121
FUNCTIONAL- COMPLETENESS CRITERIA FOR FINITE DOMINS by PETER SCHOFIELD A thesis submitted to the University of Leicester for the degree of Doctor of Philosophy

FUNCTIONAL- COMPLETENESS CRITERIA FOR FINITE

Embed Size (px)

Citation preview

FUNCTIONAL- COMPLETENESS CRITERIA

FOR FINITE DOMINS

by

PETER SCHOFIELD

A t h e s i s s u b m itte d to th e U n iv e r s i ty o f L e ic e s te r

f o r th e d e g re e o f D o c to r o f P h ilo so p h y

UMI Number: U296355

All rights reserved

INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed,

a note will indicate the deletion.

DiSËürtâtion Publishing

UMI U296355Published by ProQuest LLC 2015. Copyright in the Dissertation held by the Author.

Microform Edition © ProQuest LLC.All rights reserved. This work is protected against

unauthorized copying under Title 17, United States Code.

uesfProQuest LLC

789 East Eisenhower Parkway P.O. Box 1346

Ann Arbor, Ml 48106-1346

d a is ’»

r n , s A

\S-. ? . k l

SUMMARY

N ecessary and s u f f ic ie n t c o n d itio n s f o r the fu n c tio n a l

com pleteness o f a s e t F o f fu n c tio n s w ith v a r ia b le s and va lu es

ran g in g over N = [ 0 , 1 , . . . , n J , where n ^ 1, a re in v e s t ig a te d

and in p a r t i c u l a r , com pleteness c r i t e r i a f o r a s in g le fu n c tio n

a re determ ined#

Complete s o lu tio n s a re known in th e s p e c ia l c a se s n = 1 ,2 ,

and r e s u l t s about th e se s p e c ia l c ase s which a re o f use in

fo rm u la tin g g e n e ra l theorem s a re d iscussed#

Proceeding to the g en e ra l case some p re lim in a ry c r i t e r i a

(which presuppose t h a t c e r ta in 2 -p lace fu n c tio n s a re gen era ted

by F) fo r th e fu n c tio n a l com pleteness o f F a re d e r iv e d .

These r e s u l t s show th a t th e s e t c o n s is tin g of a l l 2 -p lace fu n c tio n s

i s complete# In th e s p e c ia l case n + 1 = p (a prim e number) th e

fu n c tio n s o f . F a re shown to have a s p e c ia l form , and t h i s i s

used in some i l l u s t r a t i o n s of complete subsets#

The v a lu e sequence o f a fu n c tio n s a t i s f y in g th e S-éupecki c o n d itio n s

( th a t i s , depending on a t l e a s t 2 o f i t s argument p la c e s , and ta k in g

a l l n + 1 v a lu es of N) i s now examined, and some p ro p e r t ie s

o f such a fu n c tio n a re found# These r e s u l t s a re th en used in

dem onstra ting th e com pleteness o f a s e t F which g e n e ra te s a l l

1-p la c e fu n c tio n s , to g e th e r w ith a fu n c tio n s a t i s fy in g th e

S-ôupecki cond itions*

Our main r e s u l t s g ive improved s u f f i c i e n t co n d itio n s f o r

th e com pleteness o f F . In p a r t i c u l a r a s e t F i s com plete

i f i t g e n e ra te s a t r i p l y t r a n s i t i v e group of perm u ta tio n s o f N

and co n ta in s e i t h e r ( i ) only a s in g le fu n c tio n o r ( i i ) a t l e a s t

one fu n c tio n s a t i s f y in g th e S-ftupecki c o n d itio n s , th e l a t t e r a p a r t

from c e r ta in e x c e p tio n a l c a s e s . A d e ta i le d in v e s t ig a t io n

shows t h a t th e se occur only when n = 2 o r when n + 1 i s

a power o f 2 and a l l fu n c tio n s o f F a re l in e a r in each

v a r ia b le , r e l a t i v e to some mapping of N a s a v e c to r space

over Zg.

F in a l ly a d i f f e r e n t mapping o f N in to i s co n sid e red ,

and i t i s shown th a t th e fu n c tio n s of . F can be g iven a unique

r e p re s e n ta t io n r e l a t i v e to t h i s mapping. This r e p re s e n ta t io n

i s th e n used to f in d some examples of complete subsets*

The work of th i s th e s i s i s claim ed a s o r ig in a l except

where re fe re n c e to a n o th e r 's work i s made.

I shou ld l ik e to thank P ro fe sso r R .L .G oodstein f o r h is

h e lp and encouragement and f o r su g g estin g th e problems

considered h e re . I am a ls o g r a t e f u l to Dr. Roy 0. D avies

f o r t r a n s la t in g v a rio u s R ussian papers on t h i s work, and f o r

h is h e lp in th e p re s e n ta tio n and c o r re c t io n o f th e ty p e s c r ip t .

CONTENTS

SUMMARY

PAGECHAPTER I . INTRODUCTION

1 .1 . Basic d e f in i t io n s 1

1.2* Composition and com pleteness 2.1.3* 1-p la c e fu n c tio n s A-

1 .4 , Ba ckground 8

CHAPTER I I . COMPLETENESS CRITERIA IN Ei AND Eg

2 .1 . In tro d u c tio n 11

2 .2 . Completeness in Ej 11

2.3« Pre-comple t e su b se ts 14

CHAPTER I I I . SIMPLE GENERAL THEOREMS

3 .1 , In tro d u c tio n 17

3 .2 . Simple Completeness Theorems 18

3.3» The s p e c ia l case n + 1 = p (a prim e number) 21

CHAPTER IV. SOME FUNDAMENTAL RESULTS

4 .1 , In tro d u c tio n 32-

4 .2 , Consequences o f th e S-êupecki c o n d itio n s 33

4*3. Fundamental com pleteness c r i t e r i a in En 39

CHAPTER V. MAIN RESULTS

5*1 • In tro d u c tio n

5*2, A u x ilia ry r e s u l t s

3 . 3, An ex ce p tio n a l case 5 6

5 «4* In tro d u c tio n o f v e c to r n o ta t io n 52

3. 3 , Main conclusions 73

82CHAPTER V I. A MAPPING CF En INTO E?

6.1» In tro d u c tio n

6 .2 , Permanent s e ts 8^

6 . 3 , Some com plete su b se ts o f En 9?

BIBLIOGRAPHY. 110

I - INTRODUCTION - I

1.1 B asic D e f in it io n s

L e t n be a n a tu ra l number n ^ 1 and l e t N = { 0 , 1 , . . . , n j .

Denote by Ep th e s e t o f a l l fu n c tio n s ^(xi ) , o f any number

o f v a r ia b le s , whose v a r ia b le s range over N and whose va lues a re

elem ents of N.

REMARK. Mary a l te r n a t iv e n o ta t io n s f o r th e se b a s ic s e ts e x i s t .

For example, Sâlomàa... [27] [18] [29] uses the s e ts N + 1 = [ l , 2 , . . .,n+1 j"A/ ^

and 1^4,1 and Y ab lonsk ii [^9] uses th e s e ts = |0 , 1 , . . . , n j and

Pn* i . We have chosen th e above f o r reaso n s o f s im p l ic i ty o f n o ta t io n

and the f a c t th a t th e zero elem ent p lays an im p o rtan t ro le in some of

ou r r e s u l t s . I t should be no ted th a t th e number o f elem ents in th e b a s ic

s e t N i s n + 1, whereas the s e t o f a l l fu n c tio n s over N i s denoted

by En.

We c a l l the sequence o f values tak en by a fu n c tio n f ( x i , . . . , 2 j < ) f 3m ;

and w r i t t e n in th e o rd e r f ( 0 , . . . , 0 , 0 ) , f ( 0 , . . . , 0 , l ) , . . . , f ( 0 , . . . , 0 , n ) ,

f ( 0 , . . . , 1 , 0 ) , , . . , f ( 0 , . . . , 1 , n ) , . . . , f ( n , . . . , n ) th e value sequence o f f .

We sometimes r e fe r to an ordered s e t of k values (% , . . . ) as a p o in t.

denoted by a s in g le l e t t e r /i, and we may then w rite f( /i) fo r f (di > • • •>/ k ) •

The t r a d i t i o n a l way o f w r it in g down th e va lu e sequence of f i s in

th e form o f a t r u t h ta b le , and i t i s w e ll known th a t ev ery k -p lace

fu n c tio n f (xi c % can be u n iq u e ly re p re se n te d by a t r u t h ta b le

o f le n g th (n + 1)^* Obviously th e re a re (n + l ) d i s t i n c t

k -p la c e fu n c tio n s f ( x i , . . . , x k ) f % . When k = 2 th e most convenient

way o f w r i t in g down th e va lue sequence o f a 2 -p lace fu n c tio n f (x ,y )e

i s in th e form o f an (n + 1) x (n + 1) m atrix*so :

i ‘( x , y ) 0 1 . . • n

0 Pel * • • Pen1 P i i • • • Pin

n Pni • • • Pnn

where f o r each p a i r o f in d ic é s i , j (O ^ i , j ^ n) = f ( i , j ) .

1 .2 . Com position and Completeness

We c a l l xj an e s s e n t ia l v a r ia b le of a k -p la c e fu n c tio nCLJ_1 ,

f (xi , . . . ) i f th e re e x is t numbers a i , • . . ,a j+ 1 , . . . ,ai< such th a tA

f ( a i , . . . , a j w i , % j , a j * i , . . . , % k ) i s n o t c o n s ta n t. Any v a r ia b le of f

which i s n o t an e s s e n t i a l v a r ia b le we c a l l an in e s s e n t ia l v a r ia b le o f

f . For example, th e 2 -p lace fu n c tio n f ( x ,y ) = g (x ) , where g(x) i s n o t a

c o n s ta n t 1-p la c e fu n c tio n , has one e s s e n t ia l v a r ia b le x , and one in e s s e n t ia l

v a r ia b le y .

L e t th e i n f i n i t e s e t o f v a r ia b le s a v a ila b le f o r the fu n c tio n s o f

En be enum erated as fo llo w s: % , % , . . .,Z|( ***» (we sometimes w r ite

x ,y f o r Xi,3fe r e s p e c t iv e ly ) . For each f i n i t e p o s i t iv e in te g e r k we s h a l l

agree to re p re s e n t every k -p lace f \in c tio n o f En by a fu n c tio n in th e

f i r s t k v a r ia b le s o f th e above enum eration; f o r example f ( x i , . . . , x k )

( th e f i r s t k v a r ia b le s can be re g a rd ed as the v a r ia b le p la ce s of f ) .

By a com position o f fu n c tio n s o f a su b se t F C i t i s p o ss ib le

to produce f u r th e r fu n c tio n s (n o t n e c e s s a r i ly d i s t i n c t from th e fu n c tio n s

o f F ) . I f we d e fin e a p ro je c t io n fu n c tio n g(%i , ) , fo r any

f ix e d in te g e r k ^ 1, to be any fu n c tio n o f the form

g(xi ) = XL, f o r some index i , 1 ^ i ^ k , then

we have the fo llow ing concise d e f in i t io n o f com position.

DEFINITION'. A su b se t o f fu n c tio n s F of ^ i s c lo sed under

com position i f f (gi f F whenever f e F and f o r each

i ( i = 1 , . . . , k ) gt f F o r gj, i s a p ro je c t io n fu n c t io n . The fu n c tio n s

g l a l l have the same number o f argument p la ce s and i f th i s number i s -6

^ (ê l f ~ > • • • , . . .,Q< (3Ql , . • • ,X^) ) .

By the closu re under com position o f a su b set F o f En we mean the

s m a lle s t s e t F say , such th a t F Ç F and F i s c lo se d under^ F g en era te s by com position a fu n c tio n ^

com position . We say a ls o t h a t a s u b s e t^ if and only i f ÿ belongs to

th e c lo su re o f F under com position .

The s u b se ts o f E whose c lo su res under com position are E i t s e l f

are of s p e c ia l i n t e r e s t , and we c a l l a su b se t F com plete i f and o n ly

i f the c lo su re o f F under com position is Ep . That i s , F i s complete

i f and on ly i f F g e n e ra te s every member o f E!n •

REMARK. In e f f e c t , the above d e f in i t io n o f com position re q u ire s F to

be c lo sed w ith re s p e c t to th e fo llo w in g types of o p e ra tio n s :

( i ) re -o rd e r in g and id e n t i fy in g th e v a r ia b le s o f a fu n c tio n f f F;

( i i ) in tro d u c in g (e lim in a tin g ) any number o f in e s s e n t ia l v a r ia b le s

in to (from) the argument p la ce s o f a fu n c tio n f € F;

( i i i ) s u b s t i tu t in g fu n c tio n s o f F in to any number o f argument p lace s

o f a fu n c tio n f f F .

1*3 1 - p lace fu n c tio n s

I f f ( x i , . . * ,x k ) € Efl has a t most one e s s e n t ia l v a r ia b le ,

x t say , th en th e re i s a 1 -p lace fu n c tio n g such th a t

f ( x i , . . . ,% k ) = g (x i)

f o r a l l v a lu e s o f . I t fo llow s t h e r e f o r e th a t th e c lo su re

under com position o f th e su b se t o f a l l 1-p la c e fu n c tio n s o f i s

th e su b se t o f Ep c o n s is tin g o f a l l fu n c tio n s w ith a t most one e s s e n t i a l

v a r ia b le . This su b se t i s n o t i t s e l f s in c e E co n ta in s fu n c tio n s

w ith a t l e a s t two e s s e n t i a l v a r ia b le s ; f o r example, f (xi ,3^ , . • .,:q< )= % +%

(modulo n + 1 ) . T herefo re the s u b se t o f a l l 1 -p lace fu n c tio n s o f Ep i s n o t

com plete.

REMARK. The q u estio n o f th e s m a lle s t s e t o f 1-p la c e fu n c tio n s vdiich

g en era tes every 1-p la c e fu n c tio n o f E has been in v e s t ig a te d by Salomaa

in [2 7 ], where he proves th a t i f n = 1 th e n a t l e a s t 2 1-p la c e-v-w

fu n c tio n s a re needed to g en e ra te a l l 1-p la c e fu n c tio n s , and i f n > 2

a t l e a s t 3#

The fo llow ing n o tio n s concerning 1-p la c e fu n c tio n s w i l l p la y an

e s s e n t i a l ro le in our r e s u l t s .

A 1 -p lace fu n c tio n o f i s s a id to be o f type

where th e a 's a re n a tu ra l numbers s a t i s fy in g a% + % + . . • + a^ = n+1,

i f i t takes ex a ctly t d is t in c t va lu es , v ^ , . . . , v ^ sa y , where v

i s tak en e x a c tly aj, tim e s . I t i s easy to see th a t a 1 -p la c e fu n c tio n

o f type [ l , 1 , . . . , l ] , w ith n + 1 u n i t s , i s a p e rm u ta tio n o f N, and

t h a t a 1 -p lace fu n c tio n o f type [n + l ] i s a c o n s ta n t fu n c tio n .

We r e c a l l th a t a s e t o f 1-p la c e fu n c tio n s G i s s a id to be i f

$-p lv t r a n s i t i v e on whenever n i , . . . , n ^ are d i s t i n c t elem ents o f

N, and ni*, ; a re d i s t i n c t elem ents o f N, th e re e x is t s a 1—p lace

fu n c tio n g(x) € G, tak in g a t l e a s t th e v a lu es ni', , . • .,n4,! , su ch th a ti s c lo sed and

g (n i) = ntV, fo r i = 1 , . . . , s . I f G^ c o n s i s t s e n t i r e ly of p e rm u ta tions

o f N, then G i s an s -p ly t r a n s i t i v e group of perm u ta tio n s o f N.

REMARK. We s h a l l make use o f v a rio u s w ell-known r e s u l t s on s -p ly

t r a n s i t i v e groups o f p e rm u ta tio n s .

For a f ix e d 1 -p la ce fu n ctio n s (x )a fu n ction f (x i , . . . , % ) €

i s sa id to be se lf-co n ju g a te under s i f and only i f

f(s(x i ) , . . . , s(xi<)) = s ( f ( x i , . . . ,%())

fo r a l l v a lu es of x i , . . . , x k *

REMARK. I f g ( x i , . . . , x ^ ) , f o r some in te g e r 6 !^ 1 , i s a p ro je c t io n

fu n c tio n , th en

g ( s ( x i ) , . . . , s ( x ^ ) ) = s (x t) f o r some in te g e r i , 1 ^ i ^ *6

= s (g(xi , . . . ,x^) ) .Hence, fo r any f ix e d 1-p la c e fu n c tio n s ev ery p r o je c t io n fu n c t io n o f

En i s s e lf -c o n ju g a te under s .

THEOREM 1 .1 , I f f (xi , . . . , 2}< ) c Ep i s se lf-c o n ju g a te under some fix ed

1-p la ce fu n ction s (x ) , then every fu n ction generated by f i s a lso

se lf-co n ju g a te under s ( x ) .

PROOF. I f 0 ( x i , . . . , x ^ ) i s g en e ra ted by f , then

= f (gl (3Q1. , . . .,X ^) , . • (xi , . . *,X^) ) ,

where each of th e -6-place fu n c tio n s gi ( i = 1 , . . . , k ) i s e i th e r

g en era ted by f o r i s a p ro je c t io n fu n c tio n . S ince <f> i s n e c e s s a r i ly

a com position in v o lv in g a f i n i t e number o f o c cu rren ces , m say , o f th e

fu n c tio n f , we s h a l l proceed by in d u c tio n on t h i s number. I f a l l th e

fu n c tio n s a re p ro je c t io n fu n c tio n s , th e n in view o f th e

above remark (f) i s s e lf -c o n ju g a te under s . This e s ta b l is h e s th e r e s u l t

f o r th e case m = 1. Suppose now th a t 1, and t h a t every fu n c tio n

g en era ted by f w ith a t most m occurrences o f f in i t s com position

sequence i s s e lf -c o n ju g a te under s . L e t (f> be any fu n c tio n generated

by f w ith m + 1 occurrences o f f in i t s com position sequence.

Then each o f th e fu n c tio n s i-s s e lf -c o n ju g a te under s , s in ce

g t,( i = 1, . . . , k ) i s e i th e r a fu n c tio n g e n e ra te d by f w ith a t most *

occurrences o f f in i t s com position sequence ( in view o f th e i n i t i a l

f in th e com position sequence o f o r a p r o je c t io n fu n c tio n .

T herefore <f> i s s e lf -c o n ju g a te under s and the r e s u l t fo llo w s .

For a f ix e d p erm u ta tion p(x) o f N a su b se t F^ o f i s s a id

to be con jugate to a su b se t F of Ep 3f and on ly i f to every fu n c tio n

f ( x i , . . . , 2 k ) € F th e re corresponds a fu n c tio n f ( ] % ,. . . , :% ) e F such th a tP P

(^1 , • • * ) = p ( f (p (^1 ) , • • *>P ) ) )>

and conversely*

I t fo llo w s from th e above d e f in i t io n t h a t F i s con jugate to

Fp under th e p e rm u ta tion p“ (x ) , and so we c a l l F , F^ a p a i r o f

con jugate su b se ts o f Ep .

THEOREM 1 .2 . I f . f o r some f ix e d perm uta tion p(x) o f N, F, F^

i s a p a i r o f con jugate su b se ts o f , th en F is com plete i f and only

i f Fp i s com plete .

PROOF. Suppose i f p o ss ib le t h a t F i s complete b u t F^ i s n o t .

Then th e re i s a fu n c tio n <^(xi , . . . ,x ^ ) e % n o t g en era ted by F^.

S ince F i s com plete, th e fu n c tio n

, . . . , x ^ ) = p“ (0 ( p ( 2i ) , . . . , p ( x ^ ) ) )

i s g en era ted by F , and so <p* a r is e s ffom a com position sequence

in v o lv in g a f i n i t e number o f fu n c tio n s , s a y , where each

g t ( i = 1 , . . . ,m ) i s e i th e r a fu n c tio n o f F o r a p r o je c t io n fu n c tio n .

I f , f o r i = 1 , . . . , m , we re p la c e each o f th e fu n c tio n s g i (xi , . • . ,x^)

by th e fu n c tio n gl * (xi , . . . ,x ^ ) = p(g(p“ (xi ) , . . . ,p “ (xj< ) )) in th e

com position sequence o f < f> ', th e n we o b ta in a com position sequence

in v o lv in g a f i n i t e number o f fu n c tio n s gi * , . . . , & *, where each o f th e

fu n c tio n s gi’( i = 1 , . . . ,m ) i s e i th e r a fu n c tio n o f F^ o r a p ro je c t io n

fu n c tio n (s in c e each p ro je c t io n fu n c tio n i s s e lf -c o n ju g a te under s ) .

The r e s u l t in g fu n c tio n i s th e re fo re gen era ted by F , b u t i t i s th eP

fu n c tio nP * (p ) , • • • > P (zk ))) = ÿ (xi , . . . , x k ) ,

8

which c o n tra d ic ts th e assum ption th a t (f> i s no t g en era ted by F^.

'T herefore F^ i s a ls o com plete. Since F i s con jugate to F^ under

(x) th e r e s u l t in the o th e r d i r e c t io n fo llow s by a s im ila r argum ent.

1 .4 - Background

In t h i s work we s h a l l be concerned m ainly w ith th e problem of f in d in g

co n v en ien t c r i t e r i a fo r a su b se t F o f Ep to be com plete, and i f

p o s s ib le deducing c r i t e r i a f o r a s in g le fu n c tio n to be com plete.

These problems a rose ou t o f th e w e ll known "S h e ffe r S troke F u n c tio n s”

o f th e P r e p o s i t io n a l C alculus ( th e se are th e 2—p lace com plete fu n c tio n s

o f E l ) . In [5 3 ], S h e ffe r dem onstrated th a t ev ery fu n c tio n of th e

P re p o s i t io n a l C alculus i s g en era ted by j u s t one o f th e se fu n c tio n s .

Follow ing th e work o f S h e ffe r , P o s t, in [2 4 ], in a complete survey o f a l l

th e d o s e d s e ts o f Ep fo r the s p e c ia l case n = 1, com ple te ly so lved the above

problem s in t h i s s p e c ia l c a se . However, such a survey has proved too la b o rio u s

to be c a r r ie d o u t in a l l b u t the s p e c ia l case n = 1.

Unaware o f P o s t 's r e s u l t s , Y ablonsk ii in [4 6 ] succeeded by a more d i r e c t

p ro o f in a t t a in in g an id e n t ic a l s e t of co n d itio n s to those o f P o s t fo r

a su b se t to be complete in Ei • He then went on to f in d in [47]

id e a l c o n d itio n s f o r a su b se t o f Ep to be com plete in th e s p e c ia l

case n = 2 . The corresponding problem fo r a s in g le fu n c tio n in th i s

s p e c ia l case was p a r t ly so lved ( f o r a 2 -p lace fu n c tio n only) by M artin in

[17] and r e c e n t ly has been com pletely so lved by W heeler in [4 3 ] .

The re le v a n t r e s u l t s o f th e se two s p e c ia l cases a re m entioned in

C hapter 2 .

P roceeding to th e g en era l case . P ost in [2 2 ] f i r s t dem onstrated th e

e x is te n c e o f complete su b se ts o f Ep fo r every f i n i t e in te g e r n ^ 1.

He proved th a t , i f n ^ 2, th e su b se t o f c o n s is t in g o f th e 1-p la c e

fu n c tio n x + 1 (modulo n + 1) to g e th e r w ith th e 2 -p lace fu n c tio n m ax(x,y),

i s com plete. In C hapter 6 we s h a l l p re se n t an unusual way o f g e n e ra lis in g

t h i s r e s u l t . Many o th e r examples o f complete su b se ts and fu n c tio n s have

s in c e been found; f o r example, by Evans and Hardy [ 5 ] , G o tlind [ 9 ] ,

Webb and Salomaa [ 2 7 ] #

However, in g en e ra l attem pts to f in d c r i t e r i a f o r a su b se t o f E

to be complete have met w ith on ly p a r t i a l su cc e ss . The f i r s t s ig n i f ic a n t

r e s u l t was due to S6upecki who proved in [3 5 ] "üie fo llo w in g :

i f n ^ 2, and a su b se t F of g en era tes th e s e t o f a l l 1-p la c e fu n c tio n s of

Ep, to g e th e r w ith a s in g le non-degenera te 2 -p lace fu n c tio n F (x ,y ) which

ta k es a l l n + 1 va lues o f N, th en F i s com plete.

He then used th i s r e s u l t in concluding th a t a s in g le 2 -p lace fu n c tio n

f ( x ,y ) which gen era tes the s e t o f a l l 1-p la c e fu n c tio n s o f Ep i s com plete.

In [4 9 ] Y ablonskii in tro d u ced th e fo llow ing g e n e r a l is a t io n of 86upecki*s

c o n d itio n s fo r th e fu n c tio n f ( x ,y ) : a fu n c tio n f ( x i , . . ) s a t i s f i e d

the S^upecki co n d itio n s i f f has a t l e a s t two e s s e n t ia l v a r ia b le s and ta k es

a l l n + 1 v a lu e s . He th e reb y e s ta b lis h e d S-6upecki's r e s u l t s f o r th e more

g en era l case when f i s a k -p la c e (k ^ 2) fu n c tio n o f E . O ther

10

improvements on S-6upecki*s r e s u lt s have been concerned w ith reducing

the c la s s o f 1-p la ce fun ction s generated by F. Such improvements

have been obtained by Y ablonskii (Theorem of [ 4 9 ] ) and Salomaa

[ 2 7 ] [ 2 8 ] [ 2 9 ] * In p a r ticu la r Salomaa has proved (Theorem 1 o f [ 2 8 ] )A/ A/ 'V Ax XIXth at fo r n ^ 4 , a su b set F i s complete i f ; i t generates, both

a fu n ction s a t is fy in g th e Sôupecki conditions and a lso the a ltern a tin g

group Ap+i of permutations o f N. This r e s u lt was used in th e proof

o f the fo llow in g (Theorem 2 o f [ 2 8 ] ) • fo r n ^ 4 , a s in g le fu n ction

f ( x i , . . . , x k ) e Efl i s complete i f and only i f i t generates the a ltern a tin g

group Ap+i • In Chapter 5 we are concerned w ith improving these two

r e s u lt s o f Salomaa s t i l l fu rth er .

