45. (110)
CrsoTTI, U.: Formule integrali relative alla meccanica dei sistemi continui. Atti
2° Congr. Un. Mat. Ital. (1940), 404-414. (216, 219)
ERTEL, H.: Ein neuer hydrodynamischer Erhaltungssatz. Naturwiss. 30, 543-544.
(130)
ER TEL, H.: Ein neuer hydrodynamischer Wirbe!satz. Meteor. Z. 59, 277-281. (130)
ERTEL, H.: Über das Verhältnis des neuen hydrodynamischen Wirbelsatzes zum
Zirkulationssatz von V. Bjerknes. Meteor. Z. 59, 385-387. (130)
ERTEL, H.: Über hydrodynamische Wirbelsätze. Physik. Z. 43, 526-529. (130)
HAY, G. E.: The finite displacement of thin rods. Trans. Amer. Math. Soc. 51,
65-102. (63A, 214)
MErXNER, J.: Reversible Bewegungen von Flüssigkeiten und Gasen. Ann. Physik
(5) 41, 409-425. (241, 259)
MorsrL, G. C.: Sur le passage des variables de Lagrange aux variables d'Euler en
hydrodynamique. Bull. Math. Soc. Roumaine Sei. 44, 55-58. (66)
SIGNORINI, A.: Deformazioni elastiche finite: elastieita di 2° grado. Atti 2° Congr.
Mat. Ital. Rome (1940), 56-71. (34)
STEWART, R. J.: The energy equation for a viscous compressible fluid. Proc.
Nat. Acad. Sei. U.S.A. 28, 161-164. (241)
TEDONE, G.: Qualehe applicazione di una proprieta di media dello stress. Bol!.
Un. Mat. ltal. (2) 4, 93-99. (220)
VERSCHAFFELT, J. E.: La thermomecanique de Ja diffusion des gaz. Bull. Sei.
Acad. Roy. Belg. (5) 28, 455-475. (159, 243)
VERSCHAFFELT, J. E.: Sur Ia thermomecanique des fluides en mouvement. Bull.
Sei. Acad. Roy. Belg. (5) 28, 476-489. (159, 243)
BoYER, C. B.: History of the measurement of heat. II. The conservation of energy.
Sei. monthly 57, 546-554. (240)
MEIXNER, J.: Zur Thermodynamik der irreveriblen Prozesse in Gasen mit chemisch
reagierenden, dissoziierenden und anregbaren Komponenten. Ann. Physik (5) 43,
244-270. (243, 245, 255, 259, 295)
MEIXNER, J.: Zur Thermodynamik der irreversiblen Prozesse. Z. physik. Chem.,
Abt. B 53, 235-263. (257, 259, 295)
NovozHILOV, V.: On a error in a hypothesis of the theory of shells. C. R. Acad.
Sei. SSSR. (Doklady) (N.S.) 38, 160-164. (213)
NovozHILOV, V., and P. FrNKELSHTEIN: On an error in the hypothesis of Kirchhof!
in the theory of shells [in Russian]. Prikl. Mat. Mekh. 7, 333-340. (213)
SrGNORINI, A.: Trasformazioni termoelastiche finite, Memoria 1 a. Ann. di Mat.
{4) 22, 33-143. (12, 27, 30, 31, 34, 37, .38, 39, 41, 45, 47, 50, 59, 210)
49*
772 C. TRUESDELL and R. TouPIN: The Classical Field Theories.
7. ToLOTTI, C.: Le equazioni lagrangiane della rneccanica dei sisterni continui in
coordinate generali. Rend. Acad. Napoli (4) 13, 1-9. (210)
8. ToNOLO, A.: Teoria tensoriale delle deforrnazioni finite dei corpi solidi. Rend. Sem.
Mat. Fadova 14, 43-117. (12, 29, 31, 34)
1944 1. CARSTOIU, 1.: Sur Ia condition des accelE\rations dans un fluide incornpressible.
Math. Tirni~oara 20, 172-173. (160)
2. CHIEN, W.-Z.: The intrinsic theory of thin shells and plates. Part I. General
theory. Quart. appl. Math. 1, 297-327. (64)
3. CrsoTTI, U.: Deforrnazione finite isotrope. Rend. Ist. Lornbardo 77 = (3) 8,
73-79- (27, 31)
4. CrsoTTI, U.: Influenza delle rotazione finite nelle deforrnazioni infinitesirne di un
solido elastico. Rend. Ist. Lornbardo 77 = (3) 8, 249-252. (55, 57)
5. DuPONT, P.: Representation triangulaire du tenseur des contraintes en elasticite.
C. R. Acad. Sei., Paris 218, 778-780. (24)
6. J ACOB, C.: Sur une interpretation de I'equation de continuite hydrodynarnique.
Bull. Math. Soc. Rournaine Sei. 46, 81-90. (160)
7. LAMM, 0.: Theorie der Diffusion ternärer Lösungen. Arkiv. Kerni Miner. Geol. 18A
(1944/1945), No. 2, 10pp. (295)
8. LEAF, B.: The principles of thermodynamics. J. Chern. Phys. 12, 89-98. (245)
9. LUNEBERG, R. K.: Mathematical Theory of Optics. Mimeographed notes of a series
of Iectures given at Brown University, Providence. (278)
10. RrvAUD, ] . : Remarques sur Je problerne de l'elasticite non lineaire. C. R. Acad.
Sei., Paris 218, 698-700. (210)
11. SETH, B. R.: Consistency equations of finite strain. Proc. Indian Acad. Sei. A 20,
336-339. (34)
12. VLASOV, V. Z.: Equations of cornpatibility of deformation in curvilinear co-ordinates
[in Russian]. Prikl. Mat. Mekh. 8, 301-306. (57)
1945 1. CHARREAU, A.: Sur des representations planes du tenseur des contraintes dans
un rnilieu continu. C. R. Acad. Sei., Paris 220, 642-643. (24)
2. CoWLING, T. G.: The electrical conductivity of an ionized gas in a rnagnetic field,
with applications to the solar atmosphere and the ionosphere. Proc. Roy. Soc.
Lond., Ser. A 183, 453-479- (295)
3. KuzMIN, R. 0.: On Maxwell's forrnulae in the theory of elasticity. C. R. Acad.
Sei. USSR. (Doklady) 49, 326-328. (227)
4. ÜNSAGER, L.: Theories and problerns of liquid diffusion. Ann. N.Y. Acad. Sei.
46, 241-265. (295)
5. REINER, M.: A rnathematical theory of dilatancy. Amer. J. Math. 67, 350-362.
(299)
6. TRUESDELL, C.: The rnembrane theory of shells of revolution. Trans. Arner. Math.
Soc. 58, 96-166. (213)
1946 1. CADY, W. G.: Piezoelectricity. London: McGraw-Hill. (312)
2. CARSTOIU, I.: Generalization des forrnules de Helmholtz et de Cauchy pour un
fluide visqueux incornpressible. C. R. Acad. Sei., Paris 223, 1095-1096. (101)
3. CARSTOIU, I.: Sur le vecteur tourbillon de I'acceleration et les fonctions qui s'y
rattachent. Bull. Cl. Sei. Acad. Roumaine 29, 207-214. (101, 109)
4. CARSTOIU, I.: Sur le rnouvernent tourbillonaire a !J = const. d'un fluide parfait
incompressible. Bull. Cl. Sei. Acad. Rournaine 28, 589-592. (138)
5. HAYES, W. D.: Transformation groups of the thermodynamic variables. Quart.
appl. Math. 4, 227-232. (251)
6. KAMPE DE FERIET, J.: Sur Ia decroissance de l'energie cim\tique d'un fluide visqueux
incompressible occupant up. dornaine plan borne. C. R. Acad. Sei., Paris 223,
1096-1098. (93, 116) .
7. LEAF, B.: Phenomenological theory of transport processes in fluids. Phys. Rev.
(2) 70, 748-758. (215, 243, 255, 295) .
8. ToRRE, C.: Über den plastischen Kqrper von Prandtl. Zur Theorie der Mohrsehen
Grenzkurve. Öst. Ing.-Arch. 1, 36- SO. (24)
9. VALcovrcr, V.: Sur une interpretation cinematique du tourbillon et sur Ia rotation
des directions principales de Ia deformation. Mathernatica (Timi~oara) 22, 57-65.
(86)
10. WEYL, H.: The Classical Groups, their Invariants and Representations. Princeton.
(268)
1947 1. CALDONAZZO, B.: Sui rnoti liberi di un mezzo continuo. Ann. di Mat. (4) 26,43-55.
(141)
2. CARSTOIU, I.: De Ia circulation dans un fluide visqueux incompressible. C. R. Acad.
Sei., Paris 224, 534-535. (136)
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
1948 1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
List of Works Cited. 773
CARSTOIU, I.: Recherehes sur Ia theorie des tourbillons. MS thesis, Univ. Paris.
(123, 129)
DEWULF, N.: Cercle de Mohr et coniques d'elasticite. Houille Elanehe 2, 481-487.
(24)
FrNCK, J. L.: On metastable states of equilibrium in thermodynamics. J. Franklin
Inst. 243, 1 -12. (253)
HICKS, B., P. GUENTHER and R. WASSERMAN: New formu]ation of the equations
for compressible flow. Quart. appl. Math. 5, 357-361. (132)
JUNG, F.: Der Culmannsche und der Mohrsehe Kreis. Öst. Ing.-Arch. 1, 408-410.
(24)
KAMPE DE FERIET, J.: On a property of the Laplacian of a function in a two dimensional bounded domain, when the first derivatives of the function vanish at the
boundary. Math. Mag. 21, 74-79. (116)
MrCHAL, A. D.: Matrix and Tensor Ca!culus. New York: John Wiley & Sons;
London: Chapman-Hall, Ltd. xiii + 132 pp. (14, 19)
MouFANG, R.: Volumentreue Verzerrungen bei endlichen Formänderung. Z. angew.
Math. Mech. 25-27, 209-214. (33)
PAILLOUX, H.: Sur !es equations du mouvement des fluides parfaits. C. R. Acad.
Sei., Paris 225, 1122-1124. (211)
PRIGOGINE, I.: Etude Thermodynamique des Phenomenes Irreversibles. Paris
and Liege: Dunod-Desoer. 143 pp. (157, 158, 215, 243, 245,255, 259, 261, 295, 306)
RrVLIN, R. S.: Hydrodynamics of non-Newtonian fluids. Nature 160, 611-613.
(299)
So!IIMERFELD, A.: Vorlesungen über theoretische Physik. Il, Mechanik deformierbarer Medien, 2nd ed. Wiesbaden. (302)
SwAINGER, K.: Stress-strain compatibility in greatly deformed engineering metals.
Phi!. Mag. (7) 38, 422-439. (33A)
SwAINGER, K. H.: Large strains and displacements in stress-strain problems.
Nature, Land. 160, 399-400. (33A)
TRUESDELL, C. A., and R. N. ScHWARTZ: The Newtonian mechanics of continua.
U.S. Nav. Ord. Lab. Mem. 9223. (293)
TRUESDELL, C., and R. PRIM: Zorawski's kinematical theorems. U.S. Nav. Ord.
Lab. Mem. 9354. (75)
BALLABH, R.: On coincidence of vortex and stream lines in ideal liquids. Ganita
1, 1-14. (112)
BrLIMOVITCH, A.: Aires et volumes velocidiques et hodographiques dans un mouvement du fluide. Acad. Serbe Sei. Pub!. Inst. Math. 2, 37-52. (115, 160)
BYUSHGENS, S.: Geometry of steady flow of an ideal incompressible fluid [in
Russian]. Isv. Akad. Nauk SSSR., Ser. Mat. 12, 481-512. (110)
CARSTOIU, I.: Sur certaines formnies integrales dans le mouvement d'un fluide.
C. R. Acad. Sei., Paris 227, 1337-1339. (160)
CASTOLDI, L.: Deduzione variazionale delle equazioni della dinamica dei continui
deformabili. Nuovo Cim. (9) 5, 140-149. (210)
CASTOLDI, L.: Superficie e linee di Bernoulli nel moto stazionario di un fluido reale.
Atti Accad. Ligure 4 (1947). 21-25. (124)
CHIEN, W.-Z.: Derivation of the equations of equilibrium of an elastic shell from
the general theory of elasticity. Sei. Rep. Tsing-Hua Univ. A 5, 240-251. (213)
CouRANT, R., and K. FRIEDRICHS: Supersonic Flow and Shock Waves. New York:
Interscience. xvi + 464 pp. (241, 253, 265)
DoRN, J. E., and A. J. LATTER: Stress-strain relations for finite elastoplastic deformations. J. appl. Mech. 15, 234-236. (33A)
EcKART, C.: The thermodynamics of irreversible processes IV: The theory of elasticity and anelasticity. Phys. Rev. (2) 73, 373-382. (34, 96)
FrNCK, J. L.: Thermodynamics, Part I: The second law from the standpoint of
the equation of state; Part II: Work, heat, and temperature concepts, and an
examination of the temperature scale. J. Franktin Inst. 245, 301-317, 365-378.
(253)
LoHR, E.: Quantenstatistik und Kontinuumsphysik. Z. Naturfosch. 3a, 625-636.
(262)
MANARINI, M.: Sulle equazioni della dinamica dei fluidi perfetti. Ball. Un. Mat.
Ital. (3) 3, 111 -114. (211)
MANARINI, M.: Sui paradossi di d' Alembert e di Brillouin nella dinamica dei fluidi.
Rend. Lincei (8) 4, 427-433. (202)
MooNEY, M.: The thermodynamics of a strained elastomer. I. General analysis.
J. appl. Phys. 19, 434-444. (33A)
774
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
1949 1.
2.
3.
4.
5.
6.
7.
C. TRUESDELL and R. TouPIN: The Classical Field Theories.
MoREAU, J.-J.: Sur deux theoremes generaux de Ia dynamique d'un milieu incompressible illimite. C. R. Acad. Sei., Paris 226, 1420-1422. (117)
NEMENYI, P., and R. PRIM: Same properties of rotational flow of a perfect gas.
Proc. Nat. Acad. Sei. U.S.A. 34, 119-124; 35, 116 (1949). (132)
NovozHILOV, V. V.: Foundations of the Nonlinear Theory of Elasticity [in Russian].
(We have seen this work only in the English translation by F. BAGEMIHL, H. KROMM
and W. SEIDEL, Rochester 1953.) (12, 34, 36, 55, 56, 210)
PACELLA, G. B.: Su una proprieta della meccanica dei corpi continui e unadeduzione
geometrica della !egge di Hooke. Rend. Lincei (8) 5, 31-37. (209)
PLATRIER, C.: Equations universelles de l'equilibre des milieux les plus generaux
a tenseur symetrique. Bull. Acad. Belg., CL Sei. 34, 274-277. (210)
PRIM, R.: Extension of Crocco's theorems to flows having a non-uniform stagnation
enthalpy. Phys. Rev. (2) 73, 186. (133)
PRIM, R. C.: On doubly-laminar flow fields having a constant velocity magnitude
along each stream-line. U.S. Naval Ord. Lab. Mem. 9762. (111)
REINER, M.: Elastieity beyond the elastic Iimit. Amer. J. Math. 70, 433-446.
(32, 33, 33A, 95, 303)
RICHTER, H.: Das isotrope Elastizitätsgesetz. Z. angew. Math. Mech. 28, 205-209.
(33, 33A)
RICHTER, H.: Bemerkung zum Maufangsehen Verzerrungsdeviator. Z. angew.
Math. Mech. 28, 126-127. (33)
RIVLIN, R. S.: The hydrodynamics of non-Newtonian fluids I. Proc. Roy. Soc.
Lond. A 193, 260-281. (299)
RIVLIN, R. S.: Large elastic deformations of isotropic materials IV. Further
developments of the generat theory. Phil. Trans. Roy. Soc. A 241, 379-397. (30)
STRANG, J. A.: Superposahle fluid motions. Comm. fac. sei. Ankara 1, 1-32. (114)
SWAINGER, K. H.: Strain energy in greatly deformed elastic or inelastic anisotropic engineering metals. J. Franklin Inst. 245, 501-516. (33A)
SwAINGER, K. H.: Large displacements with small strains in loaded structures.
J. appl. Mech. 15,45-52. (33A)
ToLMAN, R. C., and P. C. FINE: On the irreversible production of entropy. Rev.
Mod. Phys. 20, 51-7]. (257)
TRUESDELL, C.: On the differential equations of slip flow. Proc. Nat. Acad. Sei.
U.S.A. 34, 342-347. (293)
TRUESDELL, C.: A new definition of a fluid. U.S. Nav. Ord. Lab. Mem. 9487. (293)
TRUESDELL, C.: On the total vortieity of motion of a continuous medium. Phys.
Rev. (2) 73, 510-512. (131)
TRUESDELL, C.: On the transfer of energy in continuous media. Phys. Rev. (2)
73, 513-515. (241)
TRUESDELL, C.: Generalisation de Ia formule de Cauchy et des theoremes de Helmholtz au mouvement d'un milieu continu quelconque. C. R. Acad. Sei., Paris
227, 757-759. (101, 129)
TRUESDELL, C.: Une formule pour le vecteur tourbillon d'un fluide visqueux
elastique. C. R. Acad. Sei., Paris 227, 821-823. (101)
WEBER, C.: Spannungsfunktionen des dreidimensionalen Kontinuums. Z. angew.
Math. Mech. 28, 193-197. (227)
ZELMANOV, A. N.: Application of convected co-ordinates in non-relativistic mechanics [in Russian]. Dokl. Akad. Nauk SSSR. 61, 993-996. (34, 66B, 95, 150,
210)
BERKER, R.: Sur certaines proprietes du rotationnel d'un champ vectoriel qui est
nul sur la frontiere de son domaine de definition. Bull. sei. math. (2) 73, 163-176.
Abstract, C. R. Acad. Sei., Paris 228, 1630-1632. (116, 119)
BERKER, R.: Inegalite verifee par l'energie einetique d'un fluide visqueux incompressible occupant un domaine spatial borne. Bull. tech. Univ. Istanbul 2, 41-50.
Summary, C. R. Acad. Sei., Paris 228, 1327-1329. (93)
CoRBEN, C.: The transformation theory of thermodynamics. Phys. Rev. (2) 76,
166. (251)
CuRnss, C. F., and J. 0. HIRSCHFELDER: Transport properlies of multicomponent
gas mixtures. J. Chem. Phys. 17, 550-555. (295)
DoNDER, T. DE, et F. H. VAN DEN DUNGEN: Sur [es prineipes variationnels des
milieux continus. Bull. Acad. Roy. Belg., Cl. Sei. (5) 35, 841-846. (236A)
ERGUN, A. N.: Same cases of superposable fluid motions. Comm. fac. sei. Ankara
2, 48-88. (114)
ER TEL, H., u. C.-G. RossBY: Ein neuer Erhaltungssatz der Hydrodynamik.
Sitzgsber. Akad. Wiss. Berlin, math. nat. Kl. 1949, 1 -11. (130)
List of Works Cited. 775
8. ERTEL, H., and C.-G. RossBY: A new conservation theorem of hydrodynamics.
Geofisica pura appl. 14, 189-193. (130)
9. ERTEL, H., u. H. KöHLER: Ein Theorem über die stationäre Wirbelbewegung
kompressibler Flüssigkeiten. Z. angew. Math. Mech. 29, 109-113. (139, 163)
10. FINCK, J. L.: Thermodynamics: the meaning of the first law, and its relation to
the behavior of material systems. Phys. Rev. (2) 67, 166. (253)
11. FINZI, B., e M. PASTOR!: Calcolo Tensoriale e Applicazioni. Bologna: Zanichelli.
vii + 427 pp. (164, 226)
12. GLEYZAL, A.: A mathematical formulation of the general continuous deformation
problem. Quart. appl. Math. 6, 429-437. (66B, 210)
13. HENCKY, H.: Mathematical principles of rheology. Research 2, 437-443. (33A,
90)
14. HrcKs, B. L.: On the characterization of fields of diabatic flow. Quart. appl.
Math. 6, 405-416. (132)
15. KoNDO, K.: A proposal of a new theory concerning the yielding of materials based
on Riemannian geometry. J. Japan Soc. appl. Mech. 2, 123-128, 146-151. (34)
16. LAMER, V. K.: Some current misconceptions of N. L.Sadi Carnot's memoir and
cycle. Seience 109, 598. (240)
17. MANARINI, M.: Sui paradossi di d' Alembert e di Brillouin nella dinamica dei
fluidi. Boll. Un. Mat. Ital. (3) 4, 352-353. (202)
18. McVITTIE, G. C.: A systematic treatment of moving axes in hydrodynamics. Proc.
Roy. Soc. Lond. A 196, 285-300. (144, 152, 154, 211, 241)
19. MoREAU, J.-J.: Sur l'interpretation tourbillonaire des surfaces de glissement. C. R.
Acad. Sei., Paris 228, 1923-1925. (186)
20. MoREAU, J .-J.: Sur Ia dynamique d'un ecoulement rotationnel. C. R. Acad. Sei.,
Paris 229, 100-102. (117)
21. NEMENYI, P., and R. PRIM: On the steady Beltrami flow of a perfect gas. Proc.
7th Int. Congr. Appl. Mech. (1948) 2, 300-314. (132)
22. PARTINGTON, J. R.: An Advanced Treatise on Physical Chemistry, vol. 1. London,
New York and Toronto. (239, 240, 245)
23. PERETTI, G.: Significato del tensore arbitrario ehe interviene neU' integrale generale delle equazioni della statica dei continui. Atti Sem. Mat. Fis. Univ. Modena
3, 77-82. (227)
24. PRIGOGINE, I.: Le domaine de Ia validite de Ia thermodynamique des phenomenes
irreversibles. Physica, Haag 15, 272-284. (259)
25. PRIM, R.: A note on the substitution prineiple for steady gas flow. J. appl. Phys.
20, 448-450. (132)
26. RicHTER, H.: Verzerrungstensor, Verzerrungsdeviator, und Spannungstensor bei
endlichen Formänderungen. Z. angew. Math. Mech. 29, 65-75. (32, 33)
27. SIGNORINI, A.: Trasformazioni termoelastiche finite, Memoria 2a. Ann. Mat. (4)
30, 1-72. (256A)
28. STORCH!, E.: lntegrazione delle equazioni indefinite della statica dei sistemi continui su una superfieie di rotazione. Rend. Lincei (8) 7, 227-231. (229)
29. SuPINO, G.: Sul moto irrotazionale dei liquidi viscosi, I. Rend. Lincei (8) 6, 615-620.
(113)
30. SWAINGER, K. H.: Saint-Venant's and Filon's finite strains: definitions non-linear
in displacement gradients. Nature, Lond. 164, 23-24. (33A)
31. THOMAS, T. Y.: The fundamental hydrodynamical equations and shock conditions
for gases. Math. Mag. 22, 169-189. (192)
32. TONOLO, A.: Sopra una classe di deformazioni finite. Ann. Mat. (4) 29, 99-114.
(27)
33. TRUESDELL, C.: A new definition of a fluid, I. The Stokesian fluid. Proc. 7th lnt.
Congr. Appl. Mech. (1948) 2, 351-364 = U.S. Naval Res. Lab. Rep. No. P-3457.
(293, 298, 299)
34. TRUESDELL, C.: A new definition of a fluid, Il. The Maxwellian fluirj. U.S. Naval
Res. Lab. Rep. No. P-3553 = [1351, 27]. (293, 307)
35. TRUESDELL, C.: Deux formes de Ia transformation de Green. C. R. Acad. Sei.,
Paris 229, 1199-1200. (118)
36. WEISSENBERG, K.: Geometry of rheological phenomena (1946-1947). The Prineiples of Rheological Measurement, pp. 36-65. London. (32)
37. WEISSENBERG, K.: Abnormal substances and abnormal phenomena of flow. Proc.
Int. Congr. Rheology (1948), 1-29 - I-46. (42)
38. WEYL, H.: Shock waves in arbitrary fluids. Comm. pure appl. Math. N.Y. Univ.
2, 103- 122. (265)
776
39.
1950 1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
C. TRUESDELL and R. TOUPIN: The Classica! Field Theories.
ZERNA, W.: Beitrag zur allgemeinen Schalenbiegetheorie. Ing.-Arch. 17, 149-164.
(213)
ALFVEN, H.: Cosmical Electrodynamics. London. (313)
BLOKH, V.: Stress functions in the theory of elasticity [in Russian]. Prikl. Mat.
Mekh. 14, 415-422. (227, 228)
CASTOLDI, L.: Linee sostanziali inestese neUe deformazioni finite dei continui
materiali. Atti Accad. Ligure 6, 165-169. (26)
CASTOLDI, L.: Linee sostanziali nel moto un continuo deformabile e moti con linee
di f!usso (e di corrente) "sostanzialmente permanente". Rend. Ist. Lombardo (3)
14, 259-264. (75)
CRocco, L.: On a kind of stress-function for the study of non-isentropic two-dimensional motion of gases. Proc. 7th Int. Congr. Appl. Mech. London (1948) 2, 315-329.
(224)
DELVAL, J.: Le principe de Ia moindre cantrainte applique a Ia dynamique des
fluides incompressibles. Bull. Acad. Roy. Belg., Cl. Sei. (5) 36, 639-648. (130)
ERTEL, H.: Ein Theorem über asynchron-periodische Wirbelbewegungen kompressibler Flüssigkeiten. Mise. Acad. Berol. 1, 62-68. (135)
GoRDON, A. N.: A linear theory of finite strain. Nature, Land. 166, 657. (33A)
GREEN, A. E., and W. ZERNA: The equilibriurrt of thin elastic shells. Quart. J.
Mech. appl. Math. 3, 9-22. (213)
GREEN, A. E., and W. ZERNA: Theory of elasticity in general Coordinates. Phi!.
Mag. (7) 41, 313-336. (12, 34, 66B)
HöLDER, E.: Über die Variationsprinzipe der Mechanik der Kontinua. Ber. Verh.
Akad. Wiss. Leipzig 97, No. 2, 13 pp. (236A)
IRVING, J ., and J. KIRKWOOD: The statistical mechanical theory of transport
processes. IV, The equations of hydrodynamics. J. Chem. Phys. 18, 817-829. (1)
JoHNSON, M. H., and E. 0. HuLBERT: Diffusion in the ionosphere. Phys. Rev.
(2) 79, 802-807. (295)
KoNDO, K.: On the dislocation, the group of holonomy and the theory of yielding.
J. Japan Soc. appl. Mech. 3, 107-11 o. ( 34)
KoNDO, K.: On the fundamental equations of the theory of yielding. J. Japan
Soc. appl. Mech. 3, 184-188. (34)
KoNno, K.: The mathematical analyses of the yield point. J. Japan Soc. appl.
Mech. 3, 188-195; 4, 4-8, 35-38 (1951). (34)
MASON, W. P.: Piezoelectric Crystals and their Application to Ultrasonics. Toronto,
New York and London: D. van Nostrand. (312)
MoREAU, J .-J.: Relations generales directes entre !es actions aerodynamiques et
!es elements tourbillonaires. Actes Congr. lnt. Mec. Poitiers 4, 6 pp. (117)
MoRINAGA, K., and T. NöNo: On stress-functions in general Coordinates. J. sei.
Hiroshima Univ. A 14, 181-194. (227, 230)
NADAI, A.: Theory of Flow and Fracture of Solids, 1, 2nd ed. New York. (24)
Ein Kapitel aus der Vorlesung von F. NEUMANN über mechanische Wärmetheorie.
Königsberg 1854/55. Ausgearbeitet von C. NEUMANN. Herausgeg. von E. R. NEUMANN.
Abh. bayer. Akad. Wiss., math.-nat. Kl. (2), No. 59. (245)
0LDROYD, J. G.: On the formulation of rheological equations of state. Proc.
Roy. Soc. Land. A 200, 523-541. (66, 66B, 104, 150, 293)
OLDROYD, J. G.: Finite strains in an anisotropic elastic continuum. Proc. Roy.
