Extraction of Expansion Trees
 257 Downloads
Abstract
We define a new method for proof mining by CERES (cutelimination by resolution) that is concerned with the extraction of expansion trees in firstorder logic (see Miller in Stud Log 46(4):347–370, 1987) with equality. In the original CERES method expansion trees can be extracted from proofs in normal form (proofs without quantified cuts) as a postprocessing of cutelimination. More precisely they are extracted from an ACNF, a proof with at most atomic cuts. We define a novel method avoiding proof normalization and show that expansion trees can be extracted from the resolution refutation and the corresponding proof projections. We prove that the new method asymptotically outperforms the standard method (which first computes the ACNF and then extracts an expansion tree). Finally we compare an implementation of the new method with the old one; it turns out that the new method is also more efficient in our experiments.
Keywords
Cutelimination Proof mining Herbrand sequent Expansion tree1 Introduction
Proof analysis and proof mining are central mathematical activities. Extracting additional mathematical information from existing proofs plays an important role in the process of proof mining. Mathematical proofs in general are based on the structuring of reasoning by intermediate statements (lemmas). The drawback of the use of lemmas is that only their truth but not their proofs are reflected in the derivation of their endsequents. These proofs, however, may contain important mathematical information which can be extracted only from the proofs of these lemmas. One of the most important theorems in mathematical logic is Gentzen’s Hauptsatz [15]. It states that lemmas (cuts) can be eliminated from firstorder derivations, resulting in a lemmafree proof combining all subproofs of the original derivation.
The result of cutelimination is a purely combinatorial proof. These combinatorial proofs can be used to extract explicit mathematical information. Proofs can be transformed in a way such that this information becomes visible. Such a transformation for cutfree LKproofs of prenex endsequents was given by Gentzen [15] in the midsequent theorem. It basically states that a proof \(\varphi \) can be transformed into a proof \(\varphi '\) such that \(\varphi '\) contains a socalled midsequent, that splits the proof into a propositional part and a part with quantifier inferences. The midsequent is propositionally valid and contains the instantiations of the quantifiers needed to prove the endsequent. These instantiations may contain crucial mathematical information.
Cutelimination plays a key role in the analysis of mathematical proofs. A prominent example is Girard’s analysis (see in [16]) of Fürstenberg and Weiss’ topological proof [14] of van der Waerden’s theorem [27] on partitions. After cutelimination was applied to the proof of Fürstenberg and Weiss, the result was van der Waerden’s original elementary proof.
Girard’s analysis of the proof of Fürstenberg and Weiss was carried out by hand within mathematical metalanguage. However, in automated proof analysis, formal proofs are vital. Therefore the first step in automated proof analysis consists in formalizing the mathematical proofs (typically expressed in traditional mathematical language). The next steps are algorithmic cutelimination and, finally, the interpretation of the resulting formal proof.
For automated proof analysis of mathematical proofs, the cutelimination method CERES (CutElimination by RESolution) was developed (see [4, 5]). CERES substantially differs from the traditional reductive cutelimination methods a la Gentzen. In the reductive methods cuts are eliminated by stepwise reduction of cutcomplexity. These methods always identify the uppermost logical operator in the cutformula and either eliminate it directly (grade reduction) or indirectly (rank reduction). It is typical for such a method that the cut formulas are “peeled” from the outside till only atomic cuts are left. These methods are local in the sense that only a small part of the whole proof is analyzed, namely the derivation corresponding to the introduction of the uppermost logical operator. As a consequence, many types of redundancy in proofs are left undetected in the reductive methods, leading to an unfortunate computational behavior. In contrast, the method CERES to be presented in Sect. 3 is based on a structural analysis of the whole proof. Here all cutderivations in an LKproof \(\varphi \) of a sequent S are analyzed simultaneously. The interplay of binary rules, which produce ancestors of cut formulas and those which do not, defines a structure which can be represented as a set of clauses \(\mathrm{CL}(\varphi )\). \(\mathrm{CL}(\varphi )\) is always unsatisfiable and thus admits resolution refutations. A resolution refutation \(\gamma \) of \(\mathrm{CL}(\varphi )\) may serve as a skeleton of an LKproof of S with only atomic cuts. The proof itself (a CERES normal form) is obtained by replacing clauses in \(\gamma \) by socalled proof projections of \(\varphi \). To handle predicate logic with equality the calculi can be extended by equality rules; instead of resolution refutations we obtain refutations by resolution and paramodulation (for details see [2]). CERES is a semisemantical method of cutelimination (see [7]). A detailed description of CERES, a comparison with reductive methods, its extensions and a complexity analysis of the method can be found in the book [6].
CERES has been applied to real mathematical proofs. The most interesting application was the analysis of Fürstenberg’s proof of the infinitude of primes [13] where, as a result of cutelimination by CERES, Euclid’s original argument of prime construction was obtained.
The last step in automated proof analysis consists in the interpretation of the result. In this interpretation it is crucial to obtain compact and meaningful information rather than a full (and typically very long) formal proof. The relevant information can be bounds for variables that are used in the proof or even programs representing its algorithmic content. Actually, it is possible to extract functionals, based on Gödel’s dialectica interpretation [17], and construct programs from proofs in Peano arithmetic; see [8, 9] for applications to mathematical proofs.
Another structure representing explicit information are midsequents (also called Herbrand sequents). Herbrand’s theorem, [10, 18], provides one of the most fundamental insights of logic and characterizes the validity of a formula in classical firstorder logic by the existence of a propositional tautology composed of instances of that formula. Roughly speaking, Herbrand sequents are compact structures encoding the essence of proofs with prenex endsequents. Hence in mathematical proof analysis it is frequently more important to extract Herbrand sequents than full formal proofs (which may be too large to be interpreted). There are efficient algorithms for extracting Herbrand sequents from cutfree proofs, see e.g. [20]. Though every formula (and any sequent) can be transformed to prenex form such a transformation is unnatural and can have a disastrous impact on proof complexity (see [3]). Thus it is vital to extend the methods to nonprenex formulas and sequents. Miller [23] developed the structure of expansion trees (and expansion proofs) generalizing the derivation of endsequents from a midsequent in the prenex case. The socalled deep function of an expansion proof generalizes the midsequent itself. As expansion proofs abstract from propositional reasoning they provide compact and explicit information about the mathematical content of formal cutfree proofs.
The result of the method CERES is a CERES normal form, which is a (typically very long) formal proof with at most atomic cuts. This proof can then be used for further investigation and particularly for the extraction of Herbrand sequents and expansion proofs in order to obtain compact information. In the ordinary CERESmethod, expansion proofs can be extracted from an ACNF. In this paper we show that even the construction of an ACNF can be avoided in computing the expansion proofs. In particular, we prove that the expansion proof of the CERES normal form can be constructed from the partial expansion proofs of the projections obtained by CERES, after deleting the clause parts. A ground refutation of the characteristic clause set and the projections suffice for the extraction of expansion proofs making the construction of the CERES normal form itself obsolete. This improvement yields a gain in asymptotic complexity. In particular we show that the new method outperforms the old one (quadratic versus cubic) and that the complexity of the new method can never be higher than that of the old one. Finally we describe an implementation of the new method and the traditional one (both methods are implemented in the Gapt system [12]) and show how they can be used to extract expansion proofs from proofs. We compare the implementations of the two methods and it turns out that even for very small and simple proofs a visible speedup in computing time can be obtained.
2 Preliminaries
2.1 Sequents and Sequent Calculus
We define an extended version of Gentzen’s calculus LK in predicate logic with equality and arbitrary function symbols.
Definition 1
Let \(\varGamma \) and \(\varDelta \) be two multisets of formulas and \(\vdash \) be a symbol not belonging to the logical language. Then \(\varGamma \vdash \varDelta \) is called a sequent.
