The

Resolution
Principle

The Resolution Principle

 Summary

The
Resolution
Principle

Introduction [Chang-Lee Ch. 5.1]

Resolution Principle for Propositional Logic [Chang-Lee
Ch. 5.2]
 Herbrand's Theorem and refutation procedures

The
Resolution
Principle

Satisability procedures
We can build refutation procedures building on Herbrand's
Theorem.
For example Gilmore's method using DPLL for checking
satisability.
This requires the generation of sets S00 , S10 , · · · of ground
clauses.
Computaton issue: for most cases this sequence grows
exponentially.
 Computational issue

The
Resolution
Principle

Exponential grow of sequence

Consider S =
{P (x , g (x ), y , h(x , y ), z , k (x , y , z )), ¬P (u , v , e (v ), w , f (v , w ), x )}
H0 = {a} H1 = {a, g (a), h(a, a), k (a, a, a), e (a), f (a, a)}
|S00 | = 2, |S10 | = 1512
Earliest unsatisable set is S50 which has approximately
10256 elements!
 The Resolution Principle

The
Resolution
Principle

Robinson 1965
Aim: directly test unsatisability of a set of clauses S
without generating all possible associated ground clauses.
Basic idea: test whether S contains the empty clause 
If  ∈ S then S is unsatisable
Otherwise need to check whether S |= 
 Connection with Sematic trees

The
Resolution
Principle

Res. Principle and Sem. Trees

Recall: by Herbrand's Theorem (version I) S is unsatisable
i there is a nite closed semantic tree T for S .
S contains  i the corresponsing closed semantic tree T
contains only the root node.
If S does not contain  then T must contain more than
one node.
If we can reduce the number of nodes in T then we can
force  to appear.
 Inference Rules

The
Resolution
Principle Inference Rules
The resolution principle is an Inference Rule
Inference Rule: a rule that generates new clauses which are
a logical consequence of some of the existing clauses
New clauses can be used to turn some of the nodes in T
to failure nodes.
Thus number of nodes in T are reduced and  will
eventually appear.
Example (Resolution Principle and Sem. Trees)
The semantic tree for S = {¬P ∨ Q , P , ¬Q } can be reduced to
 by adding {¬P } to S .
 Resolution principle for Propositional Logic

The
Resolution
Principle
Resolution and One-Literal rule
Extension of One-Literal rule of DPLL to any pair of
clauses
Focus on a unit clause containing a literal L and look for
the complement of L in another clause. Obtain a new
clause deleting the One-Literal clause, and the complement
literal from the other clause.
Example (One-Literal and resolution)
C1 = P , C2 = ¬P ∨ Q
Applying the One-Literal rule of DPLL to {C1 , C2 } we obtain
C3 = Q
 The Resolution Principle

The
Resolution
Principle
Resolution Principle
For any two clauses C1 and C2 if there is a literal L1 in C1 that is
complementary to a literal L2 in C2 then delete L1 and L2 from
C1 and C2 and generate a new clause C3 as the disjunction of
the remaining clauses.
C3 is a resolvent for C1 and C2 .
Resolution Principle: Inference rule
L1 ∨ C10 ¬L1 ∨ C20
C10 ∨ C20
 Example

The
Resolution
Principle

Example (Resolution Principle)

Consider the following clauses C1 = P ∨ R and C2 = ¬P ∨ Q
P ∨R ¬P ∨ Q
 Example

The
Resolution
Principle

Example (Resolution Principle)

Consider the following clauses C1 = P ∨ R and C2 = ¬P ∨ Q
P ∨R ¬P ∨ Q
R ∨Q
 Example

The
Resolution
Principle

Example (Resolution Principle)

Consider the following clauses C1 = P ∨ R and C2 = ¬P ∨ Q
P ∨R ¬P ∨ Q
R ∨Q
C3 = R ∨ Q is the resolvent for C1 and C2 .
 Example II

The
Resolution
Principle

Example (Resolution Principle)

Consider the following clauses C1 = ¬P ∨ Q ∨ R and
C2 = ¬Q ∨ S
¬P ∨ Q ∨ R ¬Q ∨ S
 Example II

The
Resolution
Principle

Example (Resolution Principle)

Consider the following clauses C1 = ¬P ∨ Q ∨ R and
C2 = ¬Q ∨ S
¬P ∨ Q ∨ R ¬Q ∨ S
¬P ∨ R ∨ S
 Example II

The
Resolution
Principle

Example (Resolution Principle)

Consider the following clauses C1 = ¬P ∨ Q ∨ R and
C2 = ¬Q ∨ S
¬P ∨ Q ∨ R ¬Q ∨ S
¬P ∨ R ∨ S

C3 = ¬P ∨ R ∨ S is the resolvent for C1 and C2 .