Soc. Land. A 202, 345-358. (33A, 34)
PRIM, R., and C. TRUESDELL: A derivation of Zorawski's criterion for permanent
vector-lines. Proc. Amer. Math. Soc. 1, 32-34. (75)
SCHLÜTER, A.: Dynamik des Plasmas I. Z. Naturforsch. Sa, 72-78. (295)
SEUGLING, W. R.: Equations of compatibility for finite deformation of a continuous
medium. Amer. Math. monthly 57, 679-681. (34)
STORCH!, E.: Sulle equazioni indefinite della statica delle membrane tese su generiebe superficie. Rend. Lincei 8, 116-120. (229)
STORCH!, E.: Integrazione delle equazioni indefinite della statica dei veli tesi su
una generica superficie. Rend. Lincei 8, 326-331. (229)
SwAINGER, K. H.: Non-coaxiality of principal normal stresses and the 'strain'
ellipsoid in the classical theory an infinitesimal deformation. Nature, Land. 165,
159-160. (33A)
SwAINGER, K. H.: Reply to critism of Gordon [1950, 8]. Nature, Land. 166,
657-659. (33A)
SYNGE, J. L.: Note an the kinematics of plane viscous motion. Quart. appl. Math.
8, 107-108. (116)
32.
33.
34.
35.
36.
1951 1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
List of Works Cited. 777
TRUESDELL, C.: A new definition of a fluid, I. The Stokesian fluid. J. Math.
pures appl. (9) 29, 215-244. (293, 298, 299)
TRUESDELL, C.: Bernoulli's theorem for viscous compressible fluids. Phys. Rev.
77, 535-536. (124)
TRUESDELL, C.: On the balance between deformation and rotation in the motion
of a continuous medium. J. Washington Acad. Sei. 40, 313-317. (115, 160)
WEYL, H.: Space-time-matter (English transl. of [1921, 6], with a new preface).
New York: Dover (266, 310)
ZERNA, W.: Allgemeine Grundgleichungen der Elasticitätstheorie. Ing.-Arch. 18,
211-220. (34)
BEGHIN, H.: Sur Ia notation de travail dans Ia mecanique du continu. Ann. Inst.
Fourier Grenoble 2 (1950), 173-184. (217)
BJORGUM, 0.: On Beltrami vector fields and flows, Part I. Univ. Bergen Arbok
1951, Naturv. rekke No. 1. (112, 117)
DEDECKER, P.: Sur Je theoreme de Ia circulation de V. Bjerknes et Ia theorie
des invariants integraux. Pub!. Inst. R. Meteorol. Belg. misc. No. 36, 63 pp.
(129)
DEDECKER, P.: Sur le theoreme de Ia circulation de V. Bjerknes. Mem. Inst.
R. Meteorol. Belg. 48, 4 pp. (129)
DIRAC, P. A. M.: A new classical theory of the electron, I. II, III. Proc. Roy. Soc.
Lond. A209, 291-296; A212, 330-339; A223, 438-445. (276)
FINCK, J. L.: Thermodynamics from a Generalized Standpoint. New York. _ (253)
FrLONENKo-BoRODICH, M_ M_: The problern of the equilibrium of an elastic parallelepiped subject to assigned Ioads on its boundaries [in Russian]. Prikl. Mat. Mekh.
15, 137-148. (227)
FrLONENKO-BORODICH, M_ M.: Two problems on the equilibrium of an elastic parallelepiped [in Russian]. Prikl. Mat. Mekh. 15, 563-574. (227)
GALIMOV, K. Z.: Invariant form of the conditions of compatibility in finite deformation [in Russian]. Dokl. Akad. Nauk SSSR. 77, 577-580. (34)
GrESE, J. H.: Streamfunctions for three-dimensional flows. J. Math. Phys. 30,
31-35. (163)
HAASE, R.: Zur Thermodynamik der irreversiblen Prozesse. Z. Naturforsch. 6a,
420-437. 522-540. (245)
IRVING, J. H., and R. W. ZwANZIG: The statistical mechanical theory of transport
processes. V. Quantum hydrodynamics. J. Chem. Phys. 19, 1173-1180. (1, 2)
J ARRE, G.: Sul moto relativo nei mezzi continui. Atti Accad. Torino 85, 183-191.
(146)
J OHNSON, M. H.: Diffusion as hydrodynamic motion. Phys. Rev. (2) 84, 566-568.
(295)
LonGE, A. S.: The compatibility conditions for !arge strains. Quart. J. Mech.
appl. Math. 4, 85-93. (34)
LonGE, A. S.: On the use of convected coordinate systems in the mechanics of
continuous media. Proc. Cambridge Phi!. Soc_ 47, 57 5-584. (66, 66 B, 293)
MAZUR, P., et I._PRIGOGINE: S_ur l'hydrodynamique des melanges liquides de
He3 et He4• Physica, Haag 17, 680-693. (215, 243, 259)
MuRNAGHAN, F. D.: Finite Deformation of an Elastic Solid. New York and London. (12)
PoPov, S. G.: Remark on the integrals of Bernoulli and Lagrange (Cauchy).
[in Russian]. Moskov. Gos. Univ. Zap. 152, Mekh_ 3, 43-46. (125)
FRAGER, W., and P. G. HoDGE jr.: Theory of Perfectly Plastic Solids. New York.
(24)
PRIGOGINE, 1., et P. MAZUR: Sur deux formulations de l'hydrodynamique et Je
problerne de l'helium liquide II. Physica, Haag 17, 661-679. (215, 243, 258, 259)
RrvLIN, R. S., and D. W. SAUNDERS: Large elastic deformations of isotropic
materials, VII. Experiments on the deformation of rubber. Phi!. Trans. Roy.
Soc. Land. A 243, 251-288. (40)
ScHLÜTER, A.: Dynamik des Plasmas II. Z. Naturforsch. 6a, 73-78. (295)
SMITH-WHITE, W. B.: The Poisson-Kelvin hypothesis and the theory of dielectrics.
J. Roy. Soc. N.S. Wales 85,82-112- (283)
SYNGE, J. L.: Conditions satisfied by the expansion and vorticity of a viscous
fluid in a fixed container. Quart. appl. Math. 9, 319-322. (116)
TORRE, C.: Über die physikalische Bedeutung der Mohrsehen Hüllkurve. Z. angew.
Math. Mech. 31, 275-277. (24)
TRUESDELL, C.: A new definition of a fluid. I I. The Maxwellian fluid. J. Math.
pures appl. 30, 111-158. (293, 304, 307)
778
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
1952 1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
C. TRUESDELL and R. TouPIN: The C!assical Field Theories.
TRUESDELL, C.: Vorticity averages. Canad. J. Math. 3, 69-86. (117, 118, 131)
TRUESDELL, C.: A form of Green's transformation. Amer. J. Math. 73, 43-47.
(112, 117)
TRUESDELL, C.: Caracterisation des champs vectoriels qui s'annulent sur une
frontiere fermee. C. R. Acad. Sei., Paris 232, 1277-1279. (116)
TRUESDELL, C.: Analogue tri-dimensionnel au theoreme de M. Synge sur !es
champs vectoriels plans qui s'annulent sur une frontiere fermee. C. R. Acad. Sei.,
Paris 232, 1396-1397· (116)
TRUESDELL, C.: Vereinheitlichung und Verallgemeinerung der Wirbelsätze ebener
und rotationssymmetrischer Gasbewegungen. Z. angew. Math. Mech. 31, 65-71.
An abstract under the title "A new vorticity theorem" appears in Proc. Int. Congr.
Math. (1950), 1, 639-640 (1952). (132, 133)
TRUESDELL, C.: On Ertel's vorticity theorem. Z. angew. Math. Phys. 2, 109-114.
(130)
TRUESDELL, C.: Proof that Ertel's vorticity theorem holds in average for any
medium suffering no tangential acceleration on the boundary. Geofis. pura appl.
19, No. 3-4, 1-3. (131)
TRUESDELL, C.: On the equation of the bounding surface. Bull. Tech. Univ. Istanbul
3, 71-77. (182, 184)
TRUESDELL, C.: On the velocity of sound in fluids. J. Aeronaut. Sei. 18, 501. (297)
VAN DEN DuNGEN, F.: Note on the Hamel-Synge theorem. Quart. appl. Math.
9, 203-204. (116)
VERSCHAFFELT, J. E.: La thermomecanique des phenomenes de transport. J .Phys
Radium 12, 93-98. (159)
WHITTAKER, E. T.: A History of the Theories of Aether and Electricity, vol. 1.
New York. (266, 276, 283, 289, 302, 308)
ARZHANIKH, I. S.: Tensor stress functions of hydrodynamics [in Russian]. Dokl.
Akad. Nauk SSSR. 83, 195-198. (211, 230)
CoBURN, N.: Intrinsic relations satisfied by the vorticity and velocity vectors in
fluid flow theory. Michigan Math. J. 1, 113-130; 2, 41-44 (1953). (99, 110)
DE GROOT, S. R.: Thermodynamics of Irreversible Processes. Amsterdam: NorthHolland Pub!. Co. xvi + 242 pp. (158, 243, 245, 255, 257, 259, 261, 295, 306)
ER TEL, H. : Über die physikalische Bedeutung von Funktionen, welche in der
C!ebsch-Tra~sformation der hydrodynamischen Gleichungen auftreten. Sitzgsber.
Akad. Wiss. Berlin, Kl. Math. allg. Naturw. 1952, No. 3. (135)
F:ENYES, I.: Die Anwendung der mathematischen Prinzipien der Mechanik in der
Thermodynamik. Z. Physik 132, 140-145. (251)
F:ENYES, I.: Ergänzungen zur axiomatischen Begründung der Thermodynamik.
Z. Physik 134, 95-100. (245)
GRAD, H.: Statistical mechanics, thermodynamics, and fluid dynamics of systems
with an arbitrary number of integrals. Comm. pure appl. math. 5, 455-494. (157,
205)
GRIOLI, G.: Relazioni quantitative per lo stato tensionale di un qualunque sistema
continuo e per Ia deformazione di un corpo elastico in equilibrio. Ann. di Mat.
(4) 33, 239-246. (221, 222)
GRIOLI, G.: Integrazione del problema della statica delle piastre omogenee di spessore
qualunque. Ann. Scuola Norm. Pisa (3) 6, 31-49. (222)
HEINRICH, G.: Der Energietransport in strömenden Medien. Z. angew. Math.
Mech. 32, 286-288. (217)
HERSHEY, A. V.: A review of the definitions of finite strain. Proc. 1st U.S. Nat.
Congr. Appl. Mech. (1951), 473-478. (33A)
KrRKWOOD, J. G., and B. CRAWFORD jr.: The macroscopic equations of transport
J. Physic. Chem. 56, 1048-1051. (243, 285, 295)
MoREAU, J .-J 0: Bilan dynamique d'un ecoulement rotationnel. J 0 Math. pures
appl. (9) 31, 355-375; 32, 1-78 (1953). (117, 130, 186, 297)
NARDINI, R.: Sul valore medio dello stress per particolari sollecitazioni. Ann. Univ.
Ferrara sez. VII (2) 1, 89-91. (220)
PARTINGTON, J. R.: Advances in thermodynamics. Nature, Lond. 170, 730-732.
(245)
PRIM, R. C.: Steady rotational flow of ideal gases ( 1949). J. Rational Mech. Anal.
1, 425-497. (111, 132, 133)
RrcHTER, H.: Zur Elasticitätstheorie endlicher Verformungen. Math. Nachr. 8,
65-73. (37)
STORCH!, E.: Le superficie eccezionali nella statica delle membrane. Rev. mat.
Univ. Parma 3, 339-360. (229)
19.
20.
21.
22.
23.
24.
1953 1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
List of Works Cited. 779
SZEBEHELY, V. G.: Generalization of the dimensionless frequency parameter in
unsteady flows. David Taylor Model Basin Rep. 833. (100)
ToNoLo, A.: Sopra un problema di Darboux della meccanica dei mezzi continui.
Ann. Univ. Ferrara sez. VII (2) 1, 103-109. (38)
TRUESDELL, C.: The mechanical foundations of elasticity and fluid dynamics.
J. Rational Mech. Anal. 1, 125-300. (12, 14, 17, 29, 33A, 34, 44, 46, 50, 57, 95,
96, 210, 256A, 293, 298, 299, 303)
TRUESDELL, C.: A program of physical research in classical mechanics. Z. angew.
Math. Phys. 3, 79-95. (2, 293, 298, 299)
TRUESDELL, C.: Vorticity and the Thermodynamic State in a Gas Flow. Mem.
sei. math. No. 119, Paris. (132, 133, 297)
\VHAPLES, G.: Caratheodory's temperature equations. J. Rational Mech. Anal.
1, 301-307. (245)
BILIMOVITCH, A.: Sur l'homogenisation des equations de nature velocidique.
Acad. Serbe Sei. Pub!. Inst. Math. 5, 29-34. (115, 160)
BJ0RGUM, 0., and T. GoDAL: On Beltrami vector vields and flows, Part I I. Univ.
Bergen Arbok 1952, Naturv. rekke No. 13.
CARSTOIU, I.: Sur la detormation d'une particule dans le mouvement d'un fluide.
C. R. Acad. Sei., Paris 236, 2209-2211. (101)
CASTOLDI, L.: Teoremi di Bernoulli per fluidi comprimibili viscosi. Atti Accad.
Ligure 9 (1952), 215-222. (125)
CASTOLDI, L.: Sui moti di fluidi reali per cui si verifica una esatta linearizzazione
della equazione dinamica. Atti Accad. Ligure 9 (1952), 222-227. (122, 127)
CoRSON, E. M.: Introduction to Tensors, Spinors and Relativistic Wave-Equations
(Relation Structure). New York. (282)
DARWIN, C.: Note on hydrodynamics. Proc. Cambridge Phi!. Soc. 49, 342-354.
(71)
DEFRISE, P.: Analyse geometrique de la cinematique des milieux continus. Inst.
R. Meteorol. Belg. Pub!. ser. B, No. 16, 63 pp. (12, 152, 153)
DE GROOT, S.: Hydrodynamics and thermodynamics. Proc. Symp. appl. Math.
4, 87-99. (245)
FoRTAK, H.: Zur Bedeutung der in der Clebsch-Transformation der hydrodynamischen Gleichungen auftretenden Funktionen. Acta Hydrophys. 1, 145-150. (135)
GEIRINGER, H.: Some recent results in the theory of an ideal plastic body. Adv.
appl. Mech. 3, 197-294. (300)
GRIOLI, G.: Proprietä. di media ed equilibrio elastico. Atti 4to Congr. Un. Mat.
Ital. (1951) 1, 68-77. (221, 222)
HoWARD, L. N.: Constant speed flows. Thesis, MS in Princeton University Library.
(111)
KILCHEVSKI, N. A.: Stress, velocity, and density functions in static and dynamic
problems in the mechanics of continuous media [in Russian]. Dokl. Akad. Nauk
SSSR. 92, 895-898. (228, 230)
LANG HAAR, H.: An invariant membrane stress function for shells. J. appl. Mech.
20, 178-182. (229)
LANG HAAR, H.: The principle of complementary energy in nonlinear elasticity
theory. J. Franklin Inst. 256, 255-264. (232A)
MASUDA, H.: A new proof of Lagrange's theorem in hydrodynamics. J. Phys.
Soc. Japan 8, 390-393- (136)
McVrTTIE, G. C.: A method of solution of the equations of classical gas dynamics
using Einstein's equations. Quart. appl. Math. 11, 327-336. (225, 230)
Mr~rcu, M.: Echilibrul mediilor continue cu deformäri märi. Acad. Repub. Pop.
Romane. Stud. Cerc. Mec. Metalurgie 4, 31-53. (12)
NYBORG, W. L.: Acoustic streaming equations: laws of rotational motion for fluid
elements. J. Acoust. Soc. 25, 938-944. (170)
PLATRIER, C.: Conditions d'integrabilite du tenseur de deformation totale dans
une transformation finie d'un milieu ä. trois dimensions. Ann. Ponts Chaussees
123, 703-709. (34)
PLATRIER, C.: Conditions d'integrabilite du tenseur de detormation totale dans
une transformation finie d'un milieu ä. trois dimensions. Bull. Acad. Roy. Belg.,
CL Sei. (5) 39, 490-494. (34)
FRAGER, W.: Three-dimensional plastic flow under uniform stress. Brown univ.
tech. rep. No. 95, August = Rev. fac. sei. Univ. Istanbul 19, 23-27 (1954). (142)
PRATELLI, A.: Principi variazionali nella meccanica dei fluidi. Rend. Ist. Lombardo
(3) 17 (86), 484-500. (94, 163)
780 C. TRUESDELL and R. TouPIN: The Classical Field Theories.
25. PRATELLI, A.: Sulla stazionarieta di significativi integrali nella meccanica dei
continui. Rend. Ist. Lombardo (3) 17 (86), 714-724. (227, 235)
26. REIK, H.: Zur Theorie irreversibler Vorgänge. Ann. Physik (6) 11, 270-284,
407-419, 420-428; 13, 73-96. (245)
27. REISSNER, E.: On a variational theorem for finite elastic deformations. J. Math.
Phys. 32, 129-135. (232A)
28. ScHAEFER, H.: Die Spannungsfunktionen des dreidimensionalen Kontinuums und
des elastischen Körpers. Z. angew. Math. Mech. 33, 356-362. (226, 227, 229)
29. STORCH!, E.: Sulle membrane aventi comportamento meeeanieo eceezionale. Rend.
Ist. Lombardo (3) 17 (86), 462-483. (229)
30. SzEBEHELY, V. G.: A measure of unsteadiness of time-dependent flows. Proe.
3rd Midwest Conf. on Fluid Mech., Univ. Minn., 221-231. (100)
31. TRUESDELL, C.: Two measures of vorticity. J. Rational Mech. Anal. 2, 173-217.
(Partial abstract in Proc. Int. Congr. theor. appl. Mech. Istanbul 1952.) (91, 100,
121)
32. TRUESDELL, C.: Corrections and additions to "The mechanieal foundations of
elasticity and fluid dynamics". J. Rational Mech. Anal. 2, 593-616. (12, 14,
28, 37, 150, 304)
33. TRUESDELL, C.: Generalization of a geometrical theorem of Euler. Comm. mat.
Helv. 27, 233-234. (142)
34. TRUESDELL, C.: La velocita massima nel moto di Gromeka-Beltrami. Rend. Lineei
(8) 13 (1952), 378-379- (121)
35. WHITTAKER, E. T.: A History of the Theories of Aether and Elastieity, vol. I I.
London: Nelson. (266)
36. YosHIMURA, Y.: On the natural shearing strain. Proc. 2nd Japan Mat. Congr.
appl. Mech. (1952), 1-4. (33A)
37. YoSHIMURA, Y.: On the definition of stress in the finite deformation theory. J.
Phys. Soc. Japan 8, 669-673- (210)
1954 1. CARSTOIU, I.: Vorticity and deformation in fluid mechanics. J. Rational Mech.
Anal. 3, 691-712. (101, 160)
2. CASTOLDI, L.: Le .,condizioni di congruenza" per deformazioni infinitesime non
lineari. Atti Ist. Veneto, Cl. sei. mat. nat. 112, 41-47. (54, 57)
3. CASTOLDI, L.: Sopra un classificazione dei comportamenti elastici dei mezzi deformabili. Atti Ist. Veneto, Cl. sei. mat. nat. 112, 1 7-30. (55)
4. DEAN, W. R.: Note on the motion of an infinite cylinder in a rotating viseous
liquid. Quart. J. Mech. appl. Math. 7, 25 7-262. (207)
5. DuGAS, R.: La Mecanique au XVIIe Siecle. Ed. Griffon, Neuchatel. 620 pp. (3)
6. ERICKSEN, J. L., and R. S. RrvLrN: Large elastic deformations of homogeneaus
anisotropic materials. J. Rational Mech. Anal. 3, 281-301. (49, 62, 233)
7. GREEN, A. E., and W. ZERNA: Theoretical Elastieity. Oxford: Clarendon Press.
xiii + 442 pp. (12, 66 B)
8. GüNTHER, W.: Spannungsfunktionen und Verträglichkeitsbedingungen der Kontinuumsmechanik Abh. Braunschweig. Wiss. Ges. 6, 207-219. (227, 229)
9. HIRSCHFELDER, J. 0., C. F. CURTISS, and R. B. BIRD: Moleeu!ar Theory of Gases
and Liquids. New York: Wiley. xxvi + 1219 pp. (215, 243, 259, 295)
10. KoNDO, K.: On the theory of'the mechanieal behavior of mieroscopieally nonuniform materials. Research notes Res. Assn. appl. Geom. (Tokyo) (2), No. 4,
37 pp. (34)
11. KRÖNER, E.: Die Spannungsfunktion der dreidimensionalen isotropen Elastizitätstheorie. Z. Physik 139, 175-188. Correction, Z. Physik 143, 374 (1955). (227)
11 A. LAMM, 0.: The formal theory of diffusion, and its relation to self-diffusion, Sedimentation equilibrium, and viscosity. Acta Chem. Scand. 8, 1120-1128. (295)
12. LANGHAAR, H., and M. STIPPES: Three-dimensional stress functions. J. Franklin
Inst. 258,371-382. (227)
13. LoDE, W.: Tensoren zur Berechnung großer Formänderungen. Kolloid-Z. 138,
28-38. (33A)
14. MANACORDA, T.: Sopra un prineipio variazionale di E. Reissner per Ia statica dei
mezzi continui. Bol!. Un. Mat. Ital. (3) 9, 154-159- (232A)
15. MAZZARELLA, F.: Determinazione delle componenti di secondo ordine della deformazione riferite ad un generico sistema di coordinate eurvilinee. Rend. Accad.
Napoli (4) 21, 107-114. (26)
16. 0RNSTEIN, W.: Stress functions of Maxwell and Morera. Quart. appl. Math. 12,
198-201. (227)
17. OsBORN, H.: The existence of conservation laws. Appl. math. stat. Iab. Stanford
tech. rep. 27 (70 pp.). Also Ann. of Math. 69, 105-118 (1959). (157)
List of Works Cited. 781
18. PARKER, E. N.: Tensor virial equations. Phys. Rev. (2) 96, 1686-1689. (219)
19. REIK, H.: Die Thermodynamik irreversibler Prozesse und ihre Anwendung auf
Transportphänomene. Z. Physik 137, 333-361, 463-493. (245)
20. REINER, M.: Second order effects in elasticity and hydrodynamics. Bull. Res.
Council Israel 3, 372-379. (33A)
21. ScHaUTEN, J. A.: Ricci-calculus, 2nd ed. Berlin: Springer. xx + 516 pp. (34, 150,
152, 268)
22. SIGNORINI, A.: Una espressiva applicazione delle proprieta di medio dello stress
comuni a tutti i sistemi continui. Studies math. mech. pres. R. v. Mises. New
York: Academic Press. 274-277. (222)
23. SwAINGER, K.: Analysis of Deformation. I. Mathematical Theory. London:
xix + 285 pp. (33A)
24. TRUESDELL, C.: The Kinematics of Vorticity. Indiana Univ. sei. ser. No. 19.
(12, 66A, 72, 78, 101, 102, 108, 110, 112, 114, 118, 119, 120, 121, 124, 125, 129,
130, 132, 133, 134, 136, 137, 138, 145)
25. TRUESDELL, C.: Rational fluid mechanics 1687-1765. L. Euleri Opera Omnia
(2) 12, IX-CXXV. (66A, 120, 130, 200)
26. WHITHAM, G. B.: A note on a paper by G. C. McVittie. Quart. appl. Math. 12,
316-318. (230)
1955 1. ADKINS, J.: A note on the finite plane-strain equations for isotropic incompressible
materials. Proc. Cambridge Phil. Soc. 51, 363-367. (47)
2. BILBY, B. A., R. BuLLOUGH and E. SMITH: Continuous distributions of dislocations:
a new application of the methods of non-Riemannian geometry. Proc. Roy. Soc.
Lond., Ser. A 231, 263-273. (34)
3. BJ0RGUM, 0.: On the physical boundary conditions in fluid dynamics. Univ.
Bergen Arbok, Naturv. rekke Nr. 4, 8 pp. (105)
4. CoTTER, B., and R. S. RrvLIN: Tensors associated with time-dependent stress.
Quart. appl. Math. 13, 177-182. (UJJ, 150, 293, 305)
5. ERICKSEN, J. L.: Note concerning the number of directions which, in a given
motion, suffer no instantaneous rotation. J. Washington Acad. Sei. 45, 65-66.
(78, 91)
6. ERICKSEN, J. L.: Deformations possible in every compressible, isotropic, perfectly
elastic material. J. Math. Phys. 34, 126-128. (34)
7. ERTEL, H.: Kanonischer Algorithmus hydrodynamischer Wirbelgleichungen.
Sitzgsber. Akad. Wiss. Berlin 1954, No. 4, 11 pp. (130)
8. ERTEL, H.: Ein neues Wirbel-Theorem der Hydrodynamik. Sitzgsber. Akad. Wiss.
Berlin, math. Kl. 1954, No. 5. 12 pp. (130)
9. FINZI, L.: Sulle equazioni di Pucher nell' equilibrio delle strutture a guscio. Rend.
Ist. Lombardo (3) 88, 907-916. (229)
10. GRIOLI, G.: Limitazioni per lo stato tensionale di un qualunque sistema continuo.
Ann. di Mat. (4) 39, 255-266. (222)
11. GYORGYI, G.: Die Bewegung des Energiemittelpunktes und der Energie-ImpulsTensor des elektromagnetischen Feldes in Dielektrika. Acta phys. Hung. 4, 121-131.
(284)
12. HEINRICH, G.: Der Energiestrom in elastischen Medien. Öst. Ing.-Arch. 9,
148-156. (217)
13. HERIVEL, J.: The derivation of the equations of motion of an ideal fluid by Hamilton's principle. Proc. Cambridge Phil. Soc. 51, 344-349. (236A)
14. HUNT, F. V.: Notes on the exact equations governing the propagation of sound
in fluids. J. Acoust. Soc. Amer. 27, 1019-1039. (257)
15. ]AIN, M. K.: The motion of an infinite cylinder in rotating non-Newtonian liquid.
Z. angew. Math. Mech. 35, 379-381. (207)
16. MARGUERRE, K.: Ansätze zur Lösung der Grundgleichungen der Elastizitätstheorie.
Z. angew. Math. Mech. 35, 242-263. (228)
17. McVITTIE, G. C.: Review of [1954. 24]. Bull. Amer. Math. Soc. 61, 356-358. (99)
18. NoLL, W.: On the continuity of the solid and fluid states. J. Rational Mech. Anal.
4, 13-81. (12, 37, 66B, 95, 207, 293, 303, 304, 305)
19. NoLL, W.: Die Herleitung der Grundgleichungen der Thermomechanik der Kontinua aus der statischen Mechanik. J. Rational Mech. Anal. 4, 627-646. (1, 262)
20. RICHTER, H.: Review of [1954, 23]. Zbl. Math. 56, 175-1i6. (33A)
21. RrvLIN, R. S., and J. L. ERICKSEN: Stress-deformation relations for isotropic
materials. J. Rational Mech. Anal. 4, 323-425. (33, 104, 143, 144, 149, 293, 304,
305)
22. ScHAEFER, H.: Die Spannungsfunktion einer Dyname. Abh. Braunschweig. Wiss.
Ges. 7, 107-112. (227)
782 C. TRUESDELL and R. TouPIN: The Classical Field Theories.
23. SPEISER, A.: Einleitung. L. Euleri Opera Omnia (1) 28, VII-XLIV. (33A)
24. THOMAS, T. Y.: On the structure of the stress-strain relations. Proc. Nat. Acad.
Sei. U.S.A. 41, 716-720. (148, 293)
25. THOMAS, T. Y.: Kinematically preferred co-ordinate systems. Proc. Nat. Acad.
Sei. U.S.A. 41, 762-770. (148, 293)
26. TRUESDELL, C.: I. The first three sections of Euler's treatise on fluid mechanics
( 1 766); II. The theory of aerial sound ( 1687- 1788); III. Rational fluid mechanics
(1765-1788). L. Euleri Opera Omnia (2) 13, VII-CXVIII. (66A)
27. TRUESDELL, C.: The simplest rate theory of pure elasticity. Comm. pure appl.
Math. 8, 123-132. (142, 304)
28. TRUESDELL, C.: Hypo-elasticity. J. Rational Mech. Anal. 4, 83-133, 1019-1020.
(223, 293, 304)
29. TRUESDELL, C.: Review of [1954, 23]. Math. Rev. 16, 307. (33A)
1956 1. BERKER, R.: Sur les equations de compatibilite relativesau mouvement d'un gaz.
