Fifty years of the spectrum problem:
Survey and new results

In 1952, H. Scholz published an innocent question in the Journal of Symbolic Logic [ar:Scholz52]:

1. Ein ungelöstes Problem in der symbolischen Logik. $IK$ sei der Prädikatenkalkül der ersten Stufe mit der Identität. In $IK$ ist ein Postulatensystem $BP$ für die Boole’sche Algebra mit einer einzigen zweistelligen Prädikatenvariablen formalisierbar. $\theta$ sei die Konjunktion der Postulate von $PB$. Dann ist $\theta$ für endliche $m$ $m$-zahlig erfüllbar genau dann, wennn es ein $n>0$ gibt, sodaß $m=2^{n}$.

Hieraus ergibt sich das folgende Problem. $H$ sei ein Ausdruck des $IK$. Unter dem Spectrum von $H$soll die Menge der natürlichen Zahlen verstanden sein, für welche $H$ erfüllbar ist. $M$ sei eine beliebige Menge von natürlichen Zahlen. Gesucht ist eine hinreichende [hinrerichende] und notwendige Bedingung dafür, daß es ein $H$ gibt, sodaß $M$ das Spectrum von $H$ ist.
(Received September 19, 1951).

In English:

1. An unsolved problem in symbolic logic. Let $IK$ [Identitätskalkül] be the first order predicate calculus with identity. In $IK$ one can formalize an axiom system $BP$ [Boole’sche Postulate] for Boolean algebras with only one binary relation variable. Let $\theta$ be the conjunction of the axioms of $BP$. Then $\theta$ is satisfiable in a finite domain of $m$ elements if and only if there is an $n>0$ such that $m=2^{n}$.

From this results the following problem. Let $H$ be an expression of $IK$. We call the set of natural numbers, for which $H$ is satisfiable, the spectrum of $H$. Let $M$ be an arbitrary set of natural numbers. We look for a sufficient and necessary condition that ensures that there exists an $H$, such that $M$ is the spectrum of $H$.
(Received September 19, 1951).

This question inaugurated a new column of Problems to be published in the Journal of Symbolic Logic and edited by L. Henkin. Other questions published in the same issue were authored by G. Kreisel and L. Henkin. They deal with a question about interpretations of non-finitist proofs dealing with recursive ordinals and the no-counter-example interpretation (Kreisel), the provability of formulas asserting the provability or independence of provability assertions (Henkin), and the question whether the ordering principle is equivalent to the axiom of choice (Henkin). All in all 9 problems were published, the last in 1956.

The context in which Scholz’s question was formulated is given by the various completeness and incompleteness results for First Order Logic that were the main concern of logicians of the period. An easy consequence of Gödel’s classical completeness theorem of 1929 states that validity of first order sentences in all (finite and infinite) structures is recursively enumerable, whereas Church’s and Turing’s classical theorems state that it is not recursive. In contrast to this, the following was shown in 1950 by B. Trakhtenbrot.

_

Heinrich Scholz, a German philosopher, was born 17. December 1884 in Berlin and died 30. December 1956 in Münster. He was a student of Adolf von Harnack. He studied in Berlin and Erlangen philosophy and theology and got his habilitation in 1910 in Berin for the subjects philosophy of religion and systematic theology. He received his Ph.D. in 1913 for his thesis Schleiermacher and Goethe. A contribution to the history of German thought. In 1917 he was appointed full professor for philosophy of religion in Breslau (Wroclaw, today Poland). In 1919 he moved to Kiel, and from 1928 on he taught in Münster. From 1924 till 1928 he studied exact sciences and logic and formed in Münster a center for mathematical logic and foundational studies, later to be known as the school of Münster. His chair became in 1936 the first chair for mathematical logic and the foundations of exact sciences. His seminar underwent several administrative metamorphoses that culminated in 1950 in the creation of the Institute for mathematical logic and the foundations of exact sciences, which he led until his untimely death. Among his pupils and collaborators we find W. Ackermann, F. Bachmann, G. Hasenjäger, H. Hermes, K. Schröter and H. Schweitzer. He was also among the founders of the German society bearing the same name (DVMLG). H. Scholz was a Platonist, and he considered mathematical logic as the foundation of epistemology. He is credited for his discovery of Frege’s estate, and for making Frege’s writing accessible to a wider readership. Together with his pupil Hasenjäger he authored the monograph Grundzüge der Mathematischen Logik, published posthumously in 1961.