11

I I - COMPLETENESS CRITERIA IN Ei AND % - I I

2.1» In tro d u c tio n

In C hapter 1 we m entioned th a t th e problem o f f in d in g c r i t e r i a f o r

a su b se t (o r s in g le fu n c tio n ) to be com plete in Ep has been com pletely

so lved in th e s p e c ia l cases n = 1, 2 . However, s in ce our in te n t io n i s

to improve th e p a r t i a l r e s u l t s o f th e g en era l c a se , we s h a l l p re s e n t in

th i s c h ap te r only those r e s u l t s of th e se two s p e c ia l cases which a re

of use in fo rm u la tin g our G eneral Theorems.

In 0 2 .2 we s h a l l show t h a t every fu n c tio n f € Ei belongs to th e

r in g o f polynom ials over th e f i e ld o f re s id u e s modulo 2 , and we use th i s

r e p re s e n ta t io n in a s tra ig h tfo rw a rd account o f th e n ecessa ry and s u f f i c i e n t

co n d itio n s f o r a s in g le fu n c tio n to be complete in Ei . In § 2 .3 we

s h a l l in tro d u ce th e n o tio n o f a p re-com plete su b se t o f Ep and l i s t a l l the

p re-com plete su b se ts o f Ei and % in two theorems concerning th e

r e l a t i o n between complete and pre-com plete su b se ts in th e se s p e c ia l c a se s .

2 .2 , 'C om pleteness ' in Et

THEOREM 2 .1 , ^ E i s a su b se t o f E i , w hich g e n e ra te s th e two

2-p la c e fu n c tio n s x + y modulo 2, xy, to g e th e r w ith th e c o n s ta n ts 0, 1,

th en F i s com plete.

PRÛOF. I t i s s u f f i c i e n t to prove th a t f o r ev ery f i n i t e in te g e r k ^ 1

F g en era te s th e s e t o f a l l k -p la c e fu n c tio n s o f Ei .

n

We proceed by in d u c tio n on k s t a r t i n g w ith th e case k = 1,

which fo llo w s from th e f a c t th a t th e s e t o f a l l 1-p la c e fu n c tio n s of

El i s [ 0 ,1 ,x ,x + l] , and i t i s e a s i ly seen t h a t F g en e ra te s these

if fu n c t io n s . Assume t h a t k ^ 1, and t h a t f o r each f i n i t e in te g e r

K(1 ^ K ^ k) F g en era te s every \ p lace fu n c tio n o f Ei Then

fo r each f ix e d k + 1 - p lace fu n c tio n P ( x i , # « ,y ) % th e

fu n c tio n s f ( x i , . . . , x k , O), f ( x i , . . . , x k , l ) a re generated by F,

s in ce they have a t most k v a r ia b le s . But

f ( x i , . . . , x k ,y ) = f ( x i , . . . , x k ,0 ) ( y ) + f ( x i , • . .,xk , 1 ) (y + l) ,

vdiere a d d it io n i s c a r r ie d ou t modulo 2, and so F g en era tes every

k+1 - p la ce fu n c tio n of E i , which proves th e r e s u lt*

THEOREM 2 .2 . The neaessany and s u f f i c i e n t co n d itio n s f o r a s in g le

fu n c tio n f ( x i , . . .,%k) to be com plete in % a re th a t

^ i ) f ( x ^ . . . , x ) = X + 1.

( i i ) f (x i+ 1 , .. . ,x k -+ l) / f(x i , . . . ,X k ) + 1.

C ondition ( i i ) s t a t e s t h a t f i s n o t s e lf -c o n ju g a te under x + 1.

PROOF, (a) N ecess ity ; i f f o r some v a lu e i c iO ,lj f s a t i s f i e s

f ( i , . . . , i ) = i , th e n every 1 -p lace fu n c tio n 0 g en era ted by f must

have # ( i ) = i , and so f w i l l no t be complete fo r in p a r t i c u la r f

cannot g en era te x + 1. Now f ( x , . . . , x ) = 0 , 1 , x o r x + 1 , and the

on ly p o s s ib i l i t y i s x + 1 . T herefo re ( i ) i s n e ce ssa ry .

I f f i s s e lf -c o n ju g a te under x + 1 , th en by Theorem 1.1 eveiy

fu n c tio n gen era ted by f w i l l a lso be s e lf -c o n ju g a te under x + 1 .

15

C onsequently f i s n o t com plete, s in ce E% co n ta in s fu n c tio n s which

a re n o t s e lf -c o n ju g a te under x + 1 ; f o r example th e c o n sta n ts 0, 1•

T herefo re ( i i ) i s n e ce ssa ry .

(h) S u ff ic ie n c y ; we s h a l l show t h a t f g en era te s th e co nstan ts

0 ,1 , to g e th e r w ith th e two 2 -p lace fu n c tio n s x + y modulo 2, xy

and th e n th e r e s u l t w i l l fo llow by Theorem 2 .1 .

I f f i s n o t s e lf -c o n ju g a te under x + 1 , th en f o r some f ix e d

s e t o f v a lues « x , . . . , a |< , where fo r each index i ( l ^ i ^ k)

a i f ! 0 , l j ,

f (flti+1, . . . +1 ) ^ 1 *

C onsider th e 1-p la c e fu n c tio n

^ (x ) = f (x + K i , . . . , X + ttk ) .

By c o n d itio n ( i ) f g en era te s x + 1 , and so f g en era te s ^ , b u t

ijf s a t i s f i e s

^ (x + 1 ) / ijf ix) + 1.

Only (p = 0 o r 1 s a t i s f i e s t h i s c o n d itio n , and i t fo llow s th a t f

g en era te s th e co n stan ts 0,1 (th e y w i l l be ip, ÿ + l ) .

Suppose f i s l i n e a r , th enk

a d d itio n c a r r ie d ou t modulo 2 . I f a lso f ( x , . . . , x ) = x + 1 th e nk

n e c e s s a r i ly we have Z a& = b = 1, b u t i f t h i s was so f would beLsi

s e lf -c o n ju g a te under x + 1 c o n tra d ic t in g ( i i ) . T herefore f i s

n o n - lin e a r , and we can take th e p roduct term o f l e a s t degree to be

X i3 % ...x j. Hence f (x ,y , 1, . . . , 1 , 0 , . . .,0)j». w ith the z e r o 's a l lo c a te d

u

l a s tto th e ^ k - j p la c e s , i s a n o n - lin e a r polynom ial in x ,y , say

xy .+ A ix + Agy + Ag, and so f genera tes th e p ro d u c t xy by th e com position

xy = f (x + Ag , y + A i , 1 , , . . , 1 , 0 , . . . , 0 ) + (A@+ AiAg ) .

F in a l ly we have

X + y = ((x + l ) ( y + 1 )+ I)(x y + 1 ),

and) i t fo llow s th a t f generates x + y , which com pletes th e proof*

REMARK. Theorem 2 .2 was o r ig in a l ly proved by P o s t in [2^]* The

above p ro o f i s a s l ig h t s im p l i f ic a t io n o f th e one used by R.A. Cuninghamg -

'G reen in [3].2 -1 d is t in c t k -place

2^-'! ..-ICOROLLARY. There a re 2 ‘

complete fu n c tio n s in Ei •pk __2

IROOF. C ondition ( i ) of Theorem 2 .2 g iv e s 2 k -p la c e2^“ —1fu n c tio n s in E i , and o f th e se 2 a re s e lf -c o n ju g a te under x + di.

This g ives

k -p lace com plete fu n c tio n s in E i .

2 .3 . P re-com plete su b se ts

DEFINITION. A c lo sed su b se t F o f i s c a l le d p re-com plete

i f fo r every fu n c tio n h e E n o t con tained in F th e su b se t [h | U F

i s com plete.

This d e f in i t io n , due to K uznetsov, i s quoted by Y ablonsk ii in [49]*

Follow ing KuSnetsov we say th a t a fu n c tio n ^ (xj. , . . . , % )

conserves th e p re d ic a te P ( x i , . . . , x ^ ) i f th e fo rm ula

15

D P(0 (x^1**21*" '"**k1 ^ ^*12*^22* * *'**k2^* * *"*^^*1^*^26* * *

i s v a l id f o r a l l v a lu e s of the v a r ia b le s x^^ ( i = 1 , . . . , k , j = 1 , . . . , 6 ) .

The s e t o f a l l fu n c tio n s which conserve th e p re d ic a te P is c a l le d th e

co n se rv a tio n c-laas o f P . I t i s easy to see t h a t th e co n se rv a tio n

cLàss o f a f ix e d p re d ic a te P i s c lo se d under com position , and th is

n o ta t io n g iv es us a sim ple and compact form of re p re se n tin g pre-com plete

s u b se ts .

P o s t, in [2^, in a complete a n a ly s is o f th e c lo sed su b se ts of l a ,

subsequently , found a l l the p re-com plete su b se ts o f E%. There a re f iv e o f them,

and th ey a re l i s t e d in th e s ta tem e n t of Theorem 2.3* These f i v e pre-com plete

su b se ts were l a t e r found ind ep en d en tly by Y ablonsk ii in [4^, who then

went on to f in d in &l7] a l l th e p re-com plete su b se ts o f Eg , l i s t e d in

Theorem 2.4* These r e s u l t s le d to th e fo llo w in g theorem s on com pleteness

c r i t e r i a in Ei,Eg 1

THEOREM 2.3* The n ecessa ry and s u f f i c i e n t c o n d itio n f o r a su b se t

to be complete in Ei i s th a t i t s h a l l n o t be co n ta in ed in any one of

th e co n se rv a tio n c là s s e s o f the fo llo w in g 5 p re d ic a te s ;

X = 0 , X = 1, X / y , x + y = z + u , x ^ y .

PROOF. See e i th e r P o st [24] o r Y ab lonsk ii [46].

THEOREM 2. 4 * The n ecessa ry and s u f f i c i e n t c o n d itio n , f o r a su b se t

to be complete in Ea i s t h a t i t s h a l l n o t be co n ta in ed in any one of

th e c o n se rv a tio n c la s se s o f the fo llo w in g I 8 p re d ic a te s : '

16

x = i , x / i , x + 1 = y , x + y = z + u, x + i ^ y + i , (x = i ) f ^ ( y = i ) ,

x = y v x = i v y = i , x = y v x = z v y = Z, where i i s a parameter

taking the values 0 , 1, 2 , and add ition i s taken modulo 3»

PROOF. See Y ab lonsk ii [47] o r [49]./V'*'

REMARK. Theorems 2 .3 and 2 .4 a re th e s p e c ia l oases n = 1, 2

o f th e g e n e ra l r e s u l t :

a su b se t o f Ep i s com plete i f and only i f i t i s n o t e n t i r e ly con tained in

any one o f th e p re-com ple te su b se ts o f Ep .

A p ro o f o f t h i s r e s u l t (due to Kuznetsov) i s given by Y ablonskii in

■ [^ . However, in th e g en e ra l case , a ttem p ts to f in d a l l th e pre-com plete

su b se ts o f En have met w ith g re a t d i f f i c u l t i e s . We have e x p l ic i t e ly

s ta te d Theorems 2 .3 , 2 .4 above in view o f t h e i r a p p lic a t io n to

p rov ing "end c a se s” in l a t e r Theorems.

REMARK. In [44], W heeler d e riv e s n ecessa ry and s u f f i c i e n t co n d itio n s

f o r a s in g le fu n c tio n to be com plete in % .

17

I I I - SIMPLE GENERAL THEOREMS - I I I

3 .1 . In tro d u c tio n

Since th e s u b se t o f a l l fu n c tio n s o f En w ith a t most one e s s e n t ia l

v a r ia b le i s c lo sed under com position , a n ecessa ry c o n d itio n f o r a su b se t

o f En to be complete i s th a t i t s h a l l g en e ra te a fu n c tio n w ith a t

l e a s t two e s s e n t ia l v a r ia b le s .

In § 3 .2 we s h a l l p re se n t some p re lim in a ry c r i t e r i a f o r a su b se t F

o f En to be com plete. These c r i t e r i a presuppose th a t an e x ten s iv e

c la s s o f 2 -p lace fu n c tio n s g en era ted by F i s a v a i la b le , b u t th e

r e s u l t s proved are in te r e s t in g and v a lu a b le .

In § 3 .3 we s h a l l c o n s id e r th e s p e c ia l case when n + 1 = p

(a prime number). In th i s case we s h a l l show t h a t every fu n c tio n

f e E . belongs to th e ring : o f polynom ials over th e f i e l d of re s id u e sp-1modulo p , and we s h a l l m ention examples o f com plete and p re-com plete

fu n c tio n s and su b se ts in E . .p-1

18

3 .2 , Simple Completeness Theorems

DEFINITION# We d e fin e th e fo llo w in g p a i r o f 2 -p lace fu n c tio n s

each belonging to En :

X + y by

x + 0 = x = 0 + x ;

xy by

xO = 0 = Ox

x1 = X = 1x,

o r in term s o f th e m atrix n o ta tio n *

x + y 0 1 . . . n xy 0 1 2 . . . n

0 0 1 . . . n 0 0 0 0 . . . Ù

1 1 1^11 • • Min 1 0 1 2 . . . n

2 0 2 V2 2 • .• % , , ,

• • • • • • • • •n n Mn 1 • • Mnn n 0 n i nn

We may take as th e v a lu e o f P iJ (1 (2 ^ k , -6. ^ n)

any elem ent of N. This g iv e s us ( n + l) ^ ^ d i s t i n c t p a ir s of

2 -p lace fu n c tio n s x+y, xy.

EEMA.EK# x+y (modulo n + l) , xy (modulo n+1 ) i s a p a r t i c u la r

p a i r of the above ty p e , bu t u n le ss we s t a t e o th e rw ise , we s h a l l

take x+y, xy to mean th e fu n c tio n s d e fin e d above.

19

THEOREM 3*1* I f a su b se t F g e n e ra te s a f ix e d p a i r

o f 2-p la c e fu n c tio n s x+y, xy , t o g e th e r w ith th e s e t o f a l l 1-p la c e

fu n c tio n s of Ep, th e n F i s com ple te .

PROOF. I t i s s u f f i c i e n t to prove th a t f o r each f i n i t e

in te g e r k ^ 1 F g e n e ra te s ev ery k -p la c e fu n c t io n f (x^ , . . . ,X |< ) c Ep •

We s h a l l use in d u c tio n on k , s t a r t i n g w ith th e case k=1 w hich

fo llo w s by th e hypotheses of the Theorem. Assume th a t k ^ 1 and

th a t f o r each in te g e r K (1 ^ k) F g e n e ra te s every K -place

fu n c tio n of Ep. S ince F g e n e ra te s ev ery 1-p la c e fu n c tio n of Ep,

then in p a r t i c u la r F g en era te s th e n+1 1—p lace fu n c tio n s

Mo,Mi,*.*,Mn d e fin ed by

^ l(x ) = 0 , fo r X / i ;

M i(i) = 1,

f o r each index i (O ^ i ^ n ) .

Yve can e iç r e s s each f ix e d (k + l)-p la ce fu n c tio n f ( x ^ , . . . , X k , y ) e Ep

as fo llo w s;

f ( x i , . . . , X k , y ) = z_i f ( x i , . . . , X k , i ) ^ j^(y) (a )1=0

where a d d i t io n and m u l t ip l ic a t io n a re w ith r e s p e c t t o th e

o p e ra tio n s x+y, xy re s p e c t iv e ly ( in f a c t (A) i s ju s t a co n v en ien t

way o f w r it in g down the com position in v o lv in g x+y, xy , p ^ , . . . , ^ p ) .

By the induction hypothesis F generates the n+1 fu n ction s

f ( x i , . . . , X k , i ) , for i = 0 , 1 , . . . , n , s in ce each f ( x % , . . . , X k , i ) i s a t

most a k -p lace fu n ction o f Ep . Therefore F generates f ( x i , . . . , X k , y ) ,

and the r e s u lt fo llo w s .

20

THEOREM 3*2. Denote by the s e t o f a l l k-p la c e fu n c tio n s o f ;

th e n i s com plete i f and only i f k ^ 2 .

PROOF. I f k ^ 2 , th e n by id e n t i fy in g v a r ia b le s (o r e lim in a tin g

in e s s e n t ia l v a r ia b le s ) in s u i ta b le k -p lace fu n c tio n s o f E we can show

th a t g en era te s ^ and , and so th e r e s u l t fo llo w s by

Theorem 3*1* I f k = 1 , th e n the c lo su re o f under com position is the

s e t of a l l fu n c tio n s o f Ep w ith a t most one e s s e n t ia l v a r ia b le ,

which i s a p ro p e r su b se t o f % , and so ^ is not com plete.

REMARK. Theorem 3*2 i s due to P o s t , who p re se n te d i t in [22]in th e form; every fu n c tio n can be expressed aa a f i n i t e com position

o f th e two fu n c tio n s max(x,y) and x + 1 (modulo n + l ) . I t i s d f

fundam ental im portance in th e work and w i l l be r e f e r r e d to in l a t e r

c h a p te rs .

The fo llo w in g Theorems a re improvements on Theorem 3*1*

21

THEOREM 3»3. I f a su b se t F g e n e ra te s a f ix e d p a i r of

2-p la c e fu n c tio n s x + y , xy, to g e th e r w ith a l l c o n s ta n ts and

the n + 1 1-p la c e fu n c tio n s ^q, ju^, . . . ,jUp d efin ed by

m i x ) - O f fo r X / i j

M l(i) = 1 j

fo r each index i (O ^ i ^ n ) , th en F i s com plete.

PROOF. We s h a l l prove th a t - F generates the s e t o f a l l

1-p la ce fu n ction s of Ep , and t ie n th e r e s u lt fo lla v s by Theorem 3*1*

I f f ( x ) i s any f ix e d 1-p la ce fu n ctio n of Ep , then we can

express f in the form

f ( x ) = Z f ( i ) / u i ( x ) ,L= 0

where a d d itio n is w ith resp ec t to the operation x + y , and m u ltip lica tio n

i s w ith resp ect to xy . Therefore F generates f ( x ) and the

r e s u lt fo llo w s .

REI.ARK. I f F generates the p a r ticu la r pair o f 2-p lace

fu n c tio n s x + y (modulo n + 1 ) j xy (modulo n + 1 ) , th e n fo r

F to be complete i t i s s u f f ic ie n t only fo r F to generate a lso

the n + 1 1-p lace fu n ction s s in ce every con stan t,

c , say , can be expressed in the form

c = Z[/Jt(x) + . . . + jJi(x)] (modulo n + l ) .L-oi-------------------V ---------'

c tim es

Therefore F generates a l l constants, and i s complete by Theorem 3*3*

22

THEOREM 3*4. I f a su b se t F g e n e ra te s a f ix e d p a i r of

2-p la c e fu n c tio n s x+y, xy , to g e th e r w ith a doubly t r a n s i t i v e s e t of

1-p la c e fu n c tio n s o f N, then F i s com plete.

PROOF. We s h a l l prove t h a t F g en e ra te s th e n+1 1-p la c e

fu n c tio n s d e fin ed by

m i x ) = 0^ f o r X i j

M l(i) = 1,

f o r each index i (1 ^ i ^ n ) , to g e th e r w ith a l l c o n s ta n ts , and

th en th e r e s u l t w i l l fo llo w by Theorem 3*3*

G-iven any i (O ^ i ^ n ) , we choose n 1—p la ce fu n c tio n s

Skt where k = 0 , 1 , . . . , n (k / 1 ) , g en era ted by F such th a t

Ski (k) = 0 , S k i ( i ) = 1 .T h e n

jUj.(x) = n S k c ( x ) , Iq/i

where m u l t ip l ic a t io n i s in th e sense o f th e o p e ra tio n x ,y, and

so m i x ) i s gen era ted by F .

I t remains to show th a t F g en e ra te s a l l c o n s ta n ts .

Suppose u (x ) = u i s any c o n s ta n t fu n c tio n o f Ep, and l e t S be

a 1-p la c e fu n c tio n g e n e ra te d by F such th a t S(o) s u .

We have

Mo(m i(x )) = 0 , f o r a l l x ,u nd so

u (x ) = S im O tti(x ))) .

T herefo re F g e n e ra te s a l l c o n s ta n ts .

23

COROLLARY. I f a su b se t F g e n e ra te s a 2-p lao e fu n c tio n o f

type xy , to g e th e r w ith a doubly t r a n s i t i v e group of perm u ta tions

of N, then P i s com plete.

PROOF. The r e s u l t fo llo w s by Theorem 3*4- a f t e r we have

shown th a t F a ls o g e n e ra te s a fu n c tio n of type x+y.

Choose a perm uta tion r g en e ra ted by F such th a t

r ( 0 ) = 1 , r ( l ) = 0 ;

th en F g e n e ra te s th e 2—p lace fu n c tio n

h (x ,y ) = r - i ( r ( x ) , r ( y ) ) ,

where m u l t ip l ic a t io n i s w ith r e s p e c t to xy . We have

h (0 ,x ) = X a h ( x ,0 ) ,

and so h (x ,y ) i s o f type x+y and th e r e s u l t fo llo w s .

REMARK. In g en e ra l we canno t r ^ l a c e ’xy* by ’x+y’ in th e

C o ro lla ry , a s i s shown by th e example in w hich n = 3 , and F

c o n s is ts o f th e a l te r n a t in g group o f perm uta tions A* on | 0 , 1 , 2 , 3 j ,

a l l c o n s ta n ts , and th e p a r t i c u la r fu n c tio n o f type x+y d e fin ed

as fo llo w s ;x+y 0 1 2 3

0 0 1 2 31 1 0 3 22 2 3 0 13 3 2 1 0

L et T = [A4, a l l c o n s ta n ts ] . Then T co n ta in s a doubly t r a n s i t i v e

group o f perm utations of N, namely A 4 , and i t i s easy to v e r i f y

th a t i f r ( x ) , s ( x ) e T th e n a ls o r (x )+ s (x ) e T. This means t h a t

i f g i s any 1-p la c e fu n c t io n g e n e ra te d by com positions of x+y and

members cf T, th e n g (x ) € T. But T does n o t c o n ta in a l l 1-p la c e

fu n c tio n s of E 3, and so F i s n o t com plete.

24

3*3, The S p ec ia l Case n+1=p ( a pirime number)

THEOREM 3*6. I f n+1 = p ( a prim e number) , th en every

f unet io n f(x^ ,X|< ) e belongs to th e r in g of polynom ials

over the f i e l d o f re s id u e s modulo p#

PROOF. L et F be th e su b se t cf E^ c o n s is t in g o f the

p a r t i c u la r p a i r o f 2 -p lace fu n c tio n s x+y (modulo p ) , xy (modulo p ) ,

to g e th e r w ith a l l co n s tan ts and th e p 1-p la c e fu n c tio n s

d e fin ed by

m i x ) = 0 > f o r X / i )

M l(i) = 1,

f o r each i (O ^ i ^ p - l ) .

We can re p re s e n t Ml(x) as a polynom ial over th e f i e l d o f

re s id u e s modulo p a s fo llo w s;p—1

Ml(x) =(p - 1) n (x + a ) 5a=0

a /p -4

where a d d i t io n and m u l t ip l ic a t io n a re c a r r ie d ou t modulo p .

Hence ev ery fu n c tio n g en era ted by F belongs to the r in g of

polynom ials over th e f i e l d o f re s id u e s modulo p , b u t by Theorem 3*3

F i s com plete, and so th e r e s u l t fo llo w s .

25

REMARK. Theorem 3*6 g iv es us a conven ien t method of

e x p re ss in g fu n c tio n s belong ing to E^ We s h a l l m ention the

fo llow ing examples of com plete fu n c tio n s and su b se ts in E^ ^ .

EXAMPLE 3 .1 . ^ n+1 = p , and ÿ ( x i , . . . , x ) =P P

1 + XI , where a d d i t io n , s u b tra c t io n and

m u lt ip l ic a t io n a re c a r r ie d o u t modulo p , th e n <f) i s complete#

This simple r e s u l t i s proved by R.A.Cunningham Green in [ 3 ]•REIvlARK* Examples o f complete g e n e ra to rs over G alois

f i e l d s are to be found in [ 8 ] by R .L .G oodstein .

26

EXAMPLE 3*2. n+1 = p , and f o r f ix e d v a lu es

0? 0 (modulo p ) ) F c o n s is ts o f the p a i r of 2-p la c e

fu n c tio n s

# (x ,y ) = (x -a ) + (y -a ) + a ;

4 (x ,y ) = /9 (x -a )(y -a ) + y(^-cc) + ô (y -a ) + e ,

where a d d i t io n , s u b tra c t io n and m u lt ip l ic a t io n a re c a r r ie d o u t

modulo p , th en i f a ^ e (modulo p) F i s com plete, and i f

a s e (modulo p) F g en era te s th e p re-com plete su b se t o f Ep_^

which p re se rv e s th e p re d ic a te x = a .

We see t h a t ^ p re se rv e s th e p re d ic a te x =t a ( t h a t i s

<p{cL,a.) = a ) , and i f a s e th en ijf w i l l a ls o p re se rv e x = a .

We s h a l l make th e fo llo w in g change o f v a r ia b le ;

X = X - a .

We have

0 (x ,y ) = X + Y + a

^ (x ,y ) = /9XY + yX + SY + e ,

and i f ( x i , • • . ,Xk ) i s any fu n c tio n o f which p re se rv e s th e

p re d ic a te x = a , then

^ (x i , • • • ,Xk ) = ^ ’ (Xi , • • • , Xk )+ CL)

where Xj ‘ = x j -a f o r j = 1 , . • • ,k , and ^ ' (X , . , ]^ ) p re se rv e s

th e p re d ic a te X = 0 ( th a t i s ^ * ( 0 , . . . ,0 ) = O).

miAJîK ON SUBSTITUTION. I f ^ ( x i , . . . , X k ) , a re any

f ix e d k+1 fu n c tio n s of which p re se rv e the p re d ic a te x=a, th e n

* (^1 ) ■*" ^9

where f o r each index j ( l ^ j ^ k)

+ K «

We s h a l l need th e fo llo w in g two lemmas.

