52
THE L J J . BROWER CENTENARY SYMPOSIOM A.S. TroehRa and D. van Dalen (editors) 0 North-Holhnd PubiiShing COmpl~~y, 1982 165 THE EFFECTIVE TOPOS J.M.E. Hyland Department of Pure Mathematics, Cambridge, England. 10 Introduction. The subject of this paper is the most accessible of a series of toposes which can be constructed from notions of realizability: it is that based on the original notion of recursive realizability in Kleene C19451. Of course there are many other kinds of realiza- bility (see Kleene-Vesley C19651, Kreisel C19591, Tait C19751). All these (and even the Dialectica Interpretation) fit into a very ab- stract framework described in Hyland-Johnstone-Pitts C19801. (Since we will refer to this paper frequently, we shorten the reference to HJP C19801.) In this abstract framework one passes easily (as is becoming customary, see Fourman C19771, Makkai-Reyes C19771, Boileau- Joyal C198ll) between logical and category theoretic formulations, using whichever is most appropriate. One good example is worth a host of generalities, so it is the aim of this paper to present this abstract approach to recursive realizability in some detail. The basic strategy readily extends to other cases. Many people, most notably Beeson (see f o r example Beeson C19771, have considered realizability extended to give interpretations of complicated formal systems. The flavour of the more category theo- retic treatment is to have one think in terms of models. Thus the approach looks like sheaf models for intuitionistic logic (see Fourman-Scott C 19 79 1, where one only has natural access to the models. (This parallel between realizability and sheaf models was first made explicit, for set theory, in an untitled manuscript, by Powell.)

[Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

  • Upload
    jme

  • View
    212

  • Download
    0

Embed Size (px)

Citation preview

Page 1: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

THE L J J . B R O W E R CENTENARY SYMPOSIOM A.S. TroehRa and D. van Dalen (editors) 0 North-Holhnd PubiiShing C O m p l ~ ~ y , 1982 165

THE EFFECTIVE TOPOS

J .M.E. Hyland Department of Pure Mathematics, Cambridge, England.

10 Introduct ion.

The s u b j e c t of t h i s paper is t h e most access ib l e of a series

of toposes which can be constructed from not ions of r e a l i z a b i l i t y :

it i s t h a t based on the o r i g i n a l not ion of r ecu r s ive r e a l i z a b i l i t y

i n Kleene C19451. O f course t h e r e a r e many o t h e r kinds of r e a l i z a -

b i l i t y (see Kleene-Vesley C19651, Kreisel C19591, T a i t C19751). A l l

these (and even the Dia l ec t i ca I n t e r p r e t a t i o n ) f i t i n t o a very ab-

s t r a c t framework descr ibed i n Hyland-Johnstone-Pitts C19801. (Since

w e w i l l r e f e r t o t h i s paper f r equen t ly , w e shorten the reference t o

HJP C19801.) I n t h i s a b s t r a c t framework one passes e a s i l y ( a s i s

becoming customary, see Fourman C19771, Makkai-Reyes C19771, Boileau-

Joyal C198ll) between l o g i c a l and category t h e o r e t i c formulations,

using whichever i s most appropriate . One good example i s worth a

host of g e n e r a l i t i e s , so it i s t h e aim of t h i s paper t o p re sen t t h i s

a b s t r a c t approach t o r ecu r s ive r e a l i z a b i l i t y i n some d e t a i l . The

bas i c s t r a t e g y r e a d i l y extends t o o the r cases .

Many people, most notably Beeson (see f o r example Beeson C19771,

have considered r e a l i z a b i l i t y extended t o give i n t e r p r e t a t i o n s of

complicated formal systems. The f lavour of t he more category theo-

re t ic t reatment is t o have one think i n t e r m s of models. Thus the

approach looks l i k e sheaf models f o r i n t u i t i o n i s t i c l o g i c (see

Fourman-Scott C 1 9 79 1 , where one only has n a t u r a l access t o the models.

(This p a r a l l e l between r e a l i z a b i l i t y and sheaf models was f i r s t made

e x p l i c i t , f o r set theory, i n an u n t i t l e d manuscript, by Powell.)