If \(S_1:\varGamma \vdash \varDelta \) and \(S_2:\varPi \vdash \varLambda \) are sequents we define the concatenation of \(S_1\) and \(S_2\) (notation \(S_1 \circ S_2\)) as \(\varGamma ,\varPi \vdash \varDelta ,\varLambda \).
Definition 2
Let \(S: A_{1}, \ldots , A_{n} \vdash B_{1}, \ldots , B_{m}\) be a sequent and \(\mathcal {M}\) be an interpretation over the signature of \(\{A_{1}, \ldots , A_{n}, B_{1}, \ldots , B_{m}\}\). Then S is valid in \(\mathcal {M}\) if the formula \((A_{1} \wedge \ldots \wedge A_{n}) \rightarrow (B_{1} \vee \cdots \vee B_{m})\) is valid in \(\mathcal {M}\). S is called valid if S is valid in all interpretations.
Definition 3
Let \(S: A_{1}, \ldots , A_{n} \vdash B_{1}, \ldots , B_{m}\) be a sequent. S is called a weakly quantified sequent if there is no \(\exists \) quantifier of positive polarity in some formula \(A_i\) (\(1 \le i \le n\)) and there is no \(\forall \) quantifier of positive polarity in some formula \(B_j\) (\(1 \le j \le m\)).
Definition 4
(Calculus \(\mathbf {LK}_=\)) Basically we use Gentzen’s version of LK [15] but extend it by equality rules as in [2] and call the calculus \(\mathbf {LK}_=\). Since we consider multisets of formulas, we do not need exchange or permutation rules. There are two groups of rules, the logical and the structural ones. All rules except the cut have left and right versions, denoted by l and r, respectively. The binary rules are of multiplicative type, i.e. no autocontraction of the context is applied. In the following, A and B denote formulas whereas \(\varGamma , \varDelta , \varPi , \varLambda \) denote multisets of formulas.
where t is an arbitrary term that does not contain any variables which are bound in A and \(\alpha \) is a free variable which may not occur in \(\varGamma , \varDelta , A\). \(\alpha \) is called an eigenvariable.
where the variable conditions for \(\exists _{l}\) are the same as those for \(\forall _r\) and similarly for \(\exists _r\) and \(\forall _l\). The quantifierrules \(\forall _l,\exists _r\) are called weak, the rules \(\exists _l,\forall _r\)strong.
Note that, on atomic sequents, the rules coincide with paramodulation—under previous application of the most general unifier.
Axioms:
Any set of atomic sequents which is closed under substitution and contains the sequent \(\vdash x=x\) (and thus all sequents of the form \(\vdash t=t\) for arbitrary terms t) is admitted as an axiom set. We define the set \(\mathrm{TAUT}= \{A \vdash A \,  \, A \text{ is } \text{ an } \text{ atom }\}\) and the axiom set \(Ax = \mathrm{TAUT}\cup \{\vdash t = t \mid t \text{ a } \text{ term }\}\), which is called the standard axiom set.
An \(\mathbf {LK}_=\)proof from a set of axioms \(\mathcal{{A}}\) is a tree formed according to the rules of \(\mathbf {LK}_=\) such that all leaves are in \(\mathcal{{A}}\). The formulas in \(\varGamma , \varDelta , \varPi , \varLambda \) are called context formulas. The formulas in the upper sequents that are not context formulas are called auxiliary formulas and those in the lower sequents are called main formulas. The auxiliary formulas of a cutrule are also called cutformulas. If \(\mathcal {S}\) is a set of sequents, then an LKrefutation of \(\mathcal {S}\) is an LKtree \(\pi \) where the endsequent of \(\pi \) is the empty sequent and the leaves of \(\pi \) are either axioms of the standard axiom set or sequents in \(\mathcal {S}\).
For the proof transformations in this paper we need the concept of ancestors of nodes in a proof tree and formula occurrences within sequents occurring in proofs.
Definition 5
(Formula ancestor) Let \(\nu \) be a formula occurrence in a sequent calculus proof \(\varphi \). Then \(\nu \) is an ancestor of itself in \(\varphi \) (the relation is reflexive). If \(\nu \) is a principal formula occurrence of an inference then the occurrences of the auxiliary formula (formulas) in the premises are formula ancestors of \(\nu \). If \(\nu \) is not a principal occurrence then the corresponding occurrences in contexts of the (premise) premises are formula ancestors of \(\nu \). The formula ancestor relation is then defined as the transitive closure.
Definition 6
(Sequent ancestor) Let \(\nu \) be an occurrence of a sequent in a sequent calculus proof \(\varphi \). Then \(\nu \) is a sequent ancestor of itself (reflexivity). If \(\nu \) corresponds to the conclusion of an inference with premises \(\mu _1,\mu _2\) (\(\mu \)) then \(\mu _1,\mu _2\) (\(\mu \)) are sequent ancestors of \(\nu \) in \(\varphi \). The sequent ancestor relation is then defined as the transitive closure.
Definition 7
(Clause) A sequent \(\varGamma \vdash \varDelta \) is called a clause if \(\varGamma \) and \(\varDelta \) are multisets of atoms.
Definition 8
 1.the resolution rule: Where \(n,m \ge 1\) and \(\sigma \) is a most general unifier of \(\{A_1,\ldots ,A_m,A'_1,\ldots ,A'_n\}\). It is also required that \(\varGamma \vdash \varDelta , A\) and \(\varGamma ', A' \vdash \varDelta '\) are variable disjoint.
 2.
the paramodulation rules:
We assume that the two clauses in the premises are always variable disjoint and that \(\sigma \) is a most general unifier of \(\{s,s'\}\). for inference on the left side of the clauses and for the right side, where \(\varLambda \) denotes a position of a subterm where \(s'\) is replaced by t. We call \(s=t\) the active equation of the rules.
Definition 9
Let \(\varphi \) be a proof and \(\eta \) be some arbitrary inference in \(\varphi \). We say that \(\eta \) goes into the endsequent of \(\varphi \) if the principal formula of \(\eta \) is an ancestor of the endsequent. In this case, \(\eta \) cannot be a cut. \(\eta \) goes into a cut otherwise.
2.2 Expansion Trees
Expansion trees, first introduced in [23], are natural structures representing the instantiated variables for quantified formulas.
These structures record the substitutions for quantifiers in the original formula and the formulas resulting from instantiations. Expansion trees may contain logical connectives as well as the new connective \(+^{t}\), where t is a term. Informally, an expression of the kind \(Q x A(x) +^{t_{1}} E_{1} +^{t_2} \cdots +^{t_n} E_{n}\) is an expansion tree, where \(Q \in \{\forall , \exists \}\) and \(t_{1},\ldots ,t_{n}\) are terms such that this expansion tree represents the result when instantiating the quantified expression QxA(x) with the terms \(t_{1},\ldots ,t_{n}\) to get the structures \(E_i\). \(E_i\) is again an expansion tree representing \(A(t_{i})\) for \(i = 1,\ldots ,n\).
Our definition is a modified one as our proofs are skolemized and we do not have quantifiers with eigenvariable conditions. The definition below takes care that only trees with weak quantifiers are constructed.
Definition 10
 1.
If A is a quantifierfree formula then A is an expansion tree (and a dual expansion tree) for A and \(Sh(A) = A\).
 2.
If E is an expansion tree then \(\lnot E\) is a dual expansion tree and \({ Sh}(\lnot E) = \lnot { Sh}(E)\).
 3.
If E is a dual expansion tree then \(\lnot E\) is an expansion tree and \({ Sh}(\lnot E) = \lnot { Sh}(E)\).
 4.
If \(E_1\) and \(E_2\) are (dual) expansion trees, then \(E_{1} \wedge E_2\), \(E_{1} \vee E_2\) are (dual) expansion trees and \({ Sh}(E_{1} \wedge E_{2}) = { Sh}(E_{1}) \wedge { Sh}(E_{2})\), the same for \(\vee \).
 5.
