NP and Craig's Interpolation Theorem

NP and Craig's Interpolation Theorem

LOGIC COLLOQUIUM '82 G. Lolli. G. Long0 and A . Marqa (editors) 0 Ekevier Science Publishers B. V. (North-Holland), I984 345 NP AND C R A I G ' S IN...

553KB Sizes 9 Downloads 91 Views

LOGIC COLLOQUIUM '82 G. Lolli. G. Long0 and A . Marqa (editors) 0 Ekevier Science Publishers B. V. (North-Holland), I984

345

NP AND C R A I G ' S INTERPOLATION THEOREM

Daniele Mundici

.

L oc Romol a N .76 50060 Donnini

F1 orencc-I t a l y

The t r u t h - v a l u e s of two renowned c o n j e c t u r e s about NP (namely, P f NP and, NP i s n o t c l o s e d under complementation) depend on t h e d i f f i c u l t y i n w r i t i n g down C r a i g ' s i n t e r p o l a n t s i n s e n t e n t i a l l o g i c . The g e n e r a l connection between NP and i n t e r p o l a t i o n i s s t u d i e d by b l e n d i n g i d e a s and t e c h n i q u e s from both model t h e o r y and computation t h e o r y .

0.

Introduction.

W e f i x throughout an a1 habet

c

and r e g a r d Boolean e x p r e s s i o n s a s p a r t i c u l a r words over I f B-+C i s a tautology, then Craig's i n t e r p o l a t i o n theorem y i e l d s an i n t e r p o l a n t I , t h a t i s , a Boolean e x p r e s s i o n I such t h a t B + I and IC are t a u t o l o g i e s , and t h e v a r i a b l e s o c c u r r i n g i n I are e x a c t l y t h o s e which j o i n t l y occur i n B and C. I n Theorem 2 and C o r o l l a r y 6 w e prove t h e f o l l o w i n g r e s u l t (where TAUT&* i s t h e set of t a u t o l o g i e s and is the set of words over

2.

c*

c):

A t l e a s t one of t h e f o l l o w i n g s e n t e n c e s i s t r u e :

(I)

TAUT

i s accepted i n d e t e r m i n i s t i c polynomial t i m e (viz..P=NP);

(11)

TAUT

i s n o t a c c e p t e d i n n o n d e t e r m i n i s t i c polynomial t i m e NP i s n o t c l o s e d under complementation);

(viz., (111)

-

-

I n t e r p o l a t i o n i s polynomially i n t r a c t a b l e , v i z . , f o r every f u n c t i o n Cp if $J i s computable i n d e t e r m i n i s t i c polynomial t i m e , t h e n f o r some t a u t o l o g y B C , $(B,C) f a i l s t o be an i n t e r p o l a n t f o r B+ C.

:c*xc*

c*,

For t h e proof w e use a m i x t u r e of t e c h n i q u e s from computation and model t h e o r y . Assume now t h a t t h e upper bounds f o r computations are r e l a x e d from t h e set 3 of polynomials t o any set 3 2 9 c l o s e d und e r sum and composition. Then t h e above r e s u l t s t i l l h o l d s r e l a t i v e perhaps w i t h a d i f f e r e n t d i s t r i b u t i o n of t o t h e new upper bounds t r u t h - v a l u e s among ( 1 ) - ( 1 1 1 ) . T h i s i s proved i n Theorem 7. Thus i t might be i n t e r e s t i n g t o f i n d two sets 3 and 3' ( i f any) ass i g n i n g d i f f e r e n t t r i p l e t s of t r u t h v a l u e s t o ( I ) - ( I I I ) : as a matter of f a c t , when 3 i s t h e set of a l l f u n c t i o n s , o r even, when '5 c o n t a i n s t h e e x p o n e n t i a l , t h e n t r i v i a l l y ( I ) becomes t r u e and (11)

--

3 46

D.MUNDICI

and (111) f a l s e . A t t h e o p p o s i t e extreme, when '5 is restricted t o t h e set of p o l y n o m i a l s , i t i s w i d e l y c o n j e c t u r e d t h a t ( I ) i s f a l s e , hence, e i t h e r of (11) or (111) i s t h e n t r u e . The above r e s u l t s y i e l d j u s t one more c o n n e c t i o n between t h e modelt h e o r e t i c a l n o t i o n of i n t e r p o l a t i o n , and computation t h e o r y . For a n o t h e r such c o n n e c t i o n , i n c61 i t i s proved t h a t i f t h e r a t e of growth of t h e l e n g t h of i n t e r p o l a n t s f o r any t a u t o l o g y B + C can be kept below some polynomial i n t h e l e n g t h of B and C , t h e n every f u n c t i o n which i s computable i n d e t e r m i n i s t i c polynomial time h a s c i r c u i t d e p t h growing p r o p o r t i o n a l l y t o t h e l o g a r i t h m of t h e input l e n g t h . T h i s as w e l l as o t h e r r e s u l t s c o n c e r n i n g t h e complexity of C r a i g ' s i n t e r p o l a t i o n theorem i n s e n t e n t i a l and i n f i r s t - o r d e r l o g i c a r e surveyed i n t h e f i n a l s e c t i o n of t h i s p a p e r . For t h e g e n e r a l r o l e of C r a i g ' s i n t e r p o l a t i o n theorem i n ( a b s t r a c t ) model t h e o r y , s e e , e . g . , [I1 , [8J and [91.