Page 2: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

166 J.M.E. HYLAND

A s i n t he case of sheaves, w e w i l l f i n d ou r se lves looking a t genuine

mathematical s t r u c t u r e s (wi th t h e i r non-standard l o g i c ) when w e

i n v e s t i g a t e t r u t h i n t h e e f f e c t i v e topos . W e w i l l be p re sen t ing

" the world of e f f e c t i v e mathematics" a s it appears t o t h e c l a s s i c a l

mathematician. (Of cour se , it is p o s s i b l e t o p r e s e n t t h e i d e a s i n

t h e contex t of more o r less any mathematical ideology. )

While t h e l o g i c a l approach t o c a t e g o r i e s enables us t o work

wi th concre te s t r u c t u r e s and apply our exper ience of elementary

l o g i c , t h e ca tegory t h e o r e t i c approach t o l o g i c enab le s us t o do

away wi th much l o g i c a l c a l c u l a t i o n and t o use i n s t e a d simple f a c t s

about c a t e g o r i e s ( i n p a r t i c u l a r f a c t s about toposes and geometric

morphisms). I t has become c l e a r i n r e c e n t yea r s t h a t much of con-

s t r u c t i v e l o g i c can be t r e a t e d very e l e g a n t l y i n the con tex t of

topos theory . This is i n harmony wi th work i n t h e i n t u i t i o n i s t

t r a d i t i o n on Beth and Kripke models (see van Dalen C19781, and t h e r e

w e r e many c o n t r i b u t i o n s t o t h e Brouwer Centenary Conference i n t h i s

a r e a . This paper simply does t h e same k ind of t h i n g f o r r e a l i z a -

b i l i t y . Of course t h e r e & a s u r p r i s e he re : t h e topos of t h i s paper

i s most un l ike a Grothendieck topos , and it i s n o t i n i t i a l l y

p l a u s i b l e t h a t theory a b s t r a c t e d from no t ions of c o n t i n u i t y should

have any a p p l i c a t i o n i n t h i s most non-topological s e t t i n g .

The f i r s t t h r e e s e c t i o n s of t h e paper s e rve t o in t roduce t h e

e f f e c t i v e topos a s a world b u i l t ou t of t h e l o g i c of r e c u r s i v e

r e a l i z a b i l i t y . Much d e t a i l is omi t ted i n t h e hope of g iv ing a f e e l

f o r t he s u b j e c t . The main ca t egory - theo re t i c i d e a s axe expla ined

and i n t e r p r e t e d i n 5 84-6. I n p a r t i c u l a r we show why t h e no t ion of

a nega t ive formula a r i s e s n a t u r a l l y i n the theory of sheaves. I n

887-13, w e apply t h i s work t o a s tudy of a n a l y s i s i n t h e e f f e c t i v e

topos. W e show t h a t i n essence it i s c o n s t r u c t i v e r e a l a n a l y s i s

( i n t h e sense of Markov). I am g r a t e f u l t o P ro fes so r T r o e l s t r a f o r

some advice on t h i s t o p i c (I f i n d t h e publ i shed mater ia l unreadable:

Page 3: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 167

and i n p a r t i c u l a r f o r d e t e c t i n g an e r r o r i n an e a r l y d r a f t o f t h i s

paper. 5514-17 a r e concerned wi th f e a t u r e s of t h e e f f e c t i v e topos

where t h e power se t ma t t e r s : un i formi ty p r i n c i p l e s and p r o p e r t i e s of

j -ope ra to r s . The paper c l o s e s wi th some gene ra l remarks on the

mathematical s i g n i f i c a n c e o f t h e e f f e c t i v e topos .

F i n a l l y I would l i k e t o thank t h e o rgan ize r s o f t h e Brouwer

Centenary Conference f o r t h e oppor tun i ty t o p r e s e n t t h i s paper ( i n

such p l e a s a n t sur roundings : ) and t o apologize t o everyone f o r be ing

so long i n w r i t i n g it.