_

Thus, H. Scholz really asked whether one could prove anything meaningful about f-satisfiablity besides its undecidability.

In our notation to be used throughout the paper, H. Scholz introduced the following:

Let $\tau$ be a vocabulary, i.e., set of relation and function symbols. Let $\phi$ be a sentence in some logic with equality over a vocabulary $\tau$. Unless otherwise stated the logic will be first order logic $FOL(\tau)$. Sometimes we shall also discuss second order logic, or a fragment thereof.

We shall use $S,S_{i}$ to denote spectra. For the definition of spectra it does not matter whether we use function symbols or not. So, unless otherwise stated, vocabularies will be without function symbols. However, we shall allow function symbols when dealing with sentences of special forms.

Clearly, $\textsc{spec}({\phi})=\emptyset$ if and only if $\phi$ is not f-satisfiable. By definition of satisfiability $0$ is never part of a spectrum. Very often a spectrum is finite, cofinite or even of the form $\mathbb{N}^{+}=\mathbb{N}-\{0\}$.

As H. Scholz noted, (the set of) powers of $2$ form a spectrum, because they are the cardinalities of finite Boolean algebras. Similarly, powers of primes form a spectrum, because they are the cardinalities of finite fields. For $a,b\in\mathbb{N}^{+}$ there are many ways to construct a sentence $\phi$ with $\textsc{spec}({\phi})=a+b\mathbb{N}$, one of which consists in using one unary function symbol. With a moment of reflection, one sees that spectra have the following closure properties.

In the spirit of Question 2.3 we can also ask:

We shall answer Questions 2.3 and 2.5 in Section 3.4.

The first to publish a paper in response to H. Scholz’s problem was G. Asser [ar:Asser55]. A. Robinson’s review [misc:Robinson-mr] summarizes it as follows:

($\ldots$) The present paper is concerned with the characterisation of all representable sets [=spectra]. A rather intricate necessary and sufficient condition is stated for arithmetical function $X(n)$ to be the characteristic function of a representable set. The condition shows that such a function is elementary in the sense of Kalmar. ($\ldots$) On the other hand, the author establishes that there exist non-representable sets whose characteristic function is elementary. Examples of representable sets (some of which are by no means obvious) are given without proof and the author suggests that further research in this field is desirable.

Asser also noted that his characterization did not establish whether the complement of a spectrum is a spectrum.

About the same time, A. Mostowski [ar:Mostowski56] also considered the problem. H. Curry [misc:Curry-mr] summarizes Mostowski’s paper as follows:

($\ldots$) The author proves that for each function $f(n)$ of a class $K$ of functions, which is like the class of primitive functions except that at each step all functions are truncated above at $n$, there is a formula $H$ that has a model in a set of $n+1$ individuals if and only if $f(n)=0$. From this he deduces positive solutions to Scholz’s problem in a number of special cases.

It is usually considered that A. Mostowski really proved

The detailed definitions and contents of this theorem will be discussed in Section 4.

In the last 50 years a steady stream of papers appeared dealing with spectra of first order and higher order logics. The problem seems not too important at first sight. However, some of these papers had considerable impact on what is now called Finite Model Theory and Descriptive Complexity Theory.

Scholz’s Problem, as stated, is rather vague. He asks for a characterization of a family of subsets of the natural numbers without specifying, what kind of an answer he had in mind. The answer could be in terms of number theory, recursion theory, it could be algebraic, or in terms of something still to be developed. We shall see in the sequel many solutions to Scholz’s Problem, but we consider it still open, because further answers are still possible.

The same question can be asked for any logic, in particular second order logic SO, or fragments thereof, like monadic second order logic MSOL, fixed point logic, etc., as discussed in [bk:EbbinghausF95, bk:Libkin04].

Here the answer should be yes or no.

The corresponding problem for SO has a trivial solution. Let $\phi\in\textsc{SO}({\tau})$ with $\tau=\{R_{1},\ldots,R_{k}\}$. An integer $n$ is in $\textsc{spec}({\phi})$ iff

Then, the complement of $\textsc{spec}({\phi})$ is easily seen to be the spectrum of the SO sentence $\neg(\exists R_{1}\ldots\exists R_{k}\phi)$. In passing, note that every SO-spectrum is a $\textsc{SO}({\tau})$ over a language $\tau$ containing equality only.