 Example III

The
Resolution
Principle

Example (Resolution Principle)

Consider the following clauses C1 = ¬P ∨ Q and C2 = ¬P ∨ S
There is no resolvent in this case as no complementary pair can
be found in the clauses.
 Property of Resolution

The
Resolution
Logical consequence
Principle
Given two clauses C1 and C2 , and their resolvent C , C1 ∧ C2 |= C
(C is a logical consequence of C1 and C2 ).
Proof.
Let C1 = L ∨ C10 , C2 = ¬L ∨ C20 , C = C10 ∨ C20 where C10 and C20
are disjunctions of literals. Suppose I |= C1 ∧ C2 , we want to
show that I |= C .
Note that either I |= L or I |= ¬L.
Assume I |= ¬L
Then since I |= C1 , C10 6=  and I |= C10 .
Therefore since C = C10 ∨ C20 we have that I |= C .
Similar considerations hold for I |= L.
 Derivation of the empty clause

The
Resolution
Principle

Resolution and satisability

If C1 and C2 are unit clauses then, if there is resolvent,
that resolvent will necessary be .
If we can derive the empty clause from S , then S is
unsatisable (correctness)
If S is unsatisable using resolution we can always derive
the empty clause (completeness)
 Deduction

The
Resolution
Principle

Denition (Deduction)
Given a set of clauses S a (resolution) deduction of C from S is
a nite sequence C1 , C2 , · · · , Ck of clauses such that each Ci is
either a clause in S or a resolvent of clauses preceding Ci , and
Ck = C .
 Example I: Deduction

The
Resolution
Principle

Example (deduction)
Consider S = {C1 , C2 , C3 }, where C1 = ¬P ∨ Q C2 = P and
C3 = ¬Q . Applying resolution to C1 and C2 we have:
¬P ∨ Q , P
Q
Then applying
¬Q , Q

 Deducing the empty clause

The
Resolution
Principle Empty clause, Deduction and Unsatisability
Given S , suppose we derive  using resolution;
⇒  is a logical consequence of S ;
Since S |=  then ∀I if I |= S then I |= ;
But there is no I that can verify ;
⇒ if we derive  from S using refutation then S is
unsatisable.
Later we will show that if S is unsatisable then we can
always derive  using resolution.
Denition (Refutation)
A deduction of  is called a refutation (or a proof) of S
 Example II: Deduction

The
Resolution
Principle

Example (deduction)
Given S = {C1 , C2 , C3 , C4 } and C1 = {P ∨ Q },
C2 = {¬P ∨ Q }, C3 = {P ∨ ¬Q } and C4 = {¬P ∨ ¬Q }. Apply
resolution to C1 and C2 and obtain C 0 = {Q }.
Apply resolution to C3 and C4 and obtain C 00 = {¬Q }.
Apply resolution to C 0 and C 00 and obtain .
Hence S is unsat.
 Example II: Deduction Tree

The
Resolution
Principle

Example (deduction)
Consider S from previous example and the associated deduction
steps.
The deduction tree is:
P ∨Q , ¬P ∨Q P ∨¬Q , ¬P ∨¬Q
Q ¬Q

 Exercise

The
Resolution
Principle

Exercise
Prove that the following formulas are unsat. using the resolution
principle
1 ¬q ∨ p , ¬p ∨ ¬q , q ∨ r , ¬q ∨ ¬r , ¬p ∨ ¬r , p ∨ ¬r
2 P (a), ¬D (a)∨ L(a, a), ¬P (a)∨¬Q (a)∨¬L(a, a), D (a), Q (a)
 Exercise II

The
Resolution
Principle

Exercise
1 F1 , P → (¬Q ∨ (R ∧ S ))
2 F2 , P
3 F3 , ¬S
4 G = ¬Q
Prove using the resolution principle that F1 ∧ F2 ∧ F3 |= G