C. R. Acad. Sei., Paris 242, 342-344. (110)
2. BRESSAN, A.: Sulla possibilita di stabilire limitazioni inferiori per le componenti
intrinseche del tensore degli sforzi in coordinate generali. Rend. Sem. Mat. Padova
26, 139-147. (222)
3. CoLONNETTI, G.: L'equilibre des voiles minces hyperstatiques (Le cas des voiles
de surface minimum). C. R. Acad. Sei., Paris 243, 1087-1089, 1701-1704. (229)
4. DoRN, W. S., and A. ScHILD: A converse to the virtual work theorem for deformable solids. Quart. appl. Math. 14, 209-213. (227, 234)
5. DoYLE, T. C., and J. L. ERICKSEN: Nonlinear elasticity. Adv. appl. Mech. 4,
53-115. (12, 14, 18)
6. ELSASSER, W. M.: Hydromagnetic dynamo theory. Rev. Mod. Phys. 28, 135-163.
(313)
7. FINZI, L.: Legame fra equilibrio e congruenza e suo significato fisics, Rend. Lincei
(8) 20, 205-211, 338-342. (84, 224, 234)
8. GElS, T.: ".Ähnliche" dreidimensionale Grenzschichten. J. Rational Mech. Anal.
5, 643-686. (163)
9. GREEN, A. E.: Hypo-elasticity and plasticity. Proc. Roy. Soc. Lond., Ser. A 234,
46-59. (304)
10. GREEN, A. E.: Hypo-elasticity and plasticity, II. J. Rational Mech. Anal. 5,
725-734. (304)
11. HANIN, M., and M. REINER: On isotropic tensor-functions and the measure of
deformation. Z. angew. Math. Phys. 7, 377-393. (33A)
12. HASKIND, M. D.: Unsteady motion of a solid in an accelerated flow of an unlimited
fluid [in Russian]. Prikl. Mat. Mekh. 20, 120-123. (202)
13. IKENBERRY, E., and C. TRUESDELL: On the pressures and the flux of energy in
a gas according to Maxwell's kinetic theory, I. J. Rational Mech. Anal. 5, 1-54.
(307)
14. KoPPE, E.: Methoden der nichtlinearen Elastizitätstheorie mit Anwendung auf die
dünne Platte endlicher Durchbiegung. Z. angew. Math. Mech. 36, 455-462. (34)
15. MARGULIES, G.: Remark on kinematically preferred co-ordinate systems: Proc.
Nat. Acad. Sei. U.S.A. 42, 152-153. (148)
16. OLDROYD, J. G., and R. H. THOMAS: The motion of a cylinder in a rotating liquid
with general elastic and viscous properties. Quart. J.Mech.Appl. Math. 9, 136-139.
(207)
17. OswATITSCH, K.: Über eine Verallgemeinerung des Potentials auf Strömungen
mit Drehung. Öst. lng.-Arch. 10, 239-241. (125)
18. RAw, C. J. G., and W. YouRGRAU: "Acceleration" of chemical reactions. Nature,
Lond. 178, 809. (159A)
19. SPEISER, A.: Einleitung. L. Euleri Opera Omnia (1) 29, VII-XLII. (33A)
20. ToUPIN, R. A.: The elastic dielectric. J. Rational Mech. Anal. 5, 849-915. (14,
19, 37, 312)
21. TRUESDELL, C.: Zur Geschichte des Begriffes "Innerer Druck". Physik. BI. 12,
315- 326. (200)
22. TRUESDELL, C.: Review of [1953, 20]. Math. Rev. 17, 97. (170)
23. TRUESDELL, C.: Experience, theory, and experiment. Proc. 6th Hydraulics Conf.,
bull. 36, Univ. Iowa Stud. Engin., 3-18. (3)
1957 1. BRAND, L.: The pi theorem of dimensional analysis. Arch. Rational Mech. Anal.
1 (1957/58), 34-45. (293)
2. BRDICKA, M.: On the general form of the Beltrami equation and Papkovich's
solution of the axially symmetrical problern of the classical theory of elasticity.
Czech. J. Phys. 7, 262-274. (228)
List of Works Cited. 783
3. DuNCAN, W. R.: Analysis of a vector field and some applications to fluid motion.
Aero. Quart. 8, 207-214.
4. FILONENKo-BoRomcH, M. M.: On the problern of Lame for the parallelepiped in
the general case of surface Ioads [in Russian]. Prikl. Mat. Mekh. 21, 550-559. (227)
5. GRAIFF, F.: Sulle condizioni di congruenza per una membrana. Rend. Ist. Lombardo A 92, 33-42. (84)
6. GREEN, A. E., and R. S. RrvLIN: The mechanics of non-linear materials with
memory, Part!. Arch. Rational Mech. Anal. 1 (1957/58), 1-21. (305)
7. HAYES, W.: The vorticity jump across a gasdynamic discontinuity. J. Fluid
Mech. 2, 595-600. (179)
8. HowARD, L. N.: Divergence formulas involving vorticity. Arch. Rational Mech.
Anal. 1 (1957/58), 113-123. (119, 131)
9. MrLNE-THOMSON, L. M.: A general solution of the equations of hydrodynamics.
J. Fluid Mech. 2, 88. (230)
10. NACHBAR, W., F. WrLLIAMs and S. S. PENNER: The conservation laws for independent coexistent continua and for multicomponent reacting gas mixtures. Lockheed
aircraft corp. LMSD 2082, 15 March = Quart. appl. Math. 17, 43-54 (1959).
(215, 243)
11. NoLL, W.: On the foundations of the mechanics of continuous media. Carnegie
Inst. Tech. Rep. No. 17, Air Force Off. Sei. Res. (3, 196, 196A, 211, 293, 305)
12. NoLL, W.: On the rotation of an incompressible continuous medium in plane
motion. Quart. appl. Math. 15, 317-319. (161, 207)
13. STORCH!, E.: Una soluzione delle equazioni indefinite della meccanica dei continui
negli spazi riemanniani. Rend. Ist. Lombardo 90 (1956), 369-378. (229)
14. TING, T. W., and J. C. M. LI: Thermodynamics for elastic solids. General formulation. Phys. Rev. (2) 106, 1165-1167. (246)
15. THOMAS, T. Y.: Extended compatibility conditions for the study of surfaces of
discontinuity in continuum mechanics. J. Math. Mech. 6, 311-322, 907-908.
(175, 176, 177, 179, 180, 181)
16. TRUESDELL, C.: Sulle basi della termomeccanica. Rend. Lincei (8) 22, 33-38,
158-166. (158, 159, 215, 243, 254, 255, 259)
17. TRUESDELL, C.: General solution for the stresses in a curved membrane. Proc.
Nat. Acad. Sei. U.S.A. 43, 1070-1072. (84, 229)
18. WHITNEY, H.: Geometrie Integration Theory. Princeton Univ. Press. (269)
19. YIH, C.-S.: Stream functions in three-dimensional flows. Houille Blanche,
445-450. (164)
1958 1. ERICKSEN, J. L., and C. TRUESDELL: Exact theory of stress and strain in rods
and shells. Arch. Rational Mech. Anal. 1, 295-323. (60, 61, 63A, 64, 212, 214)
1 A. FERRARESE, G.: Sulla relazione simbolica della meccanica dei sistemi continui
vincolati. Rend. Mat. e Appl. (5) 14, 305-312. (233)
2. GoDAL, T.: On Beltrami vector fields and flows, Part III. Same considerations
on the general case. Univ. Bergen Arbok 1957, Naturv. r. Nr. 12, 28 pp. (108)
3. GRAIFF, F.: Sulle condizioni di congruenza per deformazioni anche finite. Rend.
Lincei (8) 24, 415-422. (84)
4. GüNTHER, W.: Zur Statik und Kinematik des Cosseratschen Kontinuums. Abh.
Braunschweig. Wiss. Ges. 10, 195-213. (61, 200, 227)
5. KANWAL, R. P.: Determination of the vorticity and the gradients of flow parameters behind a three-dimensional unsteady curved shock wave. Arch. Rational
Mech. Anal. 1, 225-232. (175)
6. KRZYWOBLOCKI, M. Z. v.: On the stream functions in nonsteady three-dimensional
flow. J. Aeronaut. Sei. 25, 67. (160)
7. MARSHAK, R. E.: Effect of radiation on shock wave behavior. Phys. of Fluids 1,
24-29. (241)
8. NoLL, W.: A mathematical theory of the mechanical behavior of continuous media.
Arch. Rational Mech. Anal. 2 (1958/59), 197-226. (3, 196, 200, 293, 305)
9. TouPIN, R. A.: World invariant kinematics. Arch. Rational Mech. Anal. 1 (1957/58),
181-211. (152)
10. TRUESDELL, C.: Geometrie interpretation for the reciprocal deformation tensors.
Quart. appl. Math. 15, 434-435. (29)
11. TRUESDELL, C.: Intrinsic equations of spatial gas flow. Math. Res. Center U.S.
Army, Univ. Wisconsin rep. No. 33, July. Cf. [1960, 5]. (99)
12. WASHIZU, K.: A note on the conditions of compatibility. J. Math. Phys. 36,
306-312. (34)
1959 1. BLANKFIELD, J., and G. C. McVITTIE: Einstein's equations and classical hydrodynamics. Arch. Rational Mech. Anal. 2 (1958/59), 337-354. (230)
784
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
C. TRUESDELL and R. TouPIN: The Classical Field Theories.
BLANKFIELD, J., and G. C. McVITTIE: A method of solution of the equations of
magnetohydrodynamics. Arch. Rational Mech. Anal. 2 (1958/59). 411-422. (230)
CoLEMAN, B. D., and W. NoLL: On the thermostatics of continuous media. Arch.
Rational Mech. Anal. 4 (1959/60), 97-128. (262)
DROBOT, S., and A. RYBARSKI: A variational principle of hydromechanics. Arch.
Rational Mech. Anal. 2 (1958/59), 393-410. (132)
GRAIFF, F.: Soluzione generate delle equazioni indefinite di equilibrio per una
membrana. Rend. Lincei (8) 26, 189-196. (229)
GREEN, A. E., R. S. RIVLIN and A. J. M. SPENCER: The mechanics of non-linear
materials with memory. Part li. Arch. Rational Mech. Anal. 3, 82-90. (305)
GREEN, A. R.: The equilibrium of rods.Arch. Rationa!Mech. Anal. 3, 417-421. (214)
KRÖNER, E., u. A. SEEGER: Nicht-lineare Elastizitätstheorie der Versetzungen und
Eigenspannungen. Arch. Rational Mech. Anal. 3, 97-119. (20, 34)
NoLL, W.: The foundations of classical mechanics in the light of recent advances
in continuum mechanics. The Axiomatic Method, with special reference to geometry and physics (1957). Amsterdam: North Holland Co., 266-281. (3, 196, 293)
ScHAEFER, H.: Die Spannungsfunktionen des dreidimensionalen Kontinuums;
statische Deutung und Rand werte. lng.-Arch. 28, 291-306. (227)
TRUESDELL, C. : Invariant and complete stress functions for generat continua.
Arch. Rational Mech. Anal. 4 (1959/60). 1-29. (226)
ZoLLER, K.: Die Wärmeleitgleichung bei Wärmespannungen. lng.-Arch. 28,
366-372. (256A)
1960 JA. CoLEMAN, B., and C. TRuESDELL: On the reciprocal relations of Onsager. J.
Chem. Phys. forthcoming. (245)
1. ERICKSEN, J. L.: Anisotropie fluids. Arch. Rational Mech. Anal. 4 (1959/60),
231-237- (205, 298)
2. HAYES, W. D.: Generalized Kelvin's minimum energy theorem. J. Fluid Mech.,
forthcoming. (94)
3. KRÖNER, E.: Allgemeine Kontinuumstheorie der Versetzungen und Eigenspannungen. Arch. Rational Mech. Anal. 4 (1959/60). 273-334. (20, 61, 201, 266)
4. TRUESDELL, C.: The rational mechanics of flexible or elastic bodies, 1638-1788.
L. Euleri Opera Omnia (2) 11, Part 2. (33A, 66A, 200, 212)
5. TRUESDELL, C.: lntrinsic equations of spatial gas flow. Z. angew. Math. Mech.
40, 9-14. (99)
6. TRUESDELL, C.: Mechanical aspects of diffusion, forthcoming. (295)
Additional Bibliography K:
Kinematics of special motions (geometrical theory).
1872 K 1. DuRRANDE, H.: Proprietes generales du deplacement d'une figure de forme
variable. C. R. Acad. Sei., Paris 74, 1243-1247.
K 2. DuRRANDE, H.: De l'acceleration dans Je deplacement d'un systeme de points
qui reste homographique a lui-meme. C. R. Acad. Sei., Paris 75, 1177-1180.
1873 K 1. DuRRANDE, H.: Essai sur le deplacement d'une figure de forme variable. Ann.
Ec. Norm. (2) 2, 81-121.
K 2. GROUARD: Sur le mouvement d'une figure, qui se deplace dans l'espace en
restaut semblable a elle-meme. Bull. Soc. Philomath. (6) 9, 47-49. Abstract:
L'lnstitut (2) 1, 163-164.
1874 K 1. BuRMESTER, L.: Kinematisch-geometrische Untersuchungen der Bewegung ähnlich-veränderlicher ebener Systeme. Z. Math. Phys. 19, 154-169.
K 2. Bu:RMESTER, L.: Kinematisch-geometrische Untersuchungen der Bewegung affinveränderlicher und collinear-veränderllcher ebener Systeme. Z. Math. Phys. 19,
465-492.
K 3. DuRRANDE, H.: Deplacement d'un systeme de points. Proprietes geometriques
dependant des parametres differentiels du second ordre. C. R. Acad. Sei., Paris
78, 1036-1040.
K 4. DuRRANDE, H.: Etude de l'acceleration dans le deplacement d'un systeme de
forme variable. Ann. Ec. Norm. (2) 3, 151-164.
1875 K 1. BuRMESTER, L.: Kinematisch-geometrische Untersuchungen der Bewegung gesetzmäßig-veränderlicher Systeme. Z. Math. Phys. 20, 381-422.
K 2. JORDAN, C.: Sur le mouvement des figures dans le plan et dans l'espace. Bull.
Soc. Math. France 1,144-148.
Additional Bibliography K. 785
K 3. LrGUINE, V.: Sur le lieu des points d'un systeme invariable mobile d'une maniere
generale dans l'espace, dont les accelerations du premier ordre sont constantes.
Bull. Soc. Math. France 1, 152-154.
1877 K 1. MüLLER, R.: Über Selbsthüllcurven und Selbsthüllflächen in ähnlich-veränderlichen Systemen. Z. Math. Phys. 22, 369-376.
1878 K 1. BuRMESTER, L.: Kinematisch-geometrische Theorie der Bewegung der affinveränderlichen, ähnlich veränderlichen und starren räumlichen oder ebenen
Systeme. Z. Math. Phys. 23, 108-131.
K 2. BuRMESTER, L.: Über den Beschleunigungszustand ähnlich-veränderlicher und
starrer ebener Systeme. Civilingenieur 24.
1879 K 1. BURMESTER, L.: Über die Festlegung projectiv-veränderlicher ebener Systeme.
Math. Ann. 14, 472-497.
K 2. FouRET, G.: Sur le mouvement d'un corps qui se deplace et se deforme en restant
homothetique a lui-meme. C. R. Acad. Sei., Paris 88, 227-230.
K 3. FROMENTI, C.: Movimento delle figure ehe se mantengono simili a se stesse.
Giorn. math. 17, 232-243.
K 4. GEISENHEIMER, L.: Untersuchung der Bewegung ähnlich-veränderlicher Systeme.
Z. Math. Phys. 24, 129-159-
K 5. GEISENHEIMER, L.: Die Bildung affiner Figuren durch ähnlich-veränderliche
Systeme. Z. Math. Phys. 24, 345-381.
1880 K 1. BuRMESTER, L.: Über das bifocal-veränderliche System. Math. Ann.16, 89-111.
1881 K 1. ScHUMANN, A.: Beiträge zur Kinematik ähnlich-veränderlicher und affin-veränderlicher Gebilde. Z. Math. Phys. 26, 157-179-
1883 K 1. MEHMKE, R.: Über die Geschwindigkeiten beliebiger Ordnung eines in seiner
Ebene bewegten ähnlich-veränderlichen Systems. Civilingenieur (2) 29, 487- 508.
K 2. MEHMKE, R.: Über den geometrischen Ort der Punkte ohne Normalbeschleunigung in einer Phase eines starren oder affin-veränderlichen Systems. Civilingenieur (2) 29, 581-582.
K 3. NrcoLI, F.: Intorno ad un caso di movimento di una figura piana ehe si varia
rimanendo simile a se stessa. Mem. Accad. Modena (2) 1, 59-71.
K 4. NrcoLI, F.: Intorno ad un caso di movimento di una figura piana ehe si conserva
simile a se stessa. Mem. Accad. Modena (2) 1, 171-178.
K 5. NrcoLI, F.: Intorno a due casi di movimento di una figura solida ehe rimane
simile a se stessa. Mem. Accad. Modena (2) 1, 249-260.
1885 K 1. SoMOFF, P.: Über die Bewegung ähnlich-veränderlicher ebener Systeme. Z.
Math. Phys. 30, 193-209.
1888 K 1. BuRMESTER, L.: Lehrbuch der Kinematik. Leipzig: A. Felix. xx + 942 pp. +
Atlas. See §§ 329-351.
1889 K 1. SoMov, P.: Some problems on the distribution of velocityinvariable systems
[in Russian]. Varshavskia Univ. lzv. 1889, No. 4, 32 pp.
1890 K 1. MoRLEY, F.: On the kinematics of a triangle of constant shape but varying size
(with a note). Quart. J. Math. 24, 3 59-369, 386.
K 2. SoMow, P.: On the acceleration in collinearly variable systems [in Russian].
Proceedings of the 8th meeting of Russian Natural Scientists and Physicians,
St. Petersburg, 1890. Math. and Astron. 41-44. (We have not been able to see
this reference.)
1892 K 1. SHEBUEV, G. N.: Application of the theory of quaternions to the mechanics of
similar and homogeneously variable systems [in Russian]. lzv. Fiz.-Mat. Obsh.
Kazan (2) 3, 111-160.
1894 K 1. MANNHEIM, A.: Principes et developpements de geometrie cinematique. Paris:
Gauthier-Villars. ix + 589pp. See pp. 14-53, 457-475.
1897 K 1. DE SAUSSURE, R.: Calcul geometrique regle. Amer. J. Math. 19, 329-370.
1898 K 1. DE SAUSSURE, R.: Cinematique des fluides. Mouvement d'un fluide dans un
plan. Arch. Sei. Phys. Nat. Geneve (4) 5, 497-519.
1899 K 1. CAVALLI, E.: Le figure reciproche e la trasformazione quadratica nella cinematica.
Atti Ace. Napoli (2) 9, No. 12, 29 pp.
1901 K 1. DE SAussuRE, R.: Sur le mouvement d'une droite qui possede trois degres de
liberte. C. R. Acad. Sei., Paris 133, 1283-1285.
K 2. SEILIGER, D.: On a fundamental theorem in the statics of a variable system
[in Russian]. Papers Univ. Kazan No. 718, 75-82. (We have not been able to
see this reference.)
1902 K 1. BuRMESTER, L.: Kinematisch-geometrische Theorie der Bewegung der affin-veränderlichen, ähnlich-veränderlichen und starren räumlichen oder ebenen Systeme,
Teil2. Z. Math. Phys. 47, 128-156.
Handbuch der Physik, Bd. lll/1. so
786 C. TRUESDELL and R. TouPrN: The Classical Field Theories.
K 2. CARDINAAL, J.: Over de beweging van veranderlijke stelsels. Amst. Akad. Vers!.
10, 550-566, 687-691.
K 3. CARDINAAL, J.: Over de afbeelding van de beweging van veranderlijke stelsels.
Amst. Akad. Vers!. 11, 466-471.
K 4. DE SAussuRE, R.: Theorie geometrique du mouvement des corps. Arch. Sei.
Phys. Nat. Geneve (4) 13, 425-461; 14, 14-41, 209-231; 18, 25-63 (1904);
21, 36-55, 129-133 (1906). Also issued separately, in parts, Geneva 1902,
1904, 1906.
K 5. ScHOENFLIESS, A., u. M. GRüBLER: Kinematik. Enzykl. Math. Wiss. 41, 190-278.
K 6. SoMOV, P.: On hinged members with variable elements [in Russian]. Varshavskia
Univ. Izv. 1902, Part 8, No. 3, 45 pp.
1903 K 1. SoMOFF, P.: Über einige Gelenksysteme mit ähnlich-veränderlichen oder affinveränderlichen Elementen. Z. Math. Phys. 49, 25-61.
1907 K 1. KoENIGS, G.: Sur !es deformations elastiques qui laissent invariables !es Iangueurs
d'une triple infinite de lignes droites. C. R. Acad. Sei., Paris 144, 557-560.
1910 K 1. KRAUSE, M.: Zur Theorie der ebenen ähnlich veränderlichen Systeme. Jber.
dtsch. Math.-Ver. 19, 327-339.
1911
K 2. MEHMKE, R.: Analytischer Beweis des Satzes von Herrn Reinhold Müller über
K3.
K4.
K5.
K6.
K 1.
K2.
K3.
die Erzeugung der Koppelkurve durch ähnlich-veränderliche Systeme. Z. Math.
Phys. 58, 257-259.
MüLLER, R.: Erzeugung der Koppelkurve durch ähnlich-veränderliche Systeme.
Z. Math. Phys. 58, 247-251.
MüLLER, R.: Über die Momentanbewegung eines ebenen ähnlich-veränderlichen
Systems in seiner Ebene. Jber. dtsch. Math. Ver. 10, 29-89.
SKUTSCH, R.: Über die von Herrn Reinhold Müller untersuchte besondere Bewegung eines ähnlich-veränderlichen Systems. Z. Math. Phys. 58, 252-257.
STUDY, E.: Die Kinematik der Herren de Saussure und Bricard. Jber. dtsch.
Math.-Ver. 19, 255-263.
KRAUSE, M.: Zur Theorie der affin veränderlichen ebenen Systeme. Sitzgsber.
Akad. Wiss. Leipzig 63, 271-288.
KRAUSE, M.: Über räumliche Bewegungen mit ebenen Bahn kurven. Sitzgsber.
Akad. Wiss. Leipzig 63, 515-533.
MEHMKE, R.: Beiträge zur Kinematik starrer und affin-veränderlicher Systeme,
insonderheit über die Windung der Bahnen der Systempunkte. Z. Math. Phys.
59, 90-94, 204-220, 440-442.
1912 K 1. HARTMANN, T.: Zur Theorie der Momentanbewegung eines ebenen ähnlich-veränderlichen Systems. Diss. Rostock, 144 pp.
1913 K 1. DE DoNDER, T.: Sur divers modes de croissance des milieux continus. Bull. Acad.
Sei. Belg. 614-621, 642-646.
K 2. HERRMANN, E.: Über die einförmige Bewegung des ebenen kreisverwandt-veränderlichen Systems. Diss. Tech. Hoch. Dresden, 93 pp.
1914 K 1. CARL, A.: Zur Theorie der ebenen ähnlich veränderlichen Systeme. Diss. Dresden,
125 pp.
K 2. WrNKLER, R.: Über die Bewegung affin-veränderlicher ebener Systeme. Diss.
Dresden, 73 pp.
1920 K 1. KRAUSE, M., Assisted by A. CARL: Analysis der ebenen Bewegung. Berlin. 216 pp.
1922 K 1. DELASsus, E.: Stabilite de l'equilibre sur une Iiaison finie unilaterale. Bull. Sei.
Math. (2) 46, 283-304.
K 2. GAMBIER, B.: Mecanismes transformables ou deformables. Couples de surfaces
qui s'en deduisent. J. Math. Pures Appl. (9) 1, 19-76.
1932 K 1. ABRAMEsco, N.: Le mouvement d'une figure plane variable qui reste semblable
a elle-meme. Ann. Sei. Norm. Pisa (2) 1, 155-164.
K 2. PASCAL, M.: Sul moto di un corpo deformabile ehe si mantiene simile a se stesso.
I: Formola fondamentale e proprieta ehe se ne deducono. II: Centro istantaneo
di velocita e conseguenze. Rend. Lincei (6) 15, 871-874; 16, 320-324.
1933 K 1. ABGHIRIADI, M.: Sur le mouvement d'une figure plane semblablement variable.
Mathesis 47, Suppl., 14 pp.
K 2. PASCAL, M.: Sul moto di una figura deformabile piana di area costante e ehe
rimane affine a se stessa. Rend. Napoli (4) 3, 71-77.
K 3. PASCAL, M.: Sul moto di una figura deformabile piana ehe si conserva affine
a se stessa. Rend. Napoli (4) 3, 78-82.
K 4. PASCAL, M.: Sul centro istantaneo di velocita nulla nel moto di una figura piana
di area costante e a deformate affine. Rend. Napoli (4) 3, 110-113.
K 5. PASCAL, M.: Sull'accelerazione nel moto di una figura piana di area costante
e a deformate affine. Rend. Napoli (4) 3, 123-126.
Additional Bibliography N. 787
K 6. PASCAL, M.: Sulla cinematica affine di una figura piana di area costante. Rend.
Napoli (4) 3, 142-144.
1934 K 1. Dr Nor, S.: Considerazioni geometriche sul moto di un corpo deformabile ehe si
mantiene simile a se stesso. Rend. Napoli (4) 3, 176-181.
K 2. PASCAL, M.: Sul moto di un corpo deformabile di volume costante e ehe rimane
affine a se stesso. Atti Soc. ital. Progr. Sc. A 222, 194-195.
1936 K 1. ABRAMESCO, N.: Proprietäti geometrice ale mi~cärii unei figuri plane variabile
care rämane asemenea cu ea in sä~i, cand trei drepte ale figurii trei prin trei puncte
fixe, sau cand trei puncte descriu trei drepte fixe. Gaz. Mat. Bucarest 41, 409-414.
K 2. HARMEGNIES, R.: Sur le mouvement d'une figure plane qui reste homographique
a elle-meme. C. R. Acad. Sei., Paris 202, 1323-1324.
See also the following items from the "List of works cited": DuRRANDE [1871, 4], and
the papers cited in Sects. 140-142.
Additional Bibliography N:
Non-relativistic kinematics and mechanics in generalized spaces.
1876 N 1. BELTRAMI, E.: Formules fondamentales de cinematique dans !es espaces de courhure constante. Bull. Sei. Math. (1) 11, 233-240.
1878 N 1. LE:vv, M.: Sur la cinematique des figures continues sur les surfaces courbes,
et, en general, dans les varietes planes ou courbes. C. R. Acad. Sei., Paris 86,
812-818.
N 2. LE:vv, M.: Sur les conditions que doit remplir un espace pour qu'on y puisse
deplacer un systeme invariable, a partir de l'une quelconque de ses positions,
dans une ou plusieurs directions. C. R. Acad. Si., Paris 86, 875-878.
1881 N 1. BELTRAMI, E.: Sulle equazioni generali dell'elasticitä.. Ann. Mat. (2) 10, 188-211
(1880-1882) = Opere 3, 383-407.
1884 N 1. BELTRAMI, E.: Sull'uso delle coordinate curvilinee neUe teorie del potenziale e
dell'elasticita. Mem. Accad. Sei. Bologna (4) 6, 401-488 = Opere 4, 136-179.
N 2. HEATH, R. S.: On the dynamics of a rigid body in elliptic space. Phi!. Trans.
Roy. Soc. Lond. 175, 281-324.
1885 N 1. KrLLING, W.: Die Mechanik in den nicht-Euklidischen Raumformen. J. reine
angew. Math. 89, 1-48.