However, for fixed fragments of SO, Asser’s Problem remains open. In particular

In this survey we shall describe the various solutions and attempts to solve Scholz’s and Asser’s problems, and the developments these attempts triggered. We shall emphasize more the various ways the questions were approached, and focus less on the historical order of the papers.

There are several discernible themes:

Recursion Theory

The early authors H. Asser and A. Mostowski approached the question in the language of the theory of recursive functions i.e. they looked for characterization of spectra in terms of recursion schemes, or hierarchies of recursive functions. Most prominently in terms of Kalmar’s elementary functions, the Grzegorczyk hierarchy and hierarchies of arithmetical predicates, in particular rudimentary relations. This line of thought culminates in 1962 in the thesis of J. Bennett [phd:Bennett62]¹¹1 It seems that some of Bennet’s unpublished results were rediscovered independently in China in the late 1980ties by Shaokui Mo [ar:Mo91]. We shall discuss his work in Section A.4. . Although G. Asser already characterized first order spectra in such terms, his characterization was not considered satisfactory even by himself, because it did not use standard terms and was not useful in proving that a given set of integers is (or not) a spectrum. We shall discuss the recursion theoretic approach in detail in Section 4.

Complexity Theory

In the 1970s, D. Rödding and H. Schwichtenberg of the Münster school [ar:RoeddingS72] gave a sufficient but not necessary condition: any set of integers recognizable by a deterministic linear space-bounded Turing machine is a first-order spectrum. (This is also a consequence of results of Bennett and Ritchie [phd:Bennett62, ar:Ritchie63], obtained before the emergence of complexity theory.) Further, Rödding and Schwichtenberg showed that sets of integers recognisable using larger space bounds are higher order spectra. C. Christen developed this line further [phd:Christen74, proc:Christen76], independently obtaining a number of the following results.

At the same time the spectrum problem gained renewed interest in the USA. In 1972 A. Selman and N. Jones found an exact solution to Scholz’s original question [proc:JonesS72]: a set of integers is a first order spectrum if and only if it is recognizable by a non-deterministic Turing machine in time $O(2^{c\cdot n})$.

This result was independently also obtained by R. Fagin in his thesis [phd:Fagin73], which contains an abundance of further results. Most importantly, R. Fagin studies generalized spectra, which are the projective classes of Tarski, restricted to finite structures, and really laid the foundations for Finite Model Theory and Descriptive Complexity, as can be seen in the monographs [bk:Immerman99, bk:EbbinghausF95, bk:Libkin04]. We shall discuss the complexity theoretic approach in detail in Section 5.

Images and preimages of spectra

From Proposition 2.4 it follows that, if $S$ is a first order spectrum and $p$ is a polynomial with positive coefficents, then $p(S)=\{p(m):m\in S\}$ is also a spectrum. In J. Bennett’s thesis it is essentially proved that there is a first order spectrum $S$ and an integer $k$ such that $\{n:2^{n^{k}}\in S\}$ is not a spectrum. It is natural to ask what happens to a spectrum under images and preimages of number theoretic functions. The general line of this type of results states that certain images or preimages of spectra of specific forms of sentences are or are not spectra of other specific forms of sentences.

Spectra of syntactically restricted sentences

Already in a paper by L. Löwenheim from 1915 [ar:Loewenheim15] it is noted that, what later will be called the spectrum of a sentence in monadic second order logic (MSOL) with unary relation symbols only, is finite or cofinite. The set of even numbers is the spectrum of an MSOL sentence with one binary relation symbol, and it is ultimately periodic. Further, every ultimately periodic set of positive integers is a spectrum of a first order MSOL sentence with one unary function symbol. Over the last fifty years various papers were written relating restrictions on the use of relation and function symbols, or other syntactic restrictions, to special forms of spectra. R. Fagin, in his thesis, poses the following problem

Open Question 4 (Fagin’s Problem for binary relations)

Is every first order spectrum the spectrum of a first order sentence of one binary relation symbol?

The question is even open, if restricted to any fixed vocabulary that contains at least one binary relation symbol or two unary function symbols.

Much of this line of research is motivated by attempts to solve Fagin’s problem.