1.

Preliminaries.

For

A

an a r b i t r a r y s e t ,

d e n o t e s t h e set of words o v e r

A*

t h e set of a l l f i n i t e s t r i n g s of symbols from length

For paper

of

1w1

aCA,

an

A.

For

i.e.,

A,

the

w€A*,

i s t h e number of o c c u r r e n c e s of symbols i n

w

stands f o r

aa...a

(n

times).

Throughout

w.

this

d e n o t e s t h e f o l l o w i n g set of symbols:

c

{ A , v , T p ) .( , X , 0, 1 ) .

=

c,

Boolean e x p r e s s i o n s are u n d e r s t o o d as p a r t i c u l a r words over

accord-

i n g t o t h e f a m i l i a r f o r m a t i o n rules s t u d i e d i n s e n t e n t i a l l o g i c . p o s i t i o n a l v a r i a b l e s a r e words o v e r where t h e s u b s c r i p t

bl...b

i n binary notation.

For

,

var B

E

c*

I var

and

number of e l e m e n t s i n t h i s s e t .

B

I

The v a r i a b l e s i n

t h e o r d e r g i v e n by t h e i r s u b s c r i p t s . Letting

now

XI=

we mean t h a t

B,

B

b = lvar Bl (read:

. . .bn

'

an a r b i t r a r y Boolean e x p r e s s i o n ,

r e s p e c t i v e l y d e n o t e t h e set of v a r i a b l e s o c c u r r i n g i n

bits.

Xbl

i s t h e sequence of d i g i t s of a number

n B

of t h e form

{X,O,I)

Pro-

x

Elements of and

var B {0,1)

x €{O,l}b

satisfies

and

B,

,

inherit

are c a l l e d by

B)

c o n s i d e r e d as a Boolean f u n c t i o n

the

B:{O,l}b-+{O,I),

NP and Craig's Interpolation Theorem

takes value

on i n p u t

1

.. ,xb)).

( =(xl,.

x

341

One can r e c o v e r

f a m i l i a r s e m a n t i c s of s e n t e n t i a l l o g i c upon i d e n t i f y i n g "true"

and

Mod B =

For a r b i t r a r y

tautology

terminology,

( t o the f i r s t

n

Mod B r ( f i r s t n b i t s ) of t h e models of

bits)

Mod B = { O , l } b

f o r any two Boolean e x p r e s s i o n s t a u t o l o g y , an i n t e r p o l a n t var I

Here, as u s u a l ,

logies. In case

var B

n var

C

and

C

a

$

and

either i

B

g:

-

c- p,ll3

c

f o r every

{o,i)",

=(c

1

(resp.,

polynomial t i m e . strings

with regard whether tion

T u r i n g machine i n time

P=NP

one-one

. .. , g ( c n ) ,

A s usual,

P

we s h a l l b r i e f l y

s t a n d s f o r t h e c l a s s of sets

(that is,

f o r any

whether S E A * ,

NP if

say

of

whose c h a r a c t e r i s t i c f u n c t i o n

i n d e t e r m i n i s t i c polynomial t i m e ;

o r even,

bounded

(resp., nondeterministic)

t o n o n d e t e r m i n i s t i c polynomial t i m e .

(that is,

c

i s computable by a determi-

f

nondeterministic)

which are a c c e p t e d

i s computable)

(but otherwise arbi-

,...,CJEC*.

i s computable i n d e t e r m i n i s t i c

f

i s a tautology.

n a t u r a l l y i n d u c e s a one-one map:

g

by a polynomial i n t h e l e n g t h of t h e i n p u t , that

C

r ( c ) = c o n c a t e n a t i o n of g ( c l ) ,

with

I n s t e a d of s a y i n g t h a t nistic

((iB)V(C)).

mapping t h e symbols of

o n t o t h e s e t of t r i p l e t s of b i t s ;

r :C*

a

I such

a r e tauto-

I-C

or

Throughout t h i s p a p e r w e a l s o u s e a f i x e d function

is

B+C

(never occurring i n t h i s paper) Craig's

i n t e r p o l a t i o n theorem s t a t e s t h a t

trary)

) yields,

Boolean e x p r e s s i o n

B-I

a

Mod B f $.

(see [l]

such t h a t

is

B

i s an a b b r e v i a t i o n of

B-C =

and

B

that is

I,

var B f l var C,

=

by

i s t h e set

B.

is satisfiable iff

B

;

C r a i g ' s i n t e r p o l a t i o n theorem i n s e n t e n t i a l l o g i c

that

i s defined

w e a l s o set

n,
iff

Mod B

{ x € [ O p l ] b l x k B}.

I n model-theoretical of r e d u c t s

The set

"false".

with

0

the

with

1

NP

is the

same,

I t i s n o t known

i s c l o s e d under complementaSbNP

then

A*\

S€NP).

I n t h e f o l l o w i n g theorem t h e s e problems are r e l a t e d t o t h e d e g r e e of

D.MUNDICI

3 48

d i f f i c u l t y i n w r i t i n g down

Craig's interpolants i n sentential logic.

A t l e a s t one of t h e f o l l o w i n g s t a t e m e n t s h o l d s t r u e :

2 . Theorem.

