On the Automatizability of Resolution

and Related Propositional Proof Systems?

Albert Atserias and Mara Luisa Bonet
Departament de Llenguatges i Sistemes Informatics
Universitat Politecnica de Catalunya, Barcelona
C/Jordi Girona Salgado, 1-3, Edif. C6.
08034 Barcelona - Spain.
Tel: +34 93 401 69 94
Fax: +34 93 401 70 14
fatserias,[email protected]

Abstract. We analyse the possibility that a system that simulates Res-

olution is automatizable. We call this notion "weak automatizability".
We prove that Resolution is weakly automatizable if and only if Res(2)
has feasible interpolation. In order to prove this theorem, we show that
Res(2) has polynomial-size proofs of the re ection principle of Resolution
(and of any Res(k)), which is a version of consistency. We also show that
Resolution proofs of its own re ection principle require slightly subexpo-
nential size. This gives a better lower bound for the monotone interpola-
tion of Res(2) and a better separation from Resolution as a byproduct.
Finally, the techniques for proving these results give us a new complex-
ity measure for Resolution that renes the width of Ben-Sasson and
Wigderson. The new measure and techniques suggest a new algorithm
to nd Resolution refutations, and a way to obtain a large class of exam-
ples that have small Resolution refutations but require relatively large
width. This answers a question of Alekhnovich and Razborov related to
whether Resolution is automatizable in quasipolynomial-time.

?
Partially supported by CICYT TIC2001-1577-C03-02, ALCOM-FT IST-99-14186
and HA2000-41.
2 Albert Atserias and Mara Luisa Bonet

1 Introduction
In several areas of Computer Science there has been important eorts in studying
algorithms for satisability, despite the problem is NP-complete, and also in
studying the complementary problem of verifying tautologies.
By the theorem of Cook and Reckhow [14], there is strong evidence that for
every propositional proof system there is a class of tautologies whose shortest
proofs are super-polynomial in the size of the tautologies. From this we con-
clude that given a propositional proof system S, there will not be an algorithm
that will produce S-proofs of a tautology in time polynomial in the size of the
tautology. This is because in some cases we might require exponential time just
to write down the proof. Considering this limitation of proof systems, Bonet,
Pitassi and Raz [12] proposed the following denition. A propositional proof
system S is automatizable if there exists an algorithm that, given a tautology,
it produces an S-proof of it in time polynomial in the size of the smallest S-
proof of the tautology. The idea behind this denition is that if short S-proofs
exist, an automatization algorithm for S should nd them quickly. In the sequel
of papers [24,13, 9] it was proved that no proof system that simulates AC 0 -
Frege is automatizable, unless some widely accepted cryptographic conjecture
is violated. Later, Alekhnovich and Razborov [1] proved that Resolution is not
automatizable under a reasonable assumption in parameterized complexity. The
drawback of this result is that it is weaker than the others in the sense that we
do not know whether a system that simulates Resolution can be automatizable.
This problem suggests the following denition. We say that a proof system S is
weakly automatizable if there is a proof system that polynomially simulates S
and is automatizable. At this point it is still open whether Resolution is weakly
automatizable.
In this paper we characterize the question of whether Resolution is weakly
automatizable as whether the extension of Resolution Res(2) (or Res(k) for
k constant) has feasible interpolation. This notion will be dened in Section
4. Let us say for the moment, that Resolution, Cutting Planes, Relativized
Bounded Arithmetic, Polynomial Calculus, Lovasz-Schrijver and Nullstellensatz
have feasible interpolation (see [20, 12,26, 15, 22,30,29,27]). On the other hand,
the stronger system Frege, and any system that simulates AC 0-Frege do not have
feasible interpolation under a cryptographic conjecture. To obtain this character-
ization we show that Res(2) has polynomial-size proofs of the re ection principle
of Resolution, which is a form of consistency saying that if a CNF formula is
satisable, then it does not have a Resolution refutation. We also show that
Resolution requires almost exponential size to prove its own re ection princi-
ple. As a corollary we get an almost exponential lower bound for the monotone
interpolation of Res(2) improving over the quasipolynomial lower bound in [4].
Despite the discouraging results in [1] mentioned before, there is still some
eort put in nding good algorithms for proof systems such as Resolution. The
rst implementations were variants of the Davis-Putnam procedure [18,17] for
testing unsatisability that consists of either producing a tree-like Resolution
refutation (if one exists), or giving a satisfying assignment. For various versions
 On the Automatizability of Resolution 3

of this algorithm, one can prove that is it not an automatization procedure even
for tree-like Resolution. A better algorithm for nding tree-like Resolution refu-
tations was proposed by Beame and Pitassi [5]. They give an algorithm that
works in time quasipolynomial in the size of the smallest proof of the tautol-
ogy. So tree-like Resolution is automatizable in quasipolynomial time, but the
algorithm is not a good automatization procedure for general Resolution (see
[10,6, 11]). A more ecient algorithm is the one of Ben-Sasson and Wigderson
based on the width of a refutation. This algorithm weakly automatizes tree-like
Resolution in quasipolynomial time and automatizes Resolution in subexponen-
tial time. On the other hand, Bonet and Galesi gave a class of tautologies for
which the algorithm will take subexponential time to nish, matching the up-
per bound. Using the techniques introduced in this paper, we show that this
is not an isolated example. We describe a method to produce tautologies that
have small Resolution refutations but require relatively large width, answering an
open problem of Alekhnovich and Razborov [1]. As they claim, this is a necessary
step towards proving that Resolution is not automatizable in quasipolynomial-
time. Our techniques also suggest a new complexity measure for Resolution that
renes the width of Ben-Sasson and Wigderson, and that gives rise to a new
algorithm to nd Resolution refutations.