If \(E_1\) is a dual expansion tree and \(E_2\) is an expansion tree then \(E_1 \rightarrow E_2\) is an expansion tree and \({ Sh}(E_1 \rightarrow E_2) = { Sh}(E_1) \rightarrow { Sh}(E_2)\).
 6.
If \(E_1\) is an expansion tree and \(E_2\) is a dual expansion tree then \(E_1 \rightarrow E_2\) is a dual expansion tree and \({ Sh}(E_1 \rightarrow E_2) = { Sh}(E_1) \rightarrow { Sh}(E_2)\).
 7.
Let A(x) be a formula and \(t_{1},\ldots ,t_{n}\) (\(n \ge 1\)) be a list of terms. Let \(E_{1}, \ldots , E_{n}\) be expansion trees with \({ Sh}(E_{i}) = A(t_{i})\) for \(i = 1,\ldots ,n\); then \(\exists x A(x) +^{t_1} E_{1} +^{t_2} \cdots +^{t_n} E_{n}\) is an expansion tree with \({ Sh}(\exists x A(x) +^{t_1} E_{1} +^{t_2} \cdots +^{t_n} E_{n}) = \exists x A(x)\).
 8.
Let A(x) be a formula and \(t_{1},\ldots ,t_{n}\) (\(n \ge 1\)) be a list of terms. Let \(E_{1}, \ldots , E_{n}\) be dual expansion trees with \({ Sh}(E_{i}) = A(t_{i})\) for \(i = 1,\ldots ,n\); then \(\forall x A(x) +^{t_1} E_{1} +^{t_2} \cdots +^{t_n} E_{n}\) is a dual expansion tree with \({ Sh}(\forall x A(x) +^{t_1} E_{1} +^{t_2} \cdots +^{t_n} E_{n}) = \forall x A(x)\).
Example 1
The function \({ Dp}\) (deep) maps expansion trees (and dual expansion trees) to quantifierfree formulas, their full expansion.
Definition 11
In [23] a notion of expansion proof was defined from expansion trees using the conditions acyclicity and tautology. Acyclicity ensures that there are no cycles between the strong quantifier nodes in the expansion tree. Since our formulas are skolemized and hence do not contain strong quantifiers, we do not need this condition.
Definition 12
(Expansion proof) Let ET be an expansion tree of a formula A without strong quantifiers. Then ET is called an expansion proof of A from a set of axioms \(\mathcal{{A}}\) if \(Sh(ET) = A\) and \(\mathcal{{A}}\models { Dp}(ET)\) (where \(\models \) is the consequence relation in predicate logic with equality).
Expansion proofs encode a proof of validity of the formula they represent. They can be directly translated into sequent calculus, see [23], and the transformation is based on socalled qsequents, which we refer to as sexpansion trees (sequent of expansion trees) in this paper.
Definition 13
It is also possible to read off expansion proofs from sequent calculus proofs. Note that the expansion proof of a proof \(\varphi \) is a sequent of expansion trees, which are defined to be the expansion trees of all formulas in the endsequent of \(\varphi \). An algorithm for the extraction of expansion proofs from sequent calculus proofs is presented in [23] and modified algorithms (dealing with cuts and equality) are presented in [21, 22]. There exist also algorithms for a transformation of resolutiontrees into expansiontrees, see [24].
We will use an algorithm that is briefly described in [21]. In order to show how an expansion proof is extracted from a proof in \(\mathbf {LK}_=\), we first need to define an operation on expansion trees. The Merge operator on expansion trees is defined in [23]. Intuitively, two expansion trees \(T_1\) and \(T_2\) can be merged, if \(Sh(T_1) = Sh(T_2 )\). We give a definition adapted to our concept of expansion tree.
Definition 14

If \(E_1\) is an atom then \(E_2\) is an atom too and \(E_1=E_2\); we define \(\mathrm{Merge}_t(E_1,\)\(E_2) = E_1\).

If \(E_1 = \lnot E'_1\). Then \(E_2 = \lnot E'_2\) for some \(E'_2\). Let \(\mathrm{Merge}_t(E'_1,E'_2) = E'_3\), then \(\mathrm{Merge}_t(E_1,E_2) = \lnot E'_3\).

Let \(E_1 = E_{11} \circ E_{12}\) for \(\circ \in \{\wedge ,\vee ,\rightarrow \}\). Then \(E_2 = E_{21} \circ E_{22}\) for some \(E_{21},E_{22}\). Let \(E'_1 = \mathrm{Merge}_t(E_{11},E_{21})\) and \(E'_2 = \mathrm{Merge}_t(E_{21},E_{22})\). Then \(\mathrm{Merge}_t(E_1,E_2) = E'_1 \circ E'_2\).
 Let \(E_1 = Q x.A(x) +^{t_1} E_{11} + \cdots +^{t_n} E_{1n}\). Then \(E_2\) is of the form \(Q x.A(x) +^{s_1} E_{21} + \cdots +^{s_m} E_{2m}\). Then$$\begin{aligned} \mathrm{Merge}_t(E_1,E_2) = Q x.A(x) +^{t_1} E_{11} + \cdots +^{t_n} E_{1n} +^{s_1} E_{21} + \cdots +^{s_m} E_{2m}. \end{aligned}$$
Example 2
Definition 15
(Normalized sequents) A sequent \(\varGamma \vdash \varDelta \) is called normalized if the multiplicity of all formulas occurring in \(\varGamma \) (\(\varDelta \)) is one, more precisely: if \(\varGamma = A_1,\ldots ,A_n\) and \(\varDelta = B_1,\ldots ,B_m\) then \(A_i \ne A_j\) for \(i \ne j\) (\(i,j \in \{1,\ldots ,n\}\)) and \(B_l \ne B_k\) for \(l \ne k\) (\(l,k \in \{1,\ldots ,m\}\)). Let S be an sexpansion tree then S is called normalized if \({ Seq}(S)\) is normalized.
Remark 1
Note that every LKproof \(\varphi \) of S and every sexpansion tree can be easily transformed into a proof (sexpansion tree) of a normalized sequent: just apply the rules \(c_l,c_r\) to S. In Sect. 4 we will merge sexpansion trees only if they correspond to endsequents of proofs. Therefore restricting the merge to normalized sexpansion proofs does not affect the generality of our approach.
In normalized sequents the multisets become sets which allows us to define some setbased operations on sequents:
Definition 16
Now we are ready to define the merging of normalized sexpansion trees.
Definition 17
Frequently we will write \(\mathrm{merge}_s\{S_i \mid i=1,\ldots ,n\}\) for \(\mathrm{merge}_s(S_1,\ldots ,S_n)\). If no confusion arises we will frequently write \(\mathrm{merge}\) instead of \(\mathrm{merge}_t\) and \(\mathrm{merge}_s\).
Example 3
The extraction of expansion proofs from LKproofs requires quantifierfree cuts. Due to the structure of the CERESmethod (which will be used for a efficient method of extracting expansion proofs) we consider proofs with only atomic cuts.
Definition 18
 1.
\(\varphi \) does not contain strong quantifier inferences.
 2.
All cuts in \(\varphi \) are atomic.
 3.
Equality rules are only applied to atoms.
 4.
The axiom set contains Ax.
Definition 19
(Extraction ofsexpansion trees from proofs in\(\mathbf {LK}_0\)) We define a transformation \(\mathrm{ET}\) which maps proofs in \(\mathbf {LK}_0\) to sexpansion trees. We define the transformation inductively (on the number of inferences in the proof) but the rules for \(\lnot _l , \lnot _r , \vee _l , \vee _{r_1}, \vee _{r_2}, =_{l2}, =_{r2}\) are omitted, the transformation of the these rules being obvious.
base case: \(\varphi \) is an axiom. Then \(\varphi \) is of the form \(A_1,\ldots ,A_n \vdash B_1,\ldots ,B_m\) for atoms \(A_i,B_j\) and so \(\mathrm{ET}(\varphi ) = \varphi \).