(i)

P=NP;

(ii)

NP

i s n o t c l o s e d under complementation;

( i i i ) ( I n t r a c t a b i l i t y of s e n t e n t i a l i n t e r p o l a t i o n ) :

function

:

c*x c*-

c*,

$ is

if

computable i n d e t e r m i n i s t i c

polynomial t i m e , t h e n t h e r e i s a t a u t o l o g y and

var B A var C

for

B-C.

f

!if, such t h a t

f o r every

with

B-C

B,CEX*

i s n o t an i n t e r p o l a n t

$(B,C)

For t h e proof w e p r e p a r e t h e f o l l o w i n g lemma: 3.Lemma.

For

an a r b i t r a r y nonempty set t h e f o l l o w i n g

S C{O,I}*

are e q u i v a l e n t : (a)

SENP;

(b)

there e x i s t s a function

F:{l}*

-c*

d e t e r m i n i s t i c polynomial t i m e , such t h a t , have, f o r e a c h

Proof of Lemma 3 .

1x1 and

s

Fn;

and (b)

+

(a).

i s , on i n p u t

F~

and

Cook's

as f o l l o w s :

XE{O,I)*,

XI

=yl ,

(b)

.

Fnl ;

n

bits).

f i r s t compute

f i n a l l y guess a

We s h a l l o b t a i n t h e r e q u i r e d NP-complete

(see [2J ) .

,...,h)

lal,a2,...,a k )

Let

A =

and

ak=the b l a n k

v a r i a b l e s of 1

j Gk,

p

Fn

O,
symbol.

S

i n t h e l e n g t h of t h e i n p u t .

b e t h e set of s t a t e s of

T,

with

be t h e a l p h a b e t of Fix

are, f o r each

nB 1 t,i,j,q

t h e following:

We

Theorem 7 . 3 . 9 J .

be a n o n d e t e r m i n i s t i c T u r i n g machine a c c e p t i n g

bounded by a polynomial {O

by

F

well-known argument t o p r o v e t h a t t h e s a t i s f i a b i l i -

s h a l l c l o s e l y f o l l o w t h e n o t a t i o n and t e r m i n o l o g y of [5, T

n =

y E [O,l}r

...,xn=yn.

t y problem f o r Boolean e x p r e s s i o n s i s

So l e t

~ ( i ~w )e ,

=

A f a s t nondeterministic algorithm f o r

r = lvar

(a) modifying

= Mod F n r ( f i r s t

Sn{O,l)n

then let y != Fn

such t h a t

letting

n 2 1:

l v a r Fnlgn,

accepting

which i s computable i n

in t i m e

Let

h

the halting state.

T,

with

and l e t

a =1, a = O 1

m = p(n).

(OGtgm,

2 The

O,
349

NP and Craig's Interpolation Theorem

HEAD(t,i),

SYMB(t,i,j),

(1)

and

STATE(t,q),

respectively saying t h a t a t t i m e square

i,

Consider t h e

t,

a . i s p r i n t e d on t a p e J i , and T i s i n s t a t e q.

has t a p e p o s i t i o n

T

c o n j u n c t i o n of t h e f o l l o w i n g s e n t e n c e s , which uniformly

d e s c r i b e t h e b e h a v i o r of a t each t i m e

T

over any i n p u t of l e n g t h

t , each tape square

p r i n t e d on i t ,

n:

h a s p r e c i s e l y one symbol

i

i s scanning p r e c i s e l y one s q u a r e , and

T

T

i s i n p r e c i s e l y one s t a t e ; a t e a c h time

i

symbol on

if

t,

i s n o t over s q u a r e

T

t',

T

the

halts;

t h e computation s t a r t s i n s t a t e

o v e r t h e l e f t hand

0

end of

w i t h t h e t a p e o n l y c o n t a i n i n g b l a n k symbols ( w i t h

m

t h e p o s s i b l e e x c e p t i o n of t h e s q u a r e s initially

or

then

does n o t change;

a t some t i m e

the input,

i,

each t a p e square

m

through

through m+n-I

m+n-1);

has e i t h e r

1

p r i n t e d on i t ;

0

t h e changes i n t a p e symbols, head p o s i t i o n and s t a t e obey

T's

i n s t r u c tions.

w e assume t h a t t h e i n p u t i s p r i n t e d

N o t i c e t h a t , f o l l o w i n g c5],

rn

t h e tape squares and on

through

m, h, k

and t h e i n s t r u c t i o n s of

among t h o s e d i s p l a y e d i n l y w r i t t e n down i n

our

(A)

(F)

is

m+n-I.

I t i s well-known

can be w r i t t e n down as Boolean e x p r e s s i o n s

(F)

[5

t a k e c a r e of

(D)

and

t h a t (A),(B),(C) o n l y depending

are

and whose v a r i a b l e s

indeed, such expressions a ~ e x p l i c i t -

(1):

, p.236,237, e x p r e s s i o n s ( 1 ) - ( 5 ) , ( 7 ) , ( 8 ) , ( 9 ) ] :

corresponds t o ( 7 ) A (8)A (9)

T,

on

(1)A ( 2 ) A ( 3 ) , therein

.