N 2. HrLL, M. J. M.: On some general equations which include the equations of hydrodynamics ( 1883). Trans. Cambridge Phi!. Soc. 14, 1-29.
1888 N 1. CESARO, E.: Sur une recente communication deM. Levy. C. R. Acad. Sei., Paris
107, 520-522.
1889 N 1. PADOVA, E.: La teoria di Maxwell negli spazi curvi. Rend. Lincei (4) 51, 875-880.
N 2. SoMIGLIANA, C.: Sopra la dilatazione cubica di un corpo elastico isotropo in uno
spazio di curvatura costante. Ann. Mat. (2) 16, 101-115.
1890 N 1. PADOVA, E.: I! potenzialedelleforze elastiche di mezzi isotropi. Atti Ist. Veneto
48 = (7) 1, 445-451.
1894 N 1. CESARO, E.: Sulle equazioni dell'elasticitä. negli iperspazii. Rend. Lincei (5) 32,
290-294.
1900 N 1. DE FRANCESCO, D.: Aleuni problemi di meccanica in uno spazio a tre dimensioni
di curvatura costante, I and li. Atti Accad. Napoli (2) 10, Nos. 4 (38 pp.) and
9 (33 pp.).
N 2. DE FRANCESCO, D.: Sul moto spontaneo di un corpo rigido in uno spazio di curvatura costante. Atti R. Accad. Sei. Torino 35, 34-38, 231-243.
1901 N 1. BoHLIN, K.: Sur l'extension d'une formule d'Euler et sur le calcul des moments
d'inertie principaux d'un systeme de points materiels. C. R. Acad. Sei., Paris
133, 530-532.
N 2. DE FRANCESCO, D.: Su aleuni problemi di meccanica, in uno spazio pseudosferico,
analiticamente equivalenti a problemi nello spazio ordinario. Rend. Accad. Napoli
(3a) 7, 28-38.
1902 N 1. DE F~ANCEsco, D.: Aleuneformole della meccanica dei fluidi in uno spazio a
tre dimensioni di curvatura costante, I and li. Atti Accad. Napoli (2) 12, Nos. 9
(18 pp.) and 10 (13 pp.).
N 2. STÄCKEL, P.: De ea mecanicae analyticae parte quae ad varietates complurium
dimensionum spectat. Libellus Ioannis Bolyai ... ad celebrandam memoriam ...
(Claudiopoli), 63-79.
1903 N 1. STÄCKEL, P.: Bericht über die Mechanik mehrfacher Mannigfaltigkeiten. Jber.
dtsch. Math.-Ver. 12, 469-481.
50*
788 C. TRUESDELL and R. TouPrN: The Classical Field Theories.
1907 N 1. RrQUIER, CH.: Sur !es systemes d'equations aux derivees partielles auquels conduisent: 1° l'etude des dt\formations finies d'un milieu continu dans l'espace
a n dimensions; 20 Ia determination des systemes de coordonnees curvilignes
orthogonales a n variables. C. R. Acad. Sei., Paris 145, 113 7- 1139.
1911 N 1. VESSIOT, E.: Sur Ia einematique des milieux continus an dimensions. c. R. Acad.
Sei., Paris 152, 1732-1735.
1912 N 1. DE DoNDER, T.: Sur Ia cinematique des milieux continus. Bull. Acad. Roy.
Belg. Cl. Sei., 243-251.
N 2. ZoRAWSKI, K.: Über gewisse Pfaff'sche Systeme, welche bei Bewegungen kontinuierlicher Medien invariant bleiben. Bull. Int. Acad. Sei. Cracovie A 1912,
436-461.
1913 N 1. DEL RE, A.: Sulle equazioni generali per Ia dinamica negli spazii ad n dimensioni
ed a curvatura costante, Ann. Mat. (3) 22, 63-70.
1926 N 1. SYNGE, J. L.: App!ications of the absolute differential calculus to the theory of
elastieity (1924). Proc. Lond. Math. Soc. (2) 24, 103-108.
1930 N 1. ToNoLo, A.: Une interpretation physique du tenseur de Riemann et des courbures principales d'une variete V3 • C. R. Acad. Sei., Paris 190, 787-788.
N 2. ToNOLO, A.: Equazioni intrinseche di equilibrio dell' elastieita negli spazi a curvatura costante. Rend. Sem. Mat. Fadova 1, 73-84.
1931 N 1. ToNOLO, A.: Sistemi isostatiei dei corpi elastici negli spazi a curvatura costante.
Rend. Sem. Mat. Fadova 2, 152-163.
1933 N 1. CARTAN, E.: La einematique newtonienne et Ia theorie des espaces reglees a
connexion euclidienne. Ass. Franc. Avancem. Sei. 19-20.
N 2. CARTAN, E.: La einematique newtonienne et les espaces a connexion euclidienne.
Bull. Math. Soc. Roumaine Sei. 35, 69-73.
N 3. FrNZI, B.: Equazioni intrinseche della meccanica dei sistemi continui perfettamente od imperfettamente flessibili. Ann. Mat. (4) 11, 215-245.
N 4. TEODORIA, L.: Sur Ia einematique du corps solide dans l'espace euclidien a n
dimensions. Bull. Math. Soc. Roumaine Sei. 35, 243-247.
1934 N 1. PASTOR!, M.: Sulle equazioni della meccanica dei mezzi isotropi non euclidei.
Rend. Lincei (6) 19, 566-572.
N 2. PASTOR!, M.: Sulla dissipazione di energia nei fluidi viscosi. Rend. Ist. Lombarde
(2) 67, 823-848.
193 5 N 1. WESTERGAARD, H. M.: General solution of the problern of elastostatics of an
n-dimensional homogeneaus isotropic solid in an n-dimensional space. Bull. Amer.
Math. Soc. 41, 695-699.
1936 N 1. LaTZE, A.: Die Grundgleichungen der Mechanik im elliptischen Raum. Jber.
dtsch. Math.-Ver. 46, 51-70.
1937 N 1. LAMPARIELLO, G.: Varieta sostanziali nel moto di un sistema continuo. Rend.
Lincei (6a) 15, 383-387.
1940 N 1. 0RTVAY, R.: The physical implications of some new viewpoints in mathematics
[in Hungarian]. Mat. Fiz. Lapok 47, 111-138.
N 2. SKOLEM, T.: A little study on transfinite mechanics [in Norwegian]. Norsk Mat.
Tidsskr. 22, 5-9.
1942 N 1. BLASCHKE, W.: Nicht-Euklidische Geometrie und Mechanik. Hamburg. Math.
Einzelschr. 34.
1944 N 1. BLASCHKE, W.: Nicht-Euklidische Mechanik. Sitzgsber. Akad. Wiss. Heidelberg
1943, No. 2, 10 pp.
1951 N 1. SANTALO, L. A.: On permanent vector-varieties in n dimensions. Portuga!iae
Math. 10, 125-127.
N 2. SYNGE, J. L.: On permanent vector-lines in n dimensions. Proc. Amer. Math.
Soc. 2, 370-372.
See also the following items from the "List of Works Cited": CLEBSCH [1857, 1; §§ 1-4],
FADOVA [1889, 7], ZORAWSKI [1900, 12], [1901, 17], ZERMELO [1902, 10], ZORAWSKI [1911, 13
and 14], DE DONDER [1912, 2], ZORAWSKI [1912, 8].
Additional Bibliography P: Principles of Mechanics.
Any partial bibliography of work on the concepts and axioms of mechanics from the
origins through the time of LAGRANGE would be misleading. No adequate critical history
has ever been written. The remarks on this subject given in treatises or general histories of
physics are often mendacious and usually so incomplete and inaccurate as to be totally
misinformative. Large extracts from some of the sources, along with helpful comments,
may be found in:
Additional Bibliography P: Principles of Mechanics. 789
JoUGUET, E.: Lectures de mecanique, 2vols. Paris: Gauthier-Villars 1908, 1909. X+ 210
+ 284 pp.
DuHEM, P.: Le Systeme du Monde, Histoire des doctrines cosmologiques de Platon a Copernic.
Paris: Hermann. 10 Vols., 1913-1959.
DuGAS, R.: Histoire de Ia mecanique. Neuchatel: Ed. du Griffon 1950. 649pp.
DUGAS, R.: La mecanique au xvne-siecle. Neuchatel: Ed. du Griffon 1954. 620 PP·
CLAGETT, M.: The Science of Mechanics in the Middle Ages. Madison: Univ. Wisconsin Press
1959. 711 pp.
1883 P 1. MAcH, E.: Die Mechanik in ihrer Entwicklung, historisch-kritisch dargestellt.
Leipzig: Brockhaus. There are many subsequent editions and translations of
this unreliable work.
1894 P 1. HERTZ, H.: Die Prinzipien der Mechanik in neuem Zusammenhange dargestellt.
Leipzig: Johann Ambrosius Barth. xxix + 312 pp. Trans. D. E. JoNES and
]. T. WALLEY, The principles of mechanics. London: MacMillan & Co. 1899.
xxviii + 276 PP·
1897 P 1. BoLTZMANN, L.: Vorlesungen über die Principe der Mechanik I. Leipzig. x +
241 pp. See§§ 1-12.
1905 P 1. PAINLEVE, P.: Les axiomes de Ia mecanique et Ie principe de causalite. Bull.
Soc. frany. Philos. 5, 27-50. See [1922, P1].
1906 P 1. FARKAS, ].: Beiträge zu den Grundlagen der analytischen Mechanik. J. reine
angew. Math. 131, 165-201.
1909 p 1. PAINLEVE, P.: Les axiomes de la mecanique classique, chapter in La methode
dans !es sciences. Paris: Alcan. See [ 1922, P 1].
1911 P 1. MARCOLONGO, R.: Theoretische Mechanik, 2 vol., Deutsch von H. E. TrMERDING.
Leipzig: B. G. Teubner 1911 and 1912.
1922 P 1. PAINLEVE, P.: Les axiomes de Ia mecanique. Examen critique. Note sur Ia
propagation de Ia Iumiere. Paris: Gauthier-Villars. xvii + 112 pp. (Reprint
of [1905, P 2], [1909, P 1], and other material.)
1923 P 1. DIJKSTERHUIS, E. ].: De axioma's der mechanica. Christiaan Huygens 3
(1923-1924), 87-101.
1927 P 1. HAMEL, G.: Die Axiome der Mechanik. Handbuch Physik 5, 1-42.
1928 P 1. DIJKSTERHUIS, E. ].: De historische behandelingswijze van de axiomata der
mechanica van Newton. Euclides 4, 245-255.
1934 P 1. ZAREMBA, S.: Sur Ia notion de force en mecanique. Bull. Soc. Math. France
62, 110-119.
1936 P 1. LINDSAY, R., B., and H. MARGENAU: Foundations of Physics. New York: J. Wiley
& Sons, Inc. xiii + 537 pp.
P 2. PLATRIER, C.: Les Axiomes de la Mecanique Newtonienne. Act. Sei. Indust. No.427.
Paris: Hermann.
1937 P 1. BARBILIAN, D.: Eine Axiomatisierung der klassischen Mechanik. C. R. Acad.
Sei. Roumaine 2, 9-16.
P 2. PENDSE, C. G.: A note an the definition and determination of mass in Newtonian
mechanics. Phi!. Mag. (7) 24, 1012-1022.
1938 P 1. HERMES, H.: Eine Axiomatisierung der allgemeinen Mechanik. Leipzig: Hirzel 48pp.
P 2. RosSER, B.: Review of the foregoing. J. Symb. Logic 3, 119-120.
1939 P 1. KRATZER, A.: Betrachtungen zu den Grundlagen der Mechanik. Semester-Ber.
Math. Sem. Münster 14, 1-15.
P 2. PENDSE, C. G.: A further note an the definition and determination of mass in
Newtonian mechanics. Phi!. Mag. (7) 27, 51-61.
P 3. NARLIKAR, V. V.: The concept and determination of mass in Newtonian mechanics.
Phi!. Mag. (7) 27, 33-36.
1940 P 1. PENDSE, C. G.: Onmassandforcein Newtonianmechanics. Phi!. Mag. 29,477-484.
P 2. ZAREMBA, S.: Reflexions sur !es fondaments de Ia mecanique rationnelle. Enseignement Math. 38, 59-69.
1943 P 1. BRELOT, M.: Sur !es principes mathematique de Ia mecanique classique. Ann.
Univ. Grenoble (sect. sci.-med.) 18, 24 pp.
1944 p 1. BRELOT, M.: Sur quelques points de mecanique rationnelle. Ann. Univ. Grenoble
(sect. sci.-med.) 20, 37 pp.
1947 P 1. SIMON, H. A.: Axioms of Newtonian mechanics. Phi!. Mag. (7) 36, 888-905.
1949 P 1. HAMEL, G.: Theoretische Mechanik. Berlin-Göttingen-Heidelberg: Springer. xvi
+ 796 pp. See Kap. 1.
1950 P 1. BANACH, S.: Mechanika w zakresie szk61 akademickich. Krakow, Monogr. Mat.
t. 8-9, 3rd ed., ix + 555 pp. Trans. E. ]. ScoTT, Mechanics, Warsaw 1951.
iv + 546 pp. See Chap. III, §§ 1-4.
790 C. TRUESDELL and R. TOUPIN: The Classical Field Theories.
1953 P 1. McKINSEY, J. C. C., A. C. SuGAR and P. SuPPEs: Axiomatic foundations of classical
particle mechanics. J. Rational Mech. Anal. 2, 253-272.
P 2. McKINSEY, J. C. C., and P. SuPPEs: Transformations of systems of classical
particle mechanics. J. Rational Mech. Anal. 2, 273-289.
19 54 P 1. BRELOT, M.: Les Principes Mathematiques de la Mecanique Classique. Grenoble
and Paris.
P 2. PLATRIER, C.: Mecanique Rationelle 1. Paris: Dunod. See pp. 77-180.
See also the following items from the "List of works Cited": KIRCHHOFF [1876, 2; Vor!. 1,
§§ 1-2, 4, 7], WHITTAKER [1904, 8; Ch.IIJ, }AUMANN [1905, 2], HAMEL [1908, 4],
NoLL [1957, 11], [1958, 8], [1959, 9].
Additional Bibliography R: Relativistic Continuum Theories
1904 R 1.
1909 R 1.
R 2.
R 3.
R 4.
1911 R 1.
R 2.
R 3.
R 4.
R 5.
R 6.
R 7.
1912 R1.
R2.
1914 R 1.
1915 R 1.
R 2.
R3.
1916 R 1.
1917 R 1.
1918 R1.
R 2.
1922 R 1.
1923 R 1.
R 2.
ABRAHAM, M.: Zur Theorie der Strahlung und des Strahlungsdruckes. Ann. d.
Phys. 14, 236-287.
ABRAHAM, M.: Zur Elektrodynamik bewegter Körper. Rend. Pa!. 28, 1-28.
BoRN, M.: Die Theorie des starren Elektrons in der Kinematik des Relativitätsprinzips. Ann. d. Phys. 30, 1- 56.
HERGLOTZ, G.: Über den vom Standpunkt des Relativitätsprinzips aus als "Starr"
zu bezeichnenden Körper. Ann. d. Phys. 31, 393-415.
NoETHER, F.: Zur Kinematik des starren Körpers in der Relativtheorie. Ann.
d. Phys. 31, 919-944.
BORN, M.: Elastizitästheorie und Relativitätstheorie. Phys. Z. 12, 569-57 5.
HERGLOTZ, G.: Über die Mechanik des deformierbaren Körpers vom Standpunkte
der Relativitätstheorie. Ann. Phys. (4) 36, 493-533.
lGNATOWSKY, W. v.: Zur Elastizitätstheorie vom Standpunkte des Relativitätsprinzips. Phys. Z. 12, 164- 169.
lGNATOWSKY, W. v.: Zur Hydrodynamik vom Standpunkte des Relativitätsprinzips. Phys. Z. 12, 441-442.
JüTTNER, F.: Die Dynamik eines bewegten Gases in der Relativitätstheorie.
Ann. Phys. (4) 35, 145-161.
LAuE, M. v.: Zur Dynamik der Relativitätstheorie. Ann. Phys. (4) 35, 524-542.
NoRDSTRÖM, G.: Zur Relativitätsmechanik deformierbarer Körper. Phys. Z. 12,
854-857.
LAMLA, E.: Über die Hydrodynamik des Relativitätsprinzips. Ann. Phys. (4)
37, 772-796.
LEMERAY, E. M.: Elasticite longitudinale et elasticite transversale. Soc. Fran~.
Phys. No. 29, 2.
MATTIOLI, G. D.: La dinamica di relativita dei mezzi continui dedotta della
dinamica classica colla modificazione di un solo principio. Rend. Lincei (5) 232 ,
427-432.
DANIELL, P. J.: Rotation of eiastic bodies and the principle of relativity. Phi!.
Mag. (6) 30, 7 54-761.
EINSTEIN, A.: Der Energiesatz in der allgemeinen Relativitätstheorie. Sitzgsber.
preuß. Akad. Wiss. 1915, 778-786.
MATTIOLI, G. D.: La dinamica di relativita dei mezzi continui dedotta della
dinamica classica colla modificazione di un solo principio. Nuovo Cim. (6) 9,
263-270.
EINSTEIN, A.: Die Grundlage der allgemeinen Relativitätstheorie. Ann. d. Phys.
49, 769-822.
NoRDSTRÖM, G.: Die Mechanik der Continua in der Gravitationstheorie von
Einstein. Handel. Nederl. Natuuren Geneesk. Congr. 16, 176-178.
EINSTEIN, A.: Über Gravitationswellen. Sitzgsber. preuss. Akad. Wiss. 1918,
154-167.
THIRRING, H.: Über die Wirkung rotierender ferner Massen in den Einsteinsehen
Gravitationstheorie. Phys. Z. 19, 33-39.
BAcH, R.: Neue Lösungen der Einsteinsehen Gravitationsgleichungen. A. Das
Feld in der Umgebung eines langsam rotierenden kugelähnlichen Körpers von
beliebiger Masse in 1. und 2. Annäherung. Math. Z. 13, 119-133.
CARTAN, E.: Sur !es varietes a connexion affine et la theorie de la relativite generalisee. Ann. Ec. Norm. (3) 40, 325-412; 41, 1-25 (1924); 42, 17-88 (1925).
GORDON, W.: Zur Lichtfortpflanzung nach der Relativitätstheorie. Ann. d. Phys.
72, 421-456.
R 3.
1924 R 1.
1926 R 1.
1928 R 1.
R 2.
1931 R 1.
R 2.
1932 R 1.
R 2.
1933 R 1.
R 2.
1934 R 1.
R 2.
1937 R 1.
1939 R1.
R 2.
R 3.
R 4.
R 5.
1940 R 1.
R 2.
R3.
R 4.
R 5.
1941 R 1.
R 2.
R 3.
1942 R 1.
1944 R 1.
R 2.
R 3.
1945 R 1.
R 2.
Additional Bibliography R: Relativistic Continuum Theories. 791
MARCOLONGO, R.: Relativita, 2a edizione riveduta ed ampliata. Messina: G. Principato. ix + 235 pp., esp. p. 117.
EISENHART, L. P.: Space-time continua of perfect fluids in generat relativity.
Trans. Amer. Math. Soc. 26, 205-220.
MEKSYN, D.: On equilibrium and motion of a continuous medium in four-dimensional space. Phi!. Mag. (7) 2, 994-1006.
GHOSH, J.: On the gravitational field of an ideal fluid. Bull. Calcutta Math.
Soc. 19, 67-82.
LEVI-CIVITA, T.: Fondamenti di meccanica relativistica. Bologna: Zanichelli.
vii + 185 pp., esp. p. 70.
AKELEY, E. S.: The rotating fluid in the relativity theory. Phi!. Mag. (7) 11,
330-344.
AKELEY, E. S.: The rotating fluid in the relativity theory. Phys. Rev. (2) 37,
109-110.
DE DoNDER, TH., et Y. DuPoNT: Theorie relativiste de l'elasticite et de l'electromagnetostriction. Bull. Sei. Acad. Belg. (5) 18, 680-691, 782-790, 899-910;
19, 370-378 (1933).
FINZI, B.: Meccanica relativistica ereditaria. Atti Soc. Ital. Progr. Sc. 20, 8-11.
TOLMAN, R. C.: Thermodynamics and relativity. Science (2) 77, 291-298,
313-317.
ToLMAN, R. C., and H. P. RoBERTSON: On the interpretation of heat in relativistic
thermodynamics. Phys. Rev. 43, 564-568.
SYNGE, J. L.: The energy-tensor of a continuous medium. Trans. Roy. Soc.
Canada (3) 28, 127-171.
ToLMAN, R. C.: Relativity, thermodynamics and cosmology. Oxford: Clarendon
Press. xv + 502 PP·
SYNGE, J. L.: Relativistic hydrodynamics. Proc. Lond. Math. Soc. (2) 43, 376-416.
TAMM, J.: Cerenkov radiation and symmetry of the stress tensor. J. Phys. USSR.
1, 439-
VAN DANTZIG, D.: Stresstensor and particle density in special relativity. Nature,
Lond. 143, 855-
VAN DANTZIG, D.: On the phenomenological thermodynamics of moving matter.
Physica, Haag 6, 673-704.
VAN DANTZIG, D.: On relativistic thermodynamics. Proc. Kon. Nederl. Akad.
Wetensch. 42, 601-607 = lndagat. Math. 1, 212-218.
VAN DANTZIG, D.: On relativistic gas theory. Proc. Kon. N ederl. Akad. \Vetensch.
42, 608-625 = lndagat. Math. 1, 219-236.
EcKART, C.: The thermodynamics of irreversible processes, III. Relativistic theory
of the simple fluid. Phys. Rev. (2) 58, 919-924.
HAANTJES, J.: Die Gleichberechtigung gleichförmig beschleunigter Beobachter
für die electromagnetischen Erscheinungen. Proc. Kon. Nederl. Akad. Wetensch.
43, 1288-1299.
LrcHNERowrcz, A.: Sur un theoreme d'hydrodynamique relativiste. C. R. Acarl.
Sei., Paris 211, 11 7- 119.
ROSENFELD, L.: Sur le tenseur a impulsion-energie. Aead. Roy. Belg .. Cl. Sei.
Mem. Coll. 18, fase. 6, 30 pp.
VAN DANTZIG, D.: On the thermodynamics of perfeetly perfect fluids. Kon.
Nederl. Akad. Wetensch. Proc. 43, 387-401, 609-618.
BERGMANN, P. G.: On relativistic thermodynamics. Phys. Rev. (2) 59, 928.
LICHNEROWICZ, A.: Sur des theoremes d'unicite relatifs aux equations gravitationnelles du cas interieur. Bull. Sei. Math. (2) 65, 54-72.
LICHNEROWICZ, A.: Sur !'invariant integral de l'hydrodynamique re!ativiste.
Ann. Ee. Norm. (3) 58, 285-304.
BERENDA, C. W.: The problern of the rotating disk. Phys. Rev. 62, 280-290.
CosTA DE BEAUREGARD, 0.: La Relativite Restreinte et Ia Premiere J\'lecanique
Broglienne. Paris: Gauthier-Villars. 71 pp.
COSTA DE BEAUREGARD, 0.: Sur les equations fondamentales, classiques, puis
relativistes, de la dynamique des milieux continus. J. Math. Pures Appl. 23,
211-217.
EINSTEIN, A., and V. BARGMANN: Bivector fields, I and I I. Ann. Math. 45,
1-14, 15-23.
CosTA DE BEAUREGARD, 0.: Definition covariante de la force. C. R. Acad. Sei.,
Paris 221, 743-747.
MAUTNER, F., and E. ScHRÖDINGER: Infinitesimal affine connection with twofold Einstein-Bargmann symmetry. Proc. Roy. Irish Acad. 50, 223-231.
792 C. TRUESDELL and R. TouPIN: The Classical Field Theories.
R 3. ScHRÖDINGER, E.: On distant affine connection. Proc. Roy. Irish Acad. 50,
143-154.
1946 R 1. CosTA DE BEAUREGARD, 0.: Sur la conservation de la masse propre. Sur la notion
de fluide parfait. C. R. Acad. Sei., Paris 222, 271-273.
R 2. CosTA DE BEAUREGARD, 0.: Equations generales de l'hydrodynamique des fluides
parfaits. C. R. Acad. Sei., Paris 222, 369-371.
R 3. COSTA DE BEAUREGARD, 0.: Sur la theorie des forces elastiques. c. R. Acad. Sei.,
Paris 222, 477-479·
R 4. COSTA DE BEAUREGARD, 0.: Sur la thermodynamique des fluides. C. R. Acad.
Sei., Paris 22:Z: 590-592.
R 5. CoSTA DE BEAUREGARD, 0.: Retour sur la dynamique et la thermodynamique
des milieux continus. C. R. Acad. Sei., Paris 222, 1472-1474.
R 6. CosTA DE BEAUREGARD, 0.: Dynamique relativiste des milieux continus. La
variation de la masse propre en fonction du travail des forces superficielles. J.
Math. Pures Appl. 25, 187-207.
1947 R 1. DE WET, J. S.: Symmetrie energy-momentum tensors in relativistic field theories.
Proc. Cambridge Phil. Soc. 43, 511-520.
R 2. HILL, E. L.: On the kinematics of uniformly accelerated motions and classical
electromagnetic theory. Phys. Rev. 72, 143-149.
R 3. RosEN, N.: Notes on rotation and rigid bodies in relativity theory. Phys. Rev.
71, 54.
R 4. WEYSSENHOFF, J ., and A. RoABE: Relativistic dynamics of spin fluids and spin
particles. Acta phys. Polonica 9, 7-18.
1948 R 1. TAUB, A. H.: Relativistic Rankine-Hugoniot equations. Phys. Rev. (2) 74,
328-334.
1949 R 1. CLARK, G. L.: The mechanics of continuous matter in relativity theory. Proc.
Roy. Soc. Edinburgh, Sect. A 62, 434-441.
R 2. CosTA DE BEAUREGARD, 0.: La Theorie de la Relativite Restreinte. Paris: Masson
& Cie. vi + 173 pp.
R 3. FoKKER, A. D.: On the space-time geometry of a moving rigid body. Rev. Mod.
Phys. 21, 406-408.
R 4. PAPAPETROU, A.: Non-symmetrie stress-energy-momentum tensor and spin
density. Phil. Mag. 40, 937-946.
1950 R 1. BLOCH, C.: Variational principle and conservation equations in non-local field
theory. Kgl. danske Vidensk. Selsk., mat.-fys. Medd. 26, 30 pp.
R 2. CLARK, G. L.: The extemal gravitational and electromagnetic fields of rotating
bodies. Proc. Roy. Soc. Lond., Ser. A 201, 488-509.
R 3. DE HoFFMAN, F., and E. TELLER: Magneto-hydrodynamic shocks. Phys. Rev.
(2) 80, 692-702.
R 4. RoBERTSON, H. P.: The geometries of the thermal and gravitational fields. Amer.
Math. Monthly 51, 232-245.
1951 R 1. HILL, E. L.: The deformation of moving coordinate systems in relativistic theories.
Phys. Rev. 84, 1165-1168.
R 2. LEAF, B.: The continuum in special relativity theory. Phys. Rev. 84, 345-350.
1952 R 1. GETHING, P. J. D.: Rotation and magnetism in the world-models of kinematic
relativity. Monthly Notices Roy. Astronom. Soc. London 112, 578-582.
R 2. MöLLER, C.: The Theory of Relativity. Oxford.
R 3. TAKENO, H.: On relativistic theory of rotating disk. Progr. Theor. Phys. Japan
7, 367-376.
R 4. UENO, Y., and H. TAKENO: On equivalent observers. Prog. Theor. Phys. Japan
8, 291-301.
R 5. WHITROW, G. J.: The Fitzgerald-Lorentz contraction phenomenon and theories
of the relativity of Galilean frames. Sei. Proc. Roy. Dublin Soc. 26, 37-44.
1953 R 1. G"il'RSEY, F.: Gravitation and cosmic expansion in conformal space-time. Proc.
Cambridge Phil. Soc. 49, 285-291.
R 2. KIRKWOOD, R. L.: The physical basis of gravitation. Phys. Rev. 92, 1557-1562.