51 Recursive r e a l i z a b i l i t y .

Recursive r e a l i z a b i l i t y is based on t h e p a r t i a l a p p l i c a t i v e

s t r u c t u r e (IN ,.) where a s i n H J P 119803 w e w r i t e n.m = n(m) f o r t h e

r e s u l t of apply ing t h e n ' t h p a r t i a l r e c u r s i v e func t ion t o m. (This

saves on b racke t s compared wi th t h e n o t a t i o n { n l n . ) One can de f ine

a no t ion of A-abstraction i n ( I N ,.) i n t h e usua l way from t h e com-

b i n a t o r s , and w e w i l l use it f r e e l y i n what fo l lows , so t h a t ( f o r

example) Ax.x w i l l denote an index f o r t h e i d e n t i t y func t ion . We

a l s o t ake f o r convenience a r e c u r s i v e p a i r i n g func t ion

< , >: I N x I N - > IN; (n,m) --> <n,m>,

and l e t n1,n2 be ( r e c u r s i v e i n d i c e s f o r ) t h e corresponding unpai r ing

€unc t ions .

Recursive r e a l i z a b i l i t y i s usua l ly formula ted i n terms of t h e

no t ion

e r e a l i z e s 0

where e i s a n a t u r a l number and @ i s a sen tence o f (Heyt ing ' s )

a r i t h m e t i c . The c r i t i c a l c l a u s e s i n t h e i n d u c t i v e d e f i n i t i o n a r e

imp l i ca t ion

e r e a l i z e s 4 -> $ i f f f o r a l l n , i f n real izes @ t hen e ( n ) i s

de f ined and realizes $,

Page 4: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

168 JM.E. HYLAND

un ive r sa l q u a n t i f i c a t i o n

e r e a l i z e s Vn.$(n) i f f f o r a l l n , e ( n ) is def ined and r e a l i z e s

@ ( g ) Cg t h e numeral f o r n l .

The o t h e r i nduc t ive c l a u s e s a r e

and

e r e a l i z e s $ A $ i f f n (e) r e a l i z e s $ and n 2 ( e ) r e a l i z e s $,

o r

e r e a l i z e s @vJ, i f f e i t h e r a,(e) = 0 and n 2 ( e ) r e a l i z e s $

- 1

-

o r IT^ (e) = 1 and n 2 (e) rea l izes $,

falsity

no numbers rea l ize 1,

e x i s t e n t i a l q u a n t i f i c a t i o n

e rea l izes h . $ (n) i f f n 2 (e) r e a l i z e s @ ( r l (e) ) [ n l (e) t h e - - numerical f o r I T ~ ( ~ ) I .

F i n a l l y w e g i v e t h e i n i t i a l c l a u s e f o r e q u a l i t i e s between c losed

t e r m s

e r e a l i z e s s = t i f f both s and t denote e.

For a c a r e f u l t r ea tmen t of t h e r e a l i z a b i l i t y i n t e r p r e t a t i o n of

a r i t h m e t i c t h e r eade r may c o n s u l t T r o e l s t r a C19731. W e w i l l see i n

53, t h a t t h i s is t h e i n t e r p r e t a t i o n of a r i t h m e t i c w i t h i n t h e

e f f e c t i v e topos . For an account of t h e o r i g i n a l mot iva t ion see

Kleene 119731; it i s i n t e r e s t i n g t o t r y t o understand it i n t e r m s o f

t h e p re sen t paper.

Apparently Dana S c o t t f i r s t no t i ced t h a t r e a l i z a b i l i t y could be

understood "model-theoretically ' ' i n terms of t h e t ru th -va lues

{ e ( e real izes @ I .

t ru th -va lues , and so f o r each set X , a se t Zx of non-standard pre-

d i c a t e s on X.