2 Denitions
Resolution is a refutational proof system for CNF formulas, that is, conjunctions
of clauses. The system has one inference rule, the resolution rule :
A l l B
_ : _

A B
_

where l is a literal, and A and B are clauses. The refutation nishes with the
empty clause. The size of a Resolution refutation is the number of clauses in it.
The system tree-like Resolution requires that each clause is used at most once
in the proof. When this restriction is not fullled, we say that the refutation is
in DAG form.
Following [7] the width of a refutation  is dened as the maximum number
of literals of the clauses appearing in . The main result in [7] is a relation
between the size and the width of Resolution refutations. They show that if a
set of 3-clauses has a tree-like Resolution refutation of size ST , then it has a
Resolution refutation of width log ST . Similarly, if it has a Resolution refutation
of size SR , then it has a Resolution refutation of width O( n logSR ). Ben-Sasson
p

and Wigderson used this size-width trade-o to obtain an algorithm that nds
Resolution refutations. It consists in deriving all posible clauses of increasing
width until the empty clause is found. The time of the algorithm is nO(w) where
w is the minimal width of a Resolution refutation of the initial set of clauses.
Notice that the space used by the algorithm can only be bounded by nO(w) since
all derivable clauses of width v < w are needed to obtain the clauses of width w.
Recall that the minimal width w is at most log ST in the tree-like case, where
4 Albert Atserias and Mara Luisa Bonet

ST is the minimal tree-like size to refute the initial set of clauses. Therefore, the
algorithm takes time STO(log n) in this case. Also, the minimal width w is at most
n log SR in the general case, where SR is the minimal size to refute the set of
p
p
clauses in general Resolution. This gives an nO( n log SR ) bound on the running
time.
A k-term is a conjunction of up to k literals. A k-disjunction is an (un-
bounded fan-in) disjunction of k-terms. The refutation system Res(k), dened
by Krajcek [23], works with k-disjunctions. There are three inference rules in
Res(k): Weakening, -Introduction, and Cut.
^

A A l1 B (l2 : : : ls ) A (l1 : : : ls ) B l1 : : : ls
_ _ ^ ^ _ ^ ^ _: _ _:

A B_ A B (l1 : : : ls )
_ _ ^ ^ A B _

Here A and B are k-disjunctions and the li 's are literals. As usual, if l is a
literal, l denotes the oposite literal. We also allow the axioms l l. Observe
: _:

that Res(1) is equivalent to Resolution since the axioms and the weakening rule
are easy to eliminate in this case. The size of a Res(k) refutation is the number
of k-disjunctions in it. As in Resolution, the tree-like version of Res(k) requires
each k-disjunction in the proof to be used only once.

3 Some Simple Lemmas and a New Measure

For every set of literals l1 ; : : :; ls we dene a new variable zl1 ;:::;ls meaning l1 ^

: : : ls . The following clauses dene zl1 ;:::;ls :

^

zl1 ;:::;ls li for every i

: _ 2 f 1; : : :; s g (1)
l
: 1 _ ::: _: ls zl1 ;:::;ls
_ (2)
Let be a set of clauses on the variables v1; : : :; vn. For every integer k > 0, we
C

dene k as the union of with all the dening clauses for the variables zl1 ;:::;ls
C C

for all s k.

Lemma 1. If the set of clauses C has a Res(k) refutation of size S , then Ck has
a Resolution refutation of size O(kS). Furthermore, if the Res(k) refutation is
tree-like, then the Resolution refutation is also tree-like.
Proof of Lemma 1: Let  be a Res(k) refutation of size S. To get a Resolution
refutation of k , we will rst get a clause for each k-disjunction of . The
C

translation consists in substituting each conjunction l1 : : : ls for s k in a ^ ^ 

clause of  by zl1 ;:::;ls . Also we have to make sure that we can make this new
sequence of clauses into a Resolution refutation so that if  is tree-like, then the
new refutation will also be. We have the following cases:
Case 1: In  we have the step:
C (l1 : : : ls ) D l1 : : : ls
_ ^ ^ _: _ _:

C D _
 On the Automatizability of Resolution 5

The corresponding clauses in the translation will be: C 0 zl1 ;:::;ls , D0 l1 : : :

_ _: _ _

: ls and C 0 D0 . To get a tree-like proof of C 0 D0 from the two other ones, rst
_ _

obtain zl1 ;:::;ls D0 in a tree-like way from D0 l1 : : : ls and the clauses

: _ _: _ _:

: zl1 ;:::;ls li . Finally resolve zl1 ;:::;ls D0 with C 0 zl1 ;:::;ls to get C 0 D0 .
_ : _ _ _

Case 2: In  we have the step:

C l1 D (l2 : : : ls )
_ _ ^ ^

C D (l1 : : : ls )
_ _ ^ ^

The corresponding clauses in the translation will be: C 0 l1 , D0 zl2 ;:::;ls and
_ _

C 0 D0 zl1 ;:::;ls . Notice that there is a tree-like proof of l1 zl2 ;:::;ls zl1 ;:::;ls
_ _ : _: _

from the clauses of k . Using this clause and the translation of the premises, we
C

get C 0 D0 zl1 ;:::;ls .

_ _

Case 3: The Weakening rule turns into a weakening rule for Resolution which
can be eliminated easily.
At this point we have obtained a Resolution refutation of k that may use C

axioms of the type l l. These can be eliminated easily too.