Note that, for the rules below, the auxiliary formulas of the rules are atomic.
Proposition 1
The transformation \(\mathrm{ET}\) is sound: if \(\varphi \) is a proof in \(\mathbf {LK}_0\) then \(\mathrm{ET}(\varphi )\) is an sexpansion proof.
Proof

(axiom) Let S be an axiom sequent in \(\mathcal{{A}}\). Then \(S = A_1,\ldots ,A_n \vdash B_1,\ldots ,B_m\) for atoms \(A_i,B_j\). Therefore for \(F_S:\lnot A_1 \vee \cdots \vee \lnot A_n \vee B_1 \vee \cdots B_m\) we have \({ Sh}(F_S) = { Dp}(F_S) = F_S\) and \(\mathcal{{A}}\models { Dp}(F_S)\).
 (\(\wedge _r\)) \(\varphi = \) and \(\mathrm{ET}(\pi _1 ) = \varGamma _{1}^{*}\)\(\vdash \)\(\varDelta _{1}^{*}, A^{*}\) and \(\mathrm{ET}(\pi _2 ) = \varGamma _{2}^{*}\)\(\vdash \)\(\varDelta _{2}^{*}, B^{*}\) are sexpansionproofs. Therefore, \(\lnot \varGamma _{1}^{*} \vee \varDelta _{1}^{*} \vee A^{*}\) and \(\lnot \varGamma _{2}^{*} \vee \varDelta _{2}^{*} \vee B^{*}\) are expansion proofs and \(\mathcal{{A}}\models { Dp}(\lnot \varGamma _{1}^{*} \vee \varDelta _{1}^{*} \vee A^{*})\) and \(\mathcal{{A}}\models { Dp}(\lnot \varGamma _{2}^{*} \vee \varDelta _{2}^{*} \vee B^{*})\). But then \(\mathcal{{A}}\models { Dp}(\lnot \varGamma _{1}^{*} \vee \varDelta _{1}^{*} \vee \lnot \varGamma _{2}^{*} \vee \varDelta _{2}^{*} \vee (A^{*} \wedge B^{*}))\) and \(\lnot \varGamma _{1}^{*} \vee \varDelta _{1}^{*} \vee \lnot \varGamma _{2}^{*} \vee \varDelta _{2}^{*} \vee (A^{*} \wedge B^{*})\) is an expansion proof. Therefore \(\varGamma _{1}^{*}, \varGamma _{2}^{*} \vdash \varDelta _{1}^{*}, \varDelta _{2}^{*}, A^{*} \wedge B^{*}\) (\(= \mathrm{ET}(\varphi )\)) is an sexpansionproof.
 (cut) \(\varphi = \) where \(\mathrm{ET}(\pi _1 ) = \varGamma _{1}^{*} \vdash \varDelta _{1}^{*}, A^{m}\) and \(\mathrm{ET}(\pi _1 ) = A^{n}, \varGamma _{2}^{*} \vdash \varDelta _{2}^{*}\) are sexpansionproofs. Therefore, \(\lnot \varGamma _{1}^{*} \vee \varDelta _{1}^{*} \vee A^{m}\) and \(\lnot A^{n} \vee \lnot \varGamma _{2}^{*} \vee \varDelta _{2}^{*}\) are expansion proofs and \(\mathcal{{A}}\models { Dp}(\lnot \varGamma _{1}^{*} \vee \varDelta _{1}^{*} \vee A^{m})\) and \(\mathcal{{A}}\models { Dp}(\lnot A^{n} \vee \lnot \varGamma _{2}^{*} \vee \varDelta _{2}^{*})\). But then \(\mathcal{{A}}\models { Dp}(\lnot \varGamma _{1}^{*} \vee \varDelta _{1}^{*} \vee \lnot \varGamma _{2}^{*} \vee \varDelta _{2}^{*})\) and \(\lnot \varGamma _{1}^{*} \vee \varDelta _{1}^{*} \vee \lnot \varGamma _{2}^{*} \vee \varDelta _{2}^{*}\) is an expansion proof. Therefore \(\varGamma _{1}^{*}, \varGamma _{2}^{*} \vdash \varDelta _{1}^{*}, \varDelta _{2}^{*}\) (\(= \mathrm{ET}(\varphi )\)) is an sexpansionproof.
 (\(=_{r_1}\)) \(\varphi = \) and \(\mathrm{ET}(\pi _1 ) = \varGamma _{1}^{*} \vdash \varDelta _{1}^{*}, s=t\) and \(\mathrm{ET}(\pi _1 ) = \varGamma _{2}^{*} \vdash \varDelta _{2}^{*}, A[s]_{\varLambda }\) are sexpansion proofs. Therefore \(\lnot \varGamma _{1}^{*} \vee \varDelta _{1}^{*} \vee s=t\) and \(\lnot \varGamma _{2}^{*} \vee \varDelta _{2}^{*} \vee A[s]_{\varLambda }\) are expansion proofs and hence \(\mathcal{{A}}\models { Dp}(\lnot \varGamma _{1}^{*} \vee \varDelta _{1}^{*} \vee s=t)\) and \(\mathcal{{A}}\models { Dp}(\lnot \varGamma _{2}^{*} \vee \varDelta _{2}^{*} \vee A[s]_{\varLambda })\). But then \(\mathcal{{A}}\models { Dp}(\lnot \varGamma _{1}^{*} \vee \varDelta _{1}^{*} \vee \lnot \varGamma _{2}^{*} \vee \varDelta _{2}^{*} \vee A[t]_{\varLambda })\) and \(\lnot \varGamma _{1}^{*} \vee \varDelta _{1}^{*} \vee \lnot \varGamma _{2}^{*} \vee \varDelta _{2}^{*} \vee A[t]_{\varLambda })\) is an expansion proof. Therefore, \(\varGamma _{1}^{*}, \varGamma _{2}^{*} \vdash \varDelta _{1}^{*}, \varDelta _{2}^{*}, A[t]_{\varLambda }\) (\(= \mathrm{ET}(\varphi )\)) is an sexpansionproof. \(\square \)
In case of a prenex endsequent S an expansion proof corresponds to the derivation of S from the midsequent (Herbrand sequent). The essence of Herbrand’s theorem [18] consists of the replacement of quantified formulas by instances of these formulas. This results in a quantifierfree formula which is validityequivalent to the original formula. Of course, the function \({ Dp}\) of an expansion proof corresponds to a Herbrand sequent. But if we consider proofs of prenex endsequents we can extract Herbrand sequents directly, instead of extracting expansion proofs and computing their deep functions. The method for Herbrand sequent extraction in the prenex case is based on collecting instances, and is described in [6, 20].
To illustrate construction of an sexpansion proof from an \(\mathbf {LK}_0\)proof , consider the following simple example.
Example 4
Note that S is not in prenex form. Therefore, extracting the Herbrand sequent by collecting instances is not possible. Instead we compute the expansion proof \(\mathrm{ET}(\varphi )\).
3 The Method CERES
The method CERES [4, 5], is a cutelimination method that is based on resolution. It differs from the reductive stepwise methods a la Gentzen [15] by analyzing the whole proof in a preprocessing step and extracting a formula in clausal form which forms the kernel of the cutelimination method.
CERES in predicate logic with equality roughly works as follows: The structure of a proof \(\varphi \) containing cuts is encoded in an unsatisfiable set of clauses \(\mathrm{CL}(\varphi )\) (the characteristic clause set of \(\varphi \)). A refutation of \(\mathrm{CL}(\varphi )\) by resolution and paramodulation (abbreviated as PRrefutation) then serves as a skeleton for an atomic cut normal form, a new proof which contains at most atomic cuts. The corresponding proof theoretic transformation uses socalled proof projections \(\varphi [C]\) for \(C \in \mathrm{CL}(\varphi )\), which are simple cutfree proofs extracted from \(\varphi \) (proving the endsequent S extended by the atomic sequent C). In [5] it was shown that CERES outperforms reductive methods of cutelimination (a la Gentzen or Tait) in computational complexity: there are infinite sequences of proofs where the computing time of CERES is nonelementarily faster than that of the reductive methods; on the other hand a nonelementary speedup of CERES via reductive methods is shown impossible.