27

LEM'LA. 3*1. (j> g e n e ra te s ev ery l i n e a r fu n c tio n of E^_^

which -preserves th e p re d ic a te x=a«

PROOF. F o r each f ix e d k ^ 2 we have

» . . . 1 fXk ) #. ) ) = +Xg + . . . + Xk + oc,

where Xj = x j - a , f o r j = 1 , . . . , k , and i t i s easy to s e e , by

id e n t i fy in g s u i ta b le v a r ia b le s in th e se fu n c tio n s , t h a t 0 g en era te s

eveiy l in e a r fu n c tio n which p re se rv e s th e p re d ic a te x=a.

LEIvlIA. 3 .2 . F g en e ra te s th e 2 -p lace fu n c tio n XY + — [e -a ] + a

(Note t h a t s in c e n+1 i s prime 1_ € N.)

PROOF. Lemma 3.1 ^ g e n e ra te s th e fu n c tio n

T ( x ,y ,z ) = ^ [(p - y ) X + (p - j ) Y + Z] + a ,

and so F g e n e ra te s

T (x ,y ,^ ( x ,y ) )= X Y + ^ [e -a ] + oc .

28

PROCF OF EXAMPLE 3 .2 . P a r t ( i ) I f a ^ € (modulo p ) .

By Lemma 3.1 ^ g e n e ra te s th e c o n s ta n t oc, f o r

oc~ X + . . . + X + ct.I .............. I

p tim es

We have

f p( oL, a . ) = € cl) )

and so fo r j = 0 , 1 . , , , . p —1 F g e n e ra te s th e p c o n s ta n ts

j(€ -o :) + a . Since e-tt ^ 0 (modulo p ) , th e se a re the co n sta n ts

0 , 1 , . . . , p —1 in some o rd e r .

F a ls o g en era te s th e 1-p la c e fu n c tio n s

x+j = 0 (x ,j+ a ) f o r j= 0 , 1, . . . , p - 1 ,

and hence F g e n e ra te s the p a ir of 2—place fu n c tio n s

x + y (modulo p) = <p(x + a , y + a ) - a ;

xy (modulo p) = t ( x + a , y + a , ^ (x + a , y + a ) )

- - a ) - a ,

(see Lemma 3 .2 ) .

Therefore F g en e ra te s th e r in g o f polynom ials over tl^e f i e ld

of re s id u e s modulo p , and by Theorem 3.6 t h i s i s a s u f f i c i e n t

c o n d itio n f o r F ^o be com plete.

29

P a r t ( i i ) I f a s e (modulo p '), th en every fu n c tio n

g en era ted by F w i l l p re se rv e th e p re d ic a te x = a . C onversely

we s h a l l prove th e fo llo w in g .

LEMJ.1A 3*3* ^ ^ i s any f ix e d fu n c tio n of which

p re se rv e s the p re d ic a te x=a, th en F g e n e ra te s

PROOF. I f I i s l in e a r th e n the r e s u l t fo llo w s im m ediatly

by Lemma 3*1 • Suppose th a t ^ i s n o n - l in e a r ; th en f o r a f ix e d

index 6 (-6 depending on th e number of term s in ^ ) th e re e x i s t s

a l i n e a r fu n c tio n L(xJi , . . . ,x ^ ) , vh ich p re se rv e s th e p re d ic a te x=a,

and & n o n -lin e a r fu n c tio n s d e fin ed by

Mj ( x i , . . • ,Xk ) = n (]Qn ) + ot>m= 1

where f o r each p a i r of in te g e rs j,m ( 1^ j ^ 6 , 1 ^ m ^ k)

Xj = x j - a and € [ 0 , 1 . ,p—1 j , such t h a t

^ ( x i , . . . ,X k ) = L (M i,..« ,M ^)«

I t i s s u f f i c i e n t to prove th e re fo re th a t F g en era te s every

fu n c tio n of type Mj .

In Lemma 3*X l e t /c(x,y) = T ( x , y , ^ ( x , y ) ) , and s in ce

a s f we have

K (x,y) = XY + a .

T herefore f o r each f ix e d in te g e r k = 3 ^ 3 ,# . . , F g e n e ra te s

th e k -p la c e fu n c tio n

/c(xi ,/t(xq , » . . ,^(xk-. 1 ,Xk ) . . . ) ) = X Xg # . .X% + oc ,

I t is easy to s e e , by id e n t i fy in g s u i ta b le v a r ia b le s in th e se f u n c t io n s ,

t h a t F g en era te s ev ery fu n c tio n of ty p e Mj , and so the r e s u l t fo llow s*

30

We s h a l l now prove P a r t ( i i ) o f Example 3*2. Obviously

F is not com plete s in c e E __ C ontains fu n c tio n s w hich do not

p re se rv e th e p re d ic a te x=a. On the o th e r hand by Lemma 3*3

F g e n e ra te s every fu n c tio n o f E^ which p re se rv e s x=a.

That i s F* ( th e c lo su re o f F under com position ) i s th e

co n se rv a tio n c la s s of th e p re d ic a te x s a . I f h i s any f ix e d

fu n c tio n of Ep such t h a t h / , th en h does not p re se rv e

th e p re d ic a te x = a . T herefore h ( a , . . . , a ) ^ a and th e

su b se t F^ u fh ^ o f En w i l l be com plete by a s im ila r argum ent

to P a r t ( i ) . But th i s is th e co n d itio n f o r to be p re -co m p le te ,

and so th e r e s u l t fo llo w s .

31

REMARK. In connec tion w ith p re-com plete su b se ts , i t i s

of i n t e r e s t to determ ine fo r a g iven pre-com plete su b se t P o f Ep

w hether or no t th e re i s a s in g le f u n c t io n ‘d c P such t h a t d g en era te s

P . For th e p a r t i c u la r p re—com plete su b se t of E^ which p re se rv e s

th e p re d ic a te x = a , mentioned in Example 3*2, i f p = 2 th en

d ( x ,y ,z ) = XY + Z + a i s such a fu n c tio n because

d (x ,x ,y ) = X + Y + a ;

d ( x ,x ,d ( x ,y ,x ) ) = XY + a ,

end i f p ^ 3 th en d (^ i > • * * >^p^y) ® X%Xg + (p—l)XiXg + Y + ct

i s such a fu n c tio n because

d ( x , . . . , x , y ) = X + Y + a ;

(and so <p g e n e ra te s th e c o n s ta n t a)

d ( x ,y ,a , . . . ,a') = XY + a .

This i n tu rn le a d s to th e more g en e ra l q u e s tio n ; Given any

su b se t of fu n c tio n s $ = di****;#& o f Ep , what a re the n e ce ssa ry and

s u f f ic ie n t c o n d itio n s f o r $ to co n ta in a s in g le fu n c tio n , d l say ,

such th a t d l g e n e ra te s §? This q u estio n i s e a s i ly answered i f $

c o n s is ts of only 1-p la c e fu n c tio n s o f Ep, f o r th en d i ,* '* /& 6 must

a l l be powers of d l» but g e n e ra lly i t depends upon th e fundam ental

q u es tio n ; What a re th e necessary and s u f f ic ie n t c o n d itio n s f o r

a fu n c tio n f f En to gen era te a g iven fu n c tio n g e Ep? For n=1,2 ,

by Theorems 2 .} , 2 ,4 re s p e c tiv e ly , we may deduce t h a t a n ecessa ry

c o n d itio n i s th a t e i t h e r $ co n ta in s a com plete fu n c tio n , or # i s

co n ta in ed e n t i r e ly in one o f the p re-com plete s u b s e ts .

32

IV - SOME FUNDAMENTAL RESULTS - IV

4 .1 . In tro d u c tio n

In t h i s c h ap te r we s h a l l prove some fundam ental r e s u l t s

on com plete su b se ts and fu n c tio n s in En • In f a c t , the theorem s on

complete su b se ts (Theorems 4 .4 , 4 .6 ) a re e x p re s s ly d esig n ed to

y ie ld c r i t e r i a f o r a s in g le fu n c tio n to be com plete (Theorems 4 .5 ,

4 .7 ) .

We in tro d u ce th e fo llo w in g n ecessa ry p ro p e r t ie s o f a complete

fu n c tio n .

We c a l l x j an e s s e n t ia l v a r ia b le of a k—p la ce fu n c tio n

f ( x i , . . . , X k ) i f th e re e x i s t numbers a i , . . . , a j _ i , a j + i , . . . , a k

such t h a t the 1-p la c e fu n c tio n f ( a i , . . . , a j - 1 ,x j , a j + 1 , . . . ,a k ) i s

n o t c o n s ta n t .

A fu n c tio n f ( x i , . . . , X k ) e En i s s a id to s a t i s f y th e

S^upecki c o n d itio n s i f i t ta k e s a l l n+1 va lu es of N, and

has a t l e a s t two e s s e n t i a l v a r ia b le s . I t i s easy to see th a t

a n ecessa ry co n d itio n f o r f to be com plete i s t h a t f s h a l l

s a t i s f y th e S-êupecki c o n d itio n s , fo r i f f does not take a l l

n+1 v a lu e s th en f cannot g en e ra te any fu n c tio n of En which

ta k es th e m issing v a lu e s , and i f f has a t most one e s s e n t ia l

v a r ia b le then f g e n e ra te s on ly fu n c tio n s o f Ep p o sse ss in g a t

most one e s s e n t ia l v a r ia b le .

33

4 .2 . Consequences of th e S-ê'Upecki co n d itio n s

THEOREM 4 .1 . I f n ^ 2 and f e En s a t i s f i e s th e S^upecki

c o n d it io n s , w ith Xj. a s one e s s e n t ia l v a r ia b le , th e n th e re a re p o in ts

a = = G^i, • • • ,/?k) such t h a t th e th re e v a lu es

f(oc), fC 01 ,#3 ,.#* ,K k)*

a re d i s t i n c t .

We c a l l a,/9 S’êupecki p o in ts f o r f .

PROOF. S ince x^ i s e s s e n t ia l th e re e x i s t numbers

a,nd Ys J • • • >Yk such th a t

F(y i^jY 2 > • • • ^Yk ) u , ^*(Yi ^Ys > • • • *Yk ) — ^ >

viiere u^ / u® . We s h a l l now d is t in g u is h two c a s e s .

CASE 1. Vfe have f (y i * >Ys * • • • »Yk ) = € [u^,u®j fo r some

number Y i^ . Bince one of X2 , . . . ,X k i s e s s e n t i a l , th e re e x i s t

numbers Si and S g ^ , . . . ,# ^ ^ ; such th a t

f (S i ,Sg^, • . . , 5k^ ) ^ F (S i,S g ^ , • . • ,Sk^ ) .

One o f th e se two v a lu es must be d i f f e r e n t from f ( S i ,Y s> • • • >Yk ) i we

may suppose t h a t

f (Si ,5a^ , . . . ,Sk^ ) ^ i*(Si ,Ys > • • • >Yk ) •

One o f u^ ,u^ ,u® must be d i f f e r e n t from b o th of th e s e , u*’ say .

Then th e fo llo w in g th re e v a lu es a re d i s t i n c t ;

F ( y i ** »Y2 > • • • >Yk ) > F ( S i ,Y 2 J • • • >Yk ) > f ( S i , S g ,S k ^ ) •

CASE 2 . We have f (x i ,Y 2 , . . . ,Y k ) c [u^ ,u^ ] f o r a f ( X i., L SucH tia 't

Since f ta k es a t l e a s t 5 d i s t i n c t v a lu e s th e re e x i s t n u m b e r^ S i, . . . ,Sk )

u® / [u ^ ,u ^ j . L e t f ( 8i,Y 2 ,. .* ,Y k ) = u* '(i = 1 o r 2 ) . Then th e

fo llow ing th re e v a lu es a r e d i s é i n c t j

> Y2 ;*"*yYk); ^ (S i ,Y2 j • • • >Yk ) » ^*(§1 »S2 > • • • ^^k ) .

u

REMARK. Theorem 4.1 i s im p l ic i t in Y a b lo n sk ii 's p ro o f

o f h is ’fundam ental lemma’ ([4 9 ] p . 6 9 ). The above p ro o f of

Theorem 4.1 was shown to me by Dr, Roy 0. D avies; i t i s s im p le r

than my o r ig in a l proof based on Salomaa’s fo rm u la tio n of what is h e re

Theorem 4 .2 .

I t i s easy to see t h a t i f we weaken th e S-êupecki c o n d itio n s

by re p la c in g ’f assumes a l l n+1 v a lu e s ’ by *f assum es a t

l e a s t 3 d i s t i n c t v a lu e s ’ th e n Theorem 4.1 g iv es a co n d itio n

which i s s u f f ic ie n t a s w e ll as n e ce ssa ry .

DEFINIT ION. L e t G%, ...,G k be non-empty su b se ts ©f N.

Then we denote by f ( G i , . . . ,G k ) th e s e t of va lues assumed by

f ( x i , . . . , x k ) when fo r each index i , 1 ^ i ^ k , on ly values

belong ing to a re a ss ig n ed fo r x t .

35

THEOREM 4 .2 . ^ 2 and s a t i s f i e s th e

S-6upecki c o n d itio n s , w ith a s one e s s e n t ia l v a r ia b le , th en

fo r each in te g e r j 2 ^ j < n+1, th e re a r e s e t s , i= 1 , . . . , k ,

each co n sis tin g : of a t most j v a lu e s , such th a t th e s e t f(Gt, , • . . ,G|< )

c o n s is ts of a t l e a s t j+1 v a lu e s .

Pi^OQP. By Theorei^i 4 .1 f w i l l have a p a i r of S-^upecki p o in ts

a,/9; l e t th e co rrespond ing v a lu es be u^ ,u^ ,u® . We may suppose

th a t the n+1 d i s t i n c t v a lu es of N a re u^ ,u^ ,u^ , u f , . . . ,u"'’' , and

s in ce f assumes a l l n+1 v a lu e s , f o r each i ( l ^ i ^ n+ l) th e re

e x is t s numbers X i^ ,...,X k * ’ such th a t

f ( x i^ , . . . ,X k * ’) = u ^ .

L et j be a f ix e d in te g e r , 2 ^ j < n+1, and f o r each index

i (1 ^ i ^ k) l e t th e s e t Gl = [ a i , /9 i ,xl , . . . , xl' '’'^ Î .

The s e ts G i , . . . ,G k each c o n s is t of a t most j d i s t i n c t v a lu e s ,

and the s e t f (Gi , . . . ,Gj< ) c o n s is ts of a t l e a s t the v a lu es u^ ,u^ ,u® ,u^ , • . . ,

^ j+ i^ Tliat i s , f (G% , . . . ,Gk) c o n s is ts cf a t l e a s t j+1 e lem en ts .

REMARK*. Theorem 4 .2 i s a ls o proved by Y ab lonsk ii in [49]*

For o th e r r e s u l t s on th e value sequences o f fu n c tio n s w ith a t l e a s t

two e s s e n t i a l v a r ia b le s see papers by A.Salomaa [3 0 ] and

Roy 0 . Davies [ 4 ] .

36

IffiiVIARK. We a ls o in c lu d e in t h i s § th e fo llo w in g theorem

(Theorem 4«3)* Though no t a consequence of th e S6upecki

c o n d itio n s . Theorem 4 .3 g iv es c r i t e r i a (depending on th e 1-p la c e

fu n c tio n s g en era ted ) f o r a s in g le fu n c tio n f to s a t i s f y the

S6upecki c o n d itio n s . A r e s u l t of t h i s k in d i s o f te n used by

au th o rs in th i s work ; f o r example Salomaa [27] [28] [2.9]

and Y ablonskii [4 9 ] , b u t th e proof i s seldom g iv en .

37

THEOREM 4*5* I f a s in g le fu n c tio n f c Ep g e n e ra te s

1-p la c e fu n c tio n s <f>i >• • • (6 ^ 2) such Hi a t

( i ) each cf th e n+1 v a lues of N is tak en by a t l e a s t one of

th e fu n c tio n s d i , . . . and ( i i ) th e re i s no 1-p la c e

fu n c tio n d such th a t f o r each index i ( l ^ i ^ &)

d i ( x ) = d ^ ^ x ) f o r some power k i , where d ^ K ^ ) = d ( " " * (^ ( x lL . .)

(w ith d re p ea ted kj, t im e s ) , th e n f s a t i s f i e s th e S^upecki

co n d itio n s .

PROOF. Suppose f does n o t s a t i s f y th e S-êupecki c o n d itio n s ;

i f f does no t tak e a l l n+1 v a lu e s , then f has a t l e a s t one

m issing v a lu e , u say . In t h i s case every fu n c tio n g en era ted by

f w i l l a ls o bave u a s a m issing v a lu e , in p a r t ic u la r

This c o n tra d ic ts a s s e r t io n ( i ) .

Suppose th e re fo re th a t f tak es a l l n+1 v a lu es and has a t

most one e s s e n t i a l v a r ia b le . L et f ( x , . . . , x ) = d(x)* ty a simple

in d u c tio n argument i t i s easy to show th a t every fu n c tio n g

g en era ted by f has a t most one e s s e n t i a l v a r ia b le ; and t h a t

th e 1-p la c e fu n c tio n g ( x , . . . , x ) i s a power o f d* I t fo llow s

th e re fo re t h a t every 1—p lace fu n c tio n g en era ted by f is a power

o f (p, c o n tra d ic t in g a s s e r t io n ( i i ) . The procf i s com plete.

3 8

COROLLARY 1 • I f f g e n e ra te s a l l c o n s ta n ts , th e n f

s a t i s f i e s th e S6upecki c o n d itio n s .

PROOF. To each value u e N th e re corresponds a c o n s tan t

u (x ) = u . I f d f En i s any s i n ^ e 1-p la c e fu n c tio n such

th a t d ^ (x ) = u , f o r some power k , th en n e c e s s a r i ly we must

have d (u ) = u ( i f n o t , th e n = 4>{u) / u , which i s

a c o n tra d ic t io n ) . T herefore no s in g le 1 -p lace fu n c tio n o f Ep

can g en e ra te more than one c o n s ta n t, and th e r e s u l t fo llo w s by

Theorem 4.3*

COROLLARY 2# ^ n ^ 2 and f g e n e ra te s a doubly

t r a n s i t i v e group P o f p e m u ta t io n s cf N, th en f s a t i s f i e s

th e S6upecki c o n d itio n s .

PROOF. Each member of P , being a p e im u ta tio n , ta k es

a l l n+1 v a lu es . I f d i s a s in g le p e rm u ta tion which g e n e ra te s

P th e n , s in ce P i s t r a n s i t i v e , d must be a c i r c u la r perm uta tion

of N. But ±f <p ±3 c i r c u l a r , i t canno t g e n e ra te any non—i

• id e n t ic a l perm uta tion f ix in g an elem ent o f N, and s in c e n ^ 2

P w i l l co n ta in perm uta tions o f t h i s ty p e . T herefore no s in g le

p e rm u ta tion d g a ie r a te s P , and so th e r e s u l t fo llo w s by

Theorem 4.3*

39

4*3* Fundamental oom pleteness c r i t e r i a in En

THEOREM 4 .4 . I f n ^ 2 %nd a su b se t F g e n e ra te s the s e t

o f a l l 1-p la c e fu n c tio n s of Ep , to g e th e r w ith a fu n c tio n s a t i s fy in g

th e S-êupecki c o n d itio n s , then F i s com plete#

PROOF. We may suppose t h a t i s e s s e n t ia l f o r f , and

th en by Theorem 4.1 th e re i s a p a i r o f Séupecki p o in ts l e t th e

correspond ing v a lu es be u^ ,u® ,u®. Since u^ ^ u® we must have

(%i / , and s in ce v? / u® we must have oli / f o r some index

i ^ 2 . We s h a l l need th e fo llo w in g two lemmas;

LEMMA 4 . 1. Under the hypotheses o f Theorem 4 .4 F g en e ra te s

a 2-p la c e fu n c t io n , which we s h a l l denote by x y , having

th e fo llow ing p r o p e r t ie s ;

X 0 0 = X = 0 ^ X ,

f o r a l l X e N.

PROOF. L et a (x ) be a p erm u ta tion such th a t a (u ^ ) = 2,

a(u^ ) = 0 , a(u®) = 1, and co n sid er th e k -p la c e fu n c tio n

gen era ted by F

h (x j , «. . ,Xk ) = a ( f (xji , . * * ) ) •

V/e have

h (a ) = 2 , h(/9i ,tt2 , ) = Of k(/9) = 1,

and s in ce f ta k e s a l l n+1 v a lu es o f N, f o r each in te g e r i

(3 ^ i ^ n ) , th e re e x i s t numbers Xi ,xk such th a t

h ( x i^ , .* . ,x k ^ ) = i .

^0

L et a i and a g , b e 1—p lace fu n c tio n s such th a t

Q>i ( 0 ) = ^ 1 f &i ) = (%i J &i ( 1 ) “ 3Ci f o r 1 = f 9 « • fiif

and f o r each index j (2 ^ j k)

a j (0) = ocj f a j ( 1 ) = /9j f a j ( i ) = xj f o r i =

C onsider th e 2—p lace fu n c tio n g en e ra ted by F

b (x ,y ) = h ( a i ( x ) , ag ( y ) , . . • ,a k ( y ) ) .

Y/e have

b (0 ,0 ) = 0 , b ( 0 , l ) = 1, b ( l ,0 ) = 2,

and fo r each in te g e r i ( 3 ^ i ^ n) b ( i , i ) = i .

On w ritin g b (x ,y ) in the m a trix n o ta t io n , and le a v in g any

undeterm ined v a lu e cf b b lan k , we have

We s h a l l proceed in d u c tiv e ly , denoting by (x ,y ) a fu n c tio n o f

En which has th e fo llo w in g p ro p e r t ie s :

Cfn(0 ,x ) = X = Cm(x,0 ) fo r each x (O ^ x ^ m).

Now f g e n e ra te s a fu n c tio n of ty p e C i ( x , y ) , namely d ( b ( x ,y ) ) ,

where d i s any l ^ l a c e fu n c tio n such t h a t d(o) = 0, d ( l ) = d(2)

and any fu n c tio n o f type Cn (x ,y ) can be taken a s x ^ y .

= 1 )

41

I t i s s u f f i c i e n t th e re fo re to prove th a t f o r each in te g e r m

( l i ^ m < n ) i f F g e n e ra te s a fu n c tio n o f type Cn,(x,y) th en

F g e n e ra te s a fu n c tio n o f type C m +i(x,y),

Suppose F g e n e ra te s a fu n c tio n of type Cm(x,y) (1 ^ m < n ) ,

and l e t @1,62 be the 1—p la ce fu n c tio n s d e fin e d by

e i ( o ) = 0 , 6i ( x ) = x - 1 , f o r x = 1 , . . . , n ;

63( 0 ) = 0 , 63( 1) = 1 , 63( 2) = 0 , 63 (x ) = x - 1 ,

f o r x=3 , . . . , n ;

63( 0 ) = 0 , 63( 1 ) = 1 , 63 (x ) = x+1, f o r x=2 , . . . , n ,

where a d d it io n and s u b tra c t io n a re c a r r ie d out modulo n+1 .

C onsider the two 2—p lace fu n c tio n s g en era ted by F

6 4 (x ,y ) = e s ( c m ( e i ( x ) ,e i ( y ) ) ) ;

@6 = eg(cm(e3 (x ) , e g ( y ) ) ) .

On w r it in g 64 , eg in the m a trix n o ta tio n and le av in g un im portan t

v a lu es b la n k , we have r e s p e c t iv e ly

64(x ,y )0123

m+1

n

0 1 2 5 *«ni+1 . . n0 0 1 3 . . m+1 013

m+1 m+1 m+1

n

I t i s noTf easy to v e r i fy t h a t the 2-p la c e fu n c tio n g en era ted by F

b (e 4 (x ,y ) , ©sCxjy))

is o f type C m + i(x ,y ), and so th e r e s u l t fo llo w s .

1 2

LEMMA. 4#2. Under the hypotheses o f Theorem 4 .3 , f o r

each t r i p l e o f f ix e d numbers i , j , k (O ^ i , j , k ^ n) Fk'Ugenerates th e 2-n la c e fu n c tio n rM (x ,y ) d e fin ed by

r*Lj(x,y) = 0 o th e rw ise .

PROOF. Suppose b ( l , l ) = v , say , and s e le c t a value

i € [0 ,1 ,2 j d i f f e r e n t from v . L et b(Yi ,Y2 ) = where

Yj = 0 o r 1 , f o r j = 1 ,2 . Choose 1-p la c e fu n c tio n s S i,S 2 ,S3

such th a t

5i ( l ) = Yi , 8 i(x ) = 1-^1 f o r X / 1 ;

83( 1) = Y 2 , 83 ( x ) = I-Y2 f o r X / 1 ;

s s ( i ) = 1 , 63 (x) = 0 f o r X / i .

V/e have

r i i ( x , y ) = 8 3 ( b ( s i ( x ) , S 2 ( y ) ) ) ,

and so F gen erate s r j i ( x , y ) .

For each t r i p l e of f ix e d numbers i , j , k (O ^ i , j , k ^ n)

choose 1-p la c e fu n c tio n s t i , t j , t k ' such th a t

t i ( i ) = 1, t i ( x ) / 1 fo r X / i ;

t j ( j ) = 1 , t j ( x ) / 1 fo r X / j ;

tk * (o )^ t% '( l ) = k .

Thenr i j ( x , y ) = tk’ ( r i \ ( t L ( x ) , t j ( y ) ) ) .

43

We s h a l l now r e tu r n to th e p roof of Theorem 4* 3*

I f <f>{x,y) i s any f ix e d 2—p lace fu n c t io n o f Ep then

n n0 ( x , y ) = Z Z r i j ( x , y )

1=0 J=o

where a d d i t io n i s w ith r e s p e c t to th e o p e ra tio n x y ,

and th e fu n c tio n s r y (x ,y ) a re a s d e fin ed in Lemma 4 .2 .