_: t
u

Lemma 2. If the set of clauses k has a Resolution refutation of size S , then

C C

has a Res(k) refutation of size O(kS). Furthermore, if the Resolution refutation

is tree-like, then the Res(k) refutation is also tree-like.
Proof : We rst change each clause of the Resolution refutation by a k-disjunction
of Res(k) by translating zl1 ;:::;ls by l1 : : : ls and zl1 ;:::;ls by l1 : : : ls . At
^ ^ : : _ _:

this point the rules of the Resolution refutation turn into valid rules of Res(k).
Now we only need to produce proofs of the dening clauses of the z variables
in Res(k) to nish the simulation. The clauses zl1 ;:::;ls li get translated into
: _

: l1 : : : ls li , which is a weakening of the axiom li li . The clause

_ _ : _ _ :

: l1 : : : ls zl1 ;:::;ls gets translated into l1 : : : ls (l1 : : : ls ) which

_ _: _ : _ _: _ ^ ^

can be proved form the axioms li li using the rule for the introduction of the
_:

^ . t
u

The next lemmas are essentially Proposition 1.1 and 1.2 of [21].
Lemma 3. Any Resolution refutation of width k and size S can be translated
into a tree-like Res(k) refutation of size O(kS).
Proof sketch : Let  be a Resolution refutation of width k and size S. Every non-
initial clause C of  is derived from two other clauses, say C1 and C2. Note that
the k-disjunction C1 C2 C, where Ci is the conjunction of the negated
: _: _ :

literals of Ci, has a very simple tree-like Res(k) proof. The rest of the proof goes
as in [21]. t
u

Lemma 4. ([21, 25, 19]) Any tree-like Res(k) refutation of size S can be trans-
lated into a Resolution refutation of size O(S 2 )
These lemmas suggest a renement of the width mesure that we discuss next.
Following [7], for an unsatisable set of clauses , let w( ) be the minimal width
C C
6 Albert Atserias and Mara Luisa Bonet

of the Resolution refutations of . We dene k( ) to be the minimal k such that

C C

C has a tree-like Res(k) refutation of size nk , where n is the number of variables

of . We will prove that k( ) is at most linear in w( ), and that in some cases,
C C C

k( ) is signicantly smaller than w( ).

C C

Lemma 5. k( ) = O(w( )).

C C

Proof : Let w = w( ). Then has a Resolution refutation of size nO(w) and width
C C

w since there are less than nO(w) clauses of width at most w and each clause
needs to be derived only once since we are in the dag-like case. By Lemma 3, C

a tree-like Res(w) refutation of size O(wnO(w) ). Taking k = O(w), we see that

k( ) = O(w( )).
C C t
u

Lemma 6. There are sets of 3-clauses n such that k( n) = O(1) but w( n) =

F F F

(logn= log log n).

Proof : Let n be the set of 3-clauses E-PHPmm where m0 = log m= log logm. Let
F 0

n be the number of variables of E-PHPmm . Dantchev and Riis [16] proved that n
0 F

has tree-like Resolution refutations of size 2O(m log m ) which in this case is nO(1) .
0 0

Therefore, k( n ) = O(1). On the other hand, a standard width lower bound

F

argument proves that w( n ) =

(m0 ) which in this case is
(log n= loglog n).
F

t
u

These Lemmas give rise to an algorithm to nd Resolution refutations that

improves the width algorithm of Ben-Sasson and Wigderson. Due to space lim-
itations, we omit the precise description of this algorithm (see [3] instead). In
a nutshell, the algorithm consists in using the algorithm of Beame and Pitassi
[5] to nd tree-like Resolution refutations of k of size nk for increasing val-
C

ues of k until one is found. By Lemma 6, this algorithm improves Ben-Sasson

and Wigderson in terms of space usage, and by Lemma 5 its running time is
never worse for sets of clauses with relatively small (subexponential) Resolution
refutations.

4 Re ection Principles and Weak Automatizability

Let be a refutational proof system. Following Razborov [30] (see also [28]),
S

let REF( ) be the set of pairs ( ; m), where is a CNF formula that has an
S C C

S -refutation of size m. Furthermore, let SAT  be the set of pairs ( ; m) where C

C is a satisable CNF. Observe that when m is given in unary, both REF( ) S

and SAT  are in the complexity class NP. Pudlak called (REF( ); SAT  ) the S

canonical NP-pair of . Note also that REF( ) SAT  = since is supposed

S S \ ; S

to refute unsatisable CNF formulas only. Interestingly enough, there is a tight
connection between the complexity of the canonical NP-pair of and the weak S

automatizability of . Namely, Pudlak [28] showed that is weakly automatiz-

S S

able if and only if the canonical NP-pair of is polynomially separable, which

S

means that a polynomial-time algorithm returns 0 on every input from REF( ) S

and returns 1 on every input from SAT  . We will use this connection later.
 On the Automatizability of Resolution 7

The disjointness of the canonical NP-pair for a proof system is often ex- S

pressible as a contradictory set of clauses. Suppose that one is able to write

down a CNF formula SATrn (x; z) meaning that \z encodes a truth assignment
that satises the CNF encoded by x. The CNF is of size r and the underly-
ing variables are v1; : : :; vn". Similarly, suppose that one is able to write down
a CNF formula REFr;m n (x; y) meaning that \y encodes an -refutation of theS

CNF encoded by x. The size of the refutation is m, the size of the CNF is r,
and the underlying variables are v1 ; : : :; vn". Under these two assumptions, the
disjointness of the canonical NP-pair for is expressible by the contradictions
S

REFr;mn (y; z) SAT n (x; z). This collection of CNF formulas is referred to as the
^ r n (y; z) SAT n (x; z) is a form of
Re ection Principle of . Notice that REFr;m
S r ^

consistency of S.
We turn next to the concept of Feasible Interpolation introduced by Krajicek
[22] (see also [12, 26]). Suppose that A0 (x; y0) A1 (x; y1) is a contradictory CNF
^

formula, where x, y0 , and y1 are disjoint sets of variables. Note that for every
given truth assignment a for the variables x, one of the formulas A0 (a; y0) or
A1(a; y1 ) must be contradictory by itself. We say that a proof system has the S

Interpolation Property in time T = T(m) if there exists an algorithm that, given

a truth assignment a for the common variables x, returns an i 0; 1 such that 2 f g

Ai(a; yi ) is contradictory, and the running time is bounded by T(m) where m is

the minimal size of an -refutation of A0(x; y0 ) A1 (x; y1). Whenever T(m) is
S ^

a polynomial, we say that has Feasible Interpolation.