 1.
Extraction of the characteristic clause set \(\mathrm{CL}(\varphi )\).
 2.
Construction of a PRrefutation (see Definition 8) of \(\mathrm{CL}(\varphi )\).
 3.
Extraction of a set of projections \(\pi (C)\) for every \(C \in \mathrm{CL}(\varphi )\).
 4.
Merging of refutation and projections into a proof \(\varphi ^*\) (a CERESnormal form) with only atomic cuts.
Example 5
The set of axioms \(Ax_s\) is defined as \(Ax_s = Ax \cup \{ \vdash f^2 z = gz\}\).
Intuitively, the clause set extraction consists in collecting all atomic ancestors of the cuts which occur in the axioms of the proof. The clauses are formed depending on how these atoms are related via binary inferences in the proof.
Definition 20

If \(\nu \) is an axiom, then \(\mathrm{CL}(\nu )\) contains the subsequent of \(\nu \) composed only of cutancestors.

If \(\nu \) is the result of the application of a unary rule on a sequent \(\mu \), then \(\mathrm{CL}(\nu ) = \mathrm{CL}(\mu )\)
 If \(\nu \) is the result of the application of a binary rule on sequents \(\mu _1\) and \(\mu _2\), then we distinguish two cases:

If the rule is applied to ancestors of the cut formula, then \(\mathrm{CL}(\nu ) = \mathrm{CL}(\mu _1) \cup \mathrm{CL}(\mu _2)\)

If the rule is applied to ancestors of the endsequent, then \(\mathrm{CL}(\nu ) = \mathrm{CL}(\mu _1) \times \mathrm{CL}(\mu _2)\)

If \(\nu _0\) is the root node \(\mathrm{CL}(\nu _0)\) is called the characteristic clause set of \(\varphi \).
Example 6
The characteristic clause set of our proof \(\varphi \) from Example 5 is constructed as follows:
The next step is to obtain a resolution refutation of \(CL(\varphi )\). It is thus important to show that this set is always refutable.
Theorem 1
Let \(\varphi \) be a proof of a skolemized endsequent. Then the characteristic clause set \(\mathrm{CL}(\varphi )\) is refutable by resolution and paramodulation.
Example 7
Definition 21
Each clause in the clause set will have a projection associated with it. A projection of a clause C is a derivation built from \(\varphi \) by taking the axioms in which the atoms of C occur and all the inferences that operate on endsequent ancestors. As a result, the endsequent of a projection will be the endsequent of \(\varphi \) extended by the atoms of C.
Definition 22
 (a)
\(\nu \) is a leaf in \(\varphi \). Then the sequent at \(\nu \) is an axiom and we define \(p(\nu ) = \{\nu \}\). The clause part of \(\nu \) is the subsequent \(\mathrm{CL}(\nu )\).
 (b)\(\nu \) is the conclusion of a unary rule \(\xi \) with premise \(\mu \).
 (b1)The principal formula of \(\xi \) is an ancestor of a cut. Then \( \varphi .\nu \) is of the form We define \(p(\nu ) = p(\mu )\).
 (b2)The principal formula of \(\xi \) is an ancestor of the endsequent. Then \(\varphi .\nu \) is of the form Let \(\psi \in p(\mu )\) be a proof of \(C,\varGamma _e \vdash \varDelta _e,D\) where \(C \vdash D\) is the clause part of \(\psi \). Then \(\psi ' \in p(\nu )\) for \(\psi '=\) and \(C \vdash D\) is the clause part of \(\psi '\).
 (b1)
 (c)\(\nu \) is the conclusion of a binary rule \(\xi \) with premises \(\mu _1,\mu _2\).
 (c1)The auxiliary formulas of \(\xi \) are ancestors of a cut. Then \(\varphi .\nu \) is of the form Let \(\psi \in p(\mu _1)\) such that \(\psi \) is a proof of \(C,\varGamma _e \vdash \varDelta _e,D\) where \(C \vdash D\) is the clause part of \(\psi \). Then \(\psi ' \in p(\nu )\) for \(\psi '=\) and \(C \vdash D\) is the clause part of \(\psi '\).Let \(\psi \in p(\mu _2)\) such that \(\psi \) is a proof of \(E,\varPi _e \vdash \varLambda _e,F\) where \(E \vdash F\) is the clause part of \(\psi \). Then \(\psi ' \in p(\nu )\) for \(\psi '=\) and \(E \vdash F\) is the clause part of \(\psi '\).
 (c2)The auxiliary formulas of \(\xi \) are ancestors of the endsequent. Then \(\varphi .\nu \) is of the form Let \(\psi _1 \in p(\mu _1)\) such that \(\psi _1\) is a proof of \(C,\varGamma _e \vdash \varDelta _e,D\) and \(C \vdash D\) is the clause part of \(\psi _1\); likewise let \(\psi _2 \in p(\mu _2)\) such that \(\psi _2\) is a proof of \(E,\varPi _e \vdash \varLambda _e,F\) and \(E \vdash F\) is the clause part of \(\psi _2\). Then \(\psi \in p(\nu )\) for \(\psi =\) and the clause part of \(\psi \) is \(C,E \vdash D,F\).
 (c1)
Example 8
Given the projections and a grounded PR refutation, it is possible to build a proof \(\hat{\varphi }\) of \(\varGamma \vdash \varDelta \) with only atomic cuts.
If we apply all most general unifiers in the PR proof \(\gamma \) we obtain a proof in \(\mathbf {LK}_=\) (in fact only contractions, cut and paramodulation remain). If \(\gamma \sigma \) is such a proof and we apply a substitution replacing all variables by a constant symbol we obtain a ground PR refutation. Note that after applying the most general unifiers to \(\gamma \) we obtain a derivation in \(\mathbf {LK}_=\) where the resolution rule becomes a cut rule. For a formal definition see [6].
Example 9
\(\gamma '\) can be used as a skeleton of a proof \(\varphi ^*\) with only atomic cuts of the original endsequent S. \(\varphi ^*\) is called a CERESnormal form of the original proof \(\varphi \). Below we give a formal definition. First we define a type of top normal form defined by a PRdeduction.
Definition 23

\(\varrho = C_i \vdash D_i\): then \(\varTheta (\varrho ,\mathcal{{C}},\Phi ) = \varphi _i\) and \({ top}(\varTheta (\varrho ,\mathcal{{C}},\Phi )) = \{\varphi _i\}\).
 The last inference in \(\varrho \) is R. Then \(\varrho \) is of the form Let us assume that Then we define \(\varTheta (\varrho ,\mathcal{{C}},\Phi ) =\) and \({ top}(\varTheta (\varrho ,\mathcal{{C}},\Phi )) = { top}(\varTheta (\varrho _1,\mathcal{{C}},\Phi )) \cup { top}(\varTheta (\varrho _1,\mathcal{{C}},\Phi ))\).
 The last inference in \(\varrho \) is a paramodulation rule. We consider only the case \(=_{r1}\); for the other rules the construction is analogous. Then \(\varrho \) is of the form Let us assume that Then we define \(\varTheta (\varrho ,\mathcal{{C}},\Phi ) =\) and \({ top}(\varTheta (\varrho ,\mathcal{{C}},\Phi )) = { top}(\varTheta (\varrho _1,\mathcal{{C}},\Phi )) \cup { top}(\varTheta (\varrho _1,\mathcal{{C}},\Phi ))\).
Remark 2
The function \({ top}\) collects all cutfree subproofs in a top normal form which occur at the top and thus belong to \(\Phi \).