T herefore by Lemmas 4*1, 4 .2 F g e n e ra te s ev ery 2—p lace fu n c tio n

of En, and by Theorem 3 .2 of C hapter 3 th i s i s a s u f f i c i e n t

c o n d itio n f o r F to be com plete.

REIvIAEK. Theorem 4*3 was f i r s t proved f o r th e case k = 2

by S-êupecki in [55]» The c o n d itio n n ^ 2 in Theorem 4 .1 cannot

be re la x e d , as i s shown by th e example in which n = 1 and F

c o n s is ts o f th e s e t o f a l l l i n e a r polynom ials (see Theorem 1 .2 ) .

I t i s easy to v e r ify th a t F i s c losed under com position ,and

con tains a fu n c tio n s a t i s f y in g th e S lupeck i co n d itio n s and a lso th e

s e t of a l l 1 -p lace fu n c tio n s o f (namely 0 ,1 ,x ,x + l) , b u t

con ta in s no n o n - lin e a r polynom ial of Ei^. T herefore F i s no t

com plete.

44

THEO HEM 4 . 5 . I f a s in g le fu n c tio n ^ g en era te s the

se t o f a l l 1 -p la c e fu n c tio n s of Ep , th en f i s com plete,

PROOF, % Theorem 4 .3 C o ro lla ry 1, s in ce f g e n e ra te s

a l l c o n s ta n ts , f s a t i s f i e s th e S^upecki c o n d itio n s . Hence

Theorem 4 .5 fo llo w s by Theorem 4 .4 i f n ^ 2 . I f n = 1,

th e n , by argum ents s im ila r to th o se in th e proof o f Theorem 2 .2 ,

f o r a s in g le fu n c tio n f to g en e ra te a l l 1—p lace fu n c tio n s of

Ej^ i t i s n ecessa ry th a t f ( x , . . . , x ) = x + 1 , and th a t

>i’ ) + 1 (otherwise f cannot generate the co n sta n ts).

But, by Theorem 2.2. these are s u f f ic ie n t con d ition s fo r f to

be complete.

REMARK. The proofs of Theorems 4 .3 , 4 .4 are based on

the proofs o f Theorems 5 .1 , 5*4 o f [ 2 7 ] by A.Salomaa, but

the reasoning i s more d ir e c t than Salom aa's.

CORÛLLAR.Y. I f f i s a s in g le fu n c tio n w hich g en era te s th e s e t o f

a l l .i-p la c e fu n c tio n s o f Ep ( j ^ I ) , th e n f i s com plete.

PROOF. For j = 1 th i s i s Theorem 4 .5 . I f j > 1, th en by

id e n t i fy in g v a r ia b le s in s u i ta b le j-p la c e fu n c tio n s we can show

t h a t f g en era tes the s e t o f a l l 1-p la c e fu n c tio n s o f Ep and th e

r e s u l t follow s by Theorem 4 .5 .

45

THE OEM 4 .6 . I f 2 and a su b se t P g e n e ra te s the

s e t o f a l l 1 —p lace fu n c tio n s which a re n o t perm uta tions o f N ,

to g e th e r w ith a fu n c tio n f s a t i s f y in g th e S-^upecki c o n d it io n s ,

th e n F i s com plete .

PROOF. We s h a l l prove t h a t F a lso g e n e ra te s every

p erm u ta tion of N, and co nsequen tly F g e n e ra te s th e s e t of

a l l 1-p la c e fu n c tio n s of Ep . The r e s u l t th e n fo llo w s by

Theorem 4 .4 .

V/e m%r suppose t h a t i s e s s e n t i a l f o r f , and th en

b y Theorem 4.1 th e re is a p a i r o f S*êupeclci p o in ts a,/3 for f ;

l e t th e corresponding v a lu es be u^ ,u^ ,u® . We may suppose t h a t

the n 4- 1 d i s t i n c t e lem ents o f N a re u^ ,u^ ,u® ,u^ , . . . ,

and s in ce f assumes a l l n + 1 v a lu e s ,fo r each in te g e r i

( 4 i ^ n+1 ) th e re e x i s t numbers Xj_ ^, . . . ,xi< such th a t

f (X] . ,X|< ) = u .

L e t p(x) be any f ix e d perm uta tion o f N, and choose

k 1-p la c e fu n c tio n s q i , . . . , q k d e fin ed by

Ç h (p " i(u i) ) = %i, Ç b(p ri(u 2 )) = ^ 1 , qi(p"^Iu® )) = /? i ,

fo r i = 4 , . . . , n+1;

and f o r each index 3 (2 ^ j ^ k)

qj(p"^(u^)) = a j , qj(p“ ^(u®)) = aj , qj (p“ (u® ) ) = /9j ,

qj = xj for i = 4 , . . . ,n + 1 .

4 &

Since q i , . . . , q k tak e a t most n values th e y a re n o t perm uta tions

o f N and so i h ^ a re g en era ted by P . T herefore P g en e ra te s

the 1-p la c e fu n c tio n

r ( x ) = f ( q i ( x ) ,q 2 ( x ) , . . . , q k ( x ) ) ,

and r ( x ) s a t i s f i e s

r(p ~ ^ (x ) ) = X

That i s

r ( x ) = p (x ) .

T herefore P g e n e ra te s p (x) and th e r e s u l t fo llo w s .

4 7

THEOREM 4*7* I f f is a s in g le fu n c tio n of Ep which

g e n e ra te s th e s e t of a l l 1-p la c e fu n c tio n s which a re n o t

perm utations o f N, th e n f i s com plete.

PROOF. ^ Theorem 4 .3 C o ro lla ry 1 , s in ce f g e n e ra te s

a l l c o n s ta n ts , f s a t i s f i e s th e S-^upecki c o n d itio n s .

Hence Theorem 4 .7 fo llow s from Theorem 4 .6 i f n ^ 2 .

I f n = 1 , th e n th e two 1-p la c e fu n c tio n s of E ^ which a r e not

p erm uta tions of a re th e c o n s ta n ts 0 ,1 . N ecessary

co n d itio n s f o r f to g en era te th e c o n s ta n ts 0,1 a re th a t

f 1 , and Ih a t f ( x , . . . , x ) = x+1 ( f o r i f

f ( x , . , . x ) = i then f cannot g en era te 1 - i , and i f f ( x , . . . , x ) = x

then f g e n e ra te s no o th e r 1-p la c e f u n c t io n ) . Theorem 2 .2

th e se a re a ls o s u f f i c i e n t c o n d itio n s fo r f to be com plete.

REMARKS. Theorems 4 .6 , 4 .7 were o r ig in a l ly proved in [A-9 ]

by Y ablonslcii, However, the above p roo fs do not use h is re a so n in g .

The s e t ©f a l l 1-p la c e fu n c tio n s o f i s th e f u l l tra n s fo rm a tio nsemigroup of N in to i t s e l f , the semigroup o p e ra tio n being com position .The idem potents o f th i s semigroup a re th e 1-p la c e fu n c tio n s f € such

th a t f ( f ( x ) ) In [ l l j , Howie has proved th a t th e s e t of a l l 1-p la c e

fu n c tio n s o f Ep which a re n o t p e rm u ta tions o f N,. i s generated by th e

su b se t o f a l l idempo te n ts o f th e type [ 2 , 1 , . . . , 1 ] , w ith n u n i t s . It*

fo llow s th e re fo re by Theorem 4 .7 , t h a t i f a s in g le fu n c tio n f e Ep

g en e ra te s every idem potent o f "type [ 2 , 1 , . . . , l j , w ith n u n i t s , "then

f i s com plete.

48

V - MAIN RESULTS - V

5 . 1 • In tro d u c tio n

In Theorem 4 .6 we proved th a t a s in g le fu n c tio n o f Ep which

g en e ra te s the s e t o f a l l 1-p la c e fu n c tio n s o f Ep i s com plete. In th is

C hapter we s h a l l be concerned w ith com pleteness c r i t e r i a fo r a su b se t o f

Ep which g e n e ra te s perm u ta tio n s o f N. These c r i t e r i a a re p re sen te d in

Theorems 5 .1 , 5 .2 , 5 .3 o f § 5 .5 .

In § 5 .2 we c o l le c t some f a i r l y s tra ig h t- fo rw a rd a u x i l ia r y r e s u l t s ,

w h ile in § 5 .3 and § 5 .4 we in v e s t ig a te in d e t a i l an e x ce p tio n a l s i tu a t io n

which may a r is e when n+1 i s a power o f 2 . F in a l ly in ^ 5 .5 we s t a t e

and prove our main theorem s, and compare them w ith e a r l i e r r e s u l t s .

We s h a l l make use o f v a rio u s n o tio n s and d e f in i t io n s in tro d u ced in

p rev io u s c h a p te rs . F o r example; S -^peck i co n d itio n s (C hapter 4 ) ,

s -p ly t r a n s i t i v i t y (C hapter 1) and p ro p e r t ie s o f 1-p la c e fu n c tio n s

(C hapter I ) .

5 . 2 . A u x ilia ry r e s u l t s

LEMMA 5. 1. n ^ 3 , and a su b se t F g e n e ra te s a t r i p l y t r a n s i t i v e

group o f perm uta tions of- N, to g e th e r w ith a fu n c tio n f ( x i , . . . ,% k )

s a t i s f y in g th e S ^ p e c k i c o n d itio n s , th e n F g en era te s a 1-p la c e fu n c tio n

which tak es more th a n 1 and l e s s th a n n+1 d i s t i n c t v a lu e s .

PROOF, By Theorem 4 .2 th e re i s a p o in t y (= Yi>****Yk) su ch t h a t ,

f o r s e ts S i , . . . ,S k d e fin ed by Sj, = N / îy i l , Tor i = 1 , . . . , k ,

f ( ? i , . . . > ^ ) — N.

49

From th i s we deduce th a t th e re i s a p o in t y *(* Yi * , • • • ,Yk * ), where

YL* f St fo r i = such t h a t f(Y*) = T(y )* Choose k

perm uta tions g en era ted by F such t h a t (c ) =Y l>

r t ( l ) = Y t* , f o r i = 1, . . . , k . C onsider th e 1-p la c e fu n c tio n

g en era ted by F

f (x) = f (rj. ( x ) , • • . , 1 (x) ) •

I t i s easy to see t h a t f (0 ) = f ( l ) , and so f(x ) tak es l e s s th an n+1

v a lu e s .

To complete the p ro o f i t s u f f ic e s to c o n sid e r th e case when

f (x ) i s c o n s ta n t. We may suppose t h a t Xi i s e s s e n t ia l fo r f , and

th en by Theorem 4.1 th e re i s a p a i r o f Séupecki p o in ts a , l e t the

co rrespond ing values be u^ , i f , i f . Since f i s c o n s ta n t, i t

fo llo w s th a t F g en era te s a l l c o n s ta n ts , and so F g enera tes th e

1-p la c e fu n c tio n f (x,aa , . • .yxk ) • This fu n c tio n takes th e va lues

u , i f , and th e p roo f i s com pleted i f i t does n o t a ls o tak e th e value

if . Suppose then th a t

f(oC ,R2 ,.. * . ,(%k ) = U ,

where 4= cti , ^ i . Choose a f u r th e r p o in t f (=fi ) such t h a t

4= and f o r each index i ( 2 ^ i ^ k) th e 3 v a lu es i

a re c o in c id e n t i f a t = /3 l , o r d i s t i n c t i f a t 4 P i* C onsider th e

value o f f ( f ) . I f f ( c ) = if , th e n we choose k 1-p la c e fu n c tio n s

8 i , . . . , 8 k g en era ted by F such th a t s t ( 0 ) = a t , s t ( l ) = P t ,

S t (2) = f t . Tor i = 1 , . . . , k (accord ing as a t = P t o r a t 4 P i» 81

i s a c o n s ta n t o r a p e rm u ta tion r e s p e c t iv e ly ) . C onsider the 1-p la c e

fu n c tio n gen era ted by F

so

b(x) = f ( s i ( x ) , . . . , S k ( x ) ) .

We have b (0 ) = ^ b ( l ) = b (2 ) = i f , and so b ta k es more th a n 1

and l e s s th a n n+1 values#. I f f (f ) 4 > th e n we choose a f u r th e r

perm u ta tio n s* g en era ted by F such th a t S i '( o ) = , S i* ( l) = /3 i,

8i *(2) = f 1 # C onsider th e 1-p la c e fu n c tio n gen era ted by F

b*(x) = f ( s i * ( x ) ,% ( x ) , . . . ,S k ( x ) ) .

We have b '( 0 ) = b * (l) = u® 4 t ' ( 2 ) , and so b* tak es more th a n 1 and

le s s th an n+1 d i s t i n c t v a lu e s . The p ro o f i s com pleted,

REMARK, Using th e above p ro o f Salomaa ([20 ] Lemma 1,4) d e riv e s

th e same co n c lu sio n when F g en e ra te s the a l te r n a t in g group An-Ki on

N ( to g e th e r w ith a fu n c tio n s a t i s f y in g th e S6upecki c o n d it io n s ) ,

b u t as h is p ro o f uses only th e t r i p l e t r a n s i t i v i t y o f Aq^we have a p p lie d

i t to th i s more g e n e ra l case .

The d b n d itio n n ^ 3 in Lemma 5.1 cannot be re la x e d ,a s i s shown

by th e example in w hich n = 2, and F c o n s is ts o f a l l l i n e a r fu n c tio n s ,

th a t i s , a l l fu n c tio n s e x p re ss ib le in th e form

f ( x i , . . . , 2k) = lb + ihXi + . . . + (modulo 3 ) .

I t i s e asy to v e r i f y t h a t F i s c lo sed under com position and co n ta in s

th e sym m etric group o f perm uta tions % on N, to g e th e r w ith a fu n c tio n

s a t i s f y in g th e S^upecki c o n d itio n s , fo r example x + y (modulo 3 ) , b u t

F co n ta in s no 1-p la c e fu n c tio n ta k in g e x a c tly 2 v a lu e s .

51

LEIÆMA. 5 .2 . I f n ^ 2 , and a su b se t F g e n e ra te s a t r i p l y

t r a n s i t i v e group o f p e rm u ta tio n s of N, to g e th e r w ith a 1—p la c e

fu n c tio n ta k in g e x a c tly t v a lu e s , where 1 < t < n + 1 , th e n

F g e n e ra te s a 1-p la c e fu n c tio n ta k in g e x a c tly 2 v a lu e s .

PROOF. Suppose F g e n e ra te s th e 1—p lace # (x ) ta k in g

t d i s t i n c t v a lu e s v ^ , . . . , v ^ , where 1 < t < n + 1 .

I f t = 2 th e re i s no th ing to p ro v e ; i t i s th e re fo re s u f f i c i e n t

to prove th a t i f t ^ 3 then th e hypotheses imply th a t F g e n e ra te s

a 1—p lace fu n c tio n ta k in g v a lu e s , •vdiere 1 < t* < t .

L et V*’ s [x : # (x ) = v* 'j. S ince t < n , a t l e a s t one o f

th e se s e t s , say , co n ta in s a t l e a s t two members, a,/9 say .

Let Y be any member of V®, l e t g (x ) be a p e rm u ta tion

g en era ted by F , such th a t

g ( v ^ ) = a , g(v®) = /9, g (v ® )= Y ,

and l e t 0 (x ) deno te the fu n c tio n ^ (g (< ^ (x ))). We have

e(vM = I v ^ j , e(v®) = JvM, e(v®) = [v®J,

w hile on each o f th e rem aining t —3 s e ts V * ,. . . ,V ^ th e fu n c tio n

6 ta k e s a c o n s ta n t v a lu e . Thus 6 ta k e s t* v a lu e s , where

1 < t* < t .

LEMMA 5*3. Under the c o n d itio n s o f Lemma 5 .2 , F g e n e ra te s

a l l c o n s ta n ts .

52

PROOF. Let T be any g iv en member of N* Lemma 5*2,

F g e n e ra te s a 1-p la c e fu n c tio n ^ (x ) ta k in g e x a c tly 2 v a lu e s ,

v^,v® say . At l e a s t one o f ih e se v a lu e s , v^ say , must be

assumed a t l e a s t tw ice, a t ,a® say . Let g ,h be

perm uta tions generated by F such t h a t g(v^ ) = v , and

h(v^ ) =s , h(v® ) = a® • Then th e 1 -p la c e fu n c tio n

g (^ ( h ( ^ ( x ) ) ) ) tak es th e c o n s ta n t v a lu e v ,

LSI&IA 5 .4 . ^ n ^ 1 , and a su b se t F g e n e ra te s

a doubly t r a n s i t i v e group o f perm uta tions o f N, to g e th e r

w ith a 1-p la c e fu n c tio n g ta k in g e x a c tly 2 v a lues and

a 2—p lace fu n c tio n h such th a t

h (0 ,0 ) = 0 ; h ( 0 , l ) = h ( l , l ) = h ( l ,0 ) = 1,

th en F g e n e ra te s every 1—p la c e fu n c tio n of type [ n . l l .

PROOF. Choose numbers a,/9 such th a t g (a ) ^ g (p ) .

Choose perm u ta tions r ( x ) and S i ( x ) , . . . ,Sp(x) g en era ted by F

such th a t r ( g ( a ) ) = 0 , r(g (/? )) = 1; s i (o ) = a , s i ( i ) =

f o r i = 1 , . . . , n . L et t i ( x ) = r ( g ( s ( , ( x ) ) . Then th e 1 -p la c e

fu n c tio n g en era ted by F

^ (x ) = h [ t i ( x ) , h [ t 2 ( x ) , . . • ,h [tn ^ 1 ( x ) , t p ( x ) ] • . , ] ]

assumes th e v a lu e 0 f o r x = 0 and th e v a lu e 1 o th e rw ise .

Now l e t ^ (x ) be any o th e r 1-p la c e fu n c tio n of ty p e [ n , l ] ,

assuming th e v a lu e v^ f o r x = y and th e value Vg / V;i o th e rw ise .

55

Choose p en in ita tio n s s , t gen era ted by F such th a t s ( l ) = v^ ,

s (2 ) = v^; t ( y ) “ 1 . Then *(x) = s( * ( t ( z ) ) ) , and th e re fo re

ip i s g en era ted by F*

LE&1MA 5 .5 . I T n ^ 2 , th en fo r any n a tu ra l numbers

t , b ^ , , . , , b ^ s a t i s f y in g 1 < t < n+ 1 , b^ + . . . + b^ * n ,

th e s e t o f a l l 1- p lace fu n c tio n s of ty p e , [ b ^ , . . , ,b ^ ] g e n e ra te s

every 1- p lace fu n c tio n of type [b^ + b ^ ,b ^ , , . , ,b^] ,

PROOF. L et 0 be an a r b i t r a r y f ix e d , fu n c tio n of ty p e

[b^ + b ^ ,b ^ , . . . , b ^ ] assuming th e value v^ e x a c tly b^ + b^ t im e s .g

Since t .< n + 1, 0 has a m issin g v a lu e , v say , and any fu n c tio n

(j)(x) d e fin ed by: 4>(x) = 0(x ) , excep t fo r b^ v a lu es o f x fo r

which 4»(x) = v^ ; 0(x ) = v^ , w i l l c le a r ly be o f type [b ^ ,b ^ , . , , ,b^] ,

Choose a fu n c tio n <t»* of type [ b ^ ,b ^ , . , . ,b^(| such th a t

4>*(0(x ) ) = 0(x ) ; * '(v * ) = v ^ .......................... ( i )

We can c o n s tru c t (j>' as fo llow s: s in ce t < n + 1 , fo r some

fix e d index i ( l ^ i ; ^ t ) b j ^ ^ 2 , Let th e t - 1 d i s t i n c t v a lu es

o f 0 be v ^ ,0* , . . . , 0^ ^ , and l e t (j)' ( s a t i s f y in g c o n d itio n ( i ) ) tak e

th e va lu e v^ e x a c tly b^ tim es and th e v a lu es v ^ ,0* , . . . , 0^ e x a c tly

b ^ , . . . ,b ^ _ ^ , t im e s ;re s p e c t iv e ly . S ince v i s a

m issing v a lu e o f 0, i s of re q u ire d ty p e .

F in a l ly we have

0(x ) = <j)»( * ( x ) ) ,

and th e r e s u l t fo llo w s ,

REMARK, This sim ple r e s u l t iff Lemma 1 .3 o f [gs] ,

54

LEMMA. 5*6* IT n ^ 2 , and a su b se t F g en era te s a t r i p l y

t r a n s i t i v e group o f p em iu ta tio n s of N, to g e th e r w ith a fu n c tio n

s a t i s f y in g th e S-êupecki co n d itio n s an d , f o r some

in te g e r m s a t i s f y in g 1 ^ m ^ n - 1, every 1 -p la c e fu n c tio n o f

type [n+1-m ,1, . • • ,1 ] (w ith m u n i t s ) , then F g e n e ra te s every

1-p la c e fu n c tio n o f ty p e [n -m ,1, . . . , l ] (w ith an a d d it io n a l u n i t ) .

PROOF* app ly ing Lemma 5*5 we can show th a t F g e n e ra te s

every 1-p la c e fu n c tio n which ta k e s some value a t l e a s t n+1-m tim es .

V / V V\For each v c N, th e re i s a p o in t x = (x^ , . . . ,Xk ) such

th a t f (x ^ ) = V. We may suppose w ith o u t lo s s o f g e n e r a l i ty th a t

Xi i s an e s s e n t i a l v a r ia b le f o r f , and th e n by Theorem 4*1 th e re

i s a p a ir o f S>ftupecki p o in ts a,/9; l e t th e co rrespond ing v a lu es

be u^ ,u® ,u® .

Let e (x ) be any f ix e d 1-p la c e fu n c tio n of ty p e [n -m ,1 , . . . , 1 ] ,

w ith

e ( a i ) =3 w° ( i= 1 , . . . , n - m ) ; e (b j ) = ( j = 1 , . . . ,m + l )

Choose a p e rm u ta tio n r g en era ted by F such th a t

r ( u ^ ) = w^, r(u® ) = w°, r(u® ) = w®.

C onsider th e 1-p la c e fu n c tio n s h^ and h^(& = 2 , . . . , k ) d e fin e d

a s fo llo w s;

55

h i(x ) = I

/?x fo r X rs

«1 f o r X = b i

r " i (wJ) f o r X = b a , • •

h^(x ) =

[

fo r X = 8-1 , • • • ,b%

f o r X = b2r - i (w J )

^6 f o r X = b3,*«»,bn»n*

We see t h a t each of th e se fu n c tio n s ta k e s one value a t l e a s t

n+1-m tim es, and i s th e re fo re gen era ted by F , b u t i t i s easy

to v e r i f y t h a t

e (x ) = r ( f ( h i ( x ) , h g ( x ) , . . . , h k ( x ) ) ) ,

and so e (x ) i s a lso g en e ra ted by F and th e lemma i s p roved .

COROLLARY. Under the c o n d itio n s o f Lemma 5*6 , F g e n e ra te s

every 1-p la c e fu n c t io n .

PROOF. re p e a te d a p p lic a tio n s o f Lemma 5*6, we can

deduce t h a t F g e n e ra te s every 1-p la c e fu n c tio n of type [1 ,1 ,

th a t i s , every pe rm u ta tio n , and ev e iy 1-p la c e fu n c tio n o f type

[ 2 , 1 , . , . , l ] . app ly ing Lemma 5*5, we can deduce th a t F a ls o

g e n e ra te s every 1—p lace fu n c tio n which i s n o t a p e rm u ta tio n .

LEMMA 5*7* I f n ^ 2 , th e s e t o f perm u ta tions o f N th a t

a re s e lf -c o n ju g a te under a g iven 1-p la c e fu n c tio n s (x ) ^ x i s not

doubly t r a n s i t i v e .

56

PROOF. I f Xq i s any elem ent of N fo r which s(xo ) / Xo,

and r i s any p erm u ta tio n s e lf -c o n ju g a te under s (x ) f o r which

r(xo ) = xb , th en a ls o r ( s ( x o ) ) = s ( r (x o ) ) = s(xq)*

5•3* An e x c e p tio n a l case

We s h a l l now show th a t i f n ^ 3 and F s a t i s f i e s th e

c o n d itio n s o f Lemmas 5.1 and 5*2, then i t w i l l a lso s a t i s f y

th e a d d it io n a l c o n d itio n of Lemma 5*4, u n le ss n+1 i s a power

of 2 and F has a very s p e c ia l form . We s h a l l f u r th e r

in v e s t ig a te t h i s s p e c ia l form in § 5 *4 , in p a r t i c u la r showing

in Lemma 5*10 C o ro lla ry 2 th a t i t can indeed occu r.

LEMMA. 5*8* Let n ^ 2 and suppose th a t F g e n e ra te s

a t r i p l y t r a n s i t i v e group o f perm u ta tions of N, to g e th e r w ith

a fu n c tio n f ( x ^ , . . . ,X k ) s a t i s f y in g th e S’êupecki co n d itio n s

and a 1-p la c e fu n c tio n g (x ) tak in g e x a c tly 2 v a lu e s , bu t F

does not g en era te any 2-p la c e fu n c tio n h (x ,y ) such th a t

h (0 , 0 ) = 0 ; h (0 , l ) = h ( l , l ) = h ( l , 0) = 1 . (a )

Then th e re e x i s t s a number v e N such t h a t ( i ) v / [0 ,1 ,2 }

and ( i i ) F does not g en e ra te any 1-p la c e fu n c tio n th a t on th e s e t

[ 0 , 1 , 2,v j ta k es one v a lu e e x a c tly 3 tiroes and an o th er va lue e x a c tly

once. Moreover F g e n e ra te s 1-p la c e fu n c tio n s g ^ ,* .* ,g ^ each

ta k in g th e v a lu es 0,1 and s a t i s f y in g g l(o ) = 0 ( i = 1 , . . . , j ) ,

such t h a t th e correspondence x (g^ ( x ) , . • . ,g^ (x) ) between elem ents

of N and sequences of j 0*s and 1*s i s o n e -to -o n e . In

p a r t i c u la r n+1 = 2'^.