S

The following result by Pudlak connects feasible interpolation with the re-
ection principle and weak automatizability.
Theorem 1. [28] If the re ection principle for has polynomial-size refutations
S

in a proof system that has the feasible interpolation, then the canonical NP-pair
for S is polynomially separable, and therefore S is weakly automatizable.
For the rest of this section, we will need a concrete encoding of the re ection
principle for Resolution. We start with the encoding of SATrn (x; z). The encoding
of the set of clauses by the variables in x is as follows. There are variables xe;i;j
for every e 0; 1 , i 1; : : :; n and j 1; : : :; r . The meaning of x0;i;j is
2 f g 2 f g 2 f g

that the literal vi appears in clause j, while the meaning of x1;i;j is that the
literal vi appears in clause j.
:

The encoding of the truth assignment a 0; 1 n by the variables z is as fol-

2 f g

lows. There are variables zi for every i 1; : : :; n , and ze;i;j for every e 0; 1 ,
2 f g 2 f g

i 1; : : :; n + 1 and j 1; : : :; r . The meaning of zi is that variable vi is

2 f g 2 f g

assigned true under the truth assignment. The meaning of z0;i;j is that clause j is
satised by the truth assignment due to a literal among v1 ; v1; : : :; vi 1; vi 1.
: :

Similarly, the meaning of z1;i;j is that clause j is satised by the truth assign-
ment due to a literal among v1 ; v1; : : :; vi 1; vi 1; vi . We formalize this as a
: :

set of clauses as follows:

: z0;1;j (3) z0;n+1;j (4)
z0;i;j x0;i;j zi z1;i;j (5)
_: _ _: z1;i;j x1;i;j zi z0;i+1;j (6)
_: _: _:

z0;i;j x0;i;j z1;i;j

_ _: (7) z1;i;j x1;i;j z0;i+1;j
_ _: (8)
8 Albert Atserias and Mara Luisa Bonet

The encoding of REFr;m n (x; y) is also quite standard. The encoding of the
set of clauses by the variables in x is as before. The encoding of the Resolution
refutation by the variables in y is as follows. There are variables ye;i;j for every
e 0; 1 , i 1; : : :; n , and j 1; : : :; m . The meaning of y0;i;j is that the
2 f g 2 f g 2 f g

literal vi appears in clause j of the refutation. Similarly, the meaning of y1;i;j

is that the literal vi appears in clause j of the refutation. There are variables
:

pj;k and qj;k for every j 1; : : :; m and k r; : : :; m . The meaning of pj;k

2 f g 2 f g

(of qj;k ) is that clause Ck was obtained from clause Cj and some other clause,
and Cj contains the resolved variable positively (negatively). Finally, there are
variables wi;k for every i 1; : : :; n and k r; : : :; m . The meaning of wi;k
2 f g 2 f g

is that clause Ck was obtained by resolving upon vi . We formalize this by the

following set of clauses:
: xe;i;j ye;i;j
_ (9) ye;i;m (10):

: y0;i;j y1;i;j
_: (11) p1;k : : : pk 1;k (12) _ _

q1;k : : : qk 1;k
_ _ (13) pj;k qj;k (14): _:

: pj;k pj ;k
_: 0 (15) qj;k qj ;k (16): _: 0

: pj;k wi;k y0;i;j

_: _ (17) qj;k wi;k y1;i;j (18): _: _

: pj;k wi;k ye;i;j ye;i;k (19)

_ _: _ qj;k wi;k ye;i;j ye;i;k (20) : _ _: _

w1;k : : : wn;k
_ _ (21) wi;k wi ;k (22): _: 0

Notice that this encoding has the appropriate form for the monotone interpola-
tion theorem.
Theorem 2. The re ection principle for Resolution SATrn (x; z) REFr;m n (x; y) ^

has Res(2) refutations of size (nr + nm) . O (1)

Proof : The goal is to get the following 2-disjunction

_n
Dk  (y0;i;k zi ) (y1;i;k
^ _ ^: zi )
i=1
for every k 1; : : :; m . The empty clause will follow by resolving Dm with
2 f g

(10). We distinguish two cases: k r and r < k m. Since the case k r is

  

easier but long, we leave it to Appendix A.

For the case r < k m, we show how to derive Dk from D1 ; : : :; Dk 1. First,


we derive pj;k ql;k Dk . From (18) and (11) we get ql;k wq;k y0;q;l .
: _: _ : _: _:

Resolving with Dl on y0;q;l we get

_n
ql;k
: _: wq;k (y1;q;l
_ ^: zq ) _ (y0;i;l zi ) (y1;i;l
^ _ ^: zi): (23)
i=1
i6=q

A cut with zq _: zq on y1;q;l ^: zq gives

_n
: ql;k _:wq;k _: zq _ (y0;i;l zi ) (y1;i;l
^ _ ^: zi): (24)
i=1
i6=q
 On the Automatizability of Resolution 9

Let q0 = q. A cut with zq

6 0 _: zq on y0;q ;l zq gives
0 0 ^ 0

_
: ql;k _: wq;k _: zq zq (y1;q ;l
_ 0 _ 0 ^:zq ) 0 _ (y0;i;l zi ) (y1;i;l
^ _ ^: zi ): (25)
i6=q;q 0

From (20) and (22) we get ql;k : _: wq;k _: 0 y ;q ;l y0;q ;k . Resolving with (24)
0 _ 0

on y0;q ;l zq gives
0 ^ 0

_
: ql;k _: wq;k _: zq y0;q ;k (y1;q ;l
_ 0 _ 0 ^: zq ) 0 _ (y0;i;l zi) (y1;i;l
^ _ ^: zi): (26)
i6=q;q 0

An introduction of conjunction between (25) and (26) gives

_
: ql;k _: wq;k _: zq (y0;q ;k zq ) (y1;q ;l
_ 0 ^ 0 _ 0 ^: zq )
0 _ (y0;i;l zi ) (y1;i;l
^ _ ^: zi):
i6=q;q 0

(27)
From (20) and (22) we also get ql;k : _ : wq;k _ : 1 y ;q ;l y1;q ;k . Repeating the
0 _ 0

same procedure we get

_
: ql;k _: wq;k _: zq (y0;q ;k zq ) (y1;q ;k
_ 0 ^ 0 _ 0 ^: zq )
0 _ (y0;i;l zi ) (y1;i;l
^ _ ^: zi ):
i6=q;q 0