Transfer theorems

Another way of studying spectra is given by the following result, again from Fagin’s thesis: If $S$ is a spectrum of a purely relational sentence where all the predicate symbols have arity bounded by $k$, then $S^{k}=\{m^{k}:m\in S\}$ is a spectrum of a sentence with one binary relation symbol only, or even a spectrum on simple graphs. One can view this an approach combining the study of images and preimages of spectra with either syntactically or semantically restricted spectra. Over the years quite a few results along this line were published. We shall discuss the last three approaches under the common theme of restrictions on vocabularies in detail in Section 6.

Spectra of semantically restricted classes

R. Fagin shows that Asser’s problem has a positive answer if and only if it has a positive answer if restricted to the class of simple graphs. Similarly, in order to understand Fagin’s problem better, one could consider restricted graph classes $K$, and study first order spectra restricted to graphs in $K$. One may think of graphs of bounded degree, planar graphs, trees, graphs of tree-width at most $k$, etc.

Open Question 5 (Fagin’s Problem for simple graphs)

Is every first order spectrum the spectrum of a first order sentence over simple graphs?

Open Question 6

Is every first order spectrum the spectrum of a first order sentence over planar graphs?

For restrictions to graph classes of bounded tree-width, the answer is negative. The reason for this is that spectra of graphs of bounded tree-width are ultimately periodic. In fact, this holds for a much wider class of spectra. E. Fischer and J.A. Makowsky, [ar:FischerM04], have analyzed under what conditions MSOL-spectra are ultimately periodic. We shall discuss their results in detail in Section 8.

This line of thought has not been extensively explored, this may well be a fruitful avenue for studying spectra in the future.

In the sequel of this survey we shall summarize what is known about spectra along these themes. Various solutions to Scholz’s Problem were offered in the literature, varying with the tastes of the times, but there may be still more to come. Asser’s and Fagin’s Problems are still open. Both problems are intimately related to our understanding of definability hierarchies in Descriptive Complexity Theory. They may well serve as benchmarks of our understanding.

Spectra are sets of positive natural numbers. These sets can be represented in various ways. We shall use the following:

Obvious questions are of the following type:

Coding runs of Turing machines one can easily obtain the following.

Various other partial answers to these questions will appear throughout our narrative.

An obvious question is whether the primes form a spectrum. If one gets more ambitious one can ask for special sets of primes such as Fermat primes (of the form $2^{2^{n}}+1$), Mersenne primes (of the form $2^{p}-1$ with $p$ a prime), or the set of primes $p$ such that $p+2$ is also a prime (twin primes). Even if we do not know whether such a set is finite, which is the case for twin primes, it may still be possible to prove that it is a spectrum. The answer to all these question is yes, because all these sets are easily proved to be rudimentary, see Section 4.

In the sense of the above definitions we have $\chi_{primes}$ is the charactersitic function of the set of primes, $\eta_{primes}(n)=p_{n}$, and $\gamma_{primes}(n)$ is the counting function of the primes, usually denoted by $\pi(n)$. $\delta_{primes}(n)$ is usually denoted by $d_{n}$. All these functions related to primes are subject to intensive study in the literature, see eg. [bk:Ribenboim89]. As we have said that the primes form a first order spectrum, all the features of these functions observed on primes do occur on spectra.

For instance, $\pi(n)$ is approximated by the integral logarithm $li(n)$, and it was shown by J.E. Littlewood in 1914, cf. [bk:Ribenboim89] that $\pi(n)-li(n)$ changes sign infinitely many often. For logical aspects of Littlewood’s theorem, see [ar:Kreisel52].

Let us define {gather} π^+ ={ n : π(n) - li(n) ¿ 0 } \notag
π^- ={ n : π(n) - li(n) ≥0 } \notag A less obvious question concerning spectra and primes is

Many combinatorial functions are defined by linear or polynomial recurrence relations. Among them we have the powers of $2$, factorials, the Fibonacci numbers, Bernoulli numbers, Lucas numbers, Stirling numbers and many more, cf. [bk:GrahamKP89].

The answer will be yes in all of these cases. We shall sketch a proof in Section A.3 that is based on the existence of such recurrence relations.

But these functions also allow combinatorial interpretations as counting functions: The powers of $2$ count subsets, the factorials count linear orderings, the Stirling numbers are related to counting equivalence relations. We shall see below that in these three examples the combinatorial definitions allow us to give alternative proofs that these sets of numbers are first order spectra.