57

*PROOF. We may suppose th a t i s an e s s e n t ia l v a r ia b le

f o r f , and th en by Theorem 4 .1 . th e re i s a p a i r of S-ôupecki

p o in ts a,/9; l e t th e co rrespond ing th re e v a lu es be u^ ,u® ,u® .

Since u^ / u® we must have a i ^ /9i and since u® / u® we

must have a t / ^ t Tor some index i ^ 2 .

Choose k fu n c tio n s r j ( x ) ( j = 1 , . . . , k ) gen era ted by F

such th a t r j ( l ) = a j ; r j (2 ) = /9j . We can always make t h i s

choice because i f a j then we may ta k e a s r j ( x ) a s u i ta b le

perm uta tion g en e ra ted by F, and i f aj = /9j th en s in c e , by

Lemma 5*3, F g e n e ra te s a l l c o n s ta n ts we may tak e r j (x) = a j .

Let s (x ) be a p erm uta tion g en era ted by F such th a t

s (u ^ ) = V, f o r V 3 0 , 1 , 2 .

C onsider the 2 -p lace fu n c tio n g e n e ra te d by F

T *(x,y) = s ( f ( r i ( x ) , r g ( y ) , . . . , r k ( y ) ) ) .

V/e have

f » ( 0 , 0 ) = 0 , f * ( l , 0 ) = 1 , = 2 , f » ( 0 , l ) = V ( s a y ) ,

and from th e re a so n in g so f a r on ly th e p resence o f such a fu n c tio n

need be r e ta in e d .

Since n ^ 2 th e 1-p la c e fu n c tio n g tak es one of i t s

2 v a lu es a t l e a s t tw ice , and so we may choose d i s t i n c t numbers

a,/5?Y N such th a t g (a ) / g(/9) = gCr)*

My o r ig in a l p ro o f was in ad eq u a te . I am g r a te f u l to Dr. Davies fo r p o in tin g t h i s o u t, and fo r h is h e lp in the p re se n t p to o f .

58

Suppose i f p o s s ib le t h a t v e [0 ,1 ,2 } , and l e t v = Kg,

where ,«2 a,re th e numbers 0 ,1 ,2 in some o rd e r. Choose

f [ 0,1 i so t h a t f*(/Ji,jUg) = Ko, and choose perm uta tions

b i , t g , t 3 , t 4 gen era ted by F such th a t

t i ( 0 ) = ^ 1 , t i ( l ) = 1 - tg (o ) = /ia , t g ( l ) = 1 - ^3 ;

taCao) = a, taCai) = /5, ta (ag ) = y; t4(g(a)) = 0, t 4 (g(^)) = 1.

Then i t i s easy to v e r i f y t h a t th e 2 -p lace fu n c tio n (g en e ra ted

by F)

h (x ,y ) = t 4 ( g ( t a ( f * ( t i ( x ) , t g ( y ) ) ) ) )

s a t i s f i e s ( a ) , c o n tra ry to h y p o th e s is . Thus a s s e r t io n ( i ) i s p roved .

Suppose i f p o s s ib le t h a t F g e n e ra te s a 1-p la c e fu n c tio n

d (x ) such th a t d(%o) / d ( % i ) = d ( a g ) = d ( % a ) , where Go,Gi,K2,%3

a re th e numbers 0 ,1 ,2 ,v in some o rd e r. Let S3 be a p erm u ta tion

g en era ted by F such th a t S a ( d ( a o ) ) = 0 and S a ( d ( a i ) ) = 1, choose

1 19^2 ^ [ 0 , 1 } so t h a t f *(jL/i ,/jg ) = Go, and choose p e rm u ta tions

S i ,Sg g en era ted by F such th a t

81( 0 ) = / i i , 81( 1 ) = 1 - ^ 1 ; Sg(0) = /Jg, S g (l) = 1 - /ig.

Then i t i s easy to v e r ify t h a t th e 2 -p lace fu n c tio n (g en e ra ted by F)

h (x ,y ) = 5 a (d ( f* (x ,y ) ) )

s a t i s f i e s (a ) c o n tra ry to h y p o th e s is . Thus a s s e r t io n ( i i ) i s proved.

For th e l a s t p a r t of th e p roo f i t i s convenient to fo rm u la te

an in te rm e d ia te r e s u l t .

59

ASSERTION ( i l l ) . For each in te g e r k such th a t 2 ^ ^ n+1,

F g e n e ra te s k 1-p la c e fu n c tio n s each ta k in g e x a c tly

th e v a lu es 0 ,1 and s a t i s f y in g g*’(0 ) = 0 ( i = 1 , . , , , k ) such

th a t (w ith i i , . . . , i k running over a l l sequences of k 0 ' s and 1 *s )^1 * * *

(a ) fo r each f ix e d K, 1 ^ K ^ k , the s e t s V

= [x ; g^(x) = i^ f o r 1 ^ 6 ^ K} a l l have th e same power,

(b) whenever r i s a perm u ta tio n g en era ted by F f o r which

r ( o ) , r ( l ) , r ( 2 ) e V we a ls o have r ( v ) e V ^

I t i s c l e a r t h a t th e rem ainder o f Lemma 5 w i l l fo llo w from

A sse rtio n ( i i i ) , s in c e i f j i s th e l a r g e s t in te g e r fo r which

2^ ^ n+1 th en we must have 2* = n+1.

PROOF OF ASSERTION ( i i i ) . We use in d u c tio n on k , s ta r t in g

w ith th e t r i v i a l case k = 0 , vhere i t i s understood th a t^1 • •

V = N f o r k = 0 . Suppose th e r e s u l t t r u e fo r k , and

l e t 2*^^ ^ n .

Since g^(o ) = 0 , f o r i = 1 , . . . , k , we have 0 e -(w ith k z e ro s ) . Denote by 1 ' any o th e r elem ent of , andl e t t 4 , t s be perm utations g en era ted by F such th a t

t4 (g (a ) ) = 0 , t4 (g (^ )) = 1; ts (o) = o t, tg ( 1 ' ) = ^ .

Let g' '*’^ (x ) = t 4 ( g ( t e ( x ) ) ; then g‘ '^^(x) i s g en e ra ted by F ,

tak es e x a c tly th e v a lu es 0 ,1 , and s a t i s f i e s gk+ i(o ) = 0 ,

gk+ 1, Let

y i i .* . i k ik + i _ ; g^(x ) = i^ fo r 1 < -6 ^ k+1 j .

6 0

I t w i l l be s u f f ic ie n t to prove th a t f o r a l l i i , . . . , i k th e

two s e t s jjave th e same power and th a t fo r

a l l i i , • , . ik , ik + 1 , whenever r i s a perm uta tion g en era ted by

P f o r which r ( l ) , r ( 2 ) , r ( 3 ) € we a ls o have

r ( v ) €

Let us f i r s t show th a t none o f th e s e ts

empty. Suppose i f p o s s ib le t h a t empty; then

g^* i s c o n s ta n t on Choose a perm u ta tion r i (x )

gen era ted by F such t h a t i*i(o) = 0 , r i ( l ) = 1* and i 'i ( 2 ) = u# * # ( # (

where u i s some elem ent of and. l e t r i ( v ) e v ^ ^

say . I f i ^ = i ^ fo r 1 ^ 6 ^ k then r i ( v ) c and

th e 1 -p lace fu n c tio n g* * (r^ (x ) ) ta k e s one va lue e x a c tly 3 tim es

on [ 0 ,1 ,2 ,v j , c o n tra d ic t in g ( i i ) . Hence i^ / i ^ f o r some C,

b u t th e n th e fu n c tio n g ^ ( r i ( x ) ) ta k es one value e x a c tly 3 tim es

on [ 0 ,1 ,2 ,v ] , ag a in c o n tra d ic t in g ( i i ) .

Next we s h a l l show th a t fo r a l l i i , . . . , i k th e two non-empty

s e ts ^ h ^ . . . i k i th e same power. Suppose i f

p o ss ib le th a t f o r example

= À > fi = > 0

(where c i s an a b b re v ia tio n f o r th e sequence i i . . . i k ) #

Let 0*', 1” be a r b i t r a r y elem ents of r e s p e c t iv e ly , and

denote th e elem ents of by Vq ,V i, . . .,v;\^^ where v° = 0 " .

For each i = 1 , . . . ,A ,—1 choose a p e m u ta t io n cj, g en era ted by F

such t h a t

61

c i ( 0 ) = 0", o i ( l ) = 1", c i ( 2 ) = VI.

In view o f th e in d u c tio n h y p o th esis ( b ) , we have

c i(v ) € f o r i = - 1 . Now f o r th e 1-p la c e

fu n c tio n ^(c j,(x ))w e have

g k + i ( o i ( 0 ) ) = g k + i ( c i ( 2 ) ) = 0; g k + i ( c i ( l ) ) = 1,

and we deduce from ( i i ) th at g * * ^ (c i(v )) = 1 , that i s , C[(v) €

for i = - 1 . On the other hand, c i ( l ) = 1” for a l l i ,

and s in ce X > ^ i t fo llow s th a t fo r some i ^ j we have

c t (v ) = c j (v ) = v say, iidiere v c

Choose a pe rm u ta tio n c , g en era ted by F , such t h a t

c (v ^ ) = 0” , o ( v j ) = V ', c(v®) = V®.

Consider th e two 1—p lace fu n c tio n s g e n e ra te d by F

a f * i ( o ( c i ( x ) ) ) ; g k + i ( c ( o j ( x ) ) ) .

They co incide in va lue a t th e p o in ts x = 0 ,1 ,v , b u t d i f f e r

f o r X = 2; co n seq u en tly , one of them ta k e s some value e x a c tly

3 tim es on [ 0 ,1 ,2 ,v j , co n trad ic ting ( i i ) .

F in a l ly we show th a t f o r a l l i i , . . . , i k , whenever r i s

a p erm u ta tion g en era ted by F such th a t r ( 0 ) , r ( l ) ,r-(2) e ' * *^k+1 ^

we a lso have r ( v ) € * * *^k+1 (This f i n a l c o n d itio n on ly a p p lie s

when I • • ‘^k+i I ^ ^ Suppose i f p o s s ib le th a t f o r example

r ( 0 ) , r ( l ) , r ( 2 ) e r ( v ) /

(where c" i s an a b b re v ia tio n f o r th e sequence i i # . . i k ) #

6 2

Since r ( v ) e by th e in d u c tio n hy p o th esis (b ) , we have

r ( v ) € ' ^ ^ 9 and th e 1-p la c e fu n c tio n g ^ + i( r (x ) ) ta k es one

va lue e x a c tly 3 tim es on } 0 ,1 ,2 ,v ] , con trad icting ( i i ) .

The proof i s com plete.

5 .4 . In tro d u c tio n of v e c to r n o ta tio n

In t h i s s e c tio n we assume th e conclusion of Lenma 5*8;

in p a r t i c u la r n+1 = 2^, and P g en e ra te s 1-p la c e fu n c tio n s

ta k in g the v a lu es Q,1 such th a t th e correspondence

X -* [ g ^ ( x ) , . • . ,gJ (x) j between elem ents of N and ro w -vec to rs

o f O 's and 1 's i s o n e -to -o n e . ( i t should always be c le a r

from th e co n tex t w hether [ a , b , c , . . . j deno tes a row v e c to r

o r th e s e t whose elem ents a re a , b , c , . . . . ) We s h a l l deno te by

X = [x^j =

th e row -vector corresponding to x , so t h a t x^ = g ^ (x ) .

We a ls o in tro d u ce a correspond ing n o ta t io n f o r fu n c tio n s

h ( x i , . . , ,Xk) € En ; thus h ( x i , . . . ,x k ) i s th e row -vecto r

[ h ( x i , . «. jXk) I 3 [ h ( x i , . . . ,Xk] , . . . ,h ^ (x ^ , • . • , x ^ ) ] , where

h ^ (x i , . . .,X k) = g” ( h ( x i , . . . ,X k ) ) . V/e can a ls o reg a rd h a s

a fu n c tio n h ( x i , . . . ,X k ) of th e corresponding k ro w -v ec to rs ;p

then each component h i s a fu n c tio n of th e k j v a r ia b le s

xj^(l ^ /J ^ k , 1 ^ A ^ j ) , where x^ = [3V , . T h e value0

o f each of th e se v a r ia b le s , and of h , i s 0 o r 1, and so h e Ei*

63

I t fo llow s by Theorem 2.1 o f C hapter 2 t h a t we may now w rite

h^ u n iq u e ly a s a polynom ial over th e f i e l d Zg s [ 0 , l j o f re s id u e s .

modulo 2 , th a t i s a s a sum (modulo 2 ) of d i s t i n c t term s each of

th e form

P i* '" jUr

(w ith d i s t i n c t f a c t o r s ) . Such a term i s n a tu r a l ly c a l le d l in e a r

i f r = 1 , a polynom ial i s l in e a r i f each of i t s te r n s i s l i n e a r ,

and h ( x i , . . . ,X k ) e Ep i s c a l le d l in e a r ^ i f each o f th e components

h ^ (x i , . . . ,Xk) i s a l in e a r po lynom ial. In th i s case we can w r ite

+ ho

In p a r t i c u la r i f h (x ) e En i s a 1 -p la c e l in e a r fu n c tio n then

h (x ) = ^ + ^ ,

where H 3 [h^] i s a j x j m a trix over Zg . Let us deno te

by 0 th e v e c to r [ 0 , . . . , 0 j , and b y e ^ ( l ^ ^ ^ j ) th e v e c to r

d i f f e r in g from 0 in having 1 in th e & -th p la c e ; by

0 , e i , . . . , e j we den o te th e co rrespond ing elem ents of N.

No co n fu sio n w i l l a r i s e from th i s n o ta t io n , s in ce in Lemma 5 .8

we have chosen our correspondence x -» [ g ^ (x ) , . . .,g^ ( x ) } such

th a t each g^(x) (6 s 1 , . . . , j ) s a t i s f i e s g^(o ) = 0 . Hence

th e zero elem ent of N s a t i s f i e s 0 [ 0 , . . . , 0 ] 3 0 .

t We s h a l l no t be u s in g th e word * lin e a r* in the sense o f a fu n c tio n o f th e form Zaj,xi + b modulo n , in Tdiich i t has been used f o r example in [Z9],[49].

64

I f th e 1 -p lace fu n c tio n h e En s a t i s f i e s h (o ) = 0 , then we have ^ = 0,

and in th i s case th e & -th row v ec to r of H i s sim ply th e v ec to r h (e ^ ) .

I f x ,y € N we denote by x 0 y th e elem ent z € N such t h a t z = x + jg,

th a t i s , g ^ (z ) =s g^(x ) + g^(y ) (modulo 2) f o r 6 = 1 , . . . , j . V/e s h a l l use

w ith o u t s p e c ia l m ention v a rio u s obvious p ro p e r t ie s o f t h i s a d d i t io n , f o r

example, th e com m utative, a s s o c ia t iv e and idempotency (x 0 x = O) law s.

F in a l ly we s h a l l deno te by Z th e s e t o f a l l quadruples [ft,/? y 8} of

d i s t i n c t elem ents of N f o r which a 0 /9 0 Y 0 S = 0 . Observe t h a t t h i s

i s e q u iv a le n t to 8 = a a.nd t o a + / 9 + ^ + 6 = 0 .

We now pass to th e proof o f v a rio u s r e s u l t s in whose fo rm u la tio n s

th e above n o tio n s p lay an e s s e n t i a l r o l e .

LEMT.IA. 5 •8 (a ) . Under th e hypotheses of Lemma 5*8 every 1 -p la c e

fu n c tio n s (x ) g en era ted by F s a t i s f i e s s(o) 0 s ( l ) 0 s (2 ) 0 s (v ) = 0 .

PROOF. In view of a s s e r t io n ( i i ) o f Lemma 5*8 each fu n c tio n

g ^ (s (x ) ) = 1 , . . . , j ) tak es each o f th e v a lu es 0,1 an even number

o f tim es on th e s e t [ 0 ,1 ,2 ,v ] , and so

g ^ (s (0 ) ) + g ^ ( s ( l ) ) + g^(s(2)) + g ^ (s (v ) ) = 0 (modulo 2) , 6 = 1 , . . . , j ,

which is e q u iv a le n t to th e c o n c lu s io n .

LEMIvIA 5 .8 (b ) . Under the hypo die ses o f Lan ma 5«8 every 1-p la c e

func t i on s (x ) g en era ted by F s a t i s f i e s s ( g ) f f l s(/9) s (y )(+ )8 (S ) = 0

f o r every quadruple [a ,/9 ,Y ,S l € Z .

PROOF. Choose a p erm u ta tion r , g e n e ra te d by F , such th a t r (o ) = a ,

r ( l ) = r ( 2 ) = Y* 7n view of Lemma 5 .8 (a ) we have r ( v ) = a© /90Y = 8 .

Applying Lemma 5 .8 (a ) to the fu n c tio n s ( r ( x ) ) we deduce th e re q u ire d

fo rm u la .

65

LEMMA 5*9• 1-p la c e fu n c tio n h e En i s l in e a r i f and only

i f h (g ) 0 h(j3) 0 h (y ) 0 h(S ) = 0 f o r every quadruple § j e Z .

PROOF. I f h i s l in e a r , say h(x) = xH + ho, then

h(%) + 6 % ) + & (l) + = (« + g + % + = 0 , a s re q u ire d .

•6I f h i s non—lin e a r , th e n some component h co n ta in s

a n o n - lin e a r te rm ; choose one co n ta in in g as few v a r ia b le s as

p o s s ib le . W ithout lo s s of g e n e r a l i ty we may suppose th a t i t

i s T(x ) = x^x®. . .x^ ^ 2 ) . D efine

a =

g =

Y = 01, B ■

Then [a.,/3,x,8] € Z , b u t we s h a l l show th a t

h'^(a) + h'^Cg) + h"^(^) + h'^Cg) s 1 (mod 2 ) .

In f a c t we v e r i f y a t once t h a t

T(a) + T(g) + T(y ) + T(8) 3 1 (mod 2 ) ;

b u t f o r every o th e r term S(x) o f h^ we have

S (a) + S(g) + S(;^) + S(S) = 0 (mod 2 ) .

Namely, i f S(x) i s l in e a r th i s fo llo w s from th e f i r s t p a r t of

the p ro o f , w h ile i f S(x) i s n o n - lin e a r then (by th e d e f in i t io n o f T)

66

i t co n ta in s a f a c to r x'" w ith m > 6 and th e re fo re S(x) s 0

(mod 2) f o r x = a,g,Y ,

LEMMA 5*10* The group G- of a l l 1-p la c e l in e a r fu n c tio n s

o f En which a r e a lso p e m u ta tio n s of N i s t r i p l y t r a n s i t i v e .

PROOF. A 1-p la c e l in e a r fu n c tio n h e , w ith

h (x ) = ^ , i s a perm uta tion i f and only i f th e mapping

X h (x ) i s o n e -to -o n e , and th e n ecessa ry and s u f f i c i e n t c o n d itio n

f o r t h i s i s t h a t th e m atrix H he n o n -s in g u la r . In o rd e r to

prove th e t r i p l e t r a n s i t i v i t y i t w i l l th e re fo re be s u f f ic ie n t to

show th a t g iv en any th re e d i s t i n c t numbers ao,cti, Gg € N we can

f in d ho € N and a n o n -s in g u la r m a trix H over Zg such t h a t

OH + ho = Go ; e^H + ho = gi ( i = 1 ,2 ) . fw# /V#These eq u a tio n s hold i f and on ly i f ho = Go and th e f i r s t two

rows o f H a re «i - Go and ag - Go # The l a t t e r two v e c to rs

a re l in e a r ly in d ependen t, because Go,Gi , Gg a re d i s t i n c t , and

th e re fo re we can s e le c t th e rem aining j - 2 rows so a s to c o n s tru c t

a n o n -s in g u la r m a trix H, a s re q u ire d .

COROLIARY 1 . The o rd e r o f G- i s 24^5^ (2-J - 2 I) . ---

PROOF. There a re 2^ p o ss ib le choices f o r ho, and

n ( 2J - 2 ^ ) cho ices fo r th e non ^sin g u la r m a trix H.1=0

REMARK. I have no t been ab le to dec id e f o r j > 2 w hether

G- can co n ta in any t r i p l y t r a n s i t i v e p ro p e r subgroups. I t fo llo w s

67

by a theorem of B urnside ( [ 1 ] , p . 177) t h a t th e o rd er of such

a p ro p e r subgroup would be d iv i s ib le by 2- (2^-1 )(2 ^ -2 ) • This

r e s u l t dem onstrates th e im p o s s ib i l i ty o f t r i p l y t r a n s i t i v e p ro p er

subgroups of G- f o r j = 2 , b u t g e n e ra l ly , s in c e th e o rd e r o f G-

i s g r e a te r th an 2- (2^-1 ) (2 ^ -2 ) , th e problem i s s t i l l open.

COROLIARY 2 ; The s e t L of a l l l i n e a r fu n c tio n s of Ep

s a t i s f i e s th e hypotheses of Lemma 5*8*

PROOF. L in c lu d e s a t r i p l y t r a n s i t i v e group o f

perm uta tions of N (by Lemma 5*10); a fu n c tio n s a t i s f y in g

th e S-êupecki c o n d itio n s , f o r example G-(x,y) = x 0 y ; and

a 1-p la c e fu n c tio n ta k in g e x a c tly 2 v a lu e s , f o r example th e

fu n c tio n s g ^ (x ) a re e a s i ly seen t o be l in e a r . Hov/ever, i t i s

easy to v e r i f y t h a t L i s c lo se d under com position , and th a t

no 1-p la c e l in e a r fu n c tio n i s o f ty p e [ n , l ] . I t th e re fo re

fo llo w s from Lemma 5*4 th a t L does n o t in c lu d e any fu n c t io n

h (x ,y ) s a t i s f y in g c o n d itio n s (A ), and t h i s com pletes th e

v e r i f i c a t io n o f th e hypotheses o f Lemma 5*8.

LEMMA 5*11" F C En generates a tr ip ly tr a n s it iv e

group of permutations of N, togeth er w ith a l l constants and

a non-linear fu nction h ( x i , . . . ,Xk) , then F generates

a 1-p la ce non -linear fu n c tio n .

PROOF. We may assume th a t k ^ 2 and t h a t a l l perm uta tions

gen era ted by F a re l i n e a r , because o therw ise th e re i s no th ing

68

to p rove . F or some C, h ( x i , . . . , x k ) must be a n o n - lin e a r

polynom ial, and so contain a non -lin ear term S = . . . .12

We co n sid e r two c a s e s .p

CASE 1. In h th e re i s a t l e a s t one term S in ^b ich

some two o f th e ju 's a re e q u a l. Let S re p re s e n t such a term

w ith th e l e a s t number o f f a c t o r s , and suppose f o r example th a t

=jU2 = = m = 1, w h ile jLis ^ 2 f o r s ^ i + 1. D efine

co n stan ts o (k = 2 , . . . , k ) a s fo llo w s : = 1 i f 8 co n ta in sKA A A * A *the f a c to r x ; o therw ise o = 0 . Let S* = x x . . .

be any term of h d i f f e r e n t from S b u t w ith p re c is e ly th e same

f a c to r s o f th e form x ^ : thus we may suppose t h a t = 1 and

A’s = As fo r s = 1 , . . . , i , w hile nl ^ 2 f o r s ^ i + 1 . Then

i t i s easy to see (by th e d e f in i t io n o f S) th a t i f th e co n stan ts

C3 , . . . ,C|< a re s u b s t i tu te d f o r X a , . . . , X k th e term S* v an ish es .

I t fo llo w s th e re fo re th a t th e r e s u l t in g fu n c tio n h ^ ( x ,^ , . . . ,Ck )

w i l l co n ta in th e n o n - lin e a r term x ^ i . . . x ^ ^ ( to g e th e r p o s s ib ly

with other d is t in c t non -lin ear term s), and hence h ( x , 0 2 , . . . , C k )

i s a 1-p la c e n o n - lin e a r fu n c tio n g en era ted by P.0

CASE 2 , For every term S in h , a l l the /Li’s a re

unequal. I f k ^ 3 th e n by s u b s t i tu t in g s u i ta b le co n stan ts

69

we can red u ce th i s case to one in which k=2. (This i s e a s i ly

proved by a method s im ila r to t h a t of Case 1 .) Thus we o b ta in

a n o n - lin e a r polynom ial o f degree 2 in of the form

h ^ ( x i , X g ) = XiAxs* + L ( x i , x s ) ,

■vhere L i s l i n e a r , 35® (denotes th e tra n sp o se (co lum n-vector)

o f , and A i s a non-zero m atrix over Zg.

Then = 1 fo r some Y ,8 ; and w ith o u t lo s s o f g e n e r a l i ty

we may suppose th a t th e lead in g elem ent a ^ = 1 , because i f

r , s a re ( l in e a r ) p e rm u ta tions g en e ra ted by P such th a t

r ( o ) = 0, r ( e ^ ) = e j s (o ) = 0, s(eg ) = e^

th en th e le ad in g elem ent in th e corresponding m a trix f o r th e fu n c tio n

h ( r ( x i ) , s (x g )) i s 1.

I f A is n o t symmetric then i t i s easy to v e r i f y t h a t the

1-p la c e fu n c tio n h (x ,x ) g en era ted by P i s n o n - lin e a r : i f

a . / a«,^ then h ^ (x ,x ) con ta in s the term x*^x^. Consequentlyyo o# /V isawe may suppose th a t A i s sym m etric, and in p a r t i c u la r a^g = &2 1 *

Choose a pe im u ta tio n t g en era ted by P such th a t

t(0) = 0, t(ei) = e% , t(j ) = i ^ 1 2 » j •

Then i t i s easy to v e r i f y t h a t f o r th e fu n c tio n h ( t ( x i ) ,x g ) th e

corresponding m a trix , say M, has m g = = a^g + 1*

C onsequently M is non-sym m etric, and thus h ( t ( x ) ,x ) i s n o n - l in e a r .