This g ives us a set C = P ( l N ) of non-standard

W e w r i t e $ = ($,lx E X) and J, = (QX\x E X) f o r

elements of C" and can r e fo rmula t e our ear l ie r d e f i n i t i o n f o r t h e

p ropor t iona l connec t ives by d e f i n i n g ope ra t ions poin twise on Zx as

fo l lows :

Page 5: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 169

( $ A $ ) ~ = $ x ~ $ x = {<n,m>ln E $x and m E $,I, ( $ v $ l X = $ x v $ x = t < ~ , n > l n E $ x I u t < l , n > j n E

( $ + I ) ) ~ = $x+$x = {el i f n E $,, then e ( n ) i s def ined and

e ( n ) E

= t h e empty set . IX

The r eade r may a l s o l i k e t o have

TX = IN.

There is a r e l a t i o n lk

Cx by

of en ta i lmen t ( a pre-order ) def ined on each

I-,$ i f f n { ( $ -> $),lx E XI i s non-empty.

The soundness of t h e r e a l i z a b i l i t y i n t e r p r e t a t i o n of i n t u i - t i o n i s t i c p r o p o s i t i o n a l l o g i c i s t h e fo l lowing p ropos i t i on .

Propos i t ion 1.1. (Z i s a Heyting pre-algebra: as a category

t h e preorder has f i n i t e l i m i t s ( m e e t s ) , f i n i t e c o l i m i t s ( j o i n s ) and

i s Car t e s i an c losed (Heyting i m p l i c a t i o n ) .

Proof: The s t r u c t u r e is g iven e x p l i c i t l y i n t h e d e f i n i t i o n s above.

X

W e now in t roduce t h e a b s t r a c t no t ion of q u a n t i f i c a t i o n from

c a t e g o r i c a l l o g i c . For any map f : X -> Y of sets w e de f ine

s u b s t i t u t i o n a long f , f* : C y -> CX a s composition wi th f :

( f*$) , = $ f (x) f * i s a func to r ( i n fact a map of Heyting pre-a lgebras) from

( Z , I - y ) t o ( C , I-x ) and q u a n t i f i c a t i o n a long f i s given by t h e

a d j o i n t s t o f * . As shown i n HJP C19801 t h e s e are de f ined by

r i g h t a d j o i n t

Y X

( v f . + ) y = n t f ( x ) = y -> $ X I X E x l ,

l e f t a d j o i n t

( 3 f . $ ) y = U { f ( X ) = y A $ x l X E XI,

where T , i f f ( x ) = y ,

f f ( x ) = Y]l = ' - J I T l f ( X ) = yj = t L , otherwise,

i s t h e n a t u r a l i n t e r p r e t a t i o n a s a non-standard p r e d i c a t e of

Page 6: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

170 J.M.E. HYLAND

f ( x ) = y .

n a t i v e d e f i n i t i o n of t h e l e f t a d j o i n t , r l {$x l f (x ) = y} i s = a

d e f i n i t i o n of t h e r i g h t a d j o i n t un le s s f : X --> Y i s s u r j e c t i v e .

However usua l q u a n t i f i c a t i o n i s q u a n t i f i c a t i o n a long t h e obvious

pro jec t ion , and almost a l l p r o j e c t i o n s are s u r j e c t i v e , so t h i s

nuance w i l l cause t h e r eade r (and au tho r ) no f u r t h e r t r o u b l e .

Note t h a t whi le U { ~ $ ~ l f ( x ) = y} i s a s a t i s f a c t o r y a l te r -

The r eade r w i l l see t h a t what w e have j u s t descr ibed i s an

i n t e r p r e t a t i o n of i n t u i t i o n i s t i c p r e d i c a t e l o g i c : w e have s t anda rd

func t ions and sets, a (non-standard r e p r e s e n t a t i o n o f ) s t anda rd

e q u a l i t y and a c o l l e c t i o n of non-standard p r e d i c a t e s . W e a l s o have

a "gener ic p red ica t e" namely t h e i d e n t i t y i n E L .

a l l t h i s s t r u c t u r e i n t h e fo l lowing p ropos i t i on .

P ropos i t i on 1 . 2 . The toge the r wi th t h e f * and t h e i r a d j o i n t s

3f and Vf and t h e "gene r i c p r e d i c a t e " , form a t r i p o s on t h e ca tegory

of Sets ( i n t h e sense of H J P C19801.