Definition 24
(CERESnormal form) Let \(\varphi \) be an \(\mathbf {LK}_=\) proof of a skolemized sequent S. Let \(\varrho \) be a grounded PRrefutation of \(\mathrm{CL}(\varphi )\), \(\mathcal{{C}}\) be the set of all ground instances of clauses in \(\mathrm{CL}(\varphi )\) appearing at the leaves of \(\varrho \) and \(\Phi \) be the set of all grounded projections. Then the proof \(\varTheta (\varrho ,\mathcal{{C}},\Phi )\) is called a CERES normal form of \(\varrho \). As \(\varrho \) is a refutation \(\varTheta (\varrho ,\mathcal{{C}},\Phi )\) is a proof of S with only atomic cuts.
Remark 3
Note that not all top normal forms are CERES normal forms as the set of cutfree proofs \(\Phi \) need not be projections.
Example 10
4 Extraction of Expansion Trees from Projections
The extraction of expansion proofs is usually performed after the construction of a proof in top normal form. However, only the logical parts of the proof play a role in the construction of expansion trees. These logical parts can be identified as the cutfree subproofs after removal of all cutancestors. Note that no cutancestor in such a subproof is principal formula of an inference; we identify such subsequents as passive subsequent.
Definition 25
(Passive subsequent) Let \(\varphi \) be a cutfree proof of \(S: C, \varGamma \vdash \varDelta , D\) such that \(C \vdash D\) is a clause. The subsequent \(C \vdash D\) of S is called passive in \(\varphi \) if no ancestor of \(C \vdash D\) in \(\varphi \) contains a formula which is principal formula of an inference.
Note that the passive subsequents are just the clauses used to define a top normal form. Examples of proofs with passive clause parts are proof projections in CERES:
Proposition 2
Let \(\psi \) be a cutfree proof of \(C',\varGamma \vdash \varDelta ,D'\) which is an instance of a proof projection \(\varphi [C \vdash D]\) in CERES. Then \(C' \vdash D'\) is passive in \(\psi \).
Proof
Definition 26

If \(\varphi \) is an axiom then \(\varphi = C, C' \vdash D, D'\) (note that the whole sequent is passive in \(\varphi \)). We define \(\varphi \backslash (C \vdash D) = C' \vdash D'\).
 Let \(\varphi = \) where \(C \vdash D\) is passive in \(\varphi \). Then, by definition of passive subclauses, \(\varphi '\) is a proof of \(C, \varGamma ' \vdash \varDelta ', D\) for some \(\varGamma '\) and \(\varDelta '\). Indeed, the subclause \(C \vdash D\) does not contain a formula which is principal formula of an inference. By induction we have a proof \(\varphi ' \backslash (C \vdash D)\) of \(\varGamma ' \vdash \varDelta '\) (note that \(C \vdash D\) is also passive in \(\varphi '\)) and we define \(\varphi \backslash (C \vdash D) =\)
 Let \(\varphi =\) where \(C \vdash D\) is passive in \(\varphi \). As C, D are not principal formulas of an inference we get that \(S_1 = C_1 , \varGamma _1 \vdash \varDelta _1 , D_1\), \(S_2 = C_2 , \varGamma _2 \vdash \varDelta _2 , D_2\), s.t. \(C_1 , C_2 \vdash D_1, D_2 = C \vdash D\) and \(C_1 \vdash D_1\) is passive in \(\varphi _1\), \(C_2 \vdash D_2\) is passive in \(\varphi _2\).By induction we have a proof \(\varphi _1 \backslash (C_1 \vdash D_1 )\) of \(\varGamma _1 \vdash \varDelta _1\) and a proof \(\varphi _2 \backslash (C_2 \vdash D_2)\) of \(\varGamma _2 \vdash \varDelta _2\). Then we obtain \(\varphi \backslash (C \vdash D) =\)
The function \(\mathrm{logical}(\varphi )\) for a proof in top normal form takes the cutfree proofs on top and “subtracts” from them all ancestors of passive clauses.
Definition 27
(logical\((\varphi )\)) Let \(\varphi :\varTheta (\rho ,\mathcal{{C}},\Phi )\) be a proof in top normal form s.t. \(\mathcal{{C}}= \{C_1 \vdash D_1,\ldots , C_n \vdash D_n\}\) and \(\Phi = \{\varphi _1,\ldots ,\varphi _n\}\) such that \(\varphi _i\) is a cutfree proof of \(C_i,\varGamma \vdash \varDelta ,D_i\). Assume that for all \(i=1,\ldots ,n\)\(C_i \vdash D_i\) is passive in \(\varphi _i\). For every \(\psi \in { top}(\varphi )\) and \(\psi = \varphi _i\) we define \(\psi ' = \varphi _i \backslash (C_i \vdash D_i)\) and \(\mathrm{logical}(\varphi ) = \{\psi ' \mid \psi \in { top}(\varphi )\}\).
Below we define an expansion tree \(\hat{E}(\varphi )\) which is defined by merging the expansion trees of \(\mathrm{logical}(\varphi )\). This structure will be the key for the development of an efficient algorithm for extracting expansion trees from CERES normal forms.
Definition 28
Theorem 2
Let \(\varphi :\ \varTheta (\varrho ,\mathcal{{C}},\Phi )\) be a proof of a skolemized and normalized sequent \(C,\varGamma \vdash \varDelta ,D\) in top normal form such that \(\mathcal{{C}}= \{C_1 \vdash D_1,\ldots , C_n \vdash D_n\}\) and \(\Phi = \{\varphi _1,\ldots ,\varphi _n\}\), where \(\varphi _i\) is a cutfree proof of \(C_i,\varGamma \vdash \varDelta ,D_i\). Assume that for all \(i=1,\ldots ,n\)\(C_i \vdash D_i\) is passive in \(\varphi _i\). Then \(\mathrm{ET}(\varphi ) = \hat{E}(\varphi ) \circ (C \vdash D)\).
Proof
By induction on the number of nodes in \(\varrho \).
Corollary 1
Let \(\varphi :\varTheta (\varrho ,\mathcal{{C}},\Phi )\) be a proof of a skolemized and normalized sequent \(\varGamma \vdash \varDelta \) in top normal form s.t. \(\mathcal{{C}}= \{C_1 \vdash D_1,\ldots , C_n \vdash D_n\}\) and \(\Phi = \{\varphi _1,\ldots ,\varphi _n\}\) such that \(\varphi _i\) is a cutfree proof of \(C_i,\varGamma \vdash \varDelta ,D_i\). Assume that for all \(i=1,\ldots ,n\)\(C_i \vdash D_i\) is passive in \(\varphi _i\). Then \(\mathrm{ET}(\varphi ) = \hat{E}(\varphi )\).
Proof
Immediate by Theorem 2: just define \(C \vdash D\) as the empty sequent. \(\square \)
Corollary 2
Let \(\varphi \) be an \(\mathbf {LK}_=\) proof of a skolemized and normalized sequent S. Let \(\varphi ^*:\varTheta (\varrho ,\mathcal{{C}},\Phi )\) be a CERES normal form of \(\varphi \) such that \(\varrho \) is a ground PRrefutation of \(\mathcal{{C}}\), the set of all ground instances of clauses in \(\mathrm{CL}(\varphi )\), and \(\Phi \) is the set of all grounded projections. Then \(\mathrm{ET}(\varphi ) = \hat{E}(\varphi )\).
Proof
Let \(\psi \) be a cutfree proof of \(C,\varGamma \vdash \varDelta ,D\) which is an instance of a projection of \(\varphi \). By Proposition 2\(C \vdash D\) is passive in \(\psi \). As CERES normal forms are top normal forms all conditions of Corollary 1 are fulfilled. \(\square \)
Definition 29
Theorem 3
Let \(\varphi \) be a proof of a skolemized, closed and normalized endsequent and \(\varphi ^{*}\) the CERES normal form based on a ground refutation R of \(\mathrm{CL}(\varphi )\). Then \(\mathrm{ET}(\varphi ^{*}) = T(\varphi , R)\).