( B ) i s (41, ( C )

i s (5).

and

The f o l l o w i n g Boolean e x p r e s s i o n s

(E):

m- 1

2m

STATE( 0 , O ) A HEAD(0.m) A ( i& SYMB( 0 , i ,k ) A (i=c+n SYMB(0, i ,k ) ) ;

(D)

mcn-I

i.A = m (sYMB(o,i,l)v S Y M B ( O , ~ , ~ ) ) .

Let

F,

=

(A)A

...A (F);

let

s = I v a r Fn I

-

I.

I t i s no

loss

of

D. MUNDICI

350

g e n e r a l i t y t o assume t h a t t h e first SYm(O,m,l), riable

Xr

x r t {O,l), F,;

...,SYMB(O,m+n-1,l)

of

n

r = 0,.

.. , s ,

are

Fn

i n t h e g i v e n o r d e r . Assume e a c h

i s assigned a b i t

F,

v a r i a b l e s of

va-

( i n t u i t i v e l y , a truth-value)

i n such a way t h a t

xo,

.. . , x

S

satisfies

i n symbols,

(2)

X

o'.

,x

*

t

Then i n t h e l i g h t of

we obtain a collection

(1)

" a t time

of t h e form

-

Fn

t ,

such and such s t a t e , and t a p e s q u a r e p r i n t e d on i t " ,

f o r each

of s t a t e m e n t s

h a s such and such p o s i t i o n , T i s i n

T

OGtLm

i

and

h a s such and such symbol OGiC2m

A

.

Now t h e d e f i n i -

Fn. t o g e t h e r w i t h ( 2 ) e n s u r e t h a t i s not a brute l e c t i o n OP i n c o h e r e n t s t a t e m e n t s , b u t r a t h e r d e s c r i b e s a l e g a l t i o n of

c e p t i n g computation of

m

tape squares

on some i n p u t

T

through

m+n-1

.

y E {O,lIn

colac-

p r i n t e d on t h e

To f i n d t h e symbol

y . originally J (j = 0 , . , n - l ) , observe t h a t i f x . = l p r i n t e d on t a p e s q u a r e m + j J t h e n SYMB(O,m+j,l) i s " t r u e " and, by ( 1 ) w e have t h a t y . = 1 .

..

J

x . = 0 , t h e n SYMB(O,m+j,l) J i s " t r u e " , s o t h a t by ( l ) ,

On t h e o t h e r hand, i f by ( E ) , that 0

SYMB(O,m+l,2)

a =1

1 ,... ,n-1,

and

a = 0).

Mod F~

Conversely ,i f

V

of

T

xo,.. /=

, xs

XO'...'xn-l'xn,... (3)

In definitive,

2

so t h a t

..

on i n p u t

E S

1

T.

xo

i n t i m e a t most

m,

s a f e l y assume t h a t t h e i n p u t i s p r i n t e d on t a p e s q u a r e s m+n-1, Using x

t h e whole t a p e (1)

,...,x n

S

having

Therefore, I n symbols,

t h e n t h e r e i s an a c c e p t i n g computation

... , xn-1

X0'

J

,...,xn- 1 E S. b i t s ) E s n { 0 . 1 ) ~.

n

*Xn-l

y. = x.

y . = 0 (recall J f o r each j =

i s a c c e p t e d by

implies

Fn

r (first

xo..

. ,xn-1

i s " f a l s e " and,

2m+l

many s q u a r e s ,

0.1.

and w e

m

can

through

...,2 m .

w e can unambiguously a s s i g n t r u t h v a l u e s t o all the

v a r i a b l e s of

of i n f o r m a t i o n c o n t a i n e d i n p u t a t i o n w e t h e n have, (4)

X0'.

Therefore,

xo,

.. ,xS

...,xn-1

v.

F

n ' Since t h e l a t t e r i s a

by d e f i n i t i o n of !=

E S

X O p . . . l X n-1 ' by j u s t coding t h e p i e c e s

legal

com-

Fn :

Fn

implies that there is

yo,

...,y s I=

Fn

35 1

NP and Craig's Interpolation Theorem

with

yo=xo,.

n

s

(5)

.. , Y , - ~ = X ~ - ~ . C _ Mod Fn

{O,ljn

Now ( 3 ) and ( 5 )

I n symbols,

r (first n

j o i n t l y y i e l d a f i r s t d e s i r e d c o n c l u s i o n about

(a) 3 (b)

To complete t h e proof t h a t

[5,

But t h i s i s well-known

( A ) A ( B ) A (C)A(F) ;

t o hold f o r

portional t o

p4(n),

t h e r e i n i s a t worst pro-

(1)-(9)

i s so simple.

and t h a t ( 1 ) - ( 9 )

The same

c l u s i o n h o l d s i n t h e p r e s e n t c a s e , w i t h t h e same argument. c l u d e s t h e proof t h a t

+ (b)

(a)

Assume s t a t e m e n t s

Mod I n

with

r

the

( i i i ) i n Theorem 2 a r e both

c*

I: {If*+

which i s computable

NP

(6)

s

=

I?

is satisfiable},

{ x ~ j o , ~ ) ~-'(XI l

map d e f i n e d i n s e c t i o n 1 . S i n c e t h e set of s a t i s f i a b l e Boolean e x p r e s s i o n s t h e n s o i s t h e set

( s e e r5, 7.3.51) =

In = I ( 1 " ) . w e

n21:

Proof of L e m m a 4. is in

T h i s con-

I

i n d e t e r m i n i s t i c polynomial t i m e , such t h a t , l e t t i n g have f o r each

con-

and completes t h e proof of lemma 3

( i i ) and

Then t h e r e i s a f u n c t i o n

false.

see

i s claimed t o h o l d ,

m8

where an upper bound of t h e form

p.2381

a f t e r o b s e r v i n g t h a t t h e l e n g t h of

4.Lemma.