The spectrum of a sentence $\phi$ witnesses the existence of models of $\phi$ of corresponding cardinalities. Instead, one could also ask for the number of ways the set $\{0,1,\ldots,n-1\}=[n]$ can be made into a model of $\phi$. Alternatively one could count models up to isomorphisms or up to some other equivalence relation.

Combinatorial counting functions come in different flavours;

Counting labeled and non-labeled structures has a rich literature, cf. [bk:HararyP73, bk:Wilf90]. Note that counting non-labeled non-isomorphic structures is in general much harder than the labeled case. The first connection between counting labeled structures and logic is the celebrated 0-1 Law for first order logic:

What we are interested in here, is the relationship of such counting functions to spectra. Our example of powers of $2$ shows that

where $\phi$ is an always-true first order sentence with one unary relation symbol, and $\psi$ is the conjunction of the axioms of Boolean algebras. Similarly,

where $\phi_{LIN}$ are the axioms of linear orders, and $\psi$ describes the following situation:

1. (i)

   $P$ is a unary relation and $R$ is an linear order on $P$.

2. (ii)

   $E$ is a ternary relation that is a bijection between the universe (first argument $x$) and all the linear orderings on $P$ (remaining two arguments $y,z$).

3. (iii)

   First we say that there is an $x$ that corresponds to $R$; and that for $x\neq x^{\prime}$ the orderings are different. This says that $E$ is injective. To ensure that we get all the orderings on $P$ we say that for every ordering and every transposition of two elements in this ordering, there is a corresponding ordering.

Hence, the size of the model of $\psi$ is the number of linear orderings on $P$.

Clearly, if $f_{\phi}$ is not strictly increasing, there is no $\psi$ with $f_{\phi}(n)=\eta_{\psi}(n)$. For instance, for $\phi$ which says that some function is a bijection of a part of the universe to its complement, we have

The converse question seems more complicated. For instance, as we have noted before, the primes $p_{n}$ are of the form $\eta_{\psi}$ for some first order $\psi$, but we are not aware of any labeled counting function that will produce the primes.

R. Fagin [proc:Fagin74] calls $\phi$ categorical if $f_{\phi}^{iso}(n)\leq 1$ for every $n$. For instance, $\phi_{LIN}$ is categorical. The counting function up to isomorphisms can be bounded by any finite number $m$, using disjunctions of different categorical sentences. So it may be less promising to study for which first order sentences $\phi$ there is a $\psi$ such that $f_{\phi}^{iso}(n)=\eta_{\psi}(n)$, or vice versa.

Surprisingly enough, C. Ash [ar:Ash94] has found a connection between Asser’s Problem and the behaviour of the Ash functions defined in Definition 3.4(v). We shall discuss this in Section 7.

In the light of Theorem 3.5 we note that if ${\mathit{p}rob}_{\phi}(n)$ tends to $1$ then $spec(\phi)$ is cofinite. Obviously, the converse does not hold, because there are categorical sentences with models in all finite cardinalities.

Trakhtenbrot’s Theorem says that it is undecidable whether a spectrum is empty, and Grandjean’s Theorem says that it is decidable, whether a sentence is almost always true, i.e. ${\mathit{p}rob}_{\phi}(n)$ tends to $1$. As a partial answer to Questions 2.3 and 2.5 we have:

Let $\chi_{\phi}(n)$ be the characteristic function of the spectrum of a first order sentence $\phi$. We can associate with $\phi$ and $a\in{\mathbb{Z}}$ the real number $r_{\phi}=a+\sum_{n}\chi_{\phi}(n)2^{-n}$.

As we have noted ultimately periodic sets of natural numbers are first order spectra, and correspond to rational numbers. Also every ultimately periodic spectrum can be realized by a formula with one function symbol only. We have

E. Specker, [ar:Specker49], proved that there is a real $x$ primitive recursive in base $2$ such that $3x$, $x+\frac{1}{3}$, $x^{2}$ are not primitive recursive in base $2$. A modern treatment can be found in [ar:ChenSuZheng2007]. H. Friedman [misc:Friedman06] mentioned on an internet discussion site that primitive recursive can be replaced in Specker’s Theorem by much lower complexity within the Grzegotczyk Hierarchy. J. Miller kindly provided us, [email:miller], with the more precise statement