Proof: See H J P C19801.

W e can encapsu la t e

In what w e have s a i d , w e have n o t needed t o d i s t i n g u i s h f o r -

mulae from t h e i r i n t e r p r e t a t i o n s , and w e w i l l cont inue t o b l u r t h i s

d i s t i n c t i o n as f a r as p o s s i b l e . ( W e w i l l use open f a c e b racke t s t o

i n d i c a t e an i n t e r p r e t a t i o n when necessary t o p reven t confus ion . )

W e say t h a t

$ E .Ex i s v a l i d i f f T I - x $ . By a d j o i n t n e s s w e have

9 E Zx i s v a l i d iff T l-lVX.$,

where X: X --> 1 is a unique map from X t o a one element se t . That

is, I$ i s v a l i d i f f Vx.I$(x) , t he un ive r sa l g e n e r a l i z a t i o n of $ is

v a l i d o r r e a l i z a b l e . W e w i l l use t h i s no t ion both t o desc r ibe and

s tudy t h e topos which w e can c o n s t r u c t on t h e b a s i s of ( 1 . 2 ) .

12 Desc r ip t ion of t h e e f f e c t i v e topos .

When cons t ruc t ing a topos from a t r i p o s a s i n H J P C19801, one

Page 7: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 171

must

(i) add new subobjects of t h e sets one has s t a r t e d with t o

represent t h e non-standard p red ica t e s , and

(ii) take quo t i en t s of t hese by the non-standard equivalence

r e l a t i o n s . This l eads t o the desc r ip t ion of t h e o b j e c t s of e f f ec -

t i v e topos. An ob jec t of t he e f f e c t i v e topos i s a set X with a

non-standard p red ica t e = on X x X such t h a t

symmetry x = y --> y = x

t r a n s i t i v i t y x = Y A Y = z -> x = z

are v a l i d . Note t h a t w e do no t have r e f l e x i v i t y : ( a s i s t h e case

for Heyting a r i t hme t i c ) t h e r e need be no uniform r e a l i z a t i o n of

(reason why) x = x. W e regard and w i l l w r i t e t he p red ica t e x = x

as an ex i s t ence p red ica t e , Ex, and a s a membership p red ica t e , x E X.

There is a use fu l discussion of the l o g i c of ex i s t ence p red ica t e s

i n Sco t t C19791.

Of course w e need t o consider a l l non-standard maps t o obtain

the e f f e c t i v e topos, and t o do t h a t w e a r e reduced t o considering

funct ional r e l a t i o n s . The m- from (X, =) t o ( Y , =) i n the

e f f e c t i v e topos a r e equivalence c l a s s e s of func t iona l r e l a t i o n s

where

( a ) G 6 cxXy i s a func t iona l r e l a t i o n i f f

r e l a t i o n a l G(x,y)AX = X ' A Y = y ' -> G(X' ,y ' )

s t r ic t G(x,y) -> EXAEY

single-valued G ( x , y ) ~ G ( x , y ' ) --> y = y '

t o t a l EX --> +y.G(x,y)

a re a l l v a l i d ,

(b) 6 i s equ iva len t t o H i f f

G(x,y) <--> H(xrY)

i s v a l i d . W e w i l l say t h a t G r ep resen t s t he map [GI: ( X , = ) - - > ( Y , = ) .

I t i s useful t o note t h a t i f G and H a r e both func t iona l r e l a t i o n s

Page 8: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

112 J.M.E. HYLAND

(from ( X I = ) t o ( Y , = ) , t hen t o show G and H e q u i v a l e n t , it s u f f i c e s

t o show t h a t an impl i ca t ion i n one d i r e c t i o n i s v a l i d .

Funct iona l r e l a t i o n s can be composed: i f G E Xxxy and H E C y x z

Also are func t iona l r e l a t i o n s , t hen s o is 3y.G(xly)AH(y,z) E C x x z .

= is a func t iona l r e l a t i o n from ( X , = ) t o i t s e l f . These g ive t h e

composition and i d e n t i t i e s , and so w e have a ca tegory . I n view