F.

w e must show t h a t t h e map

d e s c r i b e d above i s computable i n d e t e r m i n i s t i c polynomial

I n I----, Fn

time.

bits).

r

{ x € {O,l]*]

S

g i v e n by

is satisfiable].

-'(x)

T h i s c l e a r l y f o l l o w s from t h e d e f i n i t i o n of

-S =

assumed t o be f a l s e , t h e n t h e set

r

.

Since

is in

S

{O,l{*\

(ii) i s

NP, t o o .

Withcut l o s s of g e n e r a l i t y t h e r e are n o n d e t e r m i n i s t i c T u r i n g machines T

T'

and

-

accepts

S

and a polynomial

p

i n t i m e bounded by

exist functions

H':

H,

such t h a t p.

~I}*-b~*

(7)

s

n 1o.1)~

s n {o.i)."

accepts

n

Hn = H ( 1 ) ,

n21:

~~r

=

Mod

=

Mod H,:

=+

and

S

TI

(b)) t h e r e

which are computable i n d e t e r -

m i n i s t i c polynomial t i m e such t h a t , l e t t i n g

w e have f o r each

T

B y Lemma 3 ( ( a )

(first

n

bits),

I (first

n

bits).

and

HI

n

= H'(ln),

D. MUNDICI

352

Since

and

T

b o t h a c t i n time bounded by

T'

t i o n of t h e proof of

p , by an e a s y i n s p e c -

w e can s a f e l y s t i p u l a t e t h a t , i n

Lemma 3

ad-

dition, =

v a r H r\ v a r HA

(8)

n

[SYMB(O,m,l)

,...,SYMB(O,m+n-1,l))

w e j u s t u s e t h e same symbols f o r t h e f i r s t

HA,

t h e n rename t h e o t h e r v a r i a b l e s of

condition

is satisfied.

(8)

Hn

n

v a r i a b l e s of

w e have t h a t

xO , . . . , ~ n - l E S

and

so that

Consider now t h e Boolean e x p r e s s i o n

..., xn - l , x n ,...,x

xo,

Hn

( i f necessary)

i f t h e l a t t e r c o n j u n c t i o n were s a t i s f i a b l e

H n h HA:

t h e s i s ) , say

;

n5

(absurdum hypo-

HnA HA , t h e n by ( 7 )

satisfies

, which i s i m p o s s i b l e .

Therefore

we get: (9)

Hn-iHv

i s a tautology, f o r each

n

S i n c e w e a r e assuming t h a t

(iii) i s f a l s e , l e t

that is,

( i i i ) i n Theorem 2 :

example t o

n& 1.

$

$

be a c o u n t e r -

i s computable i n

t e r m i n i s t i c polynomial t i m e , and misses no i n t e r p o l a n t s .

,

= $(Hn

1HA);

then

So l e t

deIn

has t h e following properties:

In

(10)

H n I In

and

(11)

v a r In

{SYMB(O,~,I)

(12)

t h e mapping

In-

a r e t a u t o 1o g i e s ;

i H A

,...,S Y M B ( O , ~ + ~ - ~ , ~ ) ;) .

=

i s computable i n d e t e r m i n i s t i c

In

In

polynomial t i m e . Clause

(12)

i s a consequence of our assumptions about

w i t h t h e f a c t t h a t t h e maps

n

+ Hn

and

computable i n determin s t i c polynomial t i m e . in (13)

( 1 0 ) . and from Mod H~

Hence, by (14)

and

(8

r (first

n

(11)

bits)

C_

w e have Mod I

n

.

Mod In.

S i m i l a r l y , from t h e second t a u t o l o g y i n

(15)

a r e both

From t h e f i r s t t a u t o l o g y

(7) we get

S n{O,lJnC

and from

n w HA

together

(8),(11),(7)

s n {O,l)n

C_

we obtain

ModlIn

.

( l o ) , w r i t t e n as

HI

n

-

71

n'

NP and Craig's Interpolation Theorem From

(14)

and

(15), recalling

(16)

Mod I n = S n j 0 , l ) "

353

we get

(6)

is satisfiable

{ X € { O , I } ~ ~r - ' ( x )

=

w h i c h c o m p l e t e s t h e p r o o f o f our L e m m a .

5 . End of t h e p r o o f of Theorem 2. W e s h a l l prove t h a t if

(iii)

I:{l)**C*

To t h i s p u r p o s e , l e t

M

and

a r e b o t h f a l s e , t h e n P=NP.

(ii)

b e as g i v e n by L e m m a 4 , and l e t

b e a d e t e r m i n i s t i c T u r i n g machine computing e a c h

bounded by a p o l y n o m i a l

q

i n the length

B E E *a,s compute

(D2)

w r i t e down

n =

(D3)

using

w r i t e down

(D4)

check whether

cess

M

Ir(B)l r(B)

3lBI;

=

In; In.

c a n b e c a r r i e d o u t i n d e t e r m i n i s t i c p o l y n o m i a l time

(D4)

q ( n ) , and

r(B)

i n t h e l e n g t h OP

L e t now

B.