70

COROLLARY* Under th e hypotheses o f Lemma 3*8, F g en era tes only

l i n e a r fU n o tio n s*

PROOF. Immediate from Lemmaa 3 .8 (h ) , 5*9 3*11#

REMARK. I have n o t been ab le to decide w hether under th e hypotheses

o f Lemma 3*8 F must n e c e s s a r i ly g en e ra te a l l th e l i n e a r fu n c tio n s , t h a t

id , w hether any p roper su b se t o f L can a lso s a t i s f y th e hypotheses o f

Lemma 3*8.

LEMMA 3*12. n = 2^ and f i s a l i n e a r fu n c tio n o f % ,

th en f canno t g en e ra te a doubly t r a n s i t iv e group o f perm uta tions o f N.

PROOF. We s h a l l prove th a t f i s s e lf -c o n ju g a te under e itjier some n o n -id en tica l permutation S o f N o r some constant 1-p la ce fu n c tio n . I t i s easy to see t h a t ev ery fu n c tio n g en era ted by f th e n

has the same property, and the conclusion fo llow s by Lemma 3*7*

Since f i s l i n e a r , we can w r i te

J k= ! 2 2 X ^ i + b

A=1 ^=1

where each c o e f f ic ie n t a^ i s 0 o r 1 . We denote by B th e j x jK Im atrix w ith elem ents B .. = 2 a. , and we co n sid e r two c a se s .

^ /i=1CASE 1. The m a trix B -I i s s in g u la r . Then ih e re i s ^ ^ such

t h a t £B = o, and f i s s e lf -c o n ju g a te under th e p e rm u ta tio n de fin ed by

^ (x ) = X + jc because

s (x k )) = ! 2 2 ( / + o'^)j + b = + ^ BA fI ff=1 ^ ^

- £ (^1 ^ ~ ^ (£ (3 )* * * * ^ ))*

71

CASE 2 . The matrix B-I i s non -singu lar. Then "there i s d

such th a t dB - ,d =jb, and f i s se lf-co n ju g a te under th e constant

1-p la ce fu n ction e^ual to d because

f ( d , . . . , d) = [ ^ 2 a% d^] + b = dB + b = d.~ ~ ^ A»i ju-i ^ ~ ^ ^ ~

LEMMA 5*13* I f r i s a n o n - lin e a r p erm u ta tion of N. th en

th e re e x is t s a l i n e a r p erm uta tioh h such t h a t th e fu n c tio n r h r “ i s

n o n - l in e a r .

PROOF. In view o f Lemma 5*9, fo r some quadruple ^

we have

r ‘ (a ) 0 r (ft) & r (y ) © r " " (S) / 0 . ( l )

Choose a lin e a r permutation s such "that

8 ( f ( a ) ) = 0 , s (r“ ^ ) ) = ei , s(r"^ (y ) ) = % .

Then

s(r"^ (5)) / e i g)eb (2)

because otherw ise, observing th a t 0 , ei , % , ei @ % are d is t in c t

and

0 0 ei 0 3 0 (ei % ) = 0 ,

we could deduce, s in ce s“ i s l in e a r , th a t ( l ) i s f a l s e .

Choose a lin e a r permutation h* such "that

h' (0) = 0 , h*(e i ) = e i , h ’ (% ) = %, h ' ( s ( r » i ( a ) ) / s ( r “ ( a ) ) .

We can alw ays make th i s choice because i f s ( r “ ^ ( s ) ) , say , th e n

i t w i l l be s u f f i c i e n t to a rrange t h a t h* (eg ) = s(r“^ ( s ) ) .

This can be done: we need h*(x) = xH + b , and th e requ irem en ts a re

72

met by ta k in g b = 0 , and th e f i r s t 3 rows o f H to be

ei , ^ , ^(r*“ (S )) r e s p e c t iv e ly . These a re independent by (2 ) ,

and we can choose j —3 f u r th e r independent rows to make H

n o n -s in g u la r . I f s (r* '^ (5 )) = % th en we can s im i la r ly arrange th a t

h * (ea ) = % 0 e t j , and so on.

D efine th e l i n e a r p e rm u ta tio n h as s“ ^h*s and c o n s id e r the

1-p la c e fu n c tio n k = r h r“ . We f in d th a t k (a) = a ,

k(j9 ) = )5 , kCy) = y , b u t k (ô) / 8 . Hence k (a ) k(/3 ) (^) kCy) ® k(8 ) / 0

and s in ce f Z i t fo llow s t h a t k i s nonr-1 in e a r .

REMARK. In t h i s s e c tio n our r e s u l t s have been concerned w ith

th e c la s s L o f fu n c tio n s o f En t h a t a re l i n e a r r e l a t iv e to a

p a r t i c u l a r mapping o f N onto , where nW = 2 (namely the mapping2

X -* { ^ ( x ) , . . . , ^ ( x ) i ) . There a re obv iously 2 ! such mappings, and

each determ ines a corresponding c la s s L o f l i n e a r fu n c tio n s . These

c la s s e s a re con jugate to one an o th er under th e perm uta tions of N, and

i t i s o f i n t e r e s t to determ ine how many o f them a re d i s t i n c t . By Lemma

5.13 a l l th e c la s s e s con jugate to a g iven c la s s L under a g iven p e rm u ta tion

which i s n o t l i n e a r in th e sense o f L, w i l l be d i s t i n c t from L, whereas

s in ce ' l i n e a r i t y ' i s conserved under com position , a c la s s con jugate to

L under a l i n e a r ( in th e sense o f L) perm u ta tio n w i l l be id e n t ic a lj - i

w ith L i t s e l f . By Lemma 5*10 C o ro lla ry 1 th e re a re 2' II (2^-2^)J - i

such c la s s e s , and th e re fo re th e re a re 2 1/2^ II (2^ - 2^) d i s t i n c t

c la s s e s L o f l i n e a r fu n c tio n s .

(T his e v a lu a tio n was made in c o lla b o ra t io n w ith Roy 0 . Davies,..)

73

5*5 Main conclusions

THEOREM 5*1. I f n ^ 5 and P generates a t r ip ly

tr a n s it iv e group o f permutations' on N, to g eth er w ith a fu nction

s a t is fy in g the S6upecki co n d itio n s , then P i s com plete, u n less

n+1 = 2 and F generates only fu n ction s th a t are lin e a r

r e la t iv e to some p a rticu la r mapping o f N onto

PROOF. Ey Lemmas 5*1 and 5*2, F generates a 1-p la ce

fu n ction tak ing e x a c t ly 2 v a lu e s . I f F a lso generates

a 2-p lace fu n ctio n h (x ,y ) such th a t

h(0 ,0) = 0; h ( 0 , l ) = h ( l , 0 ) = h ( l , l ) = 1,

then by Lemma 5*4, Lemma 5*5, Lemma 5*6, C orollary, and Theorem

4*4 (chapter 4 , p. 3 9 ) F i s complete. I f F does not

generate such a fu n ction then i t s a t i s f i e s th e hypotheses of

Lemma 5*8, and th erefore by Lemma 5*11 C o ro lia iy , F generates

only fu n ctio n s th a t are lin e a r r e la t iv e to a cer ta in mapping

o f N onto z j .

REMARK. The example g iven in the Remark fo llo w in g

Lemma 5*tshows th a t th e con d ition n ^ 5 in Theoron 5*1 cannot

be relaxed .

THEOREM 5 . 2 . I f n ^ 4 and F generates a quadruply

tr a n s it iv e group o f -permutations on N, together w ith a fu n ctio n

s a tis fy in g th e S6upecki con d itio n s, then F i s complete.

7 4

PROOF* The r e s u l t fo llo w s im m ediately from Theorem 5*1,

u n le ss n+1 = 2 and F g en era te s only l i n e a r fu n c tio n s

( r e la t iv e to some mapping o f N onto 2 ^ ) ; h u t f o r n+1 = 2J ,

j > 2 , th e group of l in e a r perm u ta tions of En i s no t quadruply

t r a n s i t i v e . This i s c le a r from Lemma 5$9, which shows th a t

i f a ,/? ,Y a re d i s t i n c t e lem ents of N th en th e v a lu e o f any

1 -p la c e l in e a r fu n c tio n h e a t th e p o in t a Q /9 0 y i s

b (a ) (5> h(/?) Q h ( y ) , and i s thus determ ined by th e v a lu es

h ( a ) , h(/9), h (y ) .

REMARK. The r e s u l t f a i l s f o r n=3, because in t h i s case

th e group of l in e a r p e rm u ta tions i s th e whole symmetric group

S4 and th u s quadruply t r a n s i t i v e , so t h a t the c la s s F of

l i n e a r fu n c tio n s s a t i s f i e s the hypotheses a lthough obviously

n o t complete*

THEOREM 5*3* I f n ^ 2 and f i s a s in g le fu n c tio n o f En

which g e n e ra te s a t r i p l y t r a n s i t i v e group o f perm uta tions of N,

then f i s com plete.

PROOF. I f n ^ 3 , th e n by Theorem 4*3 C o ro lla ry 2 (C hapter 4 ,

p* 3 8 ) f must s a t i s f y th e S'&upecki c o n d itio n s , and th e r e s u l t

fo llo w s from Theorem 5*1 u n le ss n+1 =s 2 J and f g e n e ra te s only

l in e a r fu n c tio n s ( r e l a t iv e to some mapping of N onto Zg);

b u t th e n f i t s e l f i s l i n e a r , and by Lemma I 5 cannot g e n e ra te

even a doubly t r a n s i t i v e group of perm utations*

75

I t rem ains to prove th e Theorem f o r n=2. In t h i s case

th e only t r i p l y t r a n s i t i v e group o f p e rm u ta tio n s on N i s

th e symmetric group Sg, and i t i s w e ll known th a t each member

r of 8 3 can be w r i t t e n in th e form

r ( x ) = ax + b modulo 3,

fo r some c o n s ta n ts a ,b .

Since a l l th e p re-com ple te su b se ts o f Eg a re known

(th e y a r e l i s t e d in Theorem 2 .4 o f C hapter 2 ) , i t i s easy to

v e r i fy t h a t Sg i s e n t i r e ly co n ta in ed in only two of them,

nam ely, th e c o n se rv a tio n c la s s e s of th e two p re d ic a te s

( i ) X + y = 2 + u , ( i i ) x « y v x = z v y = z . I f we denote

th e se by L* and L” r e s p e c t iv e ly , th en i t w i l l be s u f f ic ie n t to

prove t h a t no s in g le fu n c tio n f belonging to e i th e r L' o r L"

g e n e ra te s 83.

I t i s easy to v e r ify th a t L* i s the s e t of a l l lin ea r

(modulo 3 ) fu n ction s o f Eg, and so i f f e L* then

f ( x i , . . , , X k ) = ao + a^Xi + . . . + a^xk (modulo 3)

f o r some c o n s ta n ts a g , a ^ , . . . ,a p .

According a s + ag + . . . + a% ^ 1 or + ag + . . . + a% e 1

(modulo 3 ) f ( x i , . . . , X k ) i s s e l f con jugate under some c o n s ta n t

fu n c tio n a € [ 0 , 1 , 2 j o r th e c i r c u la r p e rm u ta tio n x+1,

r e s p e c t iv e ly , and so i t fo llo w s by Lemma 5*7 th a t f cannot

g en e ra te 83 in t h i s c a se .

76

I f f , . . . ,%k ) L", then f o r ev ery th r e e p o in ts

a = ( % i , . . . , a k ) , /3 = ( / ? i , • • • ,/?k) , Y = (yi , • • • ,Yk) , such t h a t

«1 = V a i = Y l V = Y l , f o r i = we have

f ( a ) = f (/?) V f(a ') = f ( y ) v f(/9) = f ( y ) . I t fo llo w s th e re fo re

t h a t th e re cannot e x i s t s e t s G% , . . . ,Gk each c o n ta in irg a t most

two elem ents of N such th a t th e s e t f (G% , # . . ,Gk ) = N, and so

by Theorem 4*2 (C hap ter 4 , p*35 ) f does n o t s a t i s f y th e

S-ftupecki c o n d itio n s . Hence f has a t most one e s s e n t i a l

v a r ia b le , or has a t l e a s t one m issing v a lu e , and in e ith e ÿ c a se

i t i s e a s i ly seen th a t f cannot g e n e ra te S 3.

This completes the proof cf Theorem 3*3*

77

REMARK. In [ “S, 8 ] Salomaa has proved the fo llow ing

two r e s u l t s ;

( i ) " I f n ^ 4 and F i s a su b se t of En which g e n e ra te s

th e a l t e r n a t in g group An of p e rm u ta tio n s of N, to g e th e r

w ith a fu n c tio n s a t i s fy in g th e S-èupecki c o n d itio n s , th en

F i s com plete".

( i i ) " I f n ^ 3 and f i s a s in g le fu n c tio n o f F which

g e n e ra te s th e a l te r n a t in g group An of perm uta tions of N,

to g e th e r w ith a fu n c tio n s a t i s f y in g th e S-^upecki c o n d itio n s ,

then F i s com plete".

S ince the a l t e r n a t in g group An of p e rm u ta tio n s of N

i s (n-1 ) -p ly t r a n s i t i v e on N i t fo llo w s th a t f o r n ^ 5

Theorems 3*1, 3*3 a re improvements on th e se two r e s u l t s .

A ccording to M arshall H a ll ( [ 1 0 ] , p .6 8 ), a p a r t from th e

a l te r n a t in g and sym m etric groups o f p e rm u ta tions of N th e re

a re only fo u r groups known which a re quadruply t r a n s i t i v e .

These a r e th e M athieu groups on 11 ,12 ,23 and 24 l e t t e r s , and

i t i s no t known w hether t h ^ a re e x c e p tio n a l o r p a r t o f an

i n f i n i t e sequence o f groups w hich a re quadruply t r a n s i t i v e . ^

However, th e re e x i s t i n f i n i t e l y many t r i p l y t r a n s i t i v e groups

o f p e rm u ta tio n s ; Jo rdan [fj} showed th a t i f p i s prim e, th en

1* In [ 1 2 ] I to shows th a t i f p and q are both prime numbers,w ith p ^ 2q+1 and p > 11, then a n o n -so lv ab le t r a n s i t i v e p e rm u ta tion group on [ 0 , 1 , . . . , p - l j i s quadruply t r a n s i t i v e . However, in view o f Salom aa's improvements on Theorems 3*1,5*3 fo r the s p e c ia l case n+1=p (a prim e num ber), t h i s r e s u l t i s n o t re le v a n t t o th e p re se n t problbm."..

78

th e re e x is t s a t r i p l y t r a n s i t i v e group o f perm uta tions on

p^+1 l e t t e r s of o rd e r (p^+1 )p^(p^-1 ) , and th e groups G- of

Lemma 1 2 form a n o th e r i n f i n i t e sequence o f t r i p l y t r a n s i t iv e

g roups.

For th e s p e c ia l case n+1 = p (a prime number) i t i s

p o ss ib le to o b ta in b e t t e r com pleteness c r i t e r i a th a n Theorems

5 .1 , 3 . 3 . In [ 2. 9 ] Salomaa has proved th e fo llo w in g r e s u l t :

I f n+1 = p (a prime number ^ 3) and f is a s in g le

fu n c tio n which g en e ra te s (b u t i s no t s e lf -c o n ju g a te under)

a c i r c u la r p e rm u ta tion c (x ) on N, th e n f i s com plete.

THEOREM 3 . 4 . I f a c lo sed su b se t F ^ E r in c lu d e s

a t r i p l y t r a n s i t i v e group of perm uta tions of N and a ls o

a fu n c tio n s a t i s f y in g th e S6upecki c o n d it io n s , then F is

p re-com plete i f and only i f n+1 = 2* and F is th e s e t o f

a l l fu n c tio n s th a t a re l in e a r r e l a t i v e to some mapping N

onto Zg .

(For a d e f in i t io n of a p re-com plete su b se t o f Ep see

C hapter 2 g 3*-)

PROOF. I f F i s p re-com plete and s a t i s f i e s the o th er

co n d itio n s th e n , by Theorem 3 ,1 , n+1 = 2* and F in c lu d es

only fu n c tio n s l in e a r r e l a t iv e to some such mapping (o th erw ise

F would be com p le te ); and F must in c lu d e a l l th e l in e a r

79

fu n c tio n s because o therw ise a l in e a r fu n c tio n n o t in F

could be a d jo in e d to F w ith o u t making i t com plete (s in c e

s t i l l only l in e a r fu n c tio n s would be g e n e ra te d ) . The r e s u l t

in th e o th er d i r e c t io n fo llow s a t once from Theorem 3.1.

REMARK. F or each in te g e r n ^ 3) Y ab lonsk ii has

c o n s tru c te d in 1 ^ ,] an example o f a p re-com plete su b se t of

En which in c lu d es th e s e t of a l l 1-p la c e fu n c tio n s (and

consequen tly a t r i p l y t r a n s i t i v e group o f p e im u ta tio n s on n ) ,

but t h i s p a r t i c u la r su b se t c o n ta in s no fu n c tio n s a t i s f y in g th e

S-^upecki c o n d itio n s . He a ls o c o n s tru c ts s e v e ra l c la s s e s o f

p re-com ple te s u b s e ts , b u t th e l in e a r s e t s o f th is paper a r e n o t

among them . T herefore by th e R.emark fo llo w in g Lemma 1 3 , f o rJ - 1

n = 2* we have an a d d i t io n a l 2'^l/2'^ II (2*^—2*’) examples o fl»opre-com plete s u b s e ts .

We should l ik e to m ention th a t many of our d e f in i t io n s

and r e s u l t s concern ing l i n e a r i i y can probably be g e n e ra l is e d

to th e case when n i s th e j^ ^ power o f any prime p , bu t

th e re would not be any d i r e c t a p p l ic a t io n to th e main s u b je c t

o f th is t h e s i s .

REIAARK. In C hapter 3 (p*31 ) m entioned t h a t each of

th e p re-com plete su b se ts g iv en in Example 3*2 was g en e ra ted by

a s i n ^ e in c luded fu n c tio n . The fo llo w in g r e s u l t shows th a t

th i s is not alw ays th e case :

80

THEOREM 5*5« I f n+1 = 2^, and P i s th e p re-com plete

su b se t o f En c o n s is tin g o f a l l fu n c tio n s th a t a re l in e a r

r e l a t i v e to some mapping o f N onto Zg, then P does n o t

co n ta in a s in g le fu n c tio n (f> such th a t <p g en e ra te s F •

PROOF. I f j ^ 2 then by Lemma 3 .1 0 , F c o n ta in s

a t r i p l y t r a n s i t i v e group of perm uta tions o f N and th e

r e s u l t fo llow s im m ediately from Lemma 3 .1 2 . I f j= 1 , th en

F i s the s e t o f a l l l i n e a r fu n c tio n s of E^ and consequently

in c lu d e s the s e t of a l l 1-p la c e fu n c tio n s of E]_, namely

[0 ,1 ,x ,x + lJ . ^ argum ents s im ila r to th o se cf Theoren 2 .2

(c h a p te r 2 ,p . 1 2 ) a n ecessa ry c o n d itio n f o r a s in g le fu n c tio n

4> to g en era te th e s e t of a l l 1-p la c e fu n c tio n s of E % i s

th a t # ( x , . . . , x ) = x+1. I f in a d d itio n (p i s l i n e a r , th en

(p w i l l be s e lf -c o n ju g a te under the c i r c u la r perm uta tion x+1,

and in p a r t i c u la r <p cannot g e n e ra te th e c o n s ta h t fu n c tio n s

0 ,1 . T herefore no s in g le l in e a r fu n c tio n g e n e ra te s th e s e t

of a l l 1-p la c e fu n c tio n s of E a n d th e r e s u l t fo llo w s .

REMARK. In [A 9 ] Y ab lonsk ii has advanced th e fo llo w in g

c o n je c tu re :

each c lo sed s e t F Ç Eq has a f i n i t e b a s is (vhere a b a s is o f F

i s any s e t F ' Ç F such th a t the c lo su re of F* i s F ) .

In P o s t proves t h i s co n jec tu re f o r th e ^ e c i a l

case n=1, b u t acco rd ing to Y ab lonsk ii th e re i s no known proof

f o r the c a se n 2 . We f u r th e r d e f in e a m inim al b a s is B

81

o f a c lo sed su b se t P to be a " sm a lle s t b a s is " ( th a t i s ,

th e re i s no o th e r b a s is B* o f F such t h a t | B '| < | b | ) and

say th a t a c lo se d su b se t F i s o f ty p e ( i ) i f i i s th e o rd er

of any minimal b a s is cf F . The p re-com plete su b se ts g iv en

in Example 3*2 a re th e re fo re o f ty p e (1 ) , whereas th o se of

Theorem 5 «5 a re of type g r e a te r th a n 1* The ta s k of

d ec id in g w hether o r not a g iv en p re-com plete su b se t cf En

i s of type (1 ) ought to be f a i r l y s t r a i^ t f o r w a r d * The

problem a lso a r i s e s o f d e te rm in in g f a r a g iven in te g e r n th e

number of p re-com plete su b se ts o f type (1 ) . For exam ple, i f

n=1, th en th e re a re e x a c tly 3 p re-com ple te su b se ts of ty p e (1)^

These no tions g iv e r i s e t o th e fo llo w in g r e s u l t .

THEOREM 3*6. I f a su b se t F (c En) g e n e ra te s a p re-com plete

su b se t of En o f type ( i ) , and i f |p | < i , then F i s com plete .

PR-GOFr, F g e n e ra te s th e p re-com plete su b se t b u t i s n o t

a b a s i s . Therefore F in c lu d es a fu n c tio n n o t in th e p re-com plete

su b se t, and so F i s com plete.

8 2

V I . - A MAPPING- OF En INTO E / - V I

6 .1 . I n t r o d u c t i o n

This f i n a l c h a p te r p r e s e n t s an u n u s u a l ap p ro ach to th e

problem of g e n e r a t i n g th e com ple te s e t E p . We b e g in i n

§ 6 . 2 by d e f i n i n g n - t u p l e s of f u n c t i o n s f ^ , f s , . . . , fn £ E *

w ith s p e c i a l p r o p e r t i e s . I n § 6 .3 t h e f u n c t i o n s t& be

of are^mapped onto the se s p e c i a l t y p e s of n - t u p l e s ,tbit-

and t h i s mapping i s u sed i n § 6 .4 t o p ro v e v a r io u s p a r t i c u l a r

s u b s e t s o f Ep a r e comp].ete.

We have found i t n e c e s s a r y to a l t e r t h e n o t a t i o n u sed in

p re v io u s c h a p te r s , d e n o t i r g by x ,y (w i th i n d i c e s ) th e v a r i a b l e s

ra n g in g o v e r [ 0,1 ] ;<jf),f , g , h ( w i th i n d i c e s ) f u n c t i o n s of E]_ ;X,Y

( w i th i n d i c e s ) v a r i a b l e s ra n g in g o v e r n ( = [ 0 , 1 , . . . , n | ) Jand

e t c . ( wi t h i n d i c e s ) f u n c t io n s of Ep . V7e s h a l l a l s o

den o te a n a r b i t r a r y s u b s e t of Ep by S .

8 3

6 .2 . Perm anent s e t ' s

DEFINITION. F o r e a c h i n t e g e r n ^ 1 an n - p la c e f u n c t i o n

f ( x i , . . . , X p ) e El i s c a l l e d ( l , n ) perm anen t i f i t s a t i s f i e s

th e c o n d i t i o n

I ( Xi , . . • , Xj , . . . , Xp ) = f ( Xi , . . , Xj , • . . , Xp ) , . . . (-!-)

where f o r e a c h in d e x j (1 ^ j ^ n) Xj d e n o te s t h e p ro d u c t

Xi Xg . . . Xj •

THEOREM 6 . 1 . f ( x i , . . . , Xp) i s ( l , n ) perm anent i f and o n ly

i f t h e r e e x i s t s a f u n c t i o n <p € such t h a t

l ( x i , , , . , Xj , • . . , Xp ) = < ( X i , . . • , Xj , • . • , Xp ) . . . . ( i i )

PROOF. I f f s a t i s f i e s ( i ) , th en f s a t i s f i e s ( i i ) w i th

( f ) = f , C o n v erse ly f o r any <p ± f f s a t i s f i e s ( i i ) th e n

f ( Xi , . . . , Xj , « . . , Xp J — (f)(^ Xi , . • . , Xi Xg . . . Xj , • . . , Xi . . . Xp )

— 0 ( ^ 1 f • • • ,Xij , . • . ,Xp )

t h e r e f o r e f s a t i s f i e s ( i ) .

DEFINITION. For each p a i r o f i n t e g e r s m,n 1 th e

m n-p lace f u n c t i o n f ( x n , . . . ,X[,j , . . . ) e Ei i s c a l l e d (m,n)

8 4

permanent i f f c o n ta in s m s e t s o f n v a r i a b l e s x^j

( l ^ i ï ^ m , 1 ^ j ^ n) and s a t i s f i e s t h e c o n d i t i o n

',Aere f o r each p a i r o f i n t e g e r s i , j ( l J ^ i ^ m , 1 ^ j ^ n)

x^ j d e n o te s th e p ro d u c t x^ i x^g • • • * LJ .

THEOREM 6 . 2 , f . ,X[,j , . . , ,x^n ) (m,n) perm anen t

i f an d o n ly i f t h e r e e x i s t s 0 c such t h a t

f (x i 1 , . . . ,xL j , . . , ,XjYin ) = (^(xi 1 , . . . , Xlj , . . . ,Xmn j .

PROOF. S im i la r to Theorem 6 . 1 .

DEFINITION. An (m,n) perm anent f u n c t i o n f £ Ei i s

c a l l e d n - r e m a n e n t ( f o r an y m).

REI'/IARK. We s h a l l u se w i th o u t ^ e c i a l m en tion the f a c t

t h a t th e c l a s s o f n -perm anen t f u n c t i o n s i s c lo s e d under

a d d i t i o n and m u l t i p l i c a t i o n (modulo 2 ) .