(28)
Now, repeating this two-step procedure for every q0 = q, we get 6

_
:ql;k _: wq;k _: zq _ (y0;i;k zi ) (y1;i;l
^ _ ^: zi ): (29)
i6=q
A dual argument yould yield pj;k wq;k zq Wi6=q (y0;i;k zi ) (y1;i;k zi ).
: _: _ _
W ^ _ ^:

A cut with (29) on zq gives pj;k ql;k wq;k i6=q (y0;i;k zi ) (y1;i;k zi ).
: _: _: _ ^ _ ^:

Weakening gives then pj;k ql;k wq;k Dk . Resolving with (21) gives pj;k
: _: _: _ : _

:ql;k Dk . Coming to the end, we resolve this with (12) to get pl;k ql;k Dk .
_ _: _

Then resolve it with (14) to get ql;k Dk , and resolve it with (13) to get Dk .
: _

t
u

An immediate consequence of Theorems 2 and 1 is that if Res(2) has feasible

interpolation, then Resolution is weakly automatizable. The reverse implication
holds too.
Theorem 3. Resolution is weakly automatizable if and only if Res(2) has fea-
sible interpolation.
Proof : Suppose Resolution is weakly automatizable. Then by Corollary 10 in [28],
the NP-pair of resolution is polynomially separable. We claim that the canonical
pair of Res(2) is also polynomially separable. Here is the separation algorithm:
Given a set of clauses and a number S, we build 2 and run the separation
C C

algorithm for the canonical pair of Resolution on 2 and c 2S, where c is the C

hidden constant in Lemma 1. For the correctness, note that if has a Res(2) C

refutation of size S, then 2 has a Resolution refutation of size c 2S by Lemma 1,

C

10 Albert Atserias and Mara Luisa Bonet

and the separation algorithm for the canonical pair of Resolution will return 0
on it. On the other hand, if is satisable, so is 2 and the separation algorithm
C C

for Resolution will return 1 on it. Now, for the feasible interpolation of Res(2),
consider the following algorithm. Let A0 (x; y) A1 (x; z) be a contradictory set
^

of clauses with a Res(2) refutation  of size S. Given a truth assignment a for

the variables x, run the separation algorithm for the canonical pair of Res(2) on
inputs A0 (a; y) and S. For the correctness, observe that if A1 (a; z) is satisable,
say by z = b, then  x=a;z=b is a Res(2) refutation of A0(a; y) of size at most S
j

and the separation algorithm will return 0 on it. On the other hand, if A0 (a; y)
is satisable, the separation algorithm will return 1, which is correct. If both are
unsatisable, any answer is ne. t
u

The previous theorem works for any k constant. If k = log n, then we get that
if Resolution is weakly automatizable then Res(log) has feasible interpolation
in quasipolynomial time. The positive interpretation of these results is that to
show that Resolution is weakly automatizable, then we only have to prove that
Res(2) has feasible interpolation. The negative interpretation is that to show
that resolution is not weakly automatizable we only have to prove that Res(log)
doesn't have feasible interpolation in quasipolynomial time.
It is not clear whether Res(2) has feasible interpolation. We know, however,
that Res(2) does not have monotone feasible interpolation (see [4] and Corollary 1
in this paper). On the other hand, tree-like Res(2) has feasible interpolation (even
monotone) since Resolution polynomially simulates it by Lemma 4.
A natural question to ask is whether the re ection principle for Resolution
has Resolution refutations of moderate size. Since Resolution has feasible inter-
polation, a positive answer would imply that Resolution is weakly automatizable
by theorem 1. Unfortunately, as the next theorem shows, this will not happen.
The proof of this result uses an idea due to Pudlak.
Theorem 4. For some choice of n, r, and m of the order of a quasipolyno-
mial sO(log s) on the parameter s, every Resolution refutation of REFr;m n (x; y)^

SATrn (x; z) requires size at least 2

(s ) .
1 =4

Proof : Suppose for contradiction that there is a Resolution refutation of size

S=2 o(s1=4 )
. Let k = s1=2 , and let COLk (p; q) be the CNF formula expressing
that q encodes a k-coloring of the graph on s nodes encoded by pi;j . An f g

explicit denition
W is the following: For every i 1; : : :; s , there is a clause of
2 f g

the form kl=1 qil ; and for every i; j 1; : : :; s with i = j and l 1; : : :; k ,

2 f g 6 2 f g

there is a clause of the form qil qjl pij . Obviously, if G is k-colorable,

: _ : _ :

then COLk (G; q) is satisable, and if G contains a 2k-clique, then COLk (G; q)
is unsatisable. More importantly, if G contains a 2k-clique, then the clauses of
PHPk2k are contained in COLk (G; q). Now, for every graph G on s nodes, let
F (G) be the clauses COLk (G; q) together with all clauses dening the extension
variables for the conjunctions of up to c logk literals on the q-variables. Here, c
is a constant so that the kO(log k) upper bound on PHPk2k of [25] can be done
in Res(c log k). From its very denition and Lemma 1, if G contains a 2k-clique,
 On the Automatizability of Resolution 11

then (G) has a Resolution refutation of size kO(log k). Finally, for every graph
F

G, let x(G) be the encoding of the formula (G). With all this notation, we are
F

ready for the argument.