Proof
Note that Theorem 3 also holds for the \({ Dp}\) function of expansion proofs, i.e. \({ Dp}(\mathrm{ET}(\varphi ^{*})) = { Dp}(T(\varphi ,R))\).
Corollary 3
Let \(\varphi \) be a proof of a skolemized,closed and normalized sequent S and R be a refutation of \(\mathrm{CL}(\varphi )\). Then \(T(\varphi ,R)\) is an expansion proof of S.
Instead of computing all \(\varphi [C_{j}] \sigma ^{j}_{i}\) (obtained from the ACNF \(\varphi ^*\)) the algorithm EXP\(_{new}\) computes the \(\varphi [C_{j}]\) and extracts \(\mathrm{ET}(\varphi ^{}[C_{j}]) \equiv T_j\), which is a partial expansion proof, then constructs \(\times _{\sigma \in \varSigma (C_j)}T_j\sigma \) for all j and merges them. Example 11 illustrates the main features of the method.
Example 11
5 Complexity
In this section we prove that the algorithm \(\mathrm{EXP}_{new}\) outperforms the old algorithm \(\mathrm{EXP}\). In particular we prove that the complexity of \(\mathrm{EXP}_{new}\) is always better or equal to that of \(\mathrm{EXP}\). Then we define an infinite sequence of LKproofs \(\varphi _n\) where the complexity of \(\mathrm{EXP}\) is cubic in n while that of \(\mathrm{EXP}_{new}\) is only quadratic. This implies that the computational complexity of \(\mathrm{EXP}\) cannot be linearly bounded by that of \(\mathrm{EXP}_{new}\). Our complexity measure will be the maximal logical complexity of objects constructed by the algorithms.
Definition 30
To improve legibility we write \(\Vert F \Vert \) instead of \(\Vert F \Vert _f\) (with the exception of cases where the precise notation is essential) and use the measure \(\Vert \Vert \) also for sequents, proofs and clause sets.
Definition 31
Definition 32
(Size of an\(\mathbf {LK}_=\)proof) Let \(\varphi \) be an \(\mathbf {LK}_=\)proof. If \(\varphi \) is an axiom then \(\varphi \) consists of just one node labelled by a sequent S; here we define \(\Vert \varphi \Vert = \Vert S \Vert \).
If \(\varphi \) is not an axiom then the endsequent is a conclusion of a unary or of a binary rule. So we distinguish two cases:
Definition 33
Definition 34
In our algorithms \(\mathrm{EXP}\) and \(\mathrm{EXP}_{new}\) we do not only construct sequents, formulas and proofs, but also sets of clauses (which are finite sets of atomic sequents). If \(\mathcal{{C}}= \{C_1,\ldots ,C_n\}\) we define \(\Vert \mathcal{{C}} \Vert = \Vert C_1 \Vert + \ldots + \Vert C_n \Vert \). We call the objects produced by a proof transformation expressions.
Definition 35
(Expression) An expression is a formula, a sequent, a proof or a set of clauses.
Now we consider computations as sequences of expressions which are generated by an algorithmic proof transformation. So let A be an algorithm and \(\varphi \) be a proof serving as input to A. Then \(E_A(\varphi )\) is the sequence of all expressions generated by A on input \(\varphi \). Below we define a complexity function induced by A given by the maximal expression generated by A:
Definition 36
Theorem 4
\(C_{\mathrm{EXP}_{new}}(\varphi ) \le C_{\mathrm{EXP}}(\varphi )\) for all proofs \(\varphi \) in \(\mathbf {LK}_=\).
Proof
sketch: the first 4 steps of \(\mathrm{EXP}\) and \(\mathrm{EXP}_{new}\) are identical. The sum of the sizes of the expansion trees generated by \(\mathrm{EXP}_{new}\) is smaller or equal to the size of the CERES normal form generated by \(\mathrm{EXP}\). \(\square \)
Theorem 5
 (a)
\(C_{\mathrm{EXP}_{new}}(\varphi ) \le C_{\mathrm{EXP}}(\varphi )\) for all proofs \(\varphi \) in \(\mathbf {LK}_=\).
 (b)
There exists no constant d such that for all proofs \(\varphi \) in \(\mathbf {LK}_=\): \(C_{\mathrm{EXP}}(\varphi ) \le d*C_{\mathrm{EXP}_{new}}(\varphi )\).
Proof
Remark 4
6 Implementation and Experiments in Gapt
The algorithm \(\mathrm{EXP}_{new}\) is implemented in the Gaptsystem^{1} [12], which is a framework for implementing proof transformations written in the programming language Scala. Initially it was developed for the method CERES, but has been extended to other proof transformation algorithms (note that the methods described in this paper are available from version 2.5 on). This section explains how to run the described algorithm, followed by a discussion of results obtained by experiments with formal proofs.
Comparison of the three different methods for the extraction of expansion proofs: based on Gentzen’s reductive method, methods \(\mathrm{EXP}\) and \(\mathrm{EXP}_{new}\)
Proof  Reductive (ms)  \(\mathrm{EXP}\) (ms)  \(\mathrm{EXP}_{new}\) (ms) 

fol1.proof  68  27  17 
fol2.proof  53  30  21 
Pi2Pigeonhole.proof  1350  680  170 
Pi3Pigeonhole.proof  842  552  181 
poset.proof.cycleImpliesEqual3  1621  360  147 
poset.proof.cycleImpliesEqual4  6877  850  250 
CutIntroduction( LinearExampleProof( 4 ))  138  70  26 
CutIntroduction( LinearExampleProof( 8 ))  294  42  31 
CutIntroduction( LinearExampleProof( 10 ))  577  32  28 
CutIntroduction( LinearExampleProof( 15 ))  890  36  26 
CutIntroduction( LinearExampleProof( 18 ))  1329  54  25 
CutIntroduction( LinearExampleProof( 19 ))  1205  85  51 
CutIntroduction( LinearEqExampleProof( 2 ))  190  72  32 
CutIntroduction( LinearEqExampleProof( 5 ))  924  98  34 
CutIntroduction( LinearEqExampleProof( 10 ))  2640  113  43 
CutIntroduction( LinearEqExampleProof( 15 ))  6510  134  46 
CutIntroduction( LinearEqExampleProof( 16 ))  10,875  163  53 
CutIntroduction( LinearEqExampleProof( 18 ))  12,423  455  97 
CutIntroduction( FactorialFunctionEqualityExampleProof( 3 ))  3525  500  360 
CutIntroduction( FactorialFunctionEqualityExampleProof( 4 ))  8473  795  590 
CutIntroduction( FactorialFunctionEqualityExampleProof( 5 ))  20,006  1430  930 
We want to remark that the obtained results depend to a great degree on the operating system in use. Therefore, it is possible that the obtained results fluctuate. Nevertheless, the speedup of the method \(\mathrm{EXP}_{new}\) compared to \(\mathrm{EXP}\) is given and can be clearly recognized.
7 Conclusion
In the analysis of mathematical proofs it is usually more important to gain essential mathematical information from proofs (which typically lies in the terms), than to traverse complicated and long propositional proofs. This was our motivation for avoiding the construction of an atomic cut normal form and focus on the quantifier inferences in the final proof (represented by a Herbrand sequent or by an expansion proof). In the original CERES method, the extraction of expansion proofs is performed after the final result of CERES (a proof containing at most atomic cuts, the CERES normal form) is obtained. We first analysed ACNFs and defined a new version of ACNF where the cutinferences are the last inferences in the proof and only contractions and weakenings occur between them. Proofs of that structure are said to be in top normal form. We have shown that the logical parts of proofs in top normal form suffice to compute expansion proofs. Since proofs in CERES normal form are in top normal form, its logical parts (the proofprojections) suffice to extract expansion proofs. First we refute the characteristic clause set \(\mathrm{CL}(\varphi )\) and, from a corresponding ground refutation R, obtain a set of ground substitutions \(\varSigma (C)\) for all clauses \(C \in \mathrm{CL}(\varphi )\) that occur in R. We construct the general CERES projections \(\varphi [C]\), corresponding to the clauses \(C \in \mathrm{CL}(\varphi )\) that occur in R, and extract partial expansion proofs T[C]. Finally we instantiate the partial expansion proofs T[C] by substitutions in \(\varSigma (C)\) and combine the resulting expansion proofs by merging; the result is an expansion proof of the endsequent (which coincides with the expansion proof of the CERES normal form). Using this method we avoided the construction of an atomic cut normal form.