R 3. MATRAI, T.: A relativistic treatment of rigid motions. Nature, Land. 172,
858-859.
R 4. STEPHENSON, G., and C. W. KILMISTER: A unified theory of gravitation and
electromagnetism. Nuovo Cim. 10, 230-235.
R 5. TAYLOR, N. W.: The relativistic electromagnetic equations in a material medium.
Austral. J. Phys. 6, 1-9.
R 6. UENO, Y : On the wave theory of light in general relativity. Progr. ·Theor. Phys.
Japan 10, 442-450.
Additional Bibliography R: Relativistic Continuum Theories. 793
1954 R 1. GoMES, R. L.: The motion of rigid body in restricted relativity. Gaz. Mat. Lisboa
15, No. 58, 9-11.
R 2. KrLMISTER, C. W., and G. STEPHENSON: An axiomatic criticism of unified field
theories, I and II. Nuovo Cim. 11, Suppl. 91-105. 118-140.
R 3. KrRKWOOD, R. L.: Gravitational field equations. Phys. Rev. 95, 1051-1056.
R 4. POUNDER, J. R.: On relativistic rigid surfaces of revolution. Comm. Dublin Inst.
Adv. Study, Ser. A, No. 11, 53 pp.
R 5. SALZMAN, G., and A. H. TAUB: Born-typerigidmotion in relativity. Phys. Rev.
(6) 95, 1659-1669.
R 6. ScHRÖDINGER, E.: On distant affine connection. Proc. Roy. Irish Acad. 50,
143-154.
R 7. SYNGE, J. L. :Relativistically rigid surfaces, Studies in mathematics and mechanics
presented to Richard von Mises, pp. 217-226. NewYork: Academic Press.
R 8. TAUB, A. H.: General relativistic variational principle for perfect fluids. Phys.
Rev. 94, 1468-1470.
1955 R 1. BoLAZS, N. L.: The propagation of light rays in moving media. J. Opt. Soc. Amer.
45, 63-64.
R 2. FrNKELSTEIN, D.: Interna! structure of spinning particles. Phys. Rev. 100,
924-931.
R 3. KREMPASKY, J.: The strain tensor in space-time as a result of motion. Mat.
Fyz. Casopis Slovensk. Akad. 5, 124-131.
R 4. LrcHNERowrcz, A.: Theories Relativistes de la Gravitation et de l'Electromagnetisme. Paris: Masson & Cie.
R 5. MARX, G., u. G. GYÖRGYr: Über den Energie-Impuls-Tensor des elektromagnetischen Feldes in Dielektrika. Ann. Phys. (6) 16, 241-256.
1956 R 1. GoMES, R. L.: Kinematics of rigid bodies in relativity. Gaz. Fis. Lisboa 3, 99-107.
R 2. ScHILD, A.: On gravitational theories of Whitehead's type. Proc. Roy. Soc.
Lond., Ser. A 235, 202-209.
R 3. ScHNUTZER, E.: Zur relativistischen Elektrodynamik in beliebigen Medien. Ann.
Phys. (6) 18, 171-180.
R 4. SYNGE, J. L.: Relativity: The Special Theory. Amsterdam: North Holland
Publishing Co.; New York: Interscience Publ.
R 5. TAUB, A. H.: Isentropic hydrodynamics in plane symmetric space-times. Phys.
Rev. (2) 103, 454-467.
R 6. WrNOGRADSKI, J.: Sur Je tenseur impulsion-energie metrique et Je theoreme de
Noether. Cahiers de Phys. 67, 1-5.
19 57 R 1. ARNOWITT, R. L.: Phenomenological approach to a unified field theory. Phys.
Rev. 105, 735-742.
R 2. MARX, G.: Innere Arbeit in der relativistischen Dynamik. Acta phys. Acad. Sei.
Hungar. 6, 353-379.
R 3. SYNGE, J. L.: The Relativistic Gas. Amsterdam.
R 4. TAUB, A. H.: Singular hypersurfaces in general relativity. Illinois J. Math.
1, 370-388.
1959 R 1. SYNGE, J. L.: A theory of elasticity in general relativity. Math. Z. 12, 82-87.
R 2. TAUB, A. H.: On circulation in relativistic hydrodynamics. Arch. Rational Mech.
Anal. 3, 312-324.
Appendix.
Tensor Fields.
By
J. L. ERICKSEN.
With 2 Figures.
I. Preliminaries.
1. The nature of this appendix. This appendix is a heterogeneaus collection of results
on tensor fields. The reader is assumed familiar with the elements of vector and tensor
analysis and of matrix algebra. The author is indebted to Drs. W. NoLL, R.A. TouPIN,
C. TRUESDELL and M. K. ZoLLER for bringing to his attention many references and helping
to correct many errors. Such defects as remain are of course the responsibility of the author.
a) Notation.
2. Tensor notation. We employ the methods and notation of the tensor
calculus and, to a lesser extent, matrix algebra and GIBBS' vector analysis. Until
further notice, write xl, x2, ••• , x" for the co-ordinates of pointsinan n-dimensional
space. For abbreviation, we let a: stand forthisset of Co-ordinates. Until further
notice, italic lower case letters other than x will be used for the kernel indices
and tensor indices of tensors. Also, for a given set of tensor components a"···"'p ... q
we write a; for "the components a"···"'p ... q of the tensor a" or "the tensor having
the components a"···"'p ... q"• we write "the tensor a"--·"'p ... q"• or, more often, "the
tensor a ". W e employ the summation convention for diagonally repeated tensorial
indices and indices of Christofiel symbols. The notation a:::;" denotes the covariant
derivative, except when otherwise noted.
A square bracket enclosing m running indices indicates that the tensor is
completely antisymmetrized with respect to these indices. This is done by
permuting the indices in all possible ways, attaching a positive (negative) sign
to the tensors corresponding to even (odd) permutations, adding these tensors
and dividing by m!. Parentheses inclosing m running indices indicate that the
tensor is completely symmetrized with respect to the enclosed indices. This is
done in the same way except that the positive sign is attached to all permutations.
For example
W e use the short notation
\ IX\ = 8 (xl, x2, ... , x")
a:, - o(X1,X2, ••• ,Xn)
(2.1)
(2.2)
when the x" are given in terms of parameters XK by a mapping a:= a: (X), whether
this be regarded as a co-ordinate transformation or as a point transformation.
Unless greater or lesser generality is asserted explicitly or is evident from the
context, the space considered in this treatise is Euclidean 3-space with real
Sect. 3. Matrix and vector notations. 795
co-ordinates. The components of associated tensors are regarded as different
sets of components of the same tensor; the same kernelleHer is used for all these
components, and the context will make clear whether a stands for a particular
set of components or is intended as a general symbol for all components. In
particular, weshall often write 1 for the metric tensor gkm• its mixed components
ö!:,, and its inverse gkm_
Setting g == det gk m and assuming g > 0, we use the notations
(2.3)
where n is the dimension of the space and the s's are the usual permutation symbols, with sl...n = sl. .. n = 1 in all co-ordinate systems. The e's are examples of
axial tensors, which transform as absolute tensors under co-ordinate transformations with positive J acobian, and und er those only.
In Euclidean space, a reetangular Cartesian co-ordinate system is available.
If we write z or z1, z2, ••• , zn for the co-ordinates of points, this in itself indicates
that we are using reetangular Cartesian co-ordinates, and at the sametime we follow
the notation and the summation convention of Cartesian tensors, writing all tensor
indices as subscripts and summing on repeated indices.
Moreover, for the position vector from a fixed point to x, we write pk or pk
or p. If the fixed point has reetangular Cartesian co-ordinates 0z, then p is the
vector whose covariant and contravariant components reduce to z- 0z in reetangular Cartesian systems. If 0z = 0, then in reetangular Cartesian systems
and only in such systems we have x =p =Z. Formulae in which p occur should
be invariant when 0z is replaced by a different fixed point 0z*; in most cases,
verification of this invariance is left to the reader. In general co-ordinates,
p and x are entirely different quantities. For the vector element of arc we write
dx in general Co-ordinates, dz in formulae whose validity is limited to reetangular
Cartesian co-ordinates. Curvilinear components of p at x generally differ from
those at the fixed point. Except where otherwise noted, we use the former.
3. Matrix and vector notations. For second order tensors, we sometimes use
matrix or dyadic notation1. For example, we write c =a · b for the equation
c!:, = a~ b!;.; a-1 for the tensor, if it exists, which is related to a by a · a-1 =1,
the components of a-1 being a;;;lk; while a' denotes the transpose of a, given by
a~m =amk·
More generally, the single dot operation is defined by the equations a · b == t
and b · a = u, where the tensors t and u are given by
(3.1)
In three-dimensional spaces, also the single cross operation of GIBBS 2 may
be defined. If a is any vector, b any tensor, then a X b == c and b X a == d, where
the tensors c and d are given by
(3.2)
The cross or vector tx of a second order tensor t is the axial vector given by
tk = ekrs t = ekrs t x rs [rs]· (3 ·3)
Thus apart from algebraic sign, the contravariant components of tx are the components of 2 Vg t[rs]• not of Vg t[rs]• andin particular (grad b)x =curl b. Since a
1 Alternatively, our direct notation may be interpreted in terms of linear transformations.
Cf. e.g. HALMOS [1942, 1, Chap. II], and, for applications to kinematics and mechanics,
NoLL [1955. J].
2 [1881, J], [1884, J], [1909, 1].
796 J. L. ERICKSEN: Tensor Fields. Sects. 4, 5.
skew-symmetric !ensor of second order in a three-dimensional space has at most
three independent components, in any equation where it appears it may be
replaced by its cross. When both t and u are skew-symmetric,
(3.4)
We introduce the symbol '*'· which may be read "trident", to stand for a
scalar, vector, or tensor of any order, so long as the formulae in which it appears
have a sense. By grad '*'· div W, and curl W we mean the tensors a, b, and c
given by
a,. ........ = w ........ ,k. b"·""'= '*''""· ... ,,. or c:. ..... =e"" w............ (3.5)
where the last applies only in the three-dimensional case. These definitions are
so framed that the identities div curl W = 0 curl grad W = 0, familiar when W
is a vector, hold for a tensor of any order.
For positive integral K, the K-th power lJCK) of a vector bis the tensor b"'• .. . b mk,
The definition is extended to all integral K and put in inductive form as follows:
lJH)= 0, lJ(O) 1, lJ(K+l) = b(K) b. (J.6)
We denote by {bCKl '*'} the expression1
{b(K) W} = b(K) '*' + b(K-l) Wb+ ... + Wb(K). (3.7)
4. Projections. If n is a unit vector, we define the normal projection Wn and
the tangential profection W1 of w onto the direction of n by
(4.1)
the latter formula being restricted to three-dimensional space. When W = c,
a vector, we write Cn rather than Cn for the normal projection.
Again in 3-space, consider the special Co-ordinate transformations which leave x1 fixed:
zU= x1, x•« = x•« (x2, zll), and write 1g= det g .. p. where cx, ß have the range 2, 3. Then,
considering only those co-ordinate systems such that 1g > 0, we see that the quantities 1t
defined by
(4.2)
where t is a three-dimensional tensor, transform as components of a tensor 2 an the surface
x1 = const. The two-dimensional tensor 1t may be called the skew projection of t onto the
x1-direction. Taking the trace of Eq. (4.2) yields a relation3 between the components of tx
and the traces of the three skew projections kt:
t; = (gfkg)! k t~. (4. 3)
b) Use of complex co-ordinates.
5. Use of tensor methods in complex co-ordinates. The formalism of the
tensor calculus applies to complex as well as real co-ordinate transformations;
for problems in the Euclidean plane, complex Co-ordinates are sometimes convenient. Suppose, for example, that we introduce complex co-ordinates x1 =
z1 +iz2 , x2 =XI, where the z,. arereetangular Cartesian and the superposed bar
denotes the complex conjugate. The usual formulae of tensor transformation
1 TRUESDELL [1949, 3], [1951, 2], [1951, 3], [1954, 2, Chap. I].
2 This result and Eq. (4.2) are due in principle to CAUCHY [1841, 1, Th. VIII]. Cf. TRUESDELL [1954, 6, § 2].
3 When both tx and ,.e are referred to physical components (Sect. 13) in an orthogonal
co-ordinate system, the factor (gf1g)! disappears from Eq. (4.3).
Sects. 6, 7- Preliminary definitions. 797
yield the covariant components t""' (x) in terms of the Cartesian components t".,.:
4tu (x) = 4t22 (x) =tu-t2 2- i (t12 + t21), }
4t12 (x) =4t21 (x) =tu +t22 +i(t12 -t21). (5.1)
For the special case1 when t is the metric tensor, this yields gu(x) =g22 (x) =0,
g12 (x) = g21 (x) = ! ; hence g11 (x) = g2 2 (x) = 0, g12 (x) = g21 (x) = 2. All results
which do not depend essentially on conditions of reality carry over to complex
co-ordinates. As is clear from the above example, the inequalities gu > 0,
gug22 -g~ >0 which, in real co-ordinates, express the fact that the g""' are
coefficients of a real, positive definite quadratic fom1, do not. Since we occasionally
use such conditions, our treatment is restricted to real co-ordinates except where
otherwise noted.
6. Complex representation of plane second order Cartesian tensors. Retuming
to the example (5.1), we note that a rotation of the co-ordinates z" induces on x"
the transformation xu = xu = x1 e•'P, where cp is the angle of rotation, and that
the tensor t11 .,. (x) transforms according to the simple laws
t12 (x*) = t21 (x*) = t12 (x) = t21 (x), fu (x*) = t22 (x*) = t11 (x) e-2i'P. (6.1)
In linear elasticity, t, interpreted as the stress tensor, satifies the equilibrium equations
and compatibility conditions for generalized plane stress if and only if 2
112 (:r) = t21 (:r) = tZ = I + J. 111 (:r) = g - x2 /', (6.2)
where I and g are arbitrary functions of the complex variable x1• Similar results hold for
plane strain. In this context, Iu (:r) is sometimes called the conjugate stress deviator to indicate
that it is unaltered if t is replaced by t + K 1, where K is any scalar and 1 is the two-dimensional
unit matrix. Comparison of these results with their Cartesian analogues shows the simplicity
that can result from using complex co-ordinates. In both applications all three Cartesian
components of t may be calculated easily from the two complex components 112 (:r) and
Iu (:r), the former of which may be regarded as a scalar. It follows easily from Eq. (5.1)
that 112 (:r) = 0 if and only if I is symmetric and traceless, and that in the symmetric case
111 (:r) = 0 if and only if t = K 1, where K is a scalar.
When t is symmetric, by Eq. (5.1) 2 we see that t12 (x) is always real. As will
be shown in Sect. 48, there exists a real rotation that renders t diagonal. Letting
x* =zi +izi, from Eq. (5.1) we see that tu(x*) is real. By Eq. (6.1) we get
t11 (x) =fu(x*) i'~', and hence Eq. (5.1) yields 3
2112 tan2cp= ----.
111-122
II. Dimensionsand physical components.
a) Dimensionsofatensor and its components.
(6.3)
7. Preliminary definitions. Among the transformations to which quantities
occurring in physics are subject are those of dimensional units 4. In constructing
a system of measurement, in principle one must introduce a unit of measurement
for each type of physical quantity to be considered. It is customary to lay down
a finite set of units U0 ( a = 1, ... , n) as fundamental units, and to require that
1 For other examples, cf. GREEN and ZERNA [1954, 1].
2 MusKHELISHVILI [1953, 1, Chap. 5] gives a rigorous derivation of these results, which
he attributes to KoLOsov [ 1909, 4].
3 MAXWELL [1870, 3, PP· 194-195]. 4 Dimensional analysis remains a controversial and somewhat obscure subject. We
do not attempt a complete presentation here.
798 J. L. ERICKSEN: Tensor Fields. Sect. 8.
every unit be expressible in terms of them in the symbolic form Uf' ... U~n, where
the K's arereal exponents. A fundamental unit may be regarded as a measure
of a definite type of entity, such as length or time, the "size" of the unit being
assigned arbitrarily. To compensate for this arbitrariness, we regard any other
system of units, obtainable from the given set by changing the size of the fundamental units independently, as equivalent or indistinguishable. More precisely,
we require physical equations to be invariant under the transformations
(7.1)
hereafter called dimensional transformations, the ua being arbitrarily chosen real
numbers. A function Q (U) of the units Ul, ... , Un is called a measure number
having the dimension [Uf' ... U~n] provided that it transforms under Eq. (7.1)
according to the law Q (U*) = Q (U) Uf' ... U~n. The appropriate units for Q are
Uf' ... U~n, in accordance with the usual notion that the product of a measure
nurober by its units is unaffected by dimensional transformations. We write [1]
for [U~ ... un
Fora given group Gof Co-ordinate transformations, a set of functionstZ:::~(;r, U)
are said to be the components, relative to the co-ordinates xk and units Ua, of a
tensor of weight w having the dimension [Uf' ... U~n] provided that, under Eq. (7.1)
and an arbitrary co-ordinate transformation x*k = x*k (x) of G, these functions
transform according to the law
(7.2)
With a fixed choice of co-ordinates, each component of t is clearly a measure
nurober having the dimension [Uf' ... U~n]. Statements to the effect that physical equations are "in their simplest form" generally mean only that they reduce
to equations relating such tensors.
8. Physical dimensions of tensors. In classical mechanics, it is customary to
take G to be the orthogonal group1, to require one of the fundamental units
to be a unit L of length, and to assume that the co-ordinate differentials dzk
transform as a Cartesian vector having the dimension [LJ, in agreement with
the notion that ds2 = dzk dzk is an absolute scalar having the dimension [L2].
Using physicallaws and other devices, one determines the appropriate Cartesian
tensor law to represent the entities to be considered. The dimensions so assigned
in reetangular Cartesian systems we call physical dimensions to distinguish them
from other dimensionstobe introduced below. From this point of view the Cartesian metric tensor bkm can consistently be regarded as a dimensionless tensor,
by which is meant a tensor having the physical dimension [1].
In passing to orthogonal curvilinear co-ordinates, one assigns a dimension
to each variable introduced as a co-ordinate 2• The common procedure is to use
LAM:E's theory of orthogonal curvilinear co-ordinates 3, or some other mathematically equivalent procedure, to obtain curvilinear components of Cartesian
tensors. Viewed as scalars, these components have a common dimension, which
is the physical dimension of the Cartesian tensor which they represent. Under
co-ordinate transformations they do not transform as simply as do tensor com1 For simplicity, we restriet ourselves to time independent co-ordinate transformations. 2 In principle, this dimension may be arbitrary. In practice, each co-ordinate is a homogeneous function of the reetangular Cartesian co-ordinates, so that it is natural to assign it
the dimension [L K], where K is the degree of the homogeneaus function. 3 The method is given in [1840, 1], [1859, 1, Les;ons 1-2]; a typical application, in
[1841, 2, §§VII-VIII], [1859, 1, §§ 144-147].
Sect. 9. Absolute dimensions of tensors. 799
ponents. Since Rrccr and LEVI-CEVITA1 set down the foundations of tensor
calculus, there has been an increasing tendency to use tensor components in
place of LAME's, primarily because tensor components transform more simply
under co-ordinate transformations. In treating Cartesian tensors having dimensions, one fixes the units and transforms the components according to the appropriate tensor law. Having done so, one can assign a dimension to each component of the tensor consistent with the tensor law, the dimension assigned to
the Cartesian tensor, and the dimensions assigned to the co-ordinates. In general,
thesedimensionswill be different for different components and from those of the
Cartesian tensor. At least for an absolute tensor t, the dimension of the scalar
which we denote by the Russian letter ,Il,,
n = lit tk ... m ~t- k ... m ' (8.1)
will be unaltered by co-ordinate transformations, and it is customary to regard
this dimension as the physical dimension of t.
Apparently McCoNNELL 2 was the first to give a satisfactory explanation of
the relation between the "components" of mathematical physics and tensor
components for tensors of arbitrary order. His method might be explained as
follows. Given a tensor field referred to an orthogonal curvilinear co-ordinate
system, let its physical components 3 at a given point be defined as its corresponding
tensor components, at that point, in a reetangular Cartesian system whose axes
are parallel to the co-ordinate curves at the point. In mathematical physics,
the term "components" usually means these physical components, which henceforth we denote by t =--"-log fg""' uSm
and k =m
k=j=m, (14.6)
k=j=m.
To calculate physical components of derivatives, one has only to substitute these
formulae into
t(k ... m,P>=a:p t(k ... m) +F+ ··· +} (14.7)
+F be that portion of the surface of the sphere of radius p with center
at P which lies in the interior of t. If \1:' be such that
lim I da \I:'P"' = 0, p_.oo,P (24.1)
we write 2, respectively,
\l:'n=ö(p-m-2), \l:'t=o(p-m-2), \l:'=o(p-m-2). (24.2)
For (24.1) to hold it is clearly sufficient that each component of \l:'n, \l:'t, and \1:',
respectively, be o (p-"'-8) as P~ oo, which motivates the notation. The symbol o
may be read "lower mean order than ".
Weshall frequently wish to state boundary conditions of the following type:
1. On each finite boundary of v, \1:' = 0.
2. In any portion of v extending to \1:',
\1:' = 0 (p-m-2).
As apart of the convention of Sect. 23, we agree that henceforth
\1:' = 0 or o(p-"'-8) on ~ (24.3)
shall serve as an abbreviation for the statements 1 and 2. Analogons conventions are adopted for statements in terms of \l:'n or '*'t·
b) Circulation, flux, total,. and moments.
25. Definitions. Let \1:' be any integrable tensor field. The line integral
I dz · \1:' =I dz,. \l:'k...... (25.1) 1: 1:
is called the flow of \1:' along the curve c. When c is closed, it is called a circuit;
the integral (25.1) is then called the circulation of \1:' around c and is written
~dz· \1:'. {25.2) 1:
Two curves are reconcilable 3 in a given point set provided one can be continuously
deformed into the other while remaining in the set. A circuit which can be
1 [1929, 1, Chap. IV]. 2 The notations were introduced by TRUESDELL [1954, 2, § 4]. 3 These concepts and names derive from KELVIN [1869, 1, §58 and § 6o(a)]. Topologists
use the ward "homotopic" in place of "reconcilable", whereas the latter ward is generally
employed in mechanics. KELVIN stated that he applied RIEMANN's [1857, 1] theory of
multiple connectivity as presented by HELMHOLTZ [1858, 1], but the concept of reconcilable
curves does not appear in these papers.
Sect. 26. GREEN's transformation. 815
continuously shrunk to a point while remaining in a given set is said tobe reducible
in the set.
The surface integral
f da· \V (25.3) 4
is called the flux of \V across the surface 6. When 6 is closed and da is directed
outward, the integral (25.3) is called1 the flux of \V out of 6 and is written
~da· \V. (25 .4) 4
A closed surface which can be continuously shrunk to a point while remaining
in a given region is said to be reducible in it.
With the notation (3 .7), the volume integral
f {p(K) \V} dv (25.5)
"
is called the K-th moment of \V with respect to the origin. The zeroth moment,
f \Vdv (25.6)
" is called 2 the total \V in v-.
c) The transformations of GREEN and KELVIN.
26. GREEN' s transformation. We shall employ Green' s transformation s in the
forms
fgrad\Vdv=~da\V, fdiv\Vdv=~da·\V, fcurl\Vdv=fdax\V, (26.1) " d ". d ". d
where it is assumed that \V is single valued and continuous throughout the closure
of the finite region v-, bounded by d, that grad \V is continuous throughout a
finite number of subregions of v- whose sum is v-, and that the volume integrals
are convergent. In the terminology of Sect. 25, Eq. (26.1) 2 states that the total
divergence of a quantity in a region is equal to the flux of the quantity out of
the boundary of the region.
Another form of GREEN's transformation is 4
f [b {cp. The case K=O is due to BuRGATTI
[1931, 3].
816 J. L. ERICKSEN: Tensor Fie!ds. Sects. 27, 28.
where b and c are arbitrary continuously differentiable vector fields, p is the
position vector, and K is any integer.
27. POINCAR:E's generalization. In n dimensions, the tensor element of area
of an oriented m-dimensional surface dm, given parametrically by xk = xk ( u), is
d r ... s- I ~[r. 0. ozS] d 1 d m
S(m) - m. oul oum U • . . U ' (27.1)
which transforms as an absolute contravariant tensor of order m under co-ordinate
transformations, as an absolute scalar under parameter transformations with
positive Jacobian. Let !lm be bounded by the surface !lm_1 , given parametrically
by xk = xk (v). Let v be a vector defined on dm_1 , directed outward relative to !lm.
Arrange the parametrizations so that the vectors vk, oxkjovl, ... , axkjvm-1 and
8xkj8u1, 00.' axkjoum, in these orders, have the same orientation relative to (jm.
For any continuously differentiable covariant tensor field t of order m, we then
have 1
--- t dsru ... s = --t dsru ... s = t dsu ... s f ox' a u ... s (m) Ja ox[r u ... s] (m) ~ u ... s (m-1). (27.2)
It then follows from the transformation properties of the integrands that each
integral transforms as an absolute scalar. The underlying space need not be
Euclidean or have any other geometrical structure. In metric spaces, or affine
spaces with a symmetric connection, where covariant differentiation is defined,
we have a!ru tr ... s]=trr ... s,u]• which makes the invariance of the integrals more
obvious. Various alternative forms of Eq. (27.2) are available in the literature 2 •
In three-dimensional Euclidean space, GREEN's transformation may be deduced
from Eq. (27.2) with m = 3, KELVIN's transformation (Sect. 28) from Eq. (27.2)
with m=2.
28. KELVIN's transformation. KELVIN's transformation, commonly 3 called
" STOKEs' theorem ", is
J da. curl \j! = cji dz. \j!, (28.1) 6 e
subject to the usual right-handed screw convention connecting the sense of dz
with that of da. \j! is assumed continuously differentiable in a region whose
boundary contains 4 d. It is important to note that the field \j! need not be defined
on both sides of 6. In the terminology of Sect. 25, Eq. (28.1) states that the
circulation of a quantity araund a closed circuit equals the flux of its curl across
any surface bounded by the circuit.
It is possible to give forms of Eq. (28.1) which contain dz \j! and dzx \j!
on the right-hand side, in the style of Eq. (28.1). Since these are awkward in
dyadic notation, we write the former in Cartesian tensor notation as follows;
(28.2)
Contradingthis equation on k and m yields Eq. (28.1), while replacing km by
[km J yields the second alternative mentioned above.
1 This result is due in essence to POINCARE [1887, 1, § 2], [1895. 1, § 7]. 2 See e.g. ScHOUTEN [1951, 1, Chap. IVl. 3 We follow TRUESDELL [1954, 2, § 8] ih departing from usage. See the appendix to
this section.
4 See LICHTENSTEIN [1929, 2, Kap. 2, § 3] for references to works proving the result
under weaker assumptions.
Sect. 29. Definitions. 817
Appendix. History of Kelvin's transformation 1• The plane form of Eq. (28.1) was discovered by AMPERE [1826, E1]. The three-dimensional form was discovered by KELVIN,
who communicated it to STOKES in a letter dated July 2, 1850 {LARMOR's annotation to
the 1905 reprint of [1854, 1] ). Independent discovery, as weil as priority in publication, is
due to HANKEL [1861, E 1, § 7] (cf. also RocH [1863, E 1, § 4]). KELVIN's proofs were published later [1867, 1, § 190(j)] [1869, E1, § 60(q)], andin the second of these publications
he claimed priority. The only connection of STOKES with the matter was to set proof of the
result as an examination question [1854, T2]. Later KELVIN [1879 ed. of [1867, 1, § 190(j)]J
mentioned this fact, which had been noted earlier by MAXWELL [1873, 3, § 24], after whose
custom the common name has been adopted. It was KELVIN who first realized the significance of Eq. (28.1). showing how to use it for the proofs of important kinematical theorems.
If additional names beyond KELVIN's aretobe attached to Eq. (28.1), they should be those
of AMPERE and HANKEL.
V. Vector fields.
a) Vector lines, sheets, and tubes.