In the following, let n be the number of variables of (G), let r be the
F

number of clauses of (G), and let m = kO(log k). By assumption, the formulas
F

REFr;mn (x(G); y) SAT n (x(G); z) have Resolution refutations of size at most

^ r
S. Let C be the monotone circuit that interpolates these formulas given x(G).
The size of C is S O(1) . Moreover, if G is k-colorable, then SATrn (x(G); z) is
satisable, and C must return 0 on x(G). Also, if G contains a 2k-clique, then
REFr;mn (x(G); y) is satisable, and C must return 1 on x(G). Now, an anti-
monotone circuit for separating 2k-cliques from k-colorings can be built as fol-
lows: given a graph G, build the formula x(G) (anti-monotonically, see below for
details), and apply the monotone circuit given by the monotone interpolation.
The size of this circuit is 2o(s1=4 ) , and this contradicts Theorem 3.11 of Alon and
Boppana [2].
It remains to show how to build an anti-monotone circuit that, on input
G = puv , produces outputs of the form xe;i;j that correspond to the encoding
f g

of (G) in terms of the x-variables.

F

{ Clauses of the type Wkl qil: Let t be the numbering of this clause in (G).
=1 F

Then, its encoding in terms of the x-variables is produced by plugging the

constant 1 to the outputs x1;qi1 ;t; : : :; x1;qik;t . The rest of outputs of clause t
get plugged the constant 0.
{ Clauses of the type qil qjl pij: Let t be the numbering of this clause
: _: _:

in (G). The encoding is x0;qil ;t = 1, x0;qjl ;t = 1, x0;pij ;t = pij and the rest
F :

are zero. Notice that this encoding is anti-monotone in the pij 's. Notice also
that the encoded (G) contains some p-variables (and not only q-variables
F

as the reader might have expected) but this will not be a problem since the
main properties of (G) are preserved as we show below.
F

{ Finally, the clauses dening the conjunctions of up to c logk literals are

independent of G since only the q-variables are relevant here. Therefore, the
encoding is done as in the rst case.
The reader can easily verify that when G contains a 2k-clique, the encoded
formula contains the clauses of PHPk2k and the denitions of the conjunctions
up to c log k literals. Therefore REF(x(G); y) is satisable given that PHPk2k
has a small Res(c log k) refutation. Similarly, if G is k-colorable, the formula
SAT(x(G); z) is satisable by setting zpij = pij and qil = 1 if and only if node
i gets color l. Therefore, the main properties of (G) are preserved, and the
F

theorem follows. t
u

An immediate corollary of the last two results is that Res(2) is exponen-

tially more powerful than resolution. In fact, the proof shows a lower bound for
the monotone interpolation of Res(2) improving over the quasipolynomial lower
bound in [4].
12 Albert Atserias and Mara Luisa Bonet

Corollary 1. Monotone circuits that interpolate Res(2) refutations require size

2
(s
1 =4)
on Res(2) refutations of size sO(log s) .
Theorem 4 is in sharp contrast with the fact that an appropriate encoding
of the re ection principle for Res(2) has polynomial-size proofs in Res(2). This
encoding incorporates new z-variables for the truth values of conjunctions of
two literals, and new y-variables encoding the presence of conjunctions in the 2-
disjunctions of the proof. The resulting formula preserves the form of the feasible
interpolation. We leave the tedious details to the interested reader.
Theorem 5. The re ection principle for Res(2) has Res(2) refutations of size
(n2r + mr)O(1) . More strongly, the re ection principle for Res(k) has Res(2)
refutations of size (nk r + mr)O(1) .
We observe that there is a version of the re ection principle for Resolution
that has polynomial-size proofs in Resolution. Namely, let be the CNF formula
C

SATrn (x; z) REFr;m

^
n (y; z). Then, 2 has polynomial-size Resolution refutations
C

by Lemma 1 and Theorem 2. However, this does not imply the weak automatiz-
ability of Resolution since the set of clauses does not have the appropriate form
for the feasible interpolation theorem.

5 Short Proofs that Require Large Width

Bonet and Galesi [11] gave an example of a CNF expressed in constant width,
with small Resolution refutations, and requiring relatively large width (square
root of the number of variables). This showed that the size-width trade-o of
Ben-Sasson and Wigderson could not be improved. Also it showed that the
algorithm of Ben-Sasson and Wigderson for nding Resolution refutations could
perform very badly in the worst case. This is because their example requires
large width, and the algorithm would take almost exponential time, while we
know that there is a polynomial size Resolution refutation.
Alekhnovich and Razborov [1] posed the question of whether more of these
examples could be found. They say this is a necessary rst step for showing
that Resolution is not automatizable in quasipolynomial-time. Here we give a
way of producing such bad examples for the algorithm. Basically the idea is
nding CNFs that require suciently high width in Resolution, but that have
polynomial size Res(k) refutations for small k, say k logn. Then the example


consists of adding to the formula the clauses dening the extension variables for
all the conjunctions of at most k literals. Below we ilustrate this technique by
giving a large class of examples that have small Resolution refutations, require
large width. Moreover, deciding whether a formula is in the class is hard (no
polynomial-time algorithm is known).
Let G = (U V; E) be a bipartite graph on the sets U and V of cardinality
[

m and n respectively, where m > n. The G-PHPnm, dened by Ben-Sasson and

Wigderson [7], states that there is no matching from U into V . For every edge
 On the Automatizability of Resolution 13

(u; v) E, let xu;v be a propositional variable meaning that u is mapped to v.

2

The principle is then formalized as the conjunction of the following clauses:

xu;v1_


_ xu;vr u U; NG (u) = v1; : : :; vr
2 f g

xu;v xu ;v v V; u; u0 NG (v); u = u0:

_ 0 2 2 6

Here, NG (w) denotes the set of neighbors of w in G. Note that if G has left-degree
at most d, then the width of the initial clauses is bounded by d.
Ben-Sasson and Wigderson proved that whenever G is expanding in a sense
dened next, every Resolution refutation of G-PHPnm must contain a clause
with many literals. We observe that this result is not unique to Resolution and
holds in a more general setting. Before we state the precise result, let us recall
the denition of expansion:
Denition 1. [7] Let G = (U V; E) be a bipartite graph where U = m, and
[ j j

jV = n. For U 0 U , the boundary of U , denoted by @U 0 , is the set of vertices in

j 

V that have exactly one neighbor in U 0 ; that is, @U 0 = v V : N(v) U 0 = 1 .

f 2 j \ j g

We say that G is (m; n; r; f)-expanding if every subset U 0 U of size at most r



is such that @U 0 f U 0 .
j j 
j j

The proof of the following statement is the same as in [7] for Resolution.
Theorem 6. [7] Let be a sound refutation system with all rules having fan-in
S

at most two. Then, if G is (m; n; r; f)-expanding, every -refutation of G-PHPnm

S

must contain a formula that involves at least rf=2 distinct literals.