\In(

i s bound-

b i t s . The a b o v e p r o -

B

i n t i m e bounded by a p o l y n o m i a l

Therefore we conclude t h a t , under our assumptions,

T h i s c o m p l e t e l y provets

TAUTSZ*

t a u t o l og i e s

n

provides t h e required d e c i s i o n procedure f o r s a t i s -

(Dl)-(D4)

holds.

ensure t h a t

M

i s a sequence of

f i a b i l i t y of any Boolean e x p r e s s i o n

P=NP

on

r(B);

as c l a i m e d , s i n c e t h e p r o p e r t i e s o f e d by

of t h e i n p u t .

follows:

(Dl)

Notice t h a t

i n time

for satisfiability is,

A fast(deterministic)decision p r o c e d u r e

input

n

In

t h e Theorem.

d e n o t e t h e set o f B o o l e a n e x p r e s s i o n s w h i c h

are

. A t l e a s t o n e of t h e f o l l o w i n g s t a t e m e n t s h o l d s t r u e :

6.Corollary. (I)

TAUT

i s accepted i n d e t e r m i n i s t i c polynomial time;

(11)

TAUT

i s n o t accepted i n n o n d e t e r m i n i s t i c polynomial time;

(111)

same a s s t a t e m e n t

P r o o f . I t i s well-known

(iii) that

i n Theorem

TAUT

is i n

2.

P

iff

P=NP

(see [2]).

S i m i l a r l y , TAUT i s i n NP i f f NP i s c l o s e d u n d e r c o m p l e m e n t a t i o n (see, e.g.,

[3.

1.11

).

Now a p p l y Theorem

2.

D.MUNDICI

354

The above C o r o l l a r y i s s t a b l e under r e l a x a t i o n of t h e upper bounds for

computations, a s w e s h a l l

( d e t e r m i n i s t i c and n o n d e t e r m i n i s t i c )

7

s e e i n Theorem functions

below.

f : PI-

.A

El

%

A s usual,

3

set

C

N

5

all

i s c l o s e d under composition

N

5

i f f t h e composition of any two f u n c t i o n s i n in

t h e set of

denotes

i s still a function

c l o s u r e under sum i s s i m i l a r l y d e f i n e d .

;

7. Theorem.

9C%

Let

be an a r b i t r a r y set c o n t a i n i n g t h e polynoThen a t l e a s t one

m i a l s and c l o s e d under composition and sum.

3

( p e r h a p s depending on

)

of t h e f o l l o w i n g s t a t e m e n t s h o l d s t r u e :

TAUT i s a c c e p t e d by some d e t e r m i n i s t i c T u r i n g machine i n t i m e

(Ig)

3

bounded by a f u n c t i o n of (113)

( i n t h e l e n g t h of t h e i n p u t ) ;

i s n o t a c c e p t e d by any n o n d e t e r m i n i s t i c T u r i n g machine

TAUT

- c*

i n t i m e bounded by a f u n c t i o n of For every

(1113)

p

:c * X

c*

3.

;

,

4,

if

a

i s computable by

,

d e t e r m i n i s t i c T u r i n g machine i n time bounded by a f u n c t i o n of then t h e r e i s a tautology such t h a t

C , with

B--,

-

n var

var B

B,CEC*,

i s n o t an i n t e r p o l a n t f o r

H(B.C)

B

f @,

C

C.

For t h e proof w e modify L e m m a s 3 and 4 as f o l l o w s :

8. Lemma.

@

Assume

f

Then t h e r e e x i s t s

F:

i s accepted by a

non-

i n t i m e bounded by a f u n c t i o n

f e z

S E{0,1}*,

d e t e r m i n i s t i c T u r i n g machine (1)"-

T

and

z*

S

by a d e t e r -

which i s computable

3 ,

m i n i s t i c T u r i n g machine i n t i m e bounded by a f u n c t i o n of letting lvar Proof.

w e have f o r each

Fn = F ( I n )

F

~

n, I

and ~

s

n {o,i)"

=

Write down

Fn

T

i s now g i v e n by

and n o t e t h a t

of t h e p r e s e n t Lemma.

Fn

f

such t h a t

n>l: Mod

~~r

Argue e x a c t l y as i n t h e proof of Lemma 3

upper bound f o r

.

(first

n

bits).

( ( a ) =3 ( b ) ) ;

( i n s t e a d of

p

the

therein).

s a t i s f i e s t h e second r e q u i r e m e n t

T o see t h a t t h e mapping

n W F n

i s computable

by a d e t e r m i n i s t i c T u r i n g machine i n t i m e bounded by a f u n c t i o n of by analogy w i t h t h e f i n a l o b s e r v a t i o n i n t h e proof of Lemma 3,

'5 ,

first

note that lFnl i s a t most p r o p o r t i o n a l t o m4 , t h a t i s , p r o p o r t i o n a l 4 can be w r i t t e n down i n time n o t much g r e a t e r t o f ( n ) . Again, than

IFn)

,

F,

say f o r definiteness lFn12

.