29. Definitions. A vector line of a given vector field c is a curve everywhere
tangent to c. Parametrie equations of a vector line are the solutions x" = x" (u)
of the differential system
or
dU d.c XC =0, l !!;~" cml = 0.
(29.1)
If c is continuous in a closed region,
there exists at least one vector line
through each interior point of the Fig. 2. Vector tube and cross·sections.
region; if c also satisfies a Lipschitz condition 2, there is exactly one
through each point where c =f= 0.
A surface everywhere tangent to c is a vector sheet of c. Such surfaces f (~) =
const are non-constant solutions of
c · gradf = 0, or ,. of
c oxk =0. (29.2)
Equivalently, a parametric representation x =X (u, v) of a vector sheet satisfies
o.c o:JJ o.c o.c C·a;-Xav=O, -0u-Xav=f=O. (29.3)
A vector line can always be represented locally as the intersection of two vector
sheets. Consider a field c possessing a unique vector line through each point of
a region r, and let c be a curve which is not a vector line of c; then, locally, the
surface swept out by the vector lines intersecting c define a unique vector sheet
of c associated with c. When c is a circuit, this vector sheet is called a vector
tube of c.
Let 6 be a vector tube of c, bounded by two simple circuits c1 and c2 , each
embracing the tube once. Let 61 and 62 be two surfaces whose complete boundaries
are c1 and c2 respectively (Fig. 2); when the unit normals to 61 and 6 2 are given
senses so as to subtend an acute angle with c, d1 and 62 are cross-sections of the
tube. Assuming c is integrable on the cross-sections 61 and 6 2 , we call its flux
through them the strengths of the vector tubes at these cross-sections. Letting
63 = d + 61 + 62 , we have
~da· c = J da· c + J da· c, (29.4)
•• ., •• 1 Cf. TRUESDELL [1954, 2, § 8}. 2 See e.g. KAMKE [1930, 1, §§ 15, 16] for proof of uniqueness under weaker·conditions.
Handbuch der Physik, Bd. III/1. 52
818 J. L. ERICKSEN: Tensor Fields. Sect. 30.
since da· c =0 on ~. In Eq. (29.4), the normals to 1 and 2 are both taken outward. If c is directed inward relative to d on :1 1 and we take the unit normal on :11
tobe directed inward, we have
~da· c = J da· c- J da· c. (29.5) da d:a d1
Hence the flux of c out of a surface consisting of a vector tube of c and two crosssections equals the ditference between the strengths of the tube at these cross-sections.
Let v be the region bounded by :13 • Assuming sufficient smoothness for GREEN' s
transformation (26.1) to hold, from (29.5) we have
fdivcdv = J da ·C-J da· c. (29.6) V 62 61
H ence the total divergence of the field c in a region bounded by a vector tube of c and
two cross-sections equals the ditference between the strengths of the tube at these crosssections.
The above definitions may be extended to n-dimensional spaces which need not possess
a metric tensor. lf c is any relative or absolute contravariant vector, the vector lines of c
are the curves obtained as solutions of Eq. (29.1) 2 • As before, in a closed region where c
satisfies a Lipschitz condition there exists a unique vector line through each interior point
at which c =1= o. The vector lines intersecting an rn -1-dimensional surface c, when 1 < rn< n,
then sweep out an rn-dimensional vector sheet d of c, this being defined as an rn-dimensional
surface given parametrically by xk= xk(ul, ... , um), where
();x[k ox' --···--c•l-o 8u1 aum - ' (29.7)
or, equivalently, as the set of points satisfying fa (::~:) = const, a = 1, ... , n-rn, where the
f's are functionally independent solutions of Eq. (29.2) 2 • lf c be closed, 6 is called a vector
tube of c. For an n-dimensional vector tube of a relative vector of weight one, generalizations of Eqs. (29.4) and (29.5) are immediate, and a generalization of Eq. (29.6) is obtained
by setting tr ... s=ck Ekr ... s in (27.2).
30. Invariants of vector lines. Consider a region in which the magnitude c
of a vector field c is non-zero. Let t = cjc 'denote the unit tangent to the vector
lines of c, and put
Then
Now
so that
A =t·curlt, D = divt.
A = t · curl (cjc) = (tjc) · curl c = c- 2 c · curl c.
grad c = grad (c t) = (grad c) t + c grad t,
(30.1)
(30.2)
(30. 3)
div c = t · grad c + c div t = ~; + c D, (30.4)
curl c = grad c X t + c curl t, (30.5)
where djdt denotes differentiation with respect to arc length along the vector
line1 •
For any unit vector e and any vector v, we have the identity
v=ee·v-ex(exv), (30.6)
so that
curl c = t t · curl c - t X (t X curl c) . (30.7)
1 Here and elsewhere, we denote by dfde the directional derivative in the direction of a
unit vector e.
Sect. 31. Solenoidal fields I: Integral properties. 819
Using Eqs. (30.5) and (30.6) with e = t, v = grad c, we get
txcurlc =tx(gradcxt) +ctxcurlt, }
= grad c - t t · grad c + c t X curl t. (30.8)
Introducing the unit principal normal n and binarmal b of the vector lines, we
have
grad = t t · grad + n n · grad + b b · grad,
which enables us to rewrite Eq. (30.8) as
t X curl c = n n · grad c + b b · grad c + c t X curl t,
dc dc
= n !in + b Tb + c t x curl t,
dc dc = n dt + b Tb - c t · grad t,
( dc ) dc
=n dn- cx + bdb,
where x is the curvature of the vector line. Hence
dc ( dc) -tx(txcurlc)=libn + cx -!in b.
(30.9)
(30.10)
(30.11)
From Eqs. (30.2), (30.7), and (30.11) we then obtain an intrinsic relation for 1
curl c:
dc ( dc) curl c = c A t +Tb n + c x - dn b. (30.12)
The special case c = 1 yields
curl t = A t + x b. (30.13)
Following LEVI-CIVITA 2, we call theinvariant A the abnormality of the field,
this name being suggested by the following geometric interpretation, attributed
to BERTRAND by LECORNU 3.
Consider any regular surface with unit normal N such that N = t at some interior point P.
On this surface, let d be any region containing P and bounded by a circuit c, reducible on
the surface, and let s be the area of d. Then by KELVIN's transformation (28.1)
s-1 ~da:· t = s-1 f N· curltda. {30.14) e •
Reducing c to the point P, so that s --+0, N --+t, by using Eq. (30.2) we get
Ajp = lims- ~da:·t S------)>0 c {30.15)
lf there exists a congruence of surfaces with t as normal, we may take d to be a region on
one of these, obtaining from Eq. (30.15) A = 0. Thus A may be regarded as a measure of
the departure of c from the property of having anormal congruence of surfaces.
b) Special classes of fields.
31. Solenoidal fields 1: Integral properties. An integrable vector field c is
called solenoidal provided its flux out of every reducible closed surface 6 in its
1 MASOTTI [1927, 2]. Cf. BJ0RGUM [1951, 4, § 2.3], CoBURN [1952, 1, § 2]. GILBERT [1890, 2] (cf. also KLEITZ [1873, 3, §§ 2, 8, 40]) had given formulae for the gradient
and divergence of a tensor in terms of the radii of curvature of the members of a triply orthogonal system.
2 [1900, 1]. The names "torsion of the curve system" and "torsion of neighboring
vector lines" have also been proposed; cf. BJ0RGUM [1951, 4, § 2.4]. 3 [1919, 1].
52*
820 J. L. ERICKSEN: Tensor Fields. Sect. 31.
region of definition is zero. From Eq. (26.1) 2 follows Kelvin's 1 characterization:
A continuously differentiable field c is solenoidal if and only if its divergence vanishes
divc = o. (31.1)
There is also the characterization oj Helmholtz 2 : A field c, continuously
differentiable 3 in a closed region, is solenoidal if and only if the strength of every
vector tube is the same at all cross sections. The necessity of this condition follows
immediately from Eq. (29.6). If the strength of every vector tube is the same at
all cross sections, by Eq. (29.5) it follows easily that Eq. (31.1) holds, hence
that c is solenoidal.
In Sects. 31 and 32 we assume c is a continuously differentiable solenoidal
field. Then by Eq. (13.7) we calculate
div (cp(K+Il) = c · gradp(K+I) = {p(K) c};
hence by Eq. (26.1) 2 we obtain 4 for the moments (25.5) of c
C(KJ= f {p(Klc}dv =~da· cp(K+I) = ~dacnp(K+IJ.
.. d d
(31.2)
(31. 3)
Hence the moments of a solenot.dal field are determined by the boundary values of
its normal projecrion cn. I f cn = 0 on ~, then all moments of c over the volume u
bounded by ~ vant"sh.
More generallys,
div (c \V)= c · grad \V, (31.4)
whence follows
J c · grad \V d v = ~da· c \V = ~ da cn \V. (31. 5) .. d d
Hence, if c and \V are continuously differentiable and c is solenoidal, then the total
c · grad \V in a region u is determined by the boundary values of cn \V. I f cn = 0
on ~. then the total c · grad \V in u is zero 6•
1 [1851, 1, § 74]. 2 [1858, 1, § 2]. From the fact that the vorticity field is solenoidal, HELMHOLTZ [1858,
1, § 2] concluded that "Es folgt hieraus auch, daß ein Wirbelfaden nirgends innerhalb der
Flüssigkeit aufhören dürfe, sondern entweder ringförmig innerhalb der Flüssigkeit in sich
zurücklaufen, oder bis an die Grenzen der Flüssigkeit reichen müsse. Denn wenn ein Wirbelfaden innerhalb der Flüssigkeit irgendwo endete, würde sich eine geschlossene Fläche construiren lassen, für welche das Integral f q cos {} dw nicht den Werth Kuli hätte". TRUESDELL
[1954. 2, § 10] has noted that this statement is misleading and incomplete. It is not trivial
to reformulate it as a theorem susceptible of rigorous proof. One point glossed over in HELMHOLTz's statement is the fact that a vector tube of a field continuously differentiable in a
compact region need not approach the boundary, be closed, or end in the interior. For
example, in the plane, it may be bounded by curves having as a common limit set a limit
cycle. One can show that this particular situation cannot occur if the field be solenoidal.
Cf. Sect. 32, footnote 1. 3 TRUESDELL [1954, 2, § 10] remarks that it suffices to assume c continuous and piecewise continuously differentiable, but that the result need not hold for piecewise continuous
fields.
4 TRUESDELL [1949, 3], [1954, 2, § 10]. The case K=O was discussed by FöPPL
[1897. 1, § 4]. 5 TRUESDELL [1951, 2, § 9] has remarked that the more general result is implied by an
example due to KELVIN [1849, 1, § 7]. 6 Using KELvrN's transformation (28.1}, one can show that this result is equivalent to
the lemma 1 of WEYL [1940, 2]. A special case is given by BERKER [1949, 5, Chap. III].
We follow TRUESDELL [1954, 2, § 10].
Sect. 31. Solenoidal fields I: Integral properties. 821
Setting b =C in Eq. (26.2) and employing Eq. (31.1), we obtain
~ [da · c c p(K) - -i c2 da p(K)] l ~ = J [c {cp(K-l)}-! c2 grad p(K) + curl c X cp(K)] dv.
"
(31.6)
The case when K = 0 is PoiNCARE's 1 identity
~ [da· c c-! c2 da J = J curl c X c dv. (31.7) 6 "
Recalling the convention of Sect. 24, we formulate conditions sufficient for
the vanishing of the integral over il and conclude that if a field c which is continuously differentiable and solenoidal in a region v satisfies the boundary conditions
then
cn=O or ccn=o(p- 2)
c2 = const or c2 = o (p- 2)
J curlcxcdv = 0.
"
on il, }
on il,
Sufficient for Eq. (31.8) to hold are the simpler conditions
c = 0 or c = o (p-1) on il.
For K = 1, Eq. (31.6) becomes
~ [da · c c p - -i c2 da p J = f [ c c - ! c2 1 + curl c x c p J d v.
d "
(31.8)
(31.9)
(31.10)
(31.11)
Contrading Eq. (31.11) yields an identity of LAMB and J. J. THOMSON 2
! f c2 d v = ~ da · (! c2 p - c c · p) + f p · cur 1 c X c d v . (31. 12) V 6 V
from which it follows that i/ a field c which is continuously differentiable and
solenoidal in a region v satisfies the following conditions:
1. c=O or c2 =o(p-3), p-ccn=o(p-2) on il, (31.13)
2. Throughout v, curl c x c = 0;
then c = 0 throughout v.
Sufficient for (31.13) to hold are the simpler conditions
c = 0 or c = o (p- ;) on il.
The alterna ting part of Eq. (31.11) yields 3
~ [da-cpxc +ic2 daxp] =fpx (curlcxc)dv,
6 "
(31.14)
(31.15)
(31.16)
from which we conclude that if a field c which is continuously differentiable and
solenoidal in a region v satisfies the boundary conditions
cn = 0 and c = const or p X c cn = ö (p- 2), (31.17)
then
fpx (curlcxc) dv = 0.
Sufficient for Eq. (31.17) to hold are the simpler conditions (31.15).
1 [1893, 1, § 5]. See also TRUESDELL [1954, 2, §§ 10, 64].
(31.18)
2 [1879, 1, § 136], [1932, 6, § 153]; [1883, 1, § 6]. Cf. also TRUESDELL [1954, 2, § 10]
for the results given in italics here. 3 PoiNCARE [1893. 1, § 115]. Cf. TRUESDELL [1954, 1, § 10].
822 J.L. ERICKSEN: Tensor Fields. Sect. 32.
In Eq. (26.2), put K =1, b =p; then
j[pc+cp-p-c1+(3c+curlcxp)p]dv }
=~[da· (pc + cp)p- da (p · c)p] .
•
(31.19)
Contracting this relation yields
2fp·cdv =~da·cp , (31.20) " . while the alternating part is equivalent tol
3 J pxcdv = J px (pxcurlc) dv -~ " .
- ~ [(daxc) xp] xp .
•
(31.21)
J.J. THOMSON 9 derived two similar integral formulae which, while not restricted to solenoidal fields, are of interest mainly in the solenoidal case. First,
in Eq. (26.2) put K =0, b p, obtaining
J(curlcxp+3c+pdivc)dv )
" = T [da· (cp +pc)- da (p. c)],
= ~ [(daxc) xp- da· cp] .
•
Since div (cp) =p div c +c, application of Eq. (26.1h yields
2 J cdv = J pxcurlcdv +~ (daxc) xp. " . . Second, if we integrate the identity
curl (p2c) = p2curlc + 2pxc
over v and then apply Eq. (26.1)s, we obtain
2fpxcdv =-J p2 curlcdv + ~daxp c. ". ". .
(31.22)
(31.23)
(31.24)
(31.25)
For the various integral theorems we have given here, the region need not be
simply-connected.
32. Solenoidal fields II: Differential properties. lf c is any continuously differentiahte field, there exists an infinite number of scalar functions m such that mc
is solenoidal. To see this, one need only note that
div (mc) = 0, (32.1)
being a first order differential equation for m, with continuous coefficients, admits
an infinite nurober of solutions. Since c and mc have the same vector lines, the
vector lines of any continuously ditferentiable field c are also the vector lines of
an infinite number of solenoidalfields 3• This result may mislead in seeming to
imply that vector lines of solenoidal fields have no special properties. Distinctions can arise, for example, because of the fact that solutions m of Eq. (32.1)
1 MUNK [1941, 1, p. 95].
2 [1883, 1, §§ 4- 5]. The second of these is usually presented in a non-invariant form;
cf. e.g. LAMB [1932, 6, § 195], where it is asserted that an error in J. J. THOMSON's formula
was corrected by WELSH. 3 APPELL [1897, 2, § 5].
Sect. 32. Solenoidal fields II: Differential properties. 823
can possess singularities at points where c =0, even if c be analytic. For example,
in the plane, one can show that an isolated point where c = 0 is never a spiral
point for the vector lines of a solenoidal field c, whereas it can be if c is not
solenoidal 1.
Any continuously differentiable solenoidal vector field c may be represented
locally in EULER's form 2
c =gradFxgradG
where F and G are scalar functions which satisfy
C·gradF=O, C·gradG=O.
(32.2)
(32.3)
The surfaces F = const and G = const are thus vector sheets of c. One of the
functions, say F, may be taken as an arbitrary non-constant solution of Eq. (32.3);
the other may be taken as the function given by
G Jexc·dx =- e·gradF' (32.4)
where e is an arb itrary continuous vector field such that e · grad F =!= 0, and where
the path of integration lies on one of the surfaces F = const.
Combining the results of the two foregoing paragraphs, we see that any continuously differentiable field may be represented locally in the form
c =H gradFxgradG, (32.5)
the surfaces F = const and G = const being vector sheets of c. In Eq. (32. 5),
F and G may be chosen as arbitrary functionally independent solutions of
Eq. (32.3), and the function H is then determined uniquely. If c is solenoidal,
Eq. (32.5) yields
d. o(H,F, G)
0 = lVC = "( 1 2 3); u X, X, X (32.6)
therefore H =H(F, G). From Eqs. (32.5) and (32.6) follows Euler's theorem on
solenoidal jields 3: I f F and G are any independent functions such that the
surfaces F = const and G = const are vector sheets of a continuously ditferentiable
solenoidal field c, then c can be represented locally in the form
c =H(F, G)gradFxgradG. (32.7)
From Eq. (32.2) and the identity
curl (F grad G + grad H) = grad F X grad G, (32.8)
it follows that any continuously differentiahte solenoidal field c may be represented
locally as the curl of a vector v,
c = curlv, (32.9)
the field v being indeterminate to within an additive gradient. Conversely, if Eq. (32.9)
holds, div c =div curl v =04•
1 In connection with HELMHOLTz's characterization (Sect. 31, footnote 2), many expositians assert that the vector lines of a solenoidal field cannot end in the interior. KELLOGG
[ 1929, 1, Chap. II, § 6] has remarked that this is false. 2 [1770, 2, §§ 26, 49]. [1806, 1, § 142]. Derivations are given in works on vector analysis, e.g. BRAND [1947, 2, § 104, Chap. 3]. 3 [1757. 1, §§ 47-49].
' More generally, any analytic field v yields theinfinite sequence of solenoidal fields curlK v
(K = 1, 2, .... ). For K~ 3 we have curlK v =- 172 currK-2v, whence follow volume integrals
yielding currK-2v in terms of cur!Kv. Thesefacts were noted by RowLAND [1880, 1] and
FABRI [1892, 3] who attempted a kinematical interpretation of cur!Kv. For the case
K = 2, cf. BOGGIO-LERA [1887, 3].
824 J.L. ERICKSEN: Tensor Fields.
At points where c =!= o, Eq. (30.4) shows that div c = 0 if and only if
dlogc = _ D
dt '
which may be integrated along the vector lines of c to give
c = c0 exp [- J D dt],
Sect. 33.
{32.10)
{32.11)
this characterization being due to BJ0RGUM1• It follows immediately that a non-zero solenoidal
field is determined by its vector lines and by its magnitude at one point on each. By successive differentiation of (30.4), we conclude that if c and D are analytic functions of arc length
on a given vector line, c vanishes at one point if and only if it vanishes all along the vector
line through the point.
33. Lamellar and complex-lamellar fields. Following BJ0RGUM2, we call any
vector field proportional to a . . . field a complex - ... field, wherein any name
may be inserted for . . . From results given in Sect. 32, an arbitrary continuously
differentiable field may thus be called a complex-solenoidal field.
A field c is lamellar 3 in a region provided
~ d~ · c = ~ dJ!' c" = 0 (33-1) e e
for any reducible circuit c in the region 4• It follows that a field c, continuous
in a region v, is lamellar if and only if there exists a scalar P, called the potential
of c, such that
c =- gradP. (33.2)
The potential is single-valued if v is simply connected, not so in generat if v is multiply-connected. A lamellar field is thus everywhere normal to the equipotential
surfaces P = const. A complex-lamellar field, being by its definition representable
in the form
c = QgradP, (33-3)
where Q is a scalar, also is normal to a family of surfaces. From KELVIN's transformation (28.1) it follows that a continuously ditlerentiable field c is lamellar if
and only if
curl c = 0, or 8crkfox"'1 = 0. (33.4)
Similarly, a continuously ditlerentiable field c is complex-lamellar if and only if
c. curl c = 0, or c[k oc,j ox"l = 0. (33-5)
These results were established by KELVIN 5, a result equivalent to Eq. (33-5)
having been obtained much earlier by EULER6• From (33-5h it follows that c
is complex-lamellar if there exists a co-ordinate system in which
(33 .6)
A field c is plane if Eq. (33.6) holds in some reetangular Cartesian co-ordinate
system, rotationally symmetric if Eq. (33.6) holds in cylindrical co-ordinates, x3
l [1951, 4, § 3.1]. 2 [1951, 4]. The reader will not confuse this usage of "complex" with "complexvalued". 3 The names lamellar and complex-lamellar derive from KELVIN [1850, 1], [1851, 1,
§§ 68-69, 75]. 4 WEYL [1940, 2] generalizes the definitions of lamellar and solenoidal field given here. 5 [1851, 1, § 75]. VELTMANN also discussed these fields [1870, 1, pp. 453-456]. For
a modern proof that Eq. (33-5) is sufficient as well as necessary, see e.g. BRAND [1947, 2,
§ 105]. Therepresentation c = Q grad P + grad F(Q, P) is discussed by CASTOLDI [1955, 5]. 6 [1770. 1, § 1].
Sect. 33. Lamellar and complex-lamellar fields. 825
being the angle variable. It is essential that the covariant components of c be
used in Eqs. {33.4) 2 and (33.5h.
For any vector field c, any curve ;x; =J:(u) satisfying the equation
(33.7)
is orthogonal to the vector lines of c which it intersects. A theorem of CARATHEODORY 1 asserts that in a closed and bounded region where c is non-zero and
satisfies a l ipschitz condition, c is complex-lamellar if and only if in every neighborhood of an arbitrary interior point ;x; there exist a point ;x;* such that no curve
satisfying Eq. (33.7) foins ;x; to x*. If c is complex-lamellar, it follows from
Eqs. (33.3) and (33.7) that ;x; and ;x;* may be joined by a solution of Eq. (33.7)
if and only if they lie on the same surface P = const, from which the necessity
of this condition follows. We omit the more complicated proof of sufficiency.
With the exception of the definitions of plane and rotationally symmetric
fields, all these results are easily extended to n-dimensional spaces which need
not possess a metric tensor, c being taken to be a covariant field 2• Eqs. (3 3.1) to
(33.3), (33.4) 2 , (33.5) 2 and (33.7) require no alteration, while Eq. (33.6) is tobe
replaced by cl = f(xl, x2), c2 = g(xl, x2), Ca= ... = c,. = 0. (33.8)
It is clear from Eq. (33·3) that the magnitude of a complex-lamellar field is
in no way restricted by its vector lines. Indeed, it is immediate that a non-vanishing field is complex-lamellar if and only if its unit tangent is complex-lamellar.
However, the condition that a field be lamellar is essentially different in that it
connects the magnitude of the field with the geometric properties of its vector
lines.
In the three-dimensional case, it is easytorender the above remarks explicit.
In the first place, either directly from Eq. {30.1h or by substitution in Eq. (33.5),
or from Eq. (30.15), we conclude that a non-zero continuously ditferentiable field
is complex-lamellar if and only if its abnormality A is zero. For a lamellar field,
we have from Eq. (30.12) the much stronger necessary and sufficient conditions
of BJ0RGUM 3 : (1) A =0 when c=j=O, (2) c is constant along the vector lines of the
binormal field b, and (3)
c = c0 exp J xdn, (33.9)
the integration being performed on the vector lines of the principal normal n,
which lie on the equipotential surfaces, as do those of b. It follows that, in a
sufficiently small region, a lamellar field is determined by its vector lines and
by its magnitude at a single point on each equipotential surface.
When a field is complex-lamellar, its unit tangent t is the unit normal of the
surfaces P = const. Hence the mean curvature K of those surfaces is given by 4
K- d' t D d log c 1 d' ( ) =- lV =- =dt--c lVC. 33.10
Therefore, if we are given a set of vector lines having a normal congruence,
we may obtain all non-vanishing complex-lamellar fields having these vector
1 [1909, 5]. 2 In situations to which CARATHEODORY's sufficiency condition has been applied, it is
artificial and unnecessary to introduce a metric tensor. Cf. CARATHEODORY [1909, 5], ERICKSEN [1956, 1].
3 [1951. 4, § 3.4]. 4 Cf. e.g. BRAND [1947, 2, § 131]. The result is due to CHALLIS [1842, 1], whose derivation was criticized and corrected by TARDY [1850, 1]. The geometry of the surfaces P=
const is studied by CALDONAZZO [1924, 1], [1925, 1] and PASTORI [1927, 2].
826 J. L. ERICKSEN: Tensor Fields. Sect. 34.
lines by prescribing the scalar field div c everywhere and the value of c at one
point on each vector line. Also, for a non-vanishing complex-lamellar field, any
two of the following three conditions imply the third1 :
1. The normal surfaces are minimal surfaces.
2. The field is solenoidal.
3. The magnitude of the field is constant on each vector line.
34. Screw fields. From Eq. (33-5), a field c is complex-lamellar if and only
if it be normal to its curl. The opposite extreme is fumished by a screw field 2 ,
defined as being parallel to its curl:
equivalently,
cxcurlc = 0, curlc =j= o;
k_o c[k,m] c - , c[k,m] =j= 0.
(34.1)
(34.2)
In either of these equations covariant derivatives may be replaced by partial
derivatives. It follows from Eqs. (30.12) and (34.1) 1 that c is a screw field if and
only if
curlc=Ac=j=O, (34-3)
so the abnormality of a screw field c is the factor of proportionality between the curl
and the field. Since c and curl c have the same vector lines when Eq. (34-3)
holds, and since the abnormality depends only on the vector lines, we may replace
c by curl c in Eq. (30.2), obtaining3
A _2 l curl c · curl curl c =c C·cur C= . curl c · curl c
If m is any scalar function suchthat mc is solenoidal (cf. Sect. 32),
0 = divmc = div(~ curlc) = grad ~ · curlc =Ac· grad ~,
(34.4)
(34.5)
so the surfaces mjA =const are vector sheets of c and of curl c. Conversely, if,
for some scalar function f (x), the surfaces f = const are vector sheets of a screw
field c, we may set m=fA, where A is the abnormality; reading Eq. (34.5)
backwards, we obtain div mc = 0. These results constitute the following theorem 4 :
For a twice continuously differentiahte screw field c, a necessary and sufficient
1 A special case is due to CALDONAZZO [1924, 2, § 6]. We generalize an argument of
PRIM [1948, 1, § 3], [1952, 1, Chap. V]. Cf. also CASTOLDI [1947, 5], BYUSHGENIS [1948, 2,
§ 2.1]. 2 Much of the literature follows CrsOTTI [1923, 3] in calling these fields "Beltrami
fields", after the researches of BELTRAMI [1889, 1], but in fact nearly all of BELTRAMI's
results were included in the prior and more extensive work of GROMEKA [1881, 2]. BELTRAM! hirnself called them "helicoidal". We revert to the more descriptive term screw field
introduced by CRAIG [1880, 2, p. 225]. [1880, 3, p. 276], [1881, 5, pp. 5-6], the first
person to remark them. Earlier STOKES [1842, 2, p. 3] had concluded that Eq. (34.1h
implies C=O, but later he realized his error (footnote 1, p. 3, 1880 reprint of [1842, 2]). 3 LECORNU [1919, 1]. If we set B ==I curl clfc, then c is a screw field if and only if
B =A, the proof being immediate from Eq. (34.3); necessity was proved by APPELL [1921,
1, § 763], sufficiency by CARSTOIU [1946, 4]. A differential system for a screw field is
given as Eq. (35.4) below. Another follows by taking the curl of Eq. (34.3) and then eliminating A by Eq. (34.4) (BALLABH [1948, 3, § 4]), but a simpler one may be obtained by putting Eq. (34.4) into Eq. (34.3) directly. Other differential equations are derived by BJ0RGUM
[1951, 4, Sect. 5], [1954, 7]. 4 TRUESDELL [1954, 2, § 12]. This generalizes theorems of NEMENYI and PRIM [1949,
6, Th. 1], BELTRAMI [1889, 1] (see also MORERA [1889, 2]), and GROMEKA [1881, 2, GI. 2,
§ 9], the references being arranged in order of decreasing generality.
