The so-called weak form of Hilbert's Nullstellensatz says that a system of algebraic equations over a field, Qj(x) = 0, does not have a solution in the algebraic closure if and only if 1 is in the ideal generated by the polynomials (?,(*) • We shall prove a lower bound on the degrees of polynomials P,(x) such that £, P,(x)Qt(x) = 1. This result has the following application. The modular counting principle states that no finite set whose cardinality is not divisible by q can be partitioned into ^-element classes. For each fixed cardinality N, this principle can be expressed as a propositional formula Count^fo,...) with underlying variables xe, where e ranges over <7-element subsets of N. Ajtai [4] proved recently that, whenever p,q are two different primes, the propositional formulas Count $ n+I do not have polynomial size, constant-depth Frege proofs from substitution instances of Count/?, where m^O (modp). We give a new proof of this theorem based on the lower bound for Hilbert's Nullstellensatz. Furthermore our technique enables us to extend the independence results for counting principles to composite numbers p and q. This improved lower bound together with new upper bounds yield an exact characterization of when Count, can be proved efficiently from Countp, for all values of p and q.

The combinatorial matching principle states that there is no perfect matching on an odd number of vertices. This principle generalizes the pigeonhole principle, which states that for a fixed bi-partition of the vertices, there is no perfect matching between them. Therefore, it follows from recent lower bounds for the pigeonhole principle that the matching principle requires exponential-size bounded-depth Frege proofs. Ajtai [Ajt90] previously showed that the matching principle does not have polynomial-size bounded-depth Frege proofs even with the pigeonhole principle as an axiom schema. His proof utilizes nonstandard model theory and is nonconstructive. We improve Ajtai's lower bound from barely superpolynomial to exponential and eliminate the nonstandard model theory. Our lower bound is also related to the inherent complexity of particular search classes (see [Pap91]). In particular, oracle separations between the complexity classes PPA and PPAD, and between PPA and PPP also follow f...

In this paper we consider solutions to the dictionary problem on AC RAMs, i.e.

This paper serves two purposes. Firstly, it is an elementary introduction to the theory of P-completeness --- the branch of complexity theory that focuses on identifying the problems in the class P that are "hardest," in the sense that they appear to lack highly parallel solutions. That is, they do not have parallel solutions using time polynomial in the logarithm of the problem size and a polynomial number of processors unless all problem in P have such solutions, or equivalently, unless P = NC . Secondly, this paper is a reference work of P-complete problems. We present a compilation of the known P-complete problems, including several unpublished or new P-completeness results, and many open problems. This is a preliminary version, mainly containing the problem list. The latest version of this document is available in electronic form by anonymous ftp from thorhild.cs.ualberta.ca (129.128.4.53) as either a compressed dvi file (TR91-11.dvi.Z) or as a compressed postscript fi...

We prove that the one-dimensional sandpile prediction problem is . The previously best known upper bound on the AC -scale was AC . We also prove that it is not in AC for any constant > 0.

A basic aim of logic is to consider what axioms are used in proving various theorems of mathematics. This thesis will be concerned with such issues applied to a particular area of mathematics: combinatorics. We will consider two widely known groups of proof methods in combinatorics, namely, probabilistic methods and methods using linear algebra. We will consider certain applications of such methods, both of which are significant to Ramsey theory. The systems we choose to work in are various theories of bounded arithmetic. For the probabilistic method, the key point is that we use the weak pigeonhole principle to simulate the probabilistic reasoning. We formalize various applications of the ordinary probabilistic method and linearity of expectations, making partial progress on the Local Lemma. In the case of linearity of expectations, we show how to eliminate the weak pigeonhole principle by simulating the derandomization technique of “conditional probabilities.” We consider linear algebra methods applied to various set system theorems. We formalize some theorems using a linear algebra principle as an extra axiom. We also show how weaker results can be attained by giving alternative proofs that avoid linear algebra, and thus also avoid the extra axiom. We formalize upper and lower Ramsey bounds. For the lower bounds, both the probabilistic methods and the linear algebra methods are used. We provide a stratification of the various Ramsey lower bounds, showing that stronger bounds can be proved in stronger theories. A natural question is whether or not the axioms used are necessary. We provide “reversals” in a few cases, showing that the principle used to prove the theorem is in fact a consequence of the theorem (over some base theory). Thus this work can be seen as a (humble) beginning in the direction of developing the Reverse Mathematics of finite combinatorics.

In this paper we consider solutions to the static dictionary problem ���� � on RAMs, i.e. random access machines where the only restriction on the finite instruction set is that all computational instructions are ���� � in. Our main result is a tight upper and lower bound ���� � ���©��������������������� of on the time for answering membership queries in a set of � size when reasonable space is used for the data structure storing the set; the upper bound can be obtained using space ������ � �� � ���� �. Several variations of this result are also obtained. Among others, we show a tradeoff between time and circuit depth under the unit-cost assumption: any RAM instruction set which permits a linear space, constant query time solution to the static dictionary problem must have an instruction of depth �������©���������������©���� � , where � is the word size of the machine (and ���© � the size of the universe). This matches the depth of multiplication and integer division, used in the perfect hashing scheme by Fredman, Komlós and Szemerédi. ���� � space, and the lower bound holds even if we allow 1

This paper introduces algebraic proof systems for the propositional calculus. We present new results concerning the relative efficiency of these systems, and also survey what is currently known. Many open problems are presented. 1 Introduction A fundamental problem in logic and computer science is understanding the efficiency of propositional proof systems. It has been known for a long time that NP = coNP if and only if there exists an efficient propositional proof system, but despite 25 years of research, this problem is still not resolved. (See [46] for an excellent survey of this area.) The intention of the present article is to introduce a new algebraic approach to this problem. Our proof systems are simpler than classical proof systems, and purely algebraic. It is our hope that by studying proof complexity in this light, that new upper and lower bound techniques may emerge. The use of the Nullstellensatz for propositional refutations may have been first suggested in a paper by Lo...