But t h e f u n c t i o n

f8(n)

355

NP and Craig's Interpolation Theorem

2 !FA2 still is in 9.Lemma.

(119) and

).

r

{0,1)



*\ z

Z

be defined by

c*\TAUT

. 3Clearly,

1°C

=

p ( x ) is

{O,l>*I

r

,

is accept-

2

in time bounded by

Y

is in

.

w

we can safely assume

c*

HI : {I}*-

w'.

=

By Lemma 8

3

)

zn{o,I)n

=

Mod

zn(O,l}n

=

Mod

in

Hl;r

Since

(1113)

sum,

deterministic Turing

such a way

5

(using the

that

(first n

bits),

(first n

bits).

and

-

Arguing now as in the proof of Lemma 4 one shows that is a tautology.

W' in

there are functions

which are computable by

3

-

Z=

is closed under

machines in time bounded by the same function U E closure properties of

a

NP, hence the set

is accepted by a nondeterministic Turing machine

time bounded by a function ~ ' € 3 Since

H,

such that letting

is a tauto~ogy).

By assumption, and by definition of

ed by a nondeterministic Turing machine €unction ~

3 ,

n21

(x E {O,l}n

z c_ {o,I)*

Let

a taut 01ogy

a

which is computable by a deterministic

we have for each Mod In =

Proof.

(1119) are both false. Then there is

machine in time bounded by a function of

In = I(ln)

.

3

by the assumed closure properties of

- x*

Assume

function I: {I}* Turing

3,

Hn is assumed to be false, let

- I Hn '

$

be

computable by a deterministic Turing machine in time bounded by some function

bE

4

tautology, $(Hn

3

, 1 HA);

, with the property that whenever B -C is a (B,c) is an interpolant for B + C. Let In

the mapping

n !-+

In

Turing machine in time bounded by some function d € g tained as a suitable composition of the functions u,b some polynomial).

The mappingnI,n

(d

can be ob-

together with

is now proved to satisfy all

our requirements by the same argument as in the end of the proof Lemma

4.

This completes the proof o€ Lemma

3

Turing machine

accepting

Thus Theorem 7

is proved.

of

9.

Arguing now as in section 5 , using Lemmas 8 and 9 closure properties of

=

is computable by a deterministic

and the

, one easily produces a deterministic TAUT

in time bounded by a function of

5.

D. MUNDICI

356

10.

F u r t h e r Topics.

I n t h i s f i n a l s e c t i o n w e survey

what i s known on t h e complexity

of

e s h a l l s t a t e a number of r e s u l t s C r a i g ' s i n t e r p o l a t i o n theorem. W c o n c e r n i n g t h e r a t e of growth of i n t e r p o l a n t s , b o t h i n s e n t e n t i a l Boolean e x p r e s s i o n s for s e n t e n t i a l l o g i c

and i n f i r s t - o r d e r l o g i c .

are p a r t i c u l a r words over a l p h a b e t

c

as d e f i n e d i n s e c t i o n 1 .

Sentences of f i r s t - o r d e r l o g i c are u n d e r s t o o d as p a r t i c u l a r over some s u i t a b l e a l p h a b e t ).

( s ee [I]

rules

c' ,

words

according t o t h e f a m i l i a r formation

I n s e n t e n t i a l l o g i c t h e precise determination

of t h e r a t e of growth

of i n t e r p o l a n t s i s an open

(and i m p o r t a n t )

The Pollowing Theorem s t a t e s t h a t if s e n t e n t i a l i n t e r p o l a n t s

problem.

t u r n out t o grow polynomially, i n d e t e r m i n i s t i c polynomial

t h e n e v e r y f u n c t i o n which i s computable

(Turing) t i m e ,

has c i r c u i t depth

p o r t i o n a l t o t h e l o g a r i t h m of t h e i n p u t l e n g t h .

pro-

T h i s would p r o v i d e

a p o s i t i v e s o l u t i o n t o a c e n t r a l open problem of computation t h e o r y (see [I I] )

.

Assume t h e r e e x i s t s a polynomial

10.1 Theorem. whenever

an i n t e r p o l a n t

with

I

111

t h e r e s t r i c t i o n of

f

d e p t h Fn See [6

Recall function

,

< p ( 1BI +

F1 , F 2 ,

to

,

2.3.23)

f : {O,l}n-

I

tO.1)

,

Then f o r e v e r y f u n c t i o n

c i r c u i t s , with

d e l a y complexity of

computing

n

c>O,

.

n = 1,2,...

. that the delay

complexity

of a Boolean

i s t h e d e p t h oP t h e smallest d e p t h

(over our f i x e d b a s i s

Boolean e x p r e s s i o n

F

such t h a t , f o r some

f o r each

)

{A, V

. As u s u a l ,

any

i s r e g a r d e d as a Boolean f u n c t i o n o v e r i t s own

B

variables, via the identification

propagate t o

ICl).

... of

[O,lJn,

< c - l o g2 n

Theorem 2 . g

(from b 0

circuit for

such t h a t

which i s computable i n d e t e r m i n i s t i c polynomial

(0'1)

t i m e t h e r e i s a sequence

ProoP.

p

i s a t a u t o l o g y i n s e n t e n t i a l l o g i c , one can f i n d

B-C

f : {O,l]*-

for t h e n e c e s s a r y background:

See f l 0 , 2.21

B

1 = 'ltrue'l

i s , roughly,

and

0 = "false"; the

t h e t i m e needed f o r i n p u t s t o

t h e o u t p u t , i n t h e Pastest c i r c u i t computing

B.