Sect. 34. Screw fields. 827
condition that the surfaces
A (curl eh = _!<:url c) 2 = (curl c)3 = const mc1 mc2 mc3
(34.6) m
be vector sheets is that mc be solenoidal. It is true a fortiori that mfA is constant
an each vector line of a screw field c if mc is solenoidall. Setting m = 1, we obtain
the corollaries: The surfaces of constant abnormality of a screw field c are vector
sheets if and only if div c = 0; the abnormality of a solenoidal screw field is constant
along each vector line. Thus from the vector lines alone, without knowledge of
the magnitude of the field, we can determine whether or not a screw field is
solenoidal. More generally, by putting m = 1 in Eq. (34.5) we may derive
c-1 div c = - ~o~! AI • (34.7)
A relation between a screw field and its vector lines may be read off from
Eq. (30.12). This relation results from that given in Sect. 33 for lamellar fields
if we replace "A = 0" by "A =!=O" 2• Indeed, from Eqs. (30.4) and (34-3) follows
c A = c0 A 0 exp (- J D dt). (34.8)
BJ0RGUM 3 has shown that for a given screw field c, it is always possible to choose a
co-ordinate systemsuchthat the x3-lines are the vector lines, g13 = 0, g23 = x1 g33 , and cg33 = 1.
Conversely, in a Co-ordinate system such that g13 = 0 and g23 = x1 g33 , if a vector field c
be tangent to the x3-lines and of magnitude c so adjusted that cg33 = 1, then c is a screw
field.
From Eqs. (34-3) and (34.4) we conclude that
o < curl c · curl c = c · curl curl c, (34.9)
so a screw field and the curl of its curl always intersect at an acute angle. If
this angle is zero, curl c is also a screw field. I t follows from taking the curl
of Eq. (34.3) that the curl of a screw field c is again a screw field if and only if the
abnormality of c is uniform. Then c is solenoidal, and successive curls of c are
screw fields having the same abnormality 4•
In the case when A is uniform, that c is solenoidal follows alternatively as a corollary
of the first corollary following Eq. (34.6). Hence taking the curl of Eq. (34.3) yields
172 c +Ac= o. (34.10)
This equation was derived by GROMEKA 5, who based upon it a theory of determining a screw
field of constant abnormality from appropriate boundary conditions. The same problern
has been taken up by BJORGUM and GoDAL 6 ; besides constructing many interesting examples,
they have shown that such a field c can be represented in the form
c = A gradH x e + eA2 H + e · gradgradH, (34.11)
where e is a fixed unit vector and 172 H + A 2 H = o.
For the many more known properties of screw fields, the reader is referred to the treatise
of BJ0RGuM7,
1 TRUESDELL [1954, 6, § 527] attributes to VAN TuYL the remark that this theorem is
an immediate consequence of the fact that any two solenoidal vector fields with common
vector lines are proportional along them. 2 BJ0RGUM [1951, 4, § 3-3]. 3 [1957. 4, § 5].
'NEMENYI and PRIM [1949. 6, Th. 3]. 5 [1881, 2, GI. 2, § 9]. Cf. also STEKLOFF [1908, 2, §§ 39-52], TRKAL [1919, 3],
BALLABH [1940, 5, §§5-7]. 6 [1953, 3], [1958, 2]. Cf. also BJ0RGUM [1951, 4, § 6].
7 [1951, 7]. Cf. also TRUESDELL [1954, 2, §§ 12, 52].
828 J. L. ERICKSEN: Tensor Fields. Sects. 35. 36.
In n-dimensional metric spaces, screw fields may be defined by Eq. (34.2). There is no
immediate extension to spaces which are not metric since associated components do not
exist. One natural generalization of Eq. (34.2) is obtained by requiring a covariant vector
b and a contravariant vector c to satisfy
C!b[kfaxmJ c" = o, ob[kfoxmJ =!= o. (34.12)
Suchpairs of vectors occur in studies on relativity1 •
c) Potentials.
35. MoNGE's potentials. If c be a twice continuously differentiable field, the
field curl c, being solenoidal, has a representation of the form (32.2), namely,
curlc = gradFxgradG, (3 5.1)
from which follows curl(c-FgradG)=O. Thus c-FgradG is lamellar, so
there exists a scalar H such that
c =gradH +FgradG. (35.2)
In general, this representation is valid only locally 2• The three scalarsF, G, and H,
called Monge potentials 3 of c, are not uniquely determined, but in most
applications there is no need to specify one particular set rather than another.
From Eqs. (35.1) and (35.2) follows
I o (H, F, G) ( 5 )
c·curc= 8 (xi,xz,x2)' 3-3
whence, by Eq. (33-5), c is complex-lamellar .if and only if its three Monge potentials are functionally dependent. Directly from (35.1) we see that cislamellar
if and only if the two potentials F and G are functionally dependent.
MORERA 4 obtained differential equations to be satisfied by the Monge potentials of a
screw field. In geometrical terms, these equations assert that for c to be a screw field the
surfaces F=const and G=const, which by Eq. (35.1) are always vector sheets of curlc,
must simultaneously be vector sheets of c. Formally,
c · grad F = (grad H + F grad G) · grad F = 0, }
c · grad G = (gradH + Fgrad G) · gradG = o. (3 5 .4)
Conversely, if Eq. (35.4) hold and if F and G be functionally independent, then both c and
curl c are perpendicular to grad F and to grad G; hence they are parallel, so c is a screw field.
36. STOKEs' potentials. In a finite region v, an arbitrary vector field c has a
representation of the form 5
c = - grad S + curl v (36.1)
and hence is the sum of a lamellar and a solenoidal field. Functions S and v
satisfying Eq. (36.1) are called, respectively, a scalar potential and a vector potential of c; together, they are called the Stokes' potentials. An infinite
nurober of potentials correspond to a given field c. Let c be piecewise differentiable
in a finite region v, bounded by 6; within v, Iet c be continuous except upon a
surface 6 ', on each side of which it has finite Iimits; then a pair of potentials for c
1 E.g. VAN DANTZIG [1934,1, p. 646]. 2 Cf. HADAMARD [1903, 2, p. 80]. 3 This result was implied, but not stated explicitly, by MoNGE [1787, 1, §§XVI-XVIII,
XX] and PFAFF [1818, 1, § 4]. The above derivation is due to HANKEL [1861, 1, § 11]. 4 [1889, 2]. See also BJ0RGUM [1951, 4, § 5.1].
STOKES [1851, 2, Part I, Sect. 1, §§ 3 -8]. Proofs are given in works on vector
analysis, e.g. PHILLIPS [1933. 1, § 83]. Four stronger decomposition theorems are given by
·WEYL [1940, 2]. Cf. also BLUMENTHAL [1905, 1].
Sect. 37. Definitionsand conditions of reality. 829
is given by
S = ~1-J 4n divc_dv d + ~1-J 4n da· d [c] __ - __ 4n 1 ___ ,f.. 'j" da:.!!_ d ' l .. 6' 6
V = _1_! curl_!!_ dv + 1J_da x ~] __ 1_ ,f.. !-_a x~_
4n d 4n d 4n 'j" d ". .. "
(36.2)
where d is the distance from the point of integration to the point where S and v
are being calculated, and where the bold-face bracket denotes the fump of c
across 11',
(36-3)
c+ and c- being the limiting values of c on the two sides of 11' and the sense of
da being fixed appropriately in terms of the choice of signs + and - for the two
sides of 6'. In the case of a region extending to infinity in all directions, if c =
ö (p-2) the formulae (36.2) still hold, providing the integrals over 6 be omitted.
For suitably selected potentials, e.g., those given by Eqs. (36.2), the representation (36.1) is valid globally1 and v is solenoidal. We always assume the potentials are so selected.
Since curlv is solenoidal, we may replace it in Eq. (36.1) by an expression of the form
(32.2), so obtaining a local representation
c = - grad S + gradF X grad G.
VI. Tensors of order two.
a) Proper numbers and vectors.
(36.4)
37. Definitionsand conditions of reality. A proper number a of a second order
tensor a is a root of the equation
(37.1)
Since in n dimensions Eq. (37.1) is a polynomial equation of degree n in a, there
are always n proper numbers, which need not be real or distinct. The left and
right proper vectors of a corresponding to the proper number a are, respectively, the
non-zero vectors m and q, in general complex, such that
The directions of the vectors are the principal directions of a.
A sulficient condition that alt proper numbers of a be real is that 2
{ b is symmetric and positive definite 3 ,}
a = b-1 • c, where
c is symmetric.
Inded, if a be so expressible, its proper numbers are given by
1 HADAMARD [1903, 2, p. 80].
(37.2)
(37-3)
2 An alternative statement is: For some choice of the metric tensor (viz. gkm=bkm),
a is symmetric. In general, such a metric is not Euclidean. 3 Throughout this work we write "the tensor b is positive definite" in place of "the
components bkm are cocfficients of a positive definite quadratic form".
8)0 J. L. ERICKSEN: Tensor Fields. Sect. 37.
the assertion then follows because each root a of an equation ofthistype is real 1 •
Furtherroore, if the roots of Eq. (37.4) are a1 , ... , a,., there exist reallinearly
independent vectors q1 , •.• , q,. such that
(37.5)
whence follows -1
a"',.q~ = b"'' c,,.q~ = abq';, (3 7.6)
so qb is a right proper vector of a corresponding to the proper nurober ab. Defining the vector mb by ~,.=b,.. qb, froro Eq. (37.5) we get
-1
a", ~" = b"• c., bku qb = a4 b,. qg =ab mb,, (3 7.7)
so mb is a left proper vector of a corresponding to the proper nurober ab.
Were the above condition also necessary, then all proper nurobers' being real
would ensure the existence of n linearly independent right proper vectors. A
counter-exarople is shown by the roatrix
0 0 0
1 0 0 '
0 0 1
(3 7.8)
which has the real proper nurobers 0, 0, 1, but only two linearly independent
right proper vectors 2, e.g. (0, 1, 0) and (0, 0, 1). A necessary condition follows:
If the proper numbers a1 , ... , a,. of a are all real, and if there exist corresponding
right proper vectors q1 , ... , q,. forming a linearly independent set 3, then Eq. (37.3)
holds. To show this, we first note that without loss of generality the qb roay be
assuroed real. Since they are linearly independent, there exists a unique real
reciprocal set me such that
If we set
.. me,.q~ = t5eb• L: mb,.q~ = t5k.
b=l
.. b,.,- L: ~k ~, b=I
then b is a syroroetric positive definite tensor. Moreover,
-1
(37.9)
(37.10)
~" = b,.,q{,, q~ = b"'~,. (37.11)
Since the q4 are right proper vectors of a, froro Eqs. (37.2) 2 , (37.9) 2 , and (37.11) 2
it follows that
a~ = aP, t5k = at
b=1
~ mbk ql, =
b=1
f. ab mb,. q€ l .. -1 -1
= L: mbk ab bf>' mb, = bf>' c,,.,
b=1
(37.12)
1 See e.g. CoURANT and HILBERT [1931, 2, p. 32). The proofthat the roots of Eq. (31.4)3
are real is implicit in the work of CAUCHY [1828, 1], [1829, 1], who treated explicitly
the case b,.,= d,.,, thereby showing that the proper numbers of a symmetric tensor are always
real. HERMITE [1855, 1] extended CAUCHY's result to complex matrices, showing that the
proper numbers of a Hermitian matrix are always real. · 2 TRUESDELL [1954, 2, § 22) attributes this example to WHAPLES and gives an example
of a matrix having all its proper numbers real and all its proper vector proportional to a single
vector. Those familiar with the theory of elementary divisorswill see easily that the number
of linearly independent proper vectors equals the number of elementary divisors. 3 Sufficient for this is that the proper numbers all be distinct.
Sect. 37.
where we have set
Definitions and conditions of reality.
.. c,,. =Lab mb, mbk·
b=l
Since c,,.=c11 ,, Eq. (37.12) is the desired result.
831
(37.13)
The tensors b and c are not uniquely determined by a. Given any admissible
set of proper vectors qb, we can obtain an infinite number of different admissible
sets by multiplying each vector qb by an arbitrary non-zero scalar factor. As is
easily seen from Eqs. (37.8), (37.10), and (37.13), different sets generally will
determine different tensors b and c. It follows from Eqs. (37-9) and (37.11)
that b as defined by Eqs. (37.10) satisfies (37-5), whatever be the choice of the qb.
From Eqs. (37.11) and (37-7) we conclude that the reciprocal set ofvectors me,
defined by the condition (37-9), are in fact left proper vectors corresponding to
the proper numbers ae. The intermediate formula (37.12)a is particularly important in that it expresses the tensor a uniquely in terms of its proper numbers,
supposed real, and a set of its independent right proper vectors, supposed n in number,
and left proper vectors so chosen as to form a reciprocal set. The invariance of this
representation under possible different choices of the qe is immediate from
Eq. (37-9).
It can occur that the vectors qe are mutually orthogonal. If they are normalized so as tobe unit vectors, Eq. (37.10) implies that bkm=gkm• whence we
conclude that a =a'. That is, if all proper numbers of a are real, and if a has
n linearly independent mutually orthogonal proper vectors, then a is symmetric;
this theorem is due in principle to KELVIN and TAIT1• The example (37.8),
interpreted as being referred to reetangular Cartesian co-ordinates, shows that
the reality of all proper numbers and orthogonality of a maximal linearly independent set of right proper vectors is insufficient to imply a =a'. There must
exist n orthogonal proper vectors, which is not the case for the example.
Aside from the result of KELVIN and TAIT, there has been no need to introduce
the metric tensor if a is taken as a mixed tensor, b and c as covariant tensors.
For any field a(:r) such that Eq. (37-3) holds, it is possible to select a coordinate system such that at a given point P, the right proper vectors have the
components <5~. Indeed, let the qb (:r) constitute any linearly independent set of
right proper vectors, in any Co-ordinate system; then we need only select :r* =
:r* (:r) such that at P we have
8x*"
qk (:r*) = --qm (:r) = 15" b 8xm b b·
From Eq. (37-9) 2 it follows that me,.(:r*) =t:5ek at P. Hence
From Eqs. (37.12) 2, (37.14), and (37.15), at P we derive
k * _ 8 x* k 8 :>:5 u ) _ ~ 8 x* " 8 :>:5
a ,(:r ) - 8x" ax*' a .(:r - LJ 8x" 8x*' mbs qb u ab, s l
b=1
" =Lab 6~t:5br· b=1
(37.14)
(37.15)
(37.16)
This equation asserts that in the co-ordinate system :r*, the matrixlla",ll assumes
diagonal form with entries ab. Conversely, if there is a real, non-singular
1 [1867, 1, § 183].
832 J. L. ERICKSEN: Tensor Fields. Sect. 38.
co-ordinate transformation which diagonalizes II a" mll at a point, then a = b-1 · c,
" where band c are the tensors having at P the components !5,.m and 2; ab !5b,. !5bm•
b=1
respectively, in the system in which II a" mll is diagonal. Such a co-ordinate system,
hereafter called a principal Co-ordinate system, need not be orthogonal, which
means that lla"mll and lla,.mll need not be diagonalized by the transformation.
Combining these results, we have the following characterization: A necessary
and sulficient condition that a = b-1 · c, where b and c are symmetric and b is
positive definite, is that at an arbitrarily selected point P, there exist a real nonsingular co-ordinate transformation which diagonalizes the matrix II a" mll· In the
langnage of matrices, this result may be restated as follows: A necessary and
sutficient condition that a = b-1 • c, where b and c are symmetric and b is positive
definite, is that a be ·similar to a diagonal matrix modulo the real matrices, i.e.
that a =I· d ·l-1, where I is real and non-singular, d real and diagonal. A
restaterneut of the theorems of CAUCHY and KELVIN and TAIT is: I may be taken
as orthogonal, i.e.f' = 1-1, if and only if a = a'.
In a principal co-ordinate system we have
m 0 ak m + { m} { k m) { d} { } ak ,P = oxP kp ak - am unsumme ; 37.17
when k = m, the second term vanishes. In general, this formula holds only at
the point where the principal co-ordinate system is defined, and only for the
diagonal mixed components do covariant derivatives reduce to ordinary partial
derivatives. Even if lla"mll and llakmll also be diagonal matrices, we have a"",p=l=
oa""foxP and akk,P =!= oakkfoxP, in general. When the principal CO-ordinate system
is rea~ and orthogonal, Eq. {37.17) may be read in terms of physical components:
a- a)' (37.18} OSp OS_.
where we have used Eq. (25.6).
For the case of most interest here, namely n =3, we may make use of the
obvious but useful fact that the number of real proper vectors is fixed by the
sign of the discriminant of the left member of Eq. (37.1}. We may use also the
fact that the proper numbers of a are all real if and only if the same be true of
a+m 1, where m is any scalar, as follows immediately from Eq. (37.1}. If a
admits the decomposition {37-3), we have a+m1 = b-1 • (c+mb) as a corresponding decomposition for a +m1.
FROBENIUS 1 showed that a has a proper number which is real, positive, simple, and
greater in absolute value than any other proper number provided, for some choice of Coordinates, the numbers akm be all positive.
BANG 2 noted that in three dimensions the proper numbers of a are all real if, in some
co-ordinate system, a 12 a 23 a31 = a\ a32 a13 and a12 a 23/ a 13 , a23 a31f a21 , and a\ a12f a32 are of
the same sign.
38. Principal and related invariants. The K-th principal invariant I~KJ of a
second order tensor a is the K-th elementary symmetric function of the proper
numbers of a,
(38.1)
so that
(38.2)
1 [1908, 1]. 2 [1893, 3]. MuiR [1896, 2] extended BANG's result to n dimensions.
Sect. 38. Principal and related invariants. 833
Therefore the principal invariants determine the proper numbers uniquely up
to order. When the proper numbers are real, unique determination is effected
by the order convention ~~ a2 ;;::: • • • ~ a,., so that, lor values ol the J~K) such
that all roots ol Eq. (38.2) are real, any single-valued lunction ol proper numbers
equals a single-valued lunction ol principal invariants. Comparison of Eq. (38.2)
with Eq. (37.1) shows that J~KJ is the sum of the principal K-rowed minors of
the matrix lla"mll:
I (K)=_1_s.•l···'K am1 amx a K! umJ ..• mg r1 • • • •x • (38.3)
F or a real tensor a such that there is a real non-singular co-ordinate translormation which diagonalizes lla"mll, any single-valued absolute scalar invariant l(a) under
arbitrary co-ordinate translormations is expressible as a single-valued lunction
ol principal invariants1• For proof, we evaluate I in a principal co-ordinate system;
by Eq. (37.16), I is a function of the ab, which are real; the assertion follows
by the italicized statement in the paragraph preceding. Q.E.D.
This result does not extend to an arbitrary second order tensor a, it being necessary to
adjoin to the principal invariants WEvR's 2 characteristics of a to obtain a complete set. If a can be diagonalized, the WEYR characteristics of a are uniquely determined by the principal
invariants. It is true that if f is a scalar polynomial in the a"m• then it is always expressible
as a polynomial in the principal invariants 3 • The WEYR characteristics are not such polynomials; in fact they are not even continuous functions of a at a = o. Of more interest in
mechanics are corresponding results for scalar invariants of both akm and Ckm under arbitrary
co-ordinate transformations, or of a8 m under orthogonal transformations.
The K-th moment I~K) of a is the sum of K-th powers of the proper numbers
of a: .. -
1cK)- ~ (a )K _ am1 ams amx a = LJ b - ms m1 · · · m1 · (38.4)
b=l
If we set
(38.5)
.. where the sum runs over all sets of n non-negative integers Mb such that ~ Mb= L, .. b=l
~ b Mb = K, then an expression for the moments I~K) in terms of the principal
b=l
invariants J~Kl is"
K
J(K) =K "_L u(K,L) a L.J L Yla •
L~l
(38.6)
Except where otherwise noted, we henceforth assume n = 3 and write 5 Ia,
Ha, lila, la, IIa, lila in place of J~>, II!:>, I!:>, I~>, 1!:1, 1!:1, respectively. In
this notation, a is a proper number of a if and only if
a3 - Ia a2 + IIa a - lila = 0;
in particular, a = 1 is a proper number if and only if
la - Ha + lila = 1 .
(38.7)
(38.8)
1 For the symmetric case, RANKINE [1856, 2, § 3] refers to this result as a discovery of
CAYLEY. 2 [1885, 1] and [1890, 1]. See also MAcDuFFEE [1933, 2, § 40]. 3 For n = 3 this follows immediately from a result due to WEITZENBÖCK [1923, I,
pp. 65-66]. See also TRUESDELL [1952, 2, p. 132], [1953. 4, p. 594]. 4 BURNSIDE and PANTON [1901, 5, § 159] attribute this result to WARING. 5 In a space whose co-ordinates are Ia, lla, IIIa, BoRDONI [1955. 4] discusses the
surface where the discriminant of (38.7) is constant.
Handbuch der Physik, Bd. III/1. 53
834 J. L. ERICKSEN: Tensor Fields.
From Eq. (38.6) we have
Ha = I! - 2 Ha,
2 IIa = I! - IIa,
IIIa = I! - 3 Ia IIa + 3 IIIa, )
IIIa =t I!--! Ia IIa+ i IIIa.
The inverse of a, which exists when IIIa =j= 0, is given explicitly by
IIIa a-1 = a 2 - la a + IIa 1.
The octahedral invariant! Ua is defined by
Ua- L [j (ab- ac)J2 = i (Ila- IIa),)
b ßak;;;- = 3 a r a k >
(38.15)
(3 8.16)
1 The invariance and significance of Ua were known to MAXWELL in 1856 [1937. 10,
pp. 32-38]; it was introduced by v. MISES [1913, 2, § 1]. The name "octahedral invariant"
is usually attached to (2/V3) ut on the basis of a geometrical interpretation given by NADAI
and LODE [1933, 3, § II], [1937, 2, pp. 206-207].
2 KLEITZ [1873. 3, § 23] was the first to make an explicit study of the deviator.
3 More generally, for any a that can be diagonalized by real transformations. 4 LIPSCHITZ [ 187 5. 1]; HAMEL [ 1936, 1, § 1].
5 For the symmetric case, these results are given by MuRNAGHAN [1937. 3, § 3], SIGNORINI [1943,1, § 17), and REINER [1945, 3, §4}.
Sect, 39. Inequalities. 835
where Eq. (38.16) 5 , which follows from Eq. (38.16) 4 by Eq. (38.10), holds only
when u-1 exists. Also1, when u-1 exists, we have da= -a · da-1 · a and hence
(38.17)
39. Inequalities 2• CAUCHY 3 was first to establish bounds for the proper numbers
of matrices. Let the proper numbers of a real symmetric three-dimensional
matrix be ordered so that ~ ~ a3 • Then, if a be referred to a reetangular
Cartesian co-ordinate system, CAUCHY's results are that a1 is never less than,
while a3 is never greater than any one of the six quantities
(k =f= m) (3 9.1)
and that, for each k and m, a2 lies between the two numbers given by the expressions (39.1). Of course, one may interpret the components in (39.1) as physical
components of a in an arbitrary orthogonal co-ordinate system.
Let a be any second order tensor, real or complex, in n dimensions, set b ==
-! (a + a'), c =-! (a- a'), where the bar denotes the complex conjugate; select
any proper number a of a, and write it in the form a =P +iq, where p and q are
real. Refer a to any reetangular Cartesian co-ordinate system and let A, B, and C
be the maxima of the absolute values of the components of a, b, and c, respectively. We then have HrRSCH's inequalities 4
[a[ ~nA, IPI ~nB, [q[ ~nC. (39.2)
One corollary is HERMITE's theorem 5 : A sufficient condition that the proper numbers
of a alt be real is that a be Hermitian, i.e. that c =0. Another is WEIERSTRAss'
theorem 6 : A sufficient condition that the proper numbers of a alt be pure imaginary
is that a be skew-Hermitian, i.e., that b =0. HrRSCH7 showed also that when b
is real, q~ [!(n-1)]!C, and that, if the proper numbers of b, which in this case
HERMITE's theorem shows to be real, be ordered so that ~ • • · ~ b,., then
b 1 ~ p ~ b,.. When the proper numbers of the Hermitian tensor d = a . a' are
similarly ordered, BROW~E's inequality 8 asserts that any proper number a of
a satisfies dl~ a a~ d,..
We again restriet attention to the case when n = 3 and a is real. From
Eq. (38.15) follows
(39-3)
the equality holding if and only if a be symmetric. Further inequalities follow
from the observation that since complex proper numbers appear in conjugate
pairs, all symmetric functions of them are real, so that an inequality may be
inferred from every identity in which the square of such a function occurs. From
(38.15) we thus obtain
(39.4)
1 SrGNORINI [1943, 1, § 17].
2 Further discussion of results in this section is given by MAcDUFFEE [1933, 2, § 18].
3 [1828, J], [1830, 1, Th. 1]; in [1829, 1, Th. 1] he extended the result to n dimensions.
4 [1901, 2]. 5 [1855. 1]. For real a, this theorem was proved earlier by CAUCHY [1828, 1],
[1829, 1].
s [1879. 2]. For real a, the theoremwas proved earlier by CLEBSCH [1863, 2].
7 [1901, 2]. For real a, the results are due to BENDIXSON [1901, 3].
8 [1928, 1, Th. V].
i3*
J. L. ERICKSEN: Tensor Fields. Sect. 39.
with equality holdingifand only if a be symmetric. From Eqs. (38.9h and (38.11)
it follows that
II0 ~- 2Il0 ,
From Eq. (38.13) we get
II,a ~Ha.
In all cases of Ineqs. (39.5) and (39.6), equality holds if and only if Ia =0.
(3 9. 5)
(39.6)
When we add the condition that all proper numbers of a be real, as is the
case when a is symmetric, from the fact that then IIa ~ 0 with equality holding
if and only if a =0, a further sequence of inequalities may be inferred. From
Eqs. (38.9) and (38.13) we thus get
3 IIa ~ I! ~ 2 IIa, (39.7)
where equality holds on the left if and only if a is spherical; on the right, if and
only if a=O. From Eq. (38.11) follows
(39.8)
where equality holdingifand only if a =0. Also from Eq. (38.11), since Ua ~ 0,
we derive
I!~ 3 Ha,
with equality holding if and only if a is spherical.
When all proper numbers are non-negative_, we have obviously1
, Ia ~ o, IIa ~ o, I! ~ 27 IIIa ~ o.
(39.9)
(39.10)
For a real symmetric tensor referred to reetangular Cartesian Co-ordinates,
from Eq. (38.4) 2 we see that
(akm)2 ~ Ha •
There is also HADAMARD's inequality 2
(39.11)
(39.12)
If !Ia! ~ K and lila!~ K, Eq. (38.9h and Ineq. (39.11) imply that for a symmetric
tensor we have
(39.13)
In Ineqs. (39.11) to (39.13), akm can be interpreted as a physical component of a
in any orthogonal co-ordinate system. In any co-ordinate system, we have
(39-14)
where M is the maximum of the absolute values of the mixed components of a.
Results of WEDDERBURN 3 imply for any two symmetric tensors a and b
the inequalities
(39.15)
1 The third inequality, for the symmetric case, was given in more complicated form by
SIGNORINI [1949, 7, Chap. II, § 5]. 2 [1893. 2], ScHUR [1909, 2] gives an extension of this inequality as well as several
others. Cf. also MuiR [1930, 2, Chap. I(a)]. 3 [1925,E2].
Sects. 40, 41. Real powers of positive tensors. 837
b) Powers and matrix polynomials.
40. Integral powers. In this Part, the dimension of the underlying space is
arbitrary. When K is an integer, the K-th power of a tensor a is defined inductively
by1
(40.1)
where K is restricted to be positive when a has no inverse. It follows from
Eq. (37.2) that
(40.2)
hence the K-th power of any proper number of a is a proper number of aK, and
every right (left) proper vector of a is a right (left) proper vector of aK. To obtain
a sharper result, we note the following identities, valid for any scalar m and any
tensor a:
when K> 0,
aK- m 1 = (a-1 - m1 1) ... (a-1 - mK 1) when K < 0,
(40.3)
(40.4)
where m1 , ... , mK are the I Kl complex I Kl-th roots of m. Taking the determinant of both sides of Eqs. (40.3) and (40.4) shows that m isaproper number of
aK if and only ij at least one IKI-th root of m isaproper number 2 of a when K >O,
or of a-1 when K < 0. The proper numbers of a-1 are the reciprocals 3 of those
of a.