Now, for every bipartite graph G with m 2n, let (G) be the set of clauses
 C

dening G-PHPnm together with the clauses dening all the conjunctions up to
c log n literals, where c is a large constant.
Theorem 7. Let G be an (m; n;
(n= logm); 43 logm)-expander with m 2n 

and left-degree at most log m. Then (i) (G) has initial width logm, (ii) any
C

Resolution refutation of (G) requires width at least

(n= logn), and (iii) (G)
C C

has polynomial-size Resolution refutations.

Proof : Part (i) is obvious. For (ii), suppose for contradiction that C (G) has a
Resolution refutation of width w = o(n= logn). Then, by the proof of Lemma 2,
G-PHPnm has a Res(c log n) refutation in which every (c logn)-disjunction in-
volves at most wc log n = o(n) literals. This contradicts Theorem 6. For (iii),
recall that PHPnm has a Res(c log n) refutation of size nO(log n) by [25] since
m 2n. Now, setting to zero the appropriate variables of PHPnm, we get a


Res(c log n) refutation of G-PHPnm of the same size. By Lemma 1, (G) has a C

Resolution refutation of roughly the same size, which is polynomial in the size
of the formula. t
u

It is known that deciding whether a bipartite graph is an expander (for a

slightly dierent denition than ours) is coNP-complete [8]. Although we have
not checked the details, we suspect that deciding whether a bipartite graph
14 Albert Atserias and Mara Luisa Bonet

is an (m; n; r; f)-expander in the sense of Denition 1 is also coNP-complete.

However, we should note that the class of formulas (G) : G expander; m 2n
fC  g

is contained in (G) : G bipartite; m 2n which is decidable in polynomial-

fC  g

time, and that all formulas of this class have short Resolution refutations that
are easy to nd. This is so because the proof of PHPn2n in [25] is given explicitely.

6 Conclusions and Open Problems

We showed that the new measure k( ) introduced in section 3 is a renement of
C

the width w( ). Actually, we believe that a careful analysis in Lemma 5 could

C

even show that k( ) w( ) + 1 for sets of clauses with suciently many

C  C C

variables. On the other hand, we proved a logarithmic gap between k( ) and C

w( ) for a concrete class of 3-clauses n. We do not know if a larger gap is

C C

possible.
It is surprising that the weak pigeonhole principle PHPn2n has short Res-
olution proofs when encoded with the clauses dening the extension variables.
This suggests that to prove Resolution lower bounds that are robust, one should
prove Res(k) lower bounds for relatively large k. In fact, at this point the only
robust lower bounds we know are the ones for AC 0 -Frege.
Of course, it remains open whether Resolution is weakly automatizable, or
automatizable in quasipolynomial-time.
Acknowledgement. We are grateful to Pavel Pudlak for stimulating discussions on
the idea of Theorem 4.

References
1. M. Alekhnovich and A. A. Razborov. Resolution is not automatizable unless W[P]
is tractable. In 42nd Annual IEEE Symposium on Foundations of Computer Sci-
ence, 2001.
2. N. Alon and R. B. Boppana. The monotone circuit complexity of boolean functions.
Combinatorica, 7:1{22, 1987.
3. A. Atserias and M. L. Bonet. On the automatizability of resolution and related
propositional proof systems. ECCC TR02-010, 2002.
4. A. Atserias, M. L. Bonet, and J. L. Esteban. Lower bounds for the weak pigeon-
hole principle and random formulas beyond resolution. Accepted for publication
in Information and Computation. A preliminary version appeared in ICALP'01,
Lecture Notes in Computer Science 2076, Springer, pages 1005{1016., 2001.
5. P. Beame and T. Pitassi. Simplied and improved resolution lower bounds. In
37th Annual IEEE Symposium on Foundations of Computer Science, pages 274{
282, 1996.
6. E. Ben-Sasson, R. Impagliazzo, and A. Wigderson. Near-optimal separation of
general and tree-like resolution. To appear, 2002.
7. E. Ben-Sasson and A. Wigderson. Short proofs are narrow{resolution made simple.
J. ACM, 48(2):149{169, 2001.
 On the Automatizability of Resolution 15