The

f o l l o w i n g Theorem t h e n s t a t e s t h a t t h e t i m e needed by t h e f a s t e s t

NP and Craig's Interpolation Theorem

c i r c u i t t o compute terpolant

( t h e Boolean f u n c t i o n c o r r e s p o n d i n g t o ) any

needed t o compute e i t h e r of 10.2 Theorem. d<620) and I

C

or

B

C:

F o r i n f i n i t e l y many

t h e r e i s a tautology

d€N

(and s t a r t i n g w i t h

some

i n s e n t e n t i a l l o g i c , with

B+C

B

having t h e i r d e l a y complexity S d , such t h a t e v e r y i n t e r p o l a n t

h a s a d e l a y complexity

Proof.

in-

B - + C may happen t o be g r e a t e r t h a n t h e time

for

I

351

17,

See

d

I

>d +

(1/3)log2(d/2).

Theorem 2.51

As remarked above, i n s e n t e n i a l l o g i c t h e r e i s a t p r e s e n t no deEini-

t i v e e s t i m a t e oP t h e r a t e oP growth of ICl

, where

Theorem

a s a f u n c t i o n of IBI and

111

i s a smallest l e n g t h i n t e r p o l a n t € o r

I

(See [6,

B-C.

1.93 f o r an upper bound, and t r y t o improve i t ) .

By c o n t r a s t

i n f i r s t - o r d e r l o g i c w e have: 10.3 Theorem. ( i ) In t h e arithmetical hierarchy t h e r e i s a

TI1 - f u n c t i o n

g i v i n g an upper bound f o r t h e l e n g t h oP P i r s t - o r d e r (*)

whenever

B+C

interpolant ( i i ) No

c

1

I

-function

with

B+C

as i n

(*)

[ 6 , Theorem 3.13

Theorem

b

ICl). can g i v e an

.

Proof.

,

b(lBI+

111

( i . e . , no r e c u r s i v e f u n c t i o n )

111

[4

i n t e r p o l a n t s , i.e..

i s v a l i d i n f i r s t - o r d e r l o g i c , t h e r e i s an for

upper bound f o r ( i ) See

b: N+pl

.

(11)

T h i s c a n be e x t r a c t e d from

11.

Due t o i t s a s y m p t o t i c c h a r a c t e r , t h e above Theorem

10.3 ( i i ) g i v e s

no i n f o r m a t i o n on t h e p o s s i b l e l e n g t h s of i n t e r p o l a n t s € o r s h o r t plications.

The 'following i s a non-asymptotic r e s u l t :

We c a n w r i t e down a v a l i d i m p l i c a t i o n i n f i r s t - o r d e r

10.4 Theorem.

logic,

with

B---+C

l B ( .lC(<1145

pol a n t w e have : .*2

III> 2'

Proof.

See

im-

&,

Theorem 3 . 4 .

1

such t h a t whenever

seven t w o ' s

.

I

i s an i n t e r -

D.MUNDICI

358

References Chang C . C . and Keisler H . J . , Model Theory (North-Holland, Amsterdam, second e d i t i o n 1 9 7 7 ) . Cook S . A . , The complexity of theorem p r o v i n g p r o c e d u r e s , I n : P r o c e e d i n g s of t h e T h i r d Annual ACM Symp. on t h e Theory of Computing, Yay 1 9 7 1 , pp.151-158. Cook S . A . and Reckhow R.A., The r e l a t i v e e f f i c i e n c y of p r o p o s i t i o n a l proof s y s t e m s , J o u r n a l of Symb.Logic 44 ( 1 9 7 9 ) 36-50. Friedman H., The complexity of e x p l i c i t d e f i n i t i o n s , Advances i n Mathematics 2 0 ( 1 9 7 6 ) 18-29. Machtey M. and Young P . . An I n t r o d u c t i o n t o t h e General Theory of Algorithms (North-Holland, Amsterdam, t h i r d p r i n t i n g 1 9 7 9 ) . Mundici D . , Complexity of C r a i g ' s i n t e r p o l a t i o n , Annales SoC. Math.Pol., Series I V : Fundamenta I n f o r m a t i c a e V.3-4 (1982) 261- 278. Mundici D . , A lower bound f o r t h e complexity of C r a i g ' s i n t e r p o l a n t s i n s e n t e n t i a l l o g i c , Archiv math.Logik ( 1 9 8 3 ) t o appear. Yundici D . , D u a l i t y between l o g i c s and e q u i v a l e n c e r e l a t i o n s , T r a n s a c t i o n s A m e r Math SOC 27 0 ( 1 98 2) 1 1 1 -1 29.

.

. .

Mundici D . , Compactness, i n t e r p o l a t i o n and F r i e d m a n ' s t h i r d problem, Annals of Mathematical Logic 22 ( 1 982) 197-21 1

.

Savage J.E.,

The Complexity of Computing (Wiley, New Y o r k 1 9 7 6 ) .

Schnorr C.P., The network complexity and t h e T u r i n g machine complexity of f i n i t e f u n c t i o n s , Acta I n f o r m a t i c a 7 (1976) 95-107.