It can occur for some K that there exist proper vectors of aK which are not
proper vectors of a. E.g., if II aKmll = II ~ _ ~ II, then a 2 = 1, so that an arbitrary
vector is a proper vector of a 2, while the only right proper vectors of a are (b, 0)
and (0, c).
By the Hamilton-Cayley theorem, we may replace (ab)K by aK in Eq. (38.2),
obtaining
(40.5)
This formula makes it possible to write any power of a as a linear combination of
1, a, ... , an-I, with scalar coefficients that are polynomials in the principal invariants ij the power is positive, rational functions if it is negative. For positive K,
RANUM 4 gave the formula
N
an+K = L: H~K,L) a{L)' L~1
where H~K,L) is defined by Eq. (38.5), and where
n-L
a{L)= L (-1t-Q+l I~n-Q)aQ+L-1.
Q~O
(40.6)
(40.7)
41. Real powers of positive tensors. We call a tensor akm positive when its
proper numbers arereal and positive and it possesses n linearly independentproper
vectors. In particular, a positive definite (symmetric) tensor is positive. For
a positive tensor we have the representation (37.12h:
n
a's = L ab q(, mbs' b~I
1 STICKELBURGER [1881, 4] was first to define general powers of matrices. 2 BoRCHARDT [1846, 1], [1847, 1], when K>o.
(41.1)
3 This was known to SPOTTISWOODE [1856, 1]; the result then follows for all K=J= o. 4 [1911, 1].
838 J. L. ERICKSEN: Tensor Fields. Sect. 42.
where ab> 0 and where the qb and mb are reciprocal sets of right and left proper
vectors. For any real K, the K-th power of a positive tensor a is defined to be
K
the positive tensor aK whose components a~ are given by
(41.2)
wherein (ab)K is the positive real K-th power of ab. That is, aK is the unique
tensor having the sameproper vectors as a and having as its proper numbers the
positive K-th powers of the proper numbers of a. This definition is equivalent
to Eq. ( 40.1) when both are applicable. The usuallaws of exponents apply to aK
as defined by Eq. (41.2). For example, by Eq. (37.9) 1 we obtain
n n
L;(ab)K+Lqf,mbs= L (ao)Kq{,mbu(a,)Lq~mcs, (41.3)
b=l ~c=l
so that aK+L =aK. aL.
For a non-negative tensor, i.e., a tensor having n linearly independent proper
vectors and real proper numbers which are positive or zero1, Eq. (41.2) serves
to define a unique K-th power when K~ 0.
If we accept the possibility that aK may be complex and multivalued, by using different
determinations of the K-th powers of the several proper numbers in Eq. (41.2) we may define
various K-th powers of any matrix having a representation of the form (41.1). Such a definition is unsatisfactory in two respects. First, once one admits the possibility that aK be
multivalued, it seems preferable that al/M should represent any solution x of
xM=a (41.4)
when M is an integer, whereas the definition just mentioned excludes some solutions. Second,
there is little or no motivation for excluding tensors not representable in the form ( 41.1).
A more satisfactory definition of aK is easily obtained by regarding it as a complex multivalued function of a in the sense of CrPOLLA2 . We omit the details, as the definitions already
given are adequate for this appendix. Several writers, beginning with CA YLEY 3, have studied
the solution of Eq. (41.4).
42. Matrix functions. A matrix b given by
(42.1)
where the cGi are scalar constants, is a matrix polynomial in the variable matrix a.
Matrix polynomials may be set into one-to-one correspondence with the polynomials in a scalar variable x which are defined by the same set of constants,
and the algebra of matrix polynomials is isomorphic to the algebra of polynomials
in a scalar indeterminate.
Similarly, an infinite set of constants c<» defines a formal matrix power series
in a, corresponding to the formal scalar power series b (x) determined by the same
coefficients. The matrix power series thus obtained converges if and only if every
proper number of a lies inside or on the circle of convergence of the scalar series
1 For real tensors, this is a slight generalization of the usual definition of non-negative
tensors or linear transformations. See e.g. HALMOS [1942, J, §56]. 2 [1932, 2]. See also MAcDUFFEE [1933, 2, §SO]. 3 [1858, 2], [1872, 1]. WEITZENBÖCK [1932, 3] gave a method for determining all
solutions. AuTONNE [1902, 1], [1903, 1] showed that in the complex field, if a is nonnegative and Hermitian, then there is a unique solution which is non-negative and Hermitian.
A similar analysis shows that if a is non-negative according to the definition given above,
then there is a unique solution which is non-negative. For further references, see MAcDuFFEE
[1933. 2, §§ 48, SO] and WEDDERBURN [1934. 2, p. 171].
Sect. 42. Matrix functions. 839
b (x) and, for every proper number a of multiplicity m, the power series for the formal
m -1st derivative b(m-1) (a) converges 1•
Each proper number of a matrix polynomial b (a) is a function of a single
proper number of a, as follows from a more general and more explicit theorem
of FROBENIUS: Let r(x, ... , y) =P(x, ... , y)fq(x, ... , y) be a rational function
of the scalar indeterminates x, ... , y, p and q being polynomials; let a, ... , b be
commutative matrices such that the matrix q (a, ... , b) is non-singular; then the
proper numbers ab, ... , bn of a, ... , b can be ordered so that r (a1 , ... , b1), ... ,
r(a,., ... , b,.) are the proper numbers of r=p(a, ... , b) · q-1 (a, ... , b) =q-1 · p,
the ordering being the same for all rational /unctions 2• Loosely related to this is
SYLVESTER's assertion that, whether or not a and b commute, the proper numbers
of a · b and b · a coincide 3•
RANUM's equation (40.6) serves to reduce any matrix polynomial or power
series to the form
(42.2)
where the coefficients d(J) are, respectively, polynomials or power series 4 in the
principal invariants of a. There are many functions representable in this form
that are neither matrix polynomials nor matrixpower series: for example, b (a) =
n1l 1. Any function of the form (42.2) satisfies
b (I· a · 1-1) = 1 · b (a) · r, (42.3)
where I is an arbitrary non-singular matrix. Conversely, if each component of b
be a polynomial in the components of a and if b (a) satisfy Eq. (42.3) for arbitrary
non-singular I, then b is representable 6 in the form (42.2) with coefficients d(J)
which are polynomials in the principal invariants of a. DIRAC6 proposed Eq. (42.3),
with a, b, and I interpreted as elements of any algebra 7, as a part of the definition of b's being a function of a. He noted that Eq. (42.3) implies that I commutes with b whenever I commutes with a. TURNBULL and AITKEN8 showed
that if a and b are complex n x n matrices and if b commutes with every complex
matrix that commutes with a, then b is expressible as a linear combination of
1, a, ... , a"- 1. An analogaus result for matrices over the real field is readily
established. Consequently Eq. (42.3) implies that b is a linear combination of
1, a, ... , a"- 1 with coefficients depending on a in such a way as to be scalars
under arbitrary symmetry transformations. Under reasonably general conditions, these scalar coefficients are expressible as functions of the principal
invariants of a, as follows from the theorems given at the beginning of Sect. 38.
1 HENSEL [1926, 1]. WEYR [1887, 2] had previously treated the case where no
proper number of a lies on the circle of convergence. PHILLIPS [1919, 2] gave sufficient
conditions for the convergence of a matrixpower series in any finite set of commuting matrices.
2 According to MACDUFFER [1933, 2, p. 23], BROMWICH [1901, 4] noted that this
theorem may fail to hold when r (x, ... , y) is .not rational. We find no explicit statement
to this effect in BROMWICH's paper, though it may follow from results given in his § 3.
3 SYLVESTER [ 1883, 2] stated this without proof. MACDUFFER [ 1933, 2, p. 23] gives an
elegant proof.
4 Expressions of the type (42.2) are sometimes called "polynomials in a" even when
the d!Jj arenot polynomials in a. See e.g. MAcDUFFER [1933. 2, Th. 15.3]
5 For the case n = 3, this follows immediately from a theorem of WEITZENBÖCK [ 1923,
1, pp. 65-66]. The result for arbitrary n can be established similarly.
6 [1926, 2].
7 The elements of some, but not all algebras are representable by square matrices of finite
order. 8 [1932, 5, p. 150].
840 J. L. ERICKSEN: Tensor Fields. Sect. 43.
Various writers1 have considered the problern of inverting a matrix polynomial
to obtain a in terms of b. Theinverse a(b), when it exists 2, is in general complex
and multivalued. It can be shown that for arbitrary non-singular I we have
{42.4)
this being interpreted in the sense that for given I each value of a (b) is equal
to some value of 1-1 • [a(l · b · l-1)] ·I·
Contrary to what might be expected from the results of WEITZENBÖCK and of TuRNBULL
and AITKEN concerning Eq. (42.3). in general there are values of a(b) which arenot expressible as linear combinations of 1, ... , bn--1 ; values, that is which cannot be regarded as
matrix polynomials or power series. Recognizing this, algebraists have attempted to devise
more general definitions of matrix functions 3 to include such possibilities, still insisting upon
a correspondence between matrix functions and functions of a single scalar variable. According to these definitions, a matrix function x(a) always satisfies Eq. (42.3). this being interpreted as was Eq. (42.4) in cases where b (a) is multivalued. Such definitions are not easily
generalized to the case of matrices depending on several matrices.
The reader who has even slight familiarity with recent development in continuum mechanics will see that here we have come up against a central problern
in modern theories of materials. We now formulate lunctional delinitions
which seem particularly weil suited for application in classical mechanics, if
perhaps not so well for other fields:
1. If each component of the matrix b be a function of the components of the
matrices a, ... , c, we say that b is a matrix function of a, ... , c.
2. A matrix function b (a, ... , c) such that
b(o · a · o-1, ... , o · c · o-1) =o · b(a, ... , c) · o-1 (42.5)
for all orthogonal matrices o, i.e., for all o such that
o-1 = o', {42.6)
is an isotropic 4 matrix function of a, ... , c.
3. If each component of the matrix function b (a, ... , c) be a polynomial in
the components of a, ... , c, then b (a, ... , c) is a polynomial matrix function of
a, ... , c.
4. An isotropic polynomial matrix function is one satisfying both 2 and 3.
c) Decompositions.
43. Invariant decompositions. The group of transformations of reetangular
Cartesian co-ordinate systems decomposes the linear vector space of second order
tensors into three linear subspaces, the tensors comprising these spaces being,
respectively, the spherical tensors, the symmetric traceless tensors, and the skewsymmetric tensors. Any orthogonal transformation maps each subspace onto
itself; the only tensor common to two of them is the zero tensor; none possesses
a proper linear subspace mapped into itself by every orthogonal transformation.
In modern terminology, these three are the irreducible invariant subspaces of
1 Cf. MAcDuFFEE [1933, 2, §§ 47, 48] for references. RuTHERFORD [1932, 4] gives
certain rather explicit solutions for a fairly general class of equations. MAcDuFFEE [1933,
2, p. 94] indicates how all solutions can be obtained in the case when b is spherical.
2 WEITZENBÖCK [ 1932, 3, p. 161] gives a simple example of a case where no inverse
exists for the equation b =a2 •
3 Cf. MAcDuFFEE [1933, 2, §SO] for references. 4 The terminology agrees with that used by RIVLIN and ERICKSEN [1954, 5, § 21].
Sect. 43. Invariant decompositions. 841
the space of tensors or order two with respect to the orthogonal group1• Explicitly, we have the decomposition
- -1 J(1) + [ - -1 J(1) ] + a,.".- n a g,.". a(km) n a gkm a[km] • (43 .1)
which is, in the sense indicated above, maximal.
An arbitrary matrix a can be written in infinitely many ways as the product
of two symmetric matrices, one of which is non-singular2. This can be formulated
in the following two ways:
a" ". = b"' c,m,
a"". = d"' e,m,
b[km]- c -0 det b""'=l= 0, -[km]- •
d[km] = e[km] = 0, det e,..,. =!= 0.
(43.2)
(43-3)
As is clear from the results at the beginning of Sect. 37, b cannot always be chosen
tobe positive definite; the same applies to e.
Any non-singular matrix a may be written in the forms
a = s . o = o . s*, (43.4)
where o is orthogonal, and s and s* are symmetric and positive definite 3 ; o, s, and
s* are uniquely determined. We now give a proof ofthispolar decompositio'n
theorem4• Since a · a' is a positive definite 5 symmetric tensor, by the results
of Sect. 41, it has a unique positive square root s; since a is non-singular, so is s.
Therefore we ma y set
s= (a·a')i, o = s-1 - a. (43-5)
Now s-1 • a. a'. s-1= 1, or (s-1 • a) · (s-1 • a)' = 1; therefore o is orthogonal. From
Eq. (43.5) 2 follows Eq. (43.4)1 • In the same way we obtain a decomposition
a = o* · s*, with
s* = (a'. a)!, o* _ a-s-1 • (43.6)
1 The corresponding decomposition of the space of tensors of any given order is derived
and discussed in some detail by WEYL [1946, 3, Chap. VB], references to relevant literature
being given on pp. 310-311. 2 FROBENIUS [1910, 1]. Voss [1878, 1, p. 343] previously established this for nonsingular a. HILTON [1914, 1] characterized the matrices which can be written as the product
of two skew-symmetric matrices or as the product of a symmetric and a skew-symmetric
matrix. 3 We need to interpret this theoremalso in terms of lineartransformations; equivalently,
any second order tensor with non-vanishing determinant may be expressed as the product
of a unique orthogonal tensor by a unique positive definite symmetric tensor. An orthogonal
matrix was defined by Eq. (42.6). An orthogonal tensor ok". is defined by the property that
under the transformation vk = okpvP, w"' = o"'pwP, for arbitrary vectors v and w, the inner
product g,.".vkwm is invariant. That is,
g,.".v"w"' = g,.". okp o"'qvPwq = g,.".v"w"'. (A)
In order for this relation to hold for arbitrary v and w, o must satisfy
(B)
alternatively, o,."'o"p = IJp ·
In a reetangular Cartesian system, the matrix ok". is an orthogonal matrix. In general
co-ordinates, (B) asserts that (g · o)' = (o · g-1tl, where g = II g,.".IJ.
' An equivalent algebraic statement and all the essential ideas for an algebraic proof were
given by FINGER [1892, 4, Eq. (25)]; the first algebraic proof, by E. and F. CosSERAT [1896. 3,
§ 6] (cf. also BURGATTI [1914, 3], SIGNORINI [1930, 3]). The ideas of FINGERand the CosSERATS were put into matrix notation and extended to complex matrices by AuTONNE
[1902, T9, Lemma II], hisform of the proof being that given above. If a is singular, decompositions a = s · o = o* · s* exist but are not unique; cf. HALM OS [ 1942, 1, § 67].
5 If we set w := v · a, then, since a is non-singular, w = 0 if and only if v = o. Since
v · (a · a') · v =W • 1v, it follows that a · a' is positive definite.
842 J. L. ERICKSEN: Tensor Fields. Sect. 44.
We shall show presently that both decompositions of a are unique. Since a =
o . (o-1 . s . o), it is a consequence of this uniqueness that
s* = o-1 · s · o, o*=o. (43 .7)
To prove uniqueness1, we note that s · o = s' · o' implies o = s-1 · s' · o' and
o' =o-1 = (o')-1. s'. s-1, or o = s · (s')-1 · o'. Hence (s-1 · s'- s · (s')-1) · o' =0,
whence follows s-1 • s' =S. (s')-1, or s 2 = (s') 2 ; therefore, since s and s' are positive definite, we derive s = s', and hence o = o'. Q.E.D. By this uniqueness,
it follows also that s and o commute if and only if s = s*; that is, if and only if
a. a' =a'. a, in which case a is called normal 2• The dass of normal matrices
includes all symmetric, skew-symmetric, or orthogonal matrices.
lf a is non-singular, there exists a non-singular matrix b such that3
a = b2
and, if a is also symmetric, a matrix c such that 4
a = c'·c.
(43.8)
(43.9)
In Eqs. (43.8) and (43-9), b and c in general are complex and are not uniquely determined.
44. Certain canonical forms. For any second order tensor a, there exists a
co-ordinate transformation, which may be chosen to be unitary, reducing a at
a given point to superdiagonal form 5 :
a1 1 a12 ... a\
JJakmll = 0 a22 ... (44.1)
0 0 ... ann
When a is real and all its proper numbers are real, this transformation is real
and may be chosen tobe orthogonal. When there are complex proper numbers,
the transformation is necessarily complex.
In three dimensions a real tensor field has either three real proper numbers
or one real, two complex conjugate. In the latter case, there isareal transformation, which may be taken to be orthogonal, reducing a to the form 6
a b c
llakmll = - b d e (44.2)
o o t
When Eq. (44.1) holds in reetangular Cartesian co-ordinates, a is normal
(Sect. 43) if and only if the matrix ( 44.1) is in fact diagonal. Hence a tensor
all of whose proper numbers are real is normal if and only if it is symmetric. A
tensor of the form (44.2) is normal if and only if c = e = 0, a = d.
It is known7 that a real symmetric traceless matrix a can be transformed by
orthogonal transformations to a system in which all diagonal components of a
1 MURNAGHAN and WINTNER [1931, 5]. 2 MURNAGHAN and WINTNER [1931, 4], [1931, 5].
3 Cf. MACDUFFEE [1933, 2, §§ 35, 48]. 4 This follows immediately from the fami!iar result, used by LAGRANGE [1759, 1], that
a quadratic form can be reduced to a sum of squares by linear transformations. 5 SCHUR [1909, 2]. 6 MURNAGHAN and WINTNER [1931, 4]. There is a generalization to n dimensions.
7 LovE [1906, 1, § 16] stated this without proof. WHAPLES has shown us a proof and an
extension of the theorem to matrices defined over essentially arbitrary fields.
Sect. 44. Certain canonical forms. 843
vanish. Now consider the matrix (43.1); in every co-ordinate system the diagonal
components of 1 are all equal, those of the skew-symmetric part i (a- a') vanish,
and the remaining tensor in the matrix (43.1) is symmetric and traceless, whence
it follows by the preceding theorem that in a suitably chosenreetangular Cartesian
co-ordinate system, all diagonal components of a are equal. When n = 3, we have
a b c
a = da e
I g a
(44.3)
When the proper numbers of a are all distinct, 11 akmll can be transformed by real, but
not necessarily orthogonal transformations to the form 1
0 0
(44.4)
0 0 ... 1 0
where cK= (-1)K+l J~l. When a has multipleproper numbers it can be reduced to a direct
sum of matrices of the type (44.4). Denote by b the matrix on the right of Eq. (44.4). Then
c-1 · b · c is diagonal 2, where
C=
(al)"-1 (a2)"-1
(a1)n-2 (a2)"-2
(44.5)
This provides another proof of the result, noted in Sect. 3 7, tha t II ak m II can be red uced to
diagonal form by real transformations if its proper numbers be real and distinct.
Any matrix can be reduced by a complex transformation to a direct sum of matrices of
the form 3
a 1 0 0
0 a 1 0
(44.6)
a 1
0 0 0 a
If all the proper numbers are real, the transformation may be taken as real also. NoLL has
informed us that any real matrix can be transformed by a real transformation to a direct
sum of matrices of the two forms (44.6) and
0 1 0 .
~I a b 1 0
0 0 1 0
0 a b 0 0
0 0 0 0 0 (44.7)
0 a b 1
0 0 0 1
0 • a b
When allproper numbers are distinct, the forms (44.6) and (44.7) reduce to llall and ~~~ ~~~·
respectively. This furnishes still another proof that a matrix with real and distinct proper
numbers can be diagonalized by a real transformation.
1 Cf. MAcDuFFEE [1933, 2, § 39] for references and discussion.
2 SCHUR [1909, 3]. 3 JORDAN [1!170, 4, p. 114].
844 J. L. ERICKSEN: Tensor Fields. Sects. 45. 46.
d) Normaland shear components.
45. Definitions. Except where otherwise noted, the second order tensors considered in the remainder of this chapter are assumed to be symmetric, there
then being no distinction between right and left proper vectors. The scalar
akm vk vm, where v is any unit vector, is called the normal component of a for the
direction v. The scalar akmukvm, where u and v are unit vectors, is the shear
component of a for the directions u and v, the normal components being included
as a special case. When u and v are perpendicular unit vectors, we call akmukvm
the corresponding orthogonal shear component.
In orthogonal co-ordinates, the physical component a(kk) is the normal
component of a for the direction of the tangent to the k-th Co-ordinate curve,
while a(km), k=f=m, is the shear component for the directions of the tangents to
the k-th and m-th co-ordinate curves. In passages where a fixed orthogonal coordinate system has been laid down, the phrases "normal components" and
"shear components ", with no further qualification, mean the components
a(kk) and a(km), k =f= m, respectively.
Because of the existence of the canonical form exemplified by Eq. ( 44.3),
it is always possible to refer a to an orthogonal co-ordinate system in which the
normal components of a at a given point are alt equal. They can all be made to
vanish if and only if a is traceless. In order for the normal components to be
zero in all orthogonal co-ordinate system, it is necessary and sufficient that
a = o. Since by assumption a is symmetric, there exist orthogonal co-ordinate
systems in which llall is diagonal at a given point; that is, it is always possible
to refer a to an orthogonal co-ordinate system in which at a given point all the shear
components are zero. These Co-ordinate systems are principal co-ordinate systems
as defined in Sect. 37. In order for the shear components to vanish in allorthogonal
co-ordinate systems, it is necessary and sufficient that a be spherical; for a
spherical tensor, the shear component for the directions u and v is zero if and
only if u and v be orthogonal.
The proper numbers of a, which are real, we order as follows:
(45.1)
46. Extremal properties. Setting
C- akmukvm, (46.1)
for a fixed tensor a, we consider the problern of finding the extremes of the scalar C
when the vectors u and v are varied subject to certain constraints.
Problem 1. Let the constraints be u = v, uk uk = 1, so that C is the normal
component of a for the direction u. Referring a to a reetangular Cartesian principal co-ordinate system, we then have
C = a1 (u1) 2 + · · · +an (un) 2 , (u1) 2 + · · · + (un) 2 = 1, (46.2)
and from Eq. (45.1) it follows that
a1=a1{ (u1) 2+ · · · + (un) 2} ~aJ(u + · · · +an(un) ~a,.{ (u1) 2+ · · · + (un) 2}=a,., (46.3)
so that ~C~a,.. To see when equality can hold on the left, we set a1=C
in Eq. (46.2h and use Eq. (46.2) 2 , so obtaining
0 = a1 {(u1) 2 - 1} + a2 (u2) 2+ · ·· + a"(u,.) 2 }
= (a2 - a1) (u2)2 + ... + (a,.- a1) (u,.)2, (46.4)
Sect. 46. Extremal properties. 845
whence it follows by Eq. (45.1) that u11 =0 when k > 1 unless a11 =a1 • Therefore, letting K be the largest integer for which aK =a1 , so that a1 =a1 when
I::;;;; K, we conclude that uK+1 = · · · =u,. =0, whence it follows by direct calculation that a~um = a1 u11 • Thus the normal component of a is greatest for the directions
which are principal directions corresponding to the greatest proper number, and
the value of the greatest normal component is the greatest proper number. Sirnilary,
C takes on its minimum value a,. if and only if u be a proper vector of a corresponding to the proper nurober a,.. The remaining proper numbers of a are
extremal values of C; if b> 1, ab is the maximum value of C when u ranges over
the vectors which satisfy the constraints and are perpendicular to b - 1 mutually
orthogonal proper vectors of a corresponding to the proper numbers1 a1 , ... , ab-l·
In general C takes on these intermediate values for infinitely many vectors u
which do not satisfy the orthogonality conditions and are not proper vectors
of a.
Problem 2. Let the constraints be u,. u" = v11 v" = 1, so that C is the shear
component for the directions u and v. Extremal conditions for C are
(46.5)
where a and bare multipliers. Since u"oCfou"=v"oCfov"=C. we conclude
that a = b; hence
(46.6)
lf a has two proper numbers ± a that are numerically equal but opposite in sign,
Eq. (46.6) can be satisfied by choosing u and v to be non-parallel vectors such
that u +v and u- v are proper vectors corresponding to a and - a. This includes the limiting case where zero is a multiple proper nurober of a. Otherwise,
we must have u ± v = 0; the case u = v is that considered in Problem 1, while
the case u = -v is similar. We have shown that extrema of shear components
are always to be found among those for the coincident directions, which are the normal
components, and among those for opposite directions; for additional extrema of
the shear components to exist, it is necessary and sulficient that a have a pair of
proper numbers ± a.
Problem 3. In Problem 2, we add the further constraint u · v = 0, thus confining attention to orthogonal shear components. Clearly, the largest (smallest)
value of C attainable with this additional constraint is never greater (less) than
that attainable with the constraints of Problem 2. Extremal conditions are
(46.7)
where a, b, and c are multipliers. It follows that a = c and the corresponding
extremal value of C is a. Therefore
a,.". (um+ v"') = (a + b) (u11 + v,.), a,.m (um- v"') = (b- a) (u11 - v11). (46.8)
The constraints require that the vectors u ± v be non-null, whence it follows that
these are proper vectors corresponding to the proper numbers (b ±a) of a. This
result is the theorem of Coulomb and Hopkins 2 : The maximum and minimum
1 See, e.g., COURANT and HILBERT [1931, 2, pp. 20-23]. CAUCHY [1829, 1], [1830, 1,
Chap. 11] noted that the proper numbers of a are the (real) extremal values of akm X
uk u"'fli, u'. ·
2 The theoremwas derived in the plane case by CouLOMB [1 i76, 1, §VIII]; in the general
case, by HoPKINS [1847, 2, §§ 4, 5].
846 J. L. ERICKSEN: Tensor Fields. Sect. 46.
orthogonal shear components of a are given by
max C = t (a1 - an), min C = t (an- a1) (46.9)
respectively, these being takenon if and only if V2u=m1 +mn, V2v =m1 -m,.,
where m 1 and mn areorthogonal unit proper vectors of a corresponding to the greatest
and least proper numbers a1 and an. When a1 =an, a is spherical, so that all
orthogonal shear components vanish, as follows also from (46.9). The remaining
extremal values of C, as given by (46.8), are f(a0 -ae), b=J=e. Each of these
is a minimax unless a0 - ae = ± (a 1 - an)· The octahedral invariant l..la, defined
by Eq. ( 46.11), is thus the sum of the squares of the extremal orthogonal shear components of a.
From Eq. (46.9) we observe that
f (an- a1) ;;;;: akm uk v"' (u, u' V5 vs)-§ ;;;;: t (a 1 - an)
for all non-null orthogonal vectors u and v.
(46.10)
In a reetangular Cartesian co-ordinate system such that at a given point
n- 2 axes are principal axes of a and the remaining two are directions satisfying Eq. (46.7), a assumes the form
d 0 0
llakmll = 0 b a
0 a b
where d isadiagonal (n-2)X(n-2) matrix 1 .
(46.11)
Problem 4. We now take v as a fixed unit vector and vary u subject to the
constraints uk uk = 1, uk vk = 0. Thus we seek the greatest shear components
among all directions perpendicular to a given direction. As conditions that C
be extremal, we obtain ak".v"'=auk+bvk. (46.12)
Therefore b = ak". vk v"'; that is, bis the normal component of a for the direction v,
and a, which satisfies
(46.13)
is the extremal value of C. When v is a proper vector of a, then C = 0 for all u
satisfying the constraints. This is no more than the statement that the shear
component for a principal direction and any direction orthogonal to it is zero.
Otherwise Eq. (46.13) determine a unique value for a 2, and therefore Eq. (46.12)
determines u up to sign. It can be shown that2
a 2 - '\' (a - a )2 cos2 {} cos2 {} - L... b e b e• (46.14)
b