DEFINITION. An n - t u p l e o f f u n c t i o n s f ^ , . . . , f j , . . . , f p ,

where f o r each i n t e g e r j (1 ^ j ^ n) f j e E^ , i s c a l l e d

a perm anent n - t u p l e i f each f j s a t i s f i e s th e c o n d i t i o n

f j = f j , where f j = f^fg . . . f j .

THEOREM 6 .3 - ^ n - tu p le o f f u n c t io n s f , . . . , f j , . . . , fp

i s a perm anen t n- t u p l e i f and on ly i f t h e r e e x i s t , . . . ,0n ,

S5

where f o r e a c h inte^sier j (I ^ j ^ n) <j6j e such t h a t

f J = 0 j , where < j = 0 i < 2 • • *<Pj . ■ - .

PROOF, S im i la r t o Theorem 6 . 1 .

DEFINITION. An n - t u p l e of f u n c t i o n s f ^ , . . . , f j , . . . , fp

i s c a l l e d an (m, n) perm anent s e t i f f , . . . , f j , . . . ,fp

i s a perm anent n - t u p l e of (m, n) perm anent

f u n c t i o n s .

THEOREM 6 .4 . ^ n- t u p l e o f f u n c t io n s f , . . . , f j , . . . , fp

i s a n (m,n) p e rm a n m t s e t i f and o n ly i f t h e r e e x i s t

< i , . . . , ( p j , . . , ,0n , where f o r e ac h in d e x j ( l ^ j n )

(pj £ E l ) such t h a t

f J — (^11 , • • • j ^ l j , • • • ) •

PROOF-. B y Theorem 6 .2 a n d Theorem 6.3»

DEFINITION. An (m, n) pe rm anen t s e t i s c a l l e d

n -p e rm an en t ( f o r any m).

8 6

6 . 3 . A c o r re sp o n d e n c e between Ep and n - perm anent s e t s

C o n s id e r th e f o l lo w in g mapping

X -> Xi , . , , , x j , . . . , X p ( I )

where X e N (= [ 0 , 1 , . . , , n | ) , and f o r e ac h in d e x j

(1 ^ j ^ n) Xj = 1 i f j ^ X, Xj = 0 i f j > X.

NO'! ATI ON. Denote th e n - t u p l e o f v a lu e s x^ , . . . ,x j , . . . ,Xp

w i th t h e above p r o p e r t i e s by x . We s h a l l have n+1 such

n - t u p l e s x° , x^ , . . . ,x! .

DEFINITION. We c a l l an n - t u p l e o f v a lu e s x ^ , . . . , x j , . . .Xp,

\ w h e r e Xj ç [ 0 , l ] f o r j = 1 , . . . , n , an n -p e r n a n e a t p o i n t i f f o r

each in d e x j ( l ^ j ^ n) xj = xj .

XI t i s e a s y t o v e r i f y t h a t f o r e a c h X e N x i s an n -perm anent

p o i n t . C o n v e rse ly , i f x^ , . . . ,Xj , . . . ,Xp i s an F t-pem anen t p o in t

th e n t h e r e e x i s t s a u n iq u e X (O X ijC n) such t h a t Xj = 1

i f j ^ X, Xj = 0 i f j > X, t h a t i s x ^ , . . . , x j , . . . ,Xp i s

o f t h e f o r m x f o r some X e N. Hence we have p roved t h e

f o llovfing :

THEOiMvI 6 .5 . ( l ) i s a o n e - to -o n e mapping betw een e le m e n ts

o f N and n-perm anen t p o i n t s .

8 7

RElvIAxK. B y " th e v a lu e sequence cf a perm anent n - tu p le

^ 1 , # # #, j , m t h e v a n a h l e s i , * # * , x^j , # * #, x^ p ,

we s h a l l mean a sequence of n -perm anen t p o i n t s . F o r exam ple ,

th e v a lu e seq u en ce o f the p e r n a n e n t double d>i (x^ ,x^ ) ,^g ) IS

( 1 )

( 2 )

( 5 )

( 4 )

and i n v iew o f t h e f a c t t h a t 0g = each o f th e d o u b le s

( 1 ) , ( 2 ) , ( 3 ) , (4 ) w i l l be a 2 -perm anent p o i n t .

THEOREM 6 .6 . L e t <Pi, • • * , f j > • > • ,<Pn be any perm anent

n - t u p l e o f f u n c t i o n s i n m s e t s of n v a r i a b l e s x^j

( i = 1 , . . . , m , j = 1 , . . . , n ) and c o n s id e r th e v a lu e sequence

o f 0 1 , . . . , 0 j , . . . , 0 n . In p a r t i c u l a r c o n s id e r th e perm anent

•point (form ed by th e v a lu e s of 0 i , . . . ,0 j , . . . ,0p ) a t each o f

th e p o i n t sX, S i

(x i ...X]^ . . .Xgi } , .............. .................. \ A )

f o r a l l sequences X i , , . , ,X(,, . . . ,Xm w here e a c h X[ e N,Xr

and where X[ i s t h e n-p e rm an en t p o in t x ^ j , . . . , xlj , . . . ,X[,p

d e f in e d by x^ j = 1 i f j ^ X [, X ( , j = 0 i f j > Xq.

8 8

I f t h e pennanenb p o in t o f 0 i , . . . , 0j , . . . ,0p i s f i x e d a t ev e ry

p o i n t o f ty p e (A) , th e n th e r e e x i s t s an (m, n) perm anent s e t

f 1 , . . . , f j , . . . ,fp such t h a t f o r e a c h in d e x j = 1 , . . . , n

I j = 0 j a t e v e r y p o in t of ty p e ( a ) . M oreover, such an

(m, n) perm anent s e t i s u n iq u e .

PRO OB'. D efine th e n - tu p l e f , , f j , . . . , fp by

(^ 1 1 , • • • LJ , * • • ,^nin i — (^11 , • • • L j , • • • ) ,

f o r each in d e x j = 1 , . . . , n . S in ce 0 ^ , . . . , 0 j , . . . ,0p i s

a pe rm anen t n - t u p l e we have 0 j = 0 J f c r j = 1 , . . . , n ,

and so by Theorem 6 .4 f ^ . , f j , . . . , fp i s an (m,n) perm anent

s e t . A ls o , a t e a c h of t h e p o i n t s ( a ) , s in c e each x i ( f o r i = 1 , . . . , m )

i s an n -p e rm an en t p o in t , we have

( ^ 1 1 , • • *, J , • • • (^11 , • • • I j , • • • , i n ) ~

, / Xi Xl \— 0j (x i , . . . , x q J • • • / .

That i s

f j = 0 j , f o r j = 1 , . . . , n .

M oreover, f j i s u n iq u e , s i n c e t h e v a lu e of f j a t ev/€ry

p o in t Xi 1 . . . X^j . * .Xqip , w here each ^Lj ^ i ^ , 1 j ds g iv e n by

f \ / XCt X I Xrp \f \X ii , . . fXRj , • • • ,Xmn / — , • • • ,^L , • • • / ?

Xwhere f o r each in d e x i = 1 , . . . , m x [ i s the n -p e rm an en t p o in t

X

89

NOTATION* Denote by Ep th e s e t o f a l l n -perm anen t s e t s ,

THEOREM 6 . 7 . The mapping

J : F(Xi X l , . . . ,Xm ) ^*1, • • • 3 • • •

of Ep to Ep*, where f (Xi , . . . ,Xfj, ) e Ep and f . , f j , . . . ,fp

i s t h e (rn,n) perm anent s e t i n t h e v a r i a b l e s Xi 1 , • . . ,X(,j , . . . ,x^p

whose perm anent p o in t i s d e f in e d a t each o f th e p o i n t s

Xi^^ , . . . ,XL^*', . . . ,Xfn^ ( f o r a l l seq u en ces X , . . . ,X|_, . . . ,X^ where

each Xt e N) f j (x^^^ , . . . , x ^ ^ ^ , . . . , x ^ ^ ) = 1 i f j ^ P( X i , . . . ,Xm ) ;

f j (x i^ ^ , . . . ,X(,^'', . . . , Xm^ ) = 0 i f j > F(Xi , . . . ,X(D ) , i s o n e - to -o n e

and o n to .

PROOF* L e t F ( X i , . . , , X m ) be any f i x e d m -p lace f u n c t i o n

o f Ep and c o n s id e r th e c o rre sp o n d e n c e

J I F(Xi , * * . , X [ , . . . ,Xm ) -» 01, • * • ,0J , • * • >0n

where 0 i , . . . , 0j , . . , ,0p i s a perm anent n - t u p l e i n th e v a r i a b l e s

X i1 , . . . , X [ j , . . . ,Xmp whose perm anent p o i n t i s d e f in e d a t each of

th e p o i n t [ ^ 1^^ x^^^ ( f o r a l l Xj , . . , , X [ , . . . ,Xm e N* )

by 0j (x i . ,x^^ ^, • • • ,Xm ; = 1 i f j F(Xi , . . ,Xm J j

0 j (x i^^ , . . . , xl^ , X m ^ ) = 0 i f j > f (Xi , . . . ,Xm ) . Such a perm anent

n - t u p l e c e r t a i n l y e x i s t s , f o r i t can be fo n n e d s im p ly by f i l l i n g i n

a l l t h e u n d e f in e d v a lu e s i n th e v a lu e sequence o f 0 i , . . . , 0 j , . . . ,0p

by n-perinanen t p o in t s .

9 0

D efin e t h e (m,n) perm anent s e t f ^ . , f j , . . . , fp by

(^11 , * • • ,^LJ , • • * , ^ n ) “ (^11 , • • • j , • • • , ^ n )

f o r j = 1 , . . . , n .

Then, by Theorem 6 , 6 , f^ . , f j , . . . , fp i s u n iq u e and

J : F(Xi , . . . , X q , * , . ,Xm ) f ] _ , . » . , f j , . . , , f n »

C o n v e rse ly , th e n—perm anent p o i n t o f a g iv e n n—p e rn a n e n t s e t

f i , . . . , f j , . . , , fp a t e a c h o f th e p o i n t s o f i t s v a lu e sequence i s

u n iq u e ly d e f in e d . In p a r t i c u l a r , a t t h e p o i n t s . . . . . .

( f o r a l l s eq u en ces X , . . . ,X|_, . . . ,Xm e ) . Thus t h e r e e x i s t s

a u n ique f u n c t i o n F su ch t h a t

J I F(X], , . . . , X [ , . . . ,Xm ) f 1 , . , , f j , • • • , fn •

91

6 . 4 . Some co m p le te s u b s e t s o f En

We s h a l l now p ro ceed to a p p ly t h e o n e - to -o n e co rre sp o n d e n c e

J t o the p rob lem o f the g e n e r a t i o n of th e com ple te s e t E p .

F i r s t l y we have a re m ark on s u b s t i t u t i o n .

REMARK. Suppose t h a t f (Xi , . . . ) , G% , . . . a r e

f u n c t i o n s , each b e lo n g in g to Ep , a n d t h a t

J t E(Xi , . . . ,Xm ) f i , . . . , f j ( x i i , . . . , X L j , . . . ,Xmp ) , . . . , f p ,

and fo r each in d e x i = 1 , . . . , m ,

S l i > • «.j SLj > • • • ^SLn .

Then

J : F ( & i , . . . , G - j ) -> h i , . . . , h j , . . . , h p ,

where f o r j = 1 , . . . ,n

KSîvltlRK. I n f u t u r e , we s h a l l a lw ays assume t h a t th e symbol ->

d e n o te s t h e c o r re sp o n d e n c e J .

TERMINOLOGY. V/e say t h a t an a r b i t r a r y s u b s e t S of Ep

ge n e r a t e s a f i x e d s u b s e t T i f S g e n e r a t e s e v e r y s i n g l e f u n c t i o n

o f T.

DEFINITION. For e a c h f i x e d in d e x j (1 ^ j ^ n) we s h a l l

d e n o te th e s e t of a l l n -pe rm anen t s e t s f ^ , . . . , fp f o r w h ich

92

f j + i = . . . = f n = 0 by Cj*, and th e c o r re sp o n d in g s u b s e t

o f Sp by C j . That i s , Cj i s th e s e t of a l l f u n c t i o n s

F ( f Ep) such t h a t t h e n -perm anen t s e t f i , . . . , f p , where

F f 1 , . . . ,fp , b e lo n g s to C j* .

I t i s e a s y to v e r i f y t h a t £ j i s c lo s e d u r d e r c o m p o s i t io n

( f o r e a c h f i x e d j ) .

We s h a l l need t h e f o l lo w in g I n a u c t iv e C r i t e r i o n o f

C om ple teness .

INDUCTIVE CRITERION OF COMPLETENESS. I f a s u b s e t S

o f Ep g e n e r a te s C^, and f o r each in d e x j (1 ^ j < n) th e

s u b s e t S U Cj g e n e r a t e s th e s u b s e t Cj +i , th e n S i s co m p le te .

REMARK. B efore we p r e s e n t a p ra c td .c a l f or m o f t h i s C r i t e r i o n

(Lemma 6 . 1 ) v/e s h a l l -make some rem arks a b o u t tJrie ' g e n e r a t in g

p o w er’ cf t h e c i r c u l a r p e r m u ta t io n X + 1 modulo n + 1 .

I t i s e a s y t o v e r i f y t h a t

X + 1 Xp + 1 , Xp + Xi , . « . , Xp + Xp _

and t h a t , f o r j = 1 , . . . , n - 1 , t h e j + 1^^ power o f X + 1 i s

th e f u n c t i o n

( 9 ' ' ; f j y > f ’j + l 7 ^ j + 2 ) • • • 3 f*n )

X + j + 1 -> X + 1 - j +Xp_ j+1 , . . . +Xp_ j +1 ,X^_j+1 ,Xp-j +Xi , . . . ,Xp_ j +Xp_ (j + i)

93

I f F i s any f u n c t i o n o f Ep such t h a t

F-> f j , . « , fk f O j . . . j O j

v-i.th j -1 u n i t s , and n -k z e r o s , where 1 ^ j k ^ n^ then

f o r e a c h p a i r o f f i x e d i n t e g e r s i , -6. (O i j - 1 , 0 ^ ^ n -k )

we can move th e sequence f j , . . . , fk i p l a c e s t o t h e l e f t , o r

& pl& ces to the r i g h t , b y t h e r e s p e c t i v e s u b s t i t u t i o n s

( i ) F ' = ( f ) + n + 1 - i

, f j , « . . , f | ^ , 0 , « . . , 0 ,

w i t h j - ( i + l ) u n i t s , and n + i -k z e r o s ;

( i i ) F” = (F) + &

wi t h j+6-1 u n i t s , and n—(k+&) z e r o s .

vVe s h a . l l c a l l t h i s p ro c e s s s h u n t i n g .

LEHvttLfV 6 . 1 . I f a s u b s e t S En g e n e r a t e s , t o g e t h e r

;v i th t h e c i r c u l a r p e rm u ta t io n X+1 modulo n+1 , and f o r e a ch in d e x j

(1 (6 j < n) th e s u b s e t 3 U Cj g e n e r a t e s th e .1+1- p la c e f u n c t i o n. . » * *

G-J+ 1 ( X i , . . . ,Xj+1 ) -> X i , . . . ,x j ,xj+ 1 , 0 , . . . , 0 , \\faere f o r e a c h in d e x

i ( l i ^ j+1 ) Xj, d e n o te s t h e p ro d u c t i % g . . .X[ i , t h e n S

i s co m p le te .

PROOF. I t s u f f i c e s to p ro v e t h a t f c r e ac h in d e x j

(1 ^ j < n ) , t h e s u b s e t S U ^ g e n e r a te s t h e s u b s e t £ j + i > &nd

94

th e n t h e r e s u l t f o l l o w s by th e I n d u c t iv e C r i t e r i o n o f C o m p le ten ess .

Suppose F -> f i , . . . , f j + i , 0 , . . . , 0 i s any f i x e d f u n c t i o n

o f £ j + i and c o n s id e r th e j+1 f u n c t io n s d e f in e d

f o r i = 1 , . . . , j+1 by

F^ -» f L , 0 , . . . , 0 ,

w i th n-1 z e r o s .

Each F**, f o r i = 1 , . . . , n , b e lo n g s to C and i s t h e r e f o r e

g e n e r a te d by S. Hence, S U Cj g e n e r a t e s F , s i n c e

F = +

and t h e r e s u l t f o l lo v / s .

The n e x t Lemma i s co n ce rn e d w i th c r i t e r i a s a t i s f i e d by a n

a r b i t r a r y s u b s e t cf En w hich g e n e r a t e s C%.

LEI<IMA 6 , 2 . I f a s u b s e t S g e n e r a te s th e 4- f u n c t i o ns

a ( x ) -> 0 , 0 , . . . , 0 ;

B(x ) 1 , 0 , . . . , 0 ;

C(X,Y) -» Xi+yx , 0 , . . . , 0 ;

D(X,y ) -> y%

t o g e t h e r w i th th e n 1- p l a c e f u n c t i o n s H ' , . . . , H T d e f i n e d ,

f o r i = 1 , . . . , n , ^

H g x )

th e n S g e n e r a t e s C^.

9 5

PROOF. By Theorem 2.1 ( s e e C h ap te r 2 ) from A,B,C,D,

we can show t h a t S g e n e r a t e s e v e ry f u n c t i o n F ' e C of t h e

form

F -» f ( x x i , . . . , X | < ; x } j O , . . . , 0 ,

f o r any f i x e d in d e x k > 0.

We can t h e r e f o r e shov/ f rom A,B,C,D,H*^, . . . t h a t S

g e n e r a t e s e v e r y f u n c t i o n o f t h e form

F f ( x % 0 , . . . , 0 ,

f o r any f i x e d in d e x m > 0 ; t h a t i s , ev e ry f u n c t i o n of C .

For exam ple: l e t n=m=2. From A, B,C,D, S g e n e r a t e s e v e r y

f u n c t i o n o f th e fo rm

F ' (Xj ,Xg ,Xs ,^4 ) , an d

from A, . . . S g e n e r a t e s e v e ry f u n c t i o n o f t h e form

F ( X i , X s ) = F ' (H (X, ) ,H= (Xi ) , t f ( X s ) ,H= (Xs ) )

-*■ f ( X i i ,Xl2,X2 1, %g2) | 0* . ' ' . •

TIiFOREM 6 .8 . I f S i s a s u b s e t of Ep which g e n e r a t e s th e

c i r c u l a r p e rm u ta t io n X+1 modulo n+1 on N, t o g e t h e r w i th a 2 - p la c e*

f unet i o n X H Y d e f in e d a s f o l l o w s ;

d e f in e X A Y f o r f i x e d i n d i c e s r , s (1 r ^ s ^ n ) by

X A Y - > X x A Yi y •• • )Xr a ÿ r fYr+ i • fYs f iŸs+ i f >

9 5

where

A + y ^ , f o r 1 < ^ ^ r ,

f o r a f i x e d in d e x t ( l ^ t ^ r , t < n ) d e f in e

X* -> X i Xi_, X Xi., . where a ................. a iS any t a, . a -------- t+1 ' n ^t+1 nf i x e d i n c r e a s i n g sequence o f i n t e g e r s t+1 ^ ^ n ,

*Y -> , j^t+l " • • • ÏÉierË ( 3 i y - , / 3 ^ ± s _any

i n c r e a s i n g sequence o f i n t e g e r s 1 ^ / ? i ^ . . . ^ ^ t , a,nd

f o r f i x e d r , s , t ( l ^ t ^ r ^ s ^ n , t < n) we d e f i n e X A* Y

X A Y = X* A* Y.

Then S i s c o m p le te .

PROOF. We s h a l l p ro v e Theorem 6 . 8 i n two s t a g e s . F i r s t

we s h a l l shov/ t h a t S g e n e r a t e s 0% . Then we s h a l l show t h a t

f o r e a c h in d e x j ( l ^ j ^ n ) , the s u b s e t S g e n e r a t e s t h e* * »x

j+ 1 - p la c e f u n c t i o n G-j+i -» x% , . . . ,x j ,xj+ , 0 , . . . , 0 . The r e s u l t

c l e a r l y f o l lo w s by Lemma 6.1 from t h e s e two a s s e r t i o n s , a n d t h e

h y p o th e se s o f th e Theorem.

We s h a l l u se w i th o u t me n t i o n t h e f o l l o w i n g p r o p e r t i e s o f

th e f u n c t i o n s x y , x^y : i f , . . . i s an y sequence o f

f t in c t io n s each b e lo n g in g to and s a t i s f y i n g (pj = pj f o r

j = 1 , 2 , , . , , k , th e n f o r any p a i r o f i n d i c e s -6,m ( l ^ 6 ^ m ^ k )

h = h •

= h -

97

I n f a c t , x y , XaJ a r e t h e w e l l known 'minimum* and * maximum '

f u n c t io n s of r e s p e c t i v e l y .

S tage ( i ) . We s h a l l show t h a t 8 g e n e r a t e s t h e 4 f u n c t io n s

A(x ) -> 0 , 0 , . . « , 0 ;

B(x ) -> 1, 0 , . . . , 0 ;

C(X,y ) -> Xj^yi, 0 , . . . , 0 ;

D(X,Y) Xi+yi, 0 , . . . , 0)t o g e t h e r w i th the n 1- p l a c e f u n c t i o n s d e f in e d f o r each

i m e x i (1 i ^ n) by

H^ X) -> X(, , 0 , . . . , 0 .

I t wi 11 th e n f o l l o w by Lemma 6 .2 t h a t S g e n e r a te s 0% . We s h a l l

l a y the p ro o f o u t a s a s e r i e s o f Lemmas.

LEAft'IA 6 . 3 . Under t h e h y p o th e se s o f Theorem 6 .8 , 3 g e n e r a te s

th e two 1- p la c e f u n c t i o n s A,B d e f in e d by

a ( x) -> 0 , 0 , • • • , 0 ;

b( x ) -> 1 , 0 , . . . , 0 •

PROOF. F o r each i n t e g e r j = 1 , 2 , . . . , n—1 8 g e n e r a te s th e

j+ 1 ^^ power o f X+1, a n d so 8 g e n e r a t e s t h e 1- p l a c e f u n c t io n

i ( x ) = x + n n x + n - i n . . . n X + i n x

1, # . . , 1, Xp+ 1 f • • • f ^ > 0, . , . , 0.

98

C o n sid e r t h e seq u en ce o f 1 - p l a c e f u n c t i o n s , 1 ^ , . . . g e n e r a t e d

by S and d e f in e d by

I ^( X) = I (X) + n+1 - r ,

( s h u n t in g X r + i , . . . , x ^ r p l a c e s to th e l e f t ) and f o r e a c h

i n t e g e r u = 1 , . . . , n-1

i ^ \ x ) = i X i ^ ( x ) ) .

We have

a ( x ) = I"(X )

b ( x ) = i " ( x ) + 1 .

T h e re fo re S g e n e r a t e s A,B*

REMARK* I t i s more c o n v e n ie n t to show t h a t S g e n e r a t e s

, . . . b e f o r e C and D.

LEMMA 6 .4 . Under t h e h y p o th e se s o f Theorem 6 . 8 , S

g e n e r a t e s th e n 1- p la c e f u n c t i o n s , . . . ,H" d e f in e d f o r e ac h

in d e x i ( l ^ i h ) Uy

H'-(X) -> X L , 0 , . . . , 0 .

PROOF. V/e s h a l l p ro v e f i r s t l y t h a t S g e n e r a t e s H" ( X ) .

Suppose t h a t f o r some f i x e d in d e x i ( l < i $ C n ) S g e n e r a t e s

th e 1 - p l a ce f une t i on

dL (x) -> i , . . . , 1 , X [ _ , . . . , X p ,

w i th i -1 u n i t s . Then S g e n e r a t e s YP i n th e f o l l o w i n g way;

9 9

T 2 n + 1— Lc o n s id e r th e seq u en ce of f u n c t i o n s g m e r a t e d

by S , an d d e f in e d by

J l ( x ) = J | . ( X ) ,

and f o r each i n t e g e r u = - iu+i . . , u . .

J l .(X) = J t ( J | , ( x ) + n ) .

U + 1 -QThat i s , we o b t a in J [ from J|, by f i r s t s h u n t in g t h e sequence

xi+urt^ • • • 3 one p la c e to t h e l e f t , c o n s e q u e n t ly in t r o d u c in g

an e x t r a z e r o , a n d t h e n r e p l a c i n g xi+u.-iby a u n i t . We haven+ 1— L ^

J l ( X ) - > ,Xp , 0 , . « . , 0 ,

w i th i - 1 u n i t s ,

and S g e n e r a t e s im m ed ia te ly by the s u b s t i t u t i o n

tf’ (x) = X x ) ) + n + 1 - ( i - 1 )

( s h u n t in g Xp i - 1 p la c e s to t h e l e f t ) .

C o n s id e r now th e two 1- p l a ce f u n c t i o n s g e n e r a te d by S.

( Ai (x ) + t ) D X-> 1 , . . . , 1 ,x^^^ J • • • jXp ,Xp+ I f . ' . jXg , 0 , . . . , 0 ;

(A(x) + n) A X 1 , . . . , 1 , 1 , . . . , 1 ,Xp+ . ,Xg ,Xg+ I f . . . fXp *

A ccording a s r = n , o r r < n we have

(a (x ) + t ) a X = ( x ) , o r (a ( x ) + n) a X = Jp+ % (x ) r e s p e c t i v e l y .

I n v iew of Lemma 6 .3 , in e i t h e r c a s e , S g e n e r a t e s a f u n c t i o n o f

type J l , f o r some f i x e d index i ( l < i ^ n ) , and so ^ g e n e r a t e s

(x) by t h e above a rgum ent.