8. M. Blum, R. M. Karp, O. Vornberger, C. H. Papadimitriou, and M. Yannakakis.

The complexity of testing whether a graph is a superconcentrator. Information
Processing Letter, 13:164{167, 1981.
9. M. L. Bonet, C. Domingo, R. Gavalda, A. Maciel, and T. Pitassi. Non-
automatizability of bounded-depth Frege proofs. In 14th IEEE Conference in
Computational Complexity, pages 15{23, 1999. Accepted for publication in the
Journal of Computational Complexity.
10. M. L. Bonet, J. L. Esteban, N. Galesi, and J. Johansen. On the relative complex-
ity of resolution renements and cutting planes proof systems. SIAM Journal of
Computing, 30(5):1462{1484, 2000. A preliminary version appeared in FOCS'98.
11. M. L. Bonet and N. Galesi. Optimality of size-width trade-os for resolution.
Journal of Computational Complexity, 2001. To appear. A preliminary version
appeared in FOCS'99.
12. M. L. Bonet, T. Pitassi, and R. Raz. Lower bounds for cutting planes proofs with
small coecients. Journal of Symbolic Logic, 62(3):708{728, 1997. A preliminary
version appeared in STOC'95.
13. M. L. Bonet, T. Pitassi, and R. Raz. On interpolation and automatization for
Frege systems. SIAM Journal of Computing, 29(6):1939{1967, 2000. A preliminary
version appeared in FOCS'97.
14. S. Cook and R. Reckhow. The relative eciency of propositional proof systems.
Journal of Symbolic Logic, 44:36{50, 1979.
15. S. A. Cook and A. Haken. An exponential lower bound for the size of monotone
real circuits. Journal of Computer and System Sciences, 58:326{335, 1999.
16. S. Dantchev and S. Riis. Tree resolution proofs of the weak pigeon-hole principle.
In 16th IEEE Conference in Computational Complexity, pages 69{75, 2001.
17. M. Davis, G. Logemann, and D. Loveland. A machine program for theorem proving.
Communications of the ACM, 5:394{397, 1962.
18. M. Davis and H. Putnam. A computing procedure for quantication theory.
J. ACM, 7:201{215, 1960.
19. J. L. Esteban, N. Galesi, and J. Messner. Personal communication. Manuscript,
2001.
20. R. Impagliazzo, T. Pitassi, and A. Urquhart. Upper and lower bounds for tree-like
cutting planes proofs. In 9th IEEE Symposium on Logic in Computer Science,
pages 220{228, 1994.
21. J. Krajcek. Lower bounds to the size of constant-depth propositional proofs.
Journal of Symbolic Logic, 39(1):73{86, 1994.
22. J. Krajcek. Interpolation theorems, lower bounds for proof systems, and inde-
pendence results for bounded arithmetic. Journal of Symbolic Logic, 62:457{486,
1997.
23. J. Krajcek. On the weak pigeonhole principle. To appear in Fundamenta Mathe-
matic, 2000.
24. J. Krajcek and P. Pudlak. Some consequences of cryptographical conjectures for
s12 and ef . Information and Computation, 140(1):82{94, 1998.
25. A. Maciel, T. Pitassi, and A. R. Woods. A new proof of the weak pigeonhole
principle. In 32nd Annual ACM Symposium on the Theory of Computing, 2000.
26. P. Pudlak. Lower bounds for resolution and cutting plane proofs and monotone
computations. Journal of Symbolic Logic, 62(3):981{998, 1997.
27. P. Pudlak. On the complexity of the propositional calculus. In Sets and Proofs,
Invited Papers from Logic Colloquium '97, pages 197{218. Cambridge University
Press, 1999.
16 Albert Atserias and Mara Luisa Bonet

28. P. Pudlak. On reducibility and symmetry of disjoint NP-pairs. In 26th Inter-

national Symposium on Mathematical Foundations of Computer Science, Lecture
Notes in Computer Science, pages 621{632. Springer-Verlag, 2001.
29. P. Pudlak and J. Sgall. Algebraic models of computation and interpolation for
algebraic proof systems. In P. W. Beame and S. R. Buss, editors, Proof Complexity
and Feasible Arithmetic, volume 39 of DIMACS Series in Discrete Mathematics and
Theoretical Computer Science, pages 279{296. American Mathematical Society,
1998.
30. A. A. Razborov. Unprovability of lower bounds on circuit size in certain fragments
of bounded arithmetic. Izvestiya of the RAN, 1995.

A Appendix: Deriving Dk
We consider the case k r. We will derive Dk by successive steps as follows.


Let Dq;k
0
be the following 2-disjunction
_n
Dq;k
0
z0;q;k
 _ (y0;i;k zi ) (y1;i;k
^ _ ^: zi);
i=q
and let Dq;k
1
be the following 2-disjunction
_n
Dq;k z1;q;k (y1;q;k
1
 _ ^: zq ) _ (y0;i;k zi ) (y1;i;k
^ _ ^: zi ):
i=q+1
Observe that Dn0 +1;k is simply z0;n+1;k which is the clause (4) in SATrn (x; z).
We obtain Dq1 1;k from Dq;k0
as follows. Cut Dq;k
0
with (8) and (6) on z0;q;k to
get
_n
z1;q 1;k x1;q 1;k
_ (y0;i;k zi ) (y1;i;k zi );
_ ^ _ (30) ^:

i=q
and
_n
z1;q 1;k _: 1x ;q 1;k _: zq 1 _ (y0;i;k zi ) (y1;i;k
^ _ ^: zi ); (31)
i=q
respectively. A cut between (30) and (31) on x1;q 1;k gives
_n
z1;q 1;k _: zq 1 _ (y0;i;k zi ) (y1;i;k
^ _ ^:zi): (32)
i=q
On the other hand, a cut between (30) and (9) on x1;q 1;k gives
_n
z1;q 1;k y1;q 1;k
_ _ (y0;i;k zi ) (y1;i;k
^ _ ^: zi); (33)
i=q
 On the Automatizability of Resolution 17

Finally, an introduction of conjunction between (32) and (33) on zq 1 and :

y1;q 1;k gives Dq1 1;k as claimed. Next, we show how to get Dq;k
0
from Dq;k
1
. Cut
Dq;k with (7) and (5) on z1;q;k to get
1

_n
z0;q;k x0;q;k (y1;q;k
_ _ zq )
^: _ (y0;i;k zi ) (y1;i;k
^ _ ^: zi); (34)
i=q+1
and
_n
z0;q;k _: x0;q;k zq (y1;q;k
_ _ ^: zq ) _ (y0;i;k zi ) (y1;i;k
^ _ ^: zi ); (35)
i=q+1
respectively. A cut between (34) and (35) on x0;q;k gives
_n
z0;q;k zq (y1;q;k
_ _ ^: zq ) _ (y0;i;k zi ) (y1;i;k
^ _ ^: zi ): (36)
i=q+1
On the other hand, a cut between (34) and (9) on x0;q;k gives
_n
z0;q;k y0;q;k (y1;q;k
_ _ ^: zq ) _ (y0;i;k zi ) (y1;i;k
^ _ ^: zi ): (37)
i=q+1
Finally, an introduction of conjunction between (36) and (37) on zq and y0;q;k
gives Dq;k
0
as desired.
We have shown how to obtain D10;k . In order to obtain Dk , we only need to
cut D10;k with (3) on z0;1;k .