We also obtain an improvement in asymptotic complexity. Note that we do not compute n instances of the projection \(\varphi [C_{i}]\) with the corresponding substitution \(\sigma ^{i}_{j}\) like it is needed for the construction of the CERES normal form, instead we compute the projection once, remove clause parts and then instantiate the corresponding expansion proof (Herbrand sequent) with the corresponding substitutions. The described algorithm is implemented in the Gapt system and we describe how to use it. We show that even for small and not complicated proofs a speedup in time is obtained by the new algorithm.
Since we investigated this method for firstorder logic only, further work has to deal with a generalization of this method to higherorder logic. In higherorder logic we have to deal with a different CERESmethod (CERES\(^{\omega }\) [19]) and a resolution calculus for higherorder logic. The extraction of expansion proofs from CERESprojections in the higherorder case is a nontrivial task, since CERES\(^{\omega }\) as well as the resolution calculus for higherorder logic are much more complicated than in the firstorder case. If and how the method can be generalized to the higherorder case, future work will tell.
Footnotes
Notes
Acknowledgements
Open access funding provided by Austrian Science Fund (FWF).
Funding
Funding was provided by Austrian Science Fund (Grant No. I2671N35).
References
 1.Baaz, M., Hetzl, S., Leitsch, A., Richter, C., Spohr, H.: Cutelimination: experiments with ceres. In: International Conference on Logic for Programming Artificial Intelligence and Reasoning, pp. 481–495. Springer (2005)Google Scholar
 2.Baaz, M., Hetzl, S., Leitsch, A., Richter, C., Spohr, H.: Proof transformation by CERES. In: Proceedingsof 5th International Conference on Mathematical Knowledge Management, MKM 2006, Wokingham, UK, August 11–12, 2006, pp. 82–93 (2006)Google Scholar
 3.Baaz, M., Leitsch, A.: On skolemization and proof complexity. Fundam. Inf. 20(4), 353–379 (1994)MathSciNetzbMATHGoogle Scholar
 4.Baaz, M., Leitsch, A.: Cutelimination and redundancyelimination by resolution. J. Symb. Comput. 29(2), 149–177 (2000)MathSciNetCrossRefzbMATHGoogle Scholar
 5.Baaz, M., Leitsch, A.: Towards a clausal analysis of cutelimination. J. Symb. Comput. 41(3), 381–410 (2006)MathSciNetCrossRefzbMATHGoogle Scholar
 6.Baaz, M., Leitsch, A.: Methods of CutElimination, vol. 34. Springer, Berlin (2011)zbMATHGoogle Scholar
 7.Baaz, M., Leitsch, A.: Cutelimination: syntax and semantics. Stud. Log. 102(6), 1217–1244 (2014)MathSciNetCrossRefzbMATHGoogle Scholar
 8.Berger, U., Berghofer, S., Letouzey, P., Schwichtenberg, H.: Program extraction from normalization proofs. Stud. Log. 82(1), 25–49 (2006)MathSciNetCrossRefzbMATHGoogle Scholar
 9.Berger, U., Buchholz, W., Schwichtenberg, H.: Refined program extraction from classical proofs. Ann. Pure Appl. Log. 114(1), 3–25 (2002)MathSciNetCrossRefzbMATHGoogle Scholar
 10.Buss, S.R.: On Herbrand’s theorem. In: Leivant, D. (ed.) Logical and Computational Complexity. Selected Papers. Logic and Computational Complexity, International Workshop LCC ’94, Indianapolis, Indiana, USA, 13–16 October 1994. Lecture Notes in Computer Science, vol. 960, pp. 195–209. Springer (1994)Google Scholar
 11.Dunchev, C., Leitsch, A., Libal, T., Riener, M., Rukhaia, M., Weller, D., WoltzenlogelPaleo, B.: Prooftool: a gui for the gapt framework. arXiv preprint arXiv:1307.1942 (2013)
 12.Ebner, G., Hetzl, S., Reis, G., Riener, M., Wolfsteiner, S., Zivota, S.: System description: Gapt 2.0. In: International Joint Conference on Automated Reasoning, pp. 293–301. Springer (2016)Google Scholar
 13.Fürstenberg, H.: On the infinitude of the primes. Am. Math. Mon. 62, 353 (1955)MathSciNetCrossRefzbMATHGoogle Scholar
 14.Fürstenberg, H.: Topological dynamics and combinatorial number theory. J. Anal. Math. 34(1), 61–85 (1978)MathSciNetCrossRefzbMATHGoogle Scholar
 15.Gentzen, G.: Untersuchungen über das logische Schließen. Math. Z. 39, 176–210, 405–431 (1934–35)Google Scholar
 16.Girard, J.Y.: Proof Theory and Logical Complexity, vol. 1. Biblopolis, Napoli (1987)zbMATHGoogle Scholar
 17.Gödel, K.: Über eine bisher noch nicht benützte Erweiterung des finiten Standpunktes. Dialectica 12(3–4), 280–287 (1958)MathSciNetCrossRefzbMATHGoogle Scholar
 18.Herbrand, J.: Recherches sur la théorie de la démonstration. PhD thesis, Université de Paris (1930)Google Scholar
 19.Hetzl, S., Leitsch, A., Weller, D.: Ceres in higherorder logic. Ann. Pure Appl. Log. 162(12), 1001–1034 (2011)MathSciNetCrossRefzbMATHGoogle Scholar
 20.Hetzl, S., Leitsch, A., Weller, D., Paleo, B.W.: Herbrand sequent extraction. In: Autexier, S., Campbell, J.A., Rubio, J., Sorge, V., Suzuki, M., Wiedijk, F. (eds.) Proceedings of Intelligent Computer Mathematics, 9th International Conference, AISC 2008, 15th Symposium, Calculemus 2008, 7th International Conference, MKM 2008, Birmingham, UK, 28 July1 August 2008. Lecture Notes in Computer Science, vol. 5144, pp. 462–477. Springer (2008)Google Scholar
 21.Hetzl, S., Libal, T., Riener, M., Rukhaia, M.: Understanding resolution proofs through Herbrands theorem. In: Galmiche, D., LarcheyWendling, D. (eds.) Proceedings of Automated Reasoning with Analytic Tableaux and Related Methods  22nd International Conference, TABLEAUX 2013, Nancy, France, 16–19 September 2013. Lecture Notes in Computer Science, vol. 8123, pp. 157–171. Springer (2013)Google Scholar
 22.Hetzl, S., Weller, D.: Expansion trees with cut. arXiv preprint arXiv:1308.0428 (2013)
 23.Miller, D.A.: A compact representation of proofs. Stud. Log. 46(4), 347–370 (1987)MathSciNetCrossRefzbMATHGoogle Scholar
 24.Pfenning, F.: Analytic and nonanalytic proofs. In: 7th International Conference on Automated Deduction, pp. 394–413. Springer (1984)Google Scholar
 25.Reis, G.: Importing SMT and connection proofs as expansion trees. arXiv preprint arXiv:1507.08715 (2015)
 26.Urban, C.: Classical logic and computation. PhD thesis, University of Cambridge (2000)Google Scholar
 27.Van der Waerden, B.L.: Beweis einer Baudetschen Vermutung. Nieuw Arch. Wiskd. 15(2), 212–216 (1927)zbMATHGoogle Scholar
Copyright information
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.