100

We shfî.ll now prove t h a t f o r each u = 0 , ' l , , , . , n —1 S

g e n e r a t e s t h e u+1 1 -p la c e f u n c t i o n s i f . , î f “

I t i s e a sy to s e e t h a t t h e Lemma f o l l a v s on t a k in g u = n - 1 .

vVe p ro c e e d by i n d u c t io n on u , s t a r t i n g w i th th e case u = 0

w hich we have a l r e a d y shown to be t r u e . Assume t h a t f o r some

f i x e d i n t e g e r u (O (C u ^ n—2) S g e n e r a te s th e u+1 1- p l a ce

f u n c t i o n s i f , i f " . , l f “ th e n i t s u f f i c e s t o p ro v e t h a t

S g e n e r a t e s t h e 1—p la c e f u n c t i o n pn- ( u+ i y ^

C o n s id e r t h e 1- p l a ce f u n c t i o n g e n e r a te d by S

J ' ( X) = I f (X + n)

">■ Xj_ + 1 , 0 , * . * , 0 *

Hence S g e n e r a t e s t h e 1- p l a c e f u n c t i o n

t f ( x ) = J ' ( J ' ( X ) ) ,

and th e 2—p la c e f u n c t i o n

J(X ,Y) = H^(X n Y)

Xi A y , o , . . . , o ««1

B y th e i n d u c t io n h y p o th e s i s S g e n e r a te s t h e 1—p la c e f u n c t i o n

J"(x) = H "-" (XS1) + « 1-1

w ith (%i-1 u n i t s , t h a t is ^ + X p _ ( ^ i ) o cc u rrin g in th e p la c e .

1 0 1

Noting t h a t a( ^ + ^ - ( u + i ) ) = ^ - ( u + i ) see tha.t S

g e n e ra te s Hn-(u+i) a s f o l l o w s :

I f - 1 1)0= J ( l f “ ^(X) n J ” (X) )

LEMIÆA 6*5. Under th e h y p o th e se s o f Theorem 6 .8 , S

g e n e r a te s th e 2—p la c e f u n c t i o n s .

6(X,Y) -> ^2ÿ i , 0 , . . . , 0 ;

f O f . ' . f O .

PROCF. Ey Lam ma 6 . 4 S g e n e r a t e s th e 1—p la c e f u n c t i o n

i f (X) -> x i , 0 , . . . , 0 ,

ana so S g e n e r a t e s th e 2—p la c e f u n c t i o n

j ( x , y ) = H ^ X n Y)

Xi A y , o , « . . , o ,

ana th e 1—p la c e f u n c t i o n

J ’ ” (X) = H^(X) + n

-> Xi+1 , . . . ,Xi + 1 ,

w i th Xi + 1 r e p e a t e d n t im e s .

We have

C(X,Y) = J' " ( X ) , J " ' (T) ) ) ) ,

and D(X,Y) = C( J " ' ( c ( J " ' (X) , J ' " ( x ) ) )> J " ' ( C ( X , y ) ) ) .

Hence S g e n e ra te s C, D.

This completes the p roo f cf the f i r s t s t a g e o f Theor’em 6 . 8 .

102

S tage ( i i ) . We s h a l l r e q u i r e t h e fo l lo w in g Lemma,

LMh.A 6 , 6 . Under t h e h y p o th e se s o f Theorem 6 . 8 , S

'•enera tes t h e 1 - u la c e f u n c t i o n s

K ( X ) x% , . . . , , 0 , . . . , 0 ,

w i th Xj r e treated t+1 t i m e s ,

L (x ) 1 , . . . , 1 , Xp , . . . , ,

w ith t u n i t s #

PROOF. We s h a l l p rove f i r s t l y t h a t S g e n e r a t e s K.

There a r e two c a s e s .

Case ( i ) . I f t = r : by Lemmas 6 .3 and 6 .4 , S g e n e r a t e s

th e 1- p l a c e f u n c t i o n s A(x ) and H^(x), r e s p e c t i v e l y , and so 3

g e n e r a te s t h e 1- p l a c e f u n c t i o n s

H^(x) + n -> Xi + 1 , . . . ,Xi + 1 ,

and

K' (X) = (A (x)+n) n (H^(x)+n)

1 , . . . ,1 ,Xi + 1 , . . . ,Xi + 1 ,

w i th t u n i t s #

We have

k ( x) = k ' ( x ) + 1,

and i n t h i s c a se 3 g e n e r a t e s K.

103

Case ( i i ) . I f t < r : a s b e f o r e , S g e n e r a te s A and

, and so 3 g e n e r a te s t h e 1- p l a ce f u n c t i o n s

K"(X) = H ^(hH x ) + n) + n

w i th r e p e a t e d n t im e s ;

K' ” (X) = K’’(X) n a ( x )

w ith Xi r e p e a t e d r t i m e s .

A ccord ing a s 2 t ^ r#-1or 2 t < r —)we have

K(x ) = ( (A (x )+ r -^ 4 l |n K"(X)) + n + l + t - r ,

o r

K(x ) = (K "(x ) n (A(x)+r-f:tlJ)f n + l + t - r ,

r e s p e c t i v e l y ( r e p l a c i n g th e f i r s t x ^ ' s by u n i t s , an d

sh u n t in g the re m a in in g sequence of t+? x ^ ’ s p la c e s t o t h e l e f t )

Hence i n t h i s c a s e S g e n e r a t e s K.

F in a l . ly we have

L(x ) = K ( î f (x ) + n ) + n ,

where

H^(X) -> X p , 0 , . . . , 0 .

% Lemma 6 . 4 S g e n e r a t e s H" , and so S g e n e r a te s L .

The p ro o f of the Lemma i s com ple ted .

104

I t rem a in s to show t h a t , f o r each i n i ex j ( l ^ j < n ) , th e

s u b s e t S U Cj g e n e r a t e s t h e j+ 1 -p la c e f u n c t i o n>St *

G-J+ 1 -> Xj^, . . • , xj , x j + 2 , 0 , . . . , 0 »

T here a r e two c a s e s .

Case ( i ) . I f j ^ t : th e n the f u n c t io n s. * *

and. *

^ , . . . , X j + j _ ) -> X j + i , 0 , . . . , 0

each be long t o £ j , and a r e t h e r e f o r e g e n e r a te d by S U Cj .

By Lemma 6 . 6 S g e n e r a t e s K( x ) , and so S U £ j g e n e r a te s

P ( Xi , . . . ,XjV 1 ) = K(P' ' X i , . . . ,X ji 1 ) )jje O

-> X j + i , . . . , X j + ] _ , 0 , . . . , 0 5

w ith Xj+ 1 r e p e a t e d t +1 t im e s .

I f t < s , t h e n S U £ j g e n e r a t e s Gj+i by th e s u b s t i t u t i o n

Gj+ 1 ( X i , . . . ,Xj+ 1 ) = (Cj ( X i , . • • ) + t—j n P ( X i , . . . ,Xj+ I ) ) + n+1 + j —t ,*

That i s , we s h u n t X i , . . . ,x j t - j p l a c e s t o tine r i g h t , r e p l a c e t h e* * « *

f i r s t z e ro by x j + i , and sh u n t th e new sequence x^ , . . . , xj ,Xj+^ t - j

p la c e s t o th e l e f t .

I f t = s , th e n the 1—pla.ce f u n c t i o n

Q' ( x ) X j+1 , x j + X i , . . . ,Xj+Xj_ 1 , 0 , . . . , 0

b e lo n g s to £ j , T h e re fo re S U Cj g e n e r a t e s t h e 1- p la c e f u n c t i o n .

Q"(X) = Q' (X) + n

-» X i , . . . ,Xj ) . . ; , Xj ,

w i th Xj r e p e a t e d n—j t im e s .

105

C o n s id e r th e sequence o f 1 - p la c e f u n c t i o n s ”

g e n e r a te d by S U £ j and d e f in e d by

Q°(X) = Q"(X),

and f o r each u=0 , 1 , , . . , t - j -1

Q "*\x) =(«"(x) n A(x)) + 1 .

T hat i s , if' j < t we form by r e p l a c i n g t h e l a s t n - s

Xj ' 3 by z e ro s and s h u n t in g th e re m a in in g sequence one p l a c e t o

th e r i g h t , we th e n fo rm from Q^, f o r u=1 , , . . , t - ( j +1 ) , by

r e p l a c i n g th e l a s t Xj by a z e ro and s h u n t in g th e r e m a in in g

sequence one p la c e t o th e r i g h t .

We have

Q "(x) -*■ 1 , . . . , X]_, . . • , Xj , Xj , 0 , . . . , 0 ,

w i th e x a c t l y 2 x j ' s and n—( t + 1 ) z e r o s .

Hence

G j i i ( X i , . . . , X j i i ) = ( q* " J ( s (Xi , . . . , X j ) ) * P(Xi , . . . , X j> i ) ) +

+ n+1+ j - t ,

and so i n t h i s c a s e S U £ j g e n e r a t e s Gj+i .

Case 2. I f t < j < n : i n t h i s c a s e we s h a l l p rove

t h a t S U ^ g e n e r a te s th e two f u n c t i o n s

106

* * » >/'- * *T1 , • * • , Xp ) -> , • • • , x ^ , Xp , • • • , Xp , Xp , , . • , Xp ;

* îi« »U(Xl , • . . ,Xj+ 1 ) -> 1 ) » * * )%t+ 1 ) ^ t + l 3 * ' ' 3 ^ + l ) ^ 3 . . . , 0 .

From T and U i t f o l lo w s t h a t S U £ j g e n e r a te s Gj+ ^ , s i n c e

^j+ 1 (Xi , • • • jXj+ 1 ) = T(Xi , . . . ,Xp ) n TJ ( X i , » . . ,Xj+ 1 ) •

We s h a l l show f i r s t l y t h a t S U ^ g e n e r a t e s th e r - p l a ce

f u n c t i o n T(Xi , • . . ,Xp ) , S in ce t < j , th e f u n c t i o n* * »

T ( X , • « *, Xp ) -> Xp+1, Xp +x^ , • • • , Xp + x ^ , 0 , , « « j 0 ,

b e lo n g s t o £ j , and so S U £ j g e n e r a te s T by t h e s u b s t i t u t i o n

T ( X i , . . . , X p ) = T ' ( X i , . . . , X r . ) + n .

I t rem ains to show t h a t S U Cj g e n e r a te s t h e j+ 1 - p la c e

f u n c t i o n U. B y Lemma 6 . 6 S g e n e r a t e s t h e 1- p l a ce f u n c t i o n

L ( X ) 1 , . « « , 1 , Xp , • • • , Xp f

wi oh t u n i t s ,

ana so S g e n e r a te s t h e 1- p l a c e f u n c t i o n

v ( x ) = L(X) n X

A . ♦ •,X.j.^ « « ,Xp .

C o n sid e r t h e sequence o f 1—p la c e f u n c t i o n s

g e n e r a te d by S a n d d e f in e d by

Vi ( x ) = V(X) + 1

3 n+1 > - * « , ^ + 1 , ^ t +1 5 • • • Xn- 1 J

w i th âq,+1 r e p e a t e d t +1 t im e s ,

107

and f o r e a c h i n t e g e r u = 1 , 2 , . . . , n - ( t -*1 )

V^^(X) = V^(V^(X) )

( 3 • • • ) ^t+ 1 3 • • • 3 i 't+U >* » . . « . » « .

^ + i - u + ^ - u + ' J • • • 3 ^ + i-U + ^ - u + 1 f ' 3%n + x ^ -11+ I )

^ t+u+ i > ' ^ t+ u +2 ' ' ' )

Xn-u+ ' ^ -u + XLy+ 1 , * . . ,Xn-u+ (u+i ) •

We have

V ( i f (x) ) - > x.j- 3 , • • • ,x;-5+ X ,x-t+ 2 , • • • ,Xp ,

ana s i n c e t ^ 1 , t h e j+ 1- p l a c e f u n c t i o n

1 / X * *U \Xx J • • • ,Xj+X ) i -fc+x 3 • • • ^x!j+X J 0 , • . . , 0

b e lo n g s to C j . T h e re fo re S U Cj g e n e r a t e s U by th e

s u b s t i t u t i o n

U ( X i , . . . , X j * i ) = V ^ ( V " - R u ' ( X i , . . . , X j i i ) + t ) ) .

This com ple tes t h e p ro o f o f t h e seco n d s t a g e and Theorem 6 . 8

f o l l o w s .

1 0 8

D e fin e X n Y f o r f i x e d i n d i c e s r , s ( 0 r s < n) by

y « # • • • • • • » •X A Y -» Xi AYi , * ' * ,Xr A^r *yr+ ! > • • • > Ys jX^+iYs+i i»*» jX/iYn j

and f o r a f i x e d i n t e g e r t ( s ^ t < n , t > O) d e f in e

“X -» x^ J • • • X , • • • jX|i, where ax, *» * , ^ 4.ux Kt ■ ^

i s a n y i n c r e a s i n g sequence o f i n t e g e r s 1 ^ (%x ^ . . . ^ a ^ ^ t ;

y Y i s • • • > y J • • • ,Yg 9 where /St+ 1 ? • • • >fn

i s an y i n c r e a s i n g sequence o f i n t e g e r s t +1 ^/9.J-+x K . . . K ^ n ,

th e n f o r f i x e d r , s , t (0 ^ r ^ s ^ t < n , t > 0 )

X Â Y = “X A y".

'THEOREM 6 . 9 - ^ ^ i s a s u b s e t of En w hich g e n e r a te s

th e c i r c u l a r p e r m u ta t io n X+1 (modulo n+1) t o g e t h e r w i th

a 2- p l a c e X A Y, th e n Z i s c o m p le te #

PROOF. C o n s id e r t h e p e r m u ta t io n R(x ) of N d e f i n e d by

R( X) -> Xn+ 1 , . • • , Xn + x— 1 , . «. , Xx + 1 •

vVe s h a l l p ro v e t h a t th e f u n c t i o n X A Y i s c o n ju g a te u n d e r R(x )*

to a s u i t a b l e f u n c t i o n X A Y a s d e f in e d in Theorem 6 . 8 . The

r e s u l t w i l l th e n fo l lo w by Theorem 6 . 8 and Theorem 1 . 1 o f

C h ap te r 1 a f t e r we have shown uhat ^ g e n e r a t e s a 1—p la c e

f u n c t i o n c o n ju g a te u n d e r R (x) to th e c i r c u l a r p e rm u ta t io n X+1

(modulo n+1 )#

1 09

F o r f i x e d i n d i c e s r , s , t ( 0 ^ r ^ s ^ t < n , t > O)*

c o n s id e r t h e p a r t i c u l a r f u n c t i o n X O Y d e f in e d a s f o l l o w s : d e f i n e

(X n y ) ' by

U • ♦ • « • • ' *n 1 ) -> XxaYi , • • • ,Xp_ s AYn-s ^yn+1- s > • • • >Yn-r jXn+

and d e f in e

Yi*A • • • #X -*■ X x , . . . , x ^ _ t , Xn+X-a^3 • • • jXp+X-«1 >

y Yn+ 1—/?n 3 • • • > y n + , Yn+ i —t 3 • • • fYn 3o+ 1

th e n

X n Y = ( x * n # Y ) \

I t i s e a sy t o v e r i f y t h a t

r ( i t ^ ( x ) n IT X Y ) ) = X n Y,

and so X A Y i s c o n ju g a te u n d e r R(x ) t o a s u i t a b l e f u n c t i o n

o f ty p e X A Y .

F i n a l l y S g e n e r a t e s t h e 1 -p la c e f u n c t i o n X+n (modulo n+1 )

and we have

r (R"^(X) + 1) = X + n .

T h e re fo re X + 1 i s a l s o c o n ju g a te u n d e r R(x ) t o a f u n c t io n

g e n e r a te d by S. Hence the r e s u l t f o l l o v / s .

110

BIBLIOG-RAPHY

[1] BURNSIDE, W. The theory o f groups (Cambridge U n iv .P ress, 1911).

[ 2 ] BOCHUAR, D.A. On a th ree-va lu ed ca lcu lu s and i t s a p p lic a tio n~ to th e a n a ly s is o f th e paradoxes of th e extended

fu n c tio n a l c a lc u lu s . Matem. sb .4(1934-),281-308 ( in R ussian).

CUNING-HAME-UREEN, R.A. Ph.D. t h e s is . U n iv ersity o f L e ic e s te r , I960.

[4 ] DAVIES, R.O. Two theorems on e s s e n t ia l v a r ia b le s .~ J.London Math.Soc. 41 (1966).

[ 5 ] EVANS, T. and HARDY, L. S h effer stroke fu n c tio n s in many-valued~ lo g ic s . P ortu galiae Math. 16 ( l9 5 7 ) , 83-93*

[6] FOXLEY, E. The determ ination o f a l l S h effer fu n c tio n s in3 -valued lo g ic u sin g a lo g ic a l computer.Notre Dame J . o f Formal Logic 3 (1 9 6 2 ), 41-30*

[ 7 ] GOODSTEIN, R.L. Truth ta b le s . M ath.Gazette 4 6 (1 9 6 2 ), 18-23.

[8 ] GOODSTEIN, R.L. Polynom ial generators over G alois f i e l d s .~ J.London Math.Soc. V ol . 3 6 (1 9 6 l) , 29-32.

[ 9 ] GOTLIND, E. Some S h effer fu n ctio n s in n-valued lo g ic .~ P ortu ga liae Math. ^1 (1952), 141-149.

[hO] HALL, M. The theory of groups (M acmillan, 1959).

[ 1 1 ] HOWIE, J.M. The subsemigroup generated by the idem potentso f a f u l l transform ation semigroup.J.London Math.Soc. (forthcom ing).

[ 1 2 ] ITO, N. T ra n sitiv e perm utation groups o f degreep = 2q+1, p and q being prime numbers.Trans.Amer.Math.Soc. 116 (1 9 6 5 ), I 5 I - I 66.

[ 1 3 ] JORDAN, C. Recherches sur l e s s u b s t itu t io n s .J.M ath.Pures.Appl. ^7 ( I 8 7 2 ) , 351-367»

[ 1 4 ] KALICKI, J . A t e s t f o r the e x isten ce of ta u to lo g ie s accordingto many—valued tr u th - ta b le s . J.Sym bolic Logic1 5 ( 1 9 5 0 ) , 182-184.

111

[15] KUZNETSOV, A.V.

[1^] IÆARTIN, N.M.

[17] MARTIN, N.M.

[ j 8 ] MARTIN, N.M.

[1^] IIEIBON, R .I .

[20] PICCARD, S.

[ ^ ] PICCARD, S.

[22] POST, E.L.

[ 2 3 ] POST, E.L.

[ ^ ] POST, E.L.

[ 2 5 ] QUINE, W.V.

[26] ROSENBERG, M.I,

[ 2 7 ] SALOMAA, A.

On the problems o f id e n t ity and o f fu n c tio n a l com pleteness fo r a lg eb ra ic system s.Proc. 3rd A ll-U nion M ath.Congress,V o l.2 ,Moscow ( 1 9 5 6 ) , 145-14 6 ( in R ussian).

Some a n a lo g ies o f the S h effer stroke fu n ction s in n-valued lo g ic . Indagationes Math. 12 (1 9 5 0 ), 393-400.

A note on S h effer fu n ction s in n-valued lo g ic . Methodus 3 (1 9 5 1 ), 240-242.

The S h effer fu n c tio n s of 3-valued lo g ic .J . Symbolic Logic J9 (1954), 45-51 .

Sim plest normal tru th fu n c tio n s .J. Symbolic Logic 20 (1955), 105-108.

Sur l e s fo n ctio n s d é f in ie s dans le s ensembles f i n i s quelconques. Fund.Math. 24 (1 9 3 5 ), 298-301.

Sur l e s bases du group symétrique e t le s couples de su b s titu tio n s qui engendrent un group r é g u lie r . P a r is , L ibrarie V u ib ert, (1946).

In troduction to a gen eral theory o f elementary p ro p o sitio n s. Amer.J .Math. 43 (1 9 2 1 ), 163-185.

F in ite combinatory p ro ce sse s . Formulation I .J . Symbolic Logic 1 (1 9 3 6 ), 103-105.

The two—valued i t e r a t iv e system s o f mathematical lo g ic (Princeton U n iv .P ress, 1941).

The problem o f s im p lify in g tru th fu n c tio n s .Amer.Math.Monthly 59 (1952), 521-531.

La stru ctu re des fo n c tio n s de p lu sieu rs v a r ia b les sur un ensemble f i n i . C. R.Acad. Sc. Paris2 6 0 ( 1 9 6 5 ) , 3817-3819.

On the com position o f fu n c tio n s o f sev era l v a r ia b les ranging over a f i n i t e s e t .A nn.U niv.Turkuensis, S er.A .I .41 (l96o)

112

[ ^ ] SALOMAA, A.

[29] SALOMAA, A.

[ ^ ] SALOMAA, A.

[ 3 1 ] SCHOFIELD, P.

[ 3 2 ] SCHOFIELD, P.

[ ^ ] SHEFFER, H.M.

[ ^ ] SIERPINSKI, W.

[ ^ ] SLUPECKI, J.

[36] StUPECKI, J .

[ 3 7 ] SWIFT, J.D .

[38] WEBB, D.L.

[ 3 9 ] WEBB, D.L.

[ 4 0 ] WEBB, D.L.

Some c o m p le ten ess c r i t e r i a f o r s e t s of f u n c t i o n s o v e r a f i n i t e dom ain . I .A n n .U n iv .T u rk u e n s is . S e r .A . I . ^ (1962).

Some co m p le ten e ss c r i t e r i a f o r s e t s o f f u n c t i o n s o v e r a f i n i t e dom ain. I I .Ann.UniV. T u r k u e n s i s . S e r . A . I . 6^ (1963).

On e s s e n t i a l v a r i a b l e s o f f u n c t i o n s , e s p e c i a l l y i n th e a l g e b r a o f lo g ic # Ann.A c a d .S c i .P e n n .S e r .A . I .Math. 339 ( 1 9 6 3 ) , 3 -11 .

On a c o r re sp o n d en ce be tw een many—v a lu e d andtwo-va lu ed l o g i c s . Z eit . f .m a t h .L o g i k 10 ( l 964),265-274.

Complete s u b s e t s o f mappings ove r a f i n i t e domain. Proc.Cam biddge P h i l o s o p h i c a l S oc . ( i n t h e p r e s s %

A s e t o f f i v e in d e p e n d e n t p o s t u l a t e s f o r Boolean a l g e b r a s , w i th a p p l i c a t i o n to l o g i c a l c o n s ta n t s . T r a n s . Amer.M ath .S o c . 14 ( l9 1 3 ) , 481-488.

Sur l e s s u i t e s i n f i n i e s de f o n c t i o n s d é f i n i e s dans l e s ensem bles q u e lc o n q u e s . F u n d .M ath. 24 (1935),209- 212 .

K ry te r iu m p e ^ n o sc i w ie lo w ar to sc io w y ch systemow l o g i k i zdan . C .R .S o c .S c i .V a r s o v ie . 3 2 (1 939),1 0 2 -1 0 9 .

Dowéd aksjornatyzowalndsci p&^nych systemow wie-lowartdsciowych rachunku zd^n. C .R .S o c .S c i. V arsovie 3g (1 9 3 9 ), 110-128.

A lg e b ra ic p r o p e r t i e s o f N -va lued p r e p o s i t i o n a l c a l c u l i .. Amer.M ath . Monthly 59 (1952), 612-621.

G e n e ra t io n o f an y n -v a lu e d lo g ic by one b i n a r y o p e r a t i o n . P r o c .N a t .A c a d .S c i . U .S .A . 21 (1 9 3 5 ),252- 254 .

The a l g e b r a o f N -v a lu ed l o g i c . C .R .S o c .S c i .V a rso v ie 29 (1 9 3 6 ), 153—168.

D e f i n i t i o n o f P o s t ’ s g e n e r a l i z e d n e g a t i v e and maximum i n te rm s o f one b i n a r y o p e r a t i o n .Amer. J . Math.( 1 9 3 6 ) , 193-194.

113

[41] WERNICK, W. Complete s e ts o f lo g ic a l fu n c tio n s .Trans.Amer.Math.Soc. 51 (1941 ) , 117-152.

[4 2 ] WHEEIER, R.F. Complete p r e p o s it io n a l co n n ectiv es.

[4 3 ] WHEELER, R.F.

[4 4 ] WHEEIER, R.F.

[ 4 5 ] WIELAI'®T, H.

[46] YABLONSKII, S.V.

[4 7 ] YABLONSKII, S.V.

[48] YABLONSKII, S.V.

[4 9 ] YABLONSKII,S.V.

[ 5 0 ] ZHEGAnCEN,I.I.

[ 5 1 ] ZHEGALKIN,I.I.

[ 5 2 ] ZYLINSKI, E.

Z e it s c h r if t fu r math.Logik und Grundlagen Math. 7( 1 9 6 1 ) , 185-198. ~

An asym ptotic formula fo r the number of complete p r ep o s it io n a l c o n n e c tiv e s . Z e it .fü r math.Logik 8( 1962) , 1^ .

Complete connectives fo r the 3-valued p re p o s it io n a l c a lc u lu s . Proc.London Math.Soc. ^ ( l 966;, 167-192.

F in ite permutation groups (Academic P ress ,New York 1964).

On su p erp o sitio n s o f fu n c tio n s o f the algebra o f lo g ic . Mat.Sbornik 30(72) (1 9 5 2 ), 329-348 ( in R ussianyi

On fu n c tio n a l com pleteness in the th ree-va lu ed c a lc u lu s . Doklady Akad.Nauk. 3 .S .S .R . 95 (1 9 5 4 ), 1 1 5 3 - 1 1 5 5 ( in R ussian). ^

Functional co n stru ctio n s in many-valued lo g ic s .Proc. 3 ^ A ll-U nion M ath.Congress, v o l . 2 .Moscow ( 1 9 5 6 ) , 71-73 ( in R ussian).

Functional con stru ction s in a k-valued lo g ic .Trudy M at.In st.S tek lov ^1 (1958), 5-142 ( in R ussian)

On the technique of ev a lu a tin g p ro p o sitio n s in symbolic lo g ic . Matem.Sb. 34 (1 9 2 7 ), 9-28 ( in R u ssian ).,

The a r ith m etiza tio n o f sym bolic lo g ic .Mat.Sbornik 35 (1 9 2 8 ), 311-378 ( in R ussian).

Some remarks concerning th e theory of d ed u ction . Fund.Math. 7 (1 9 2 5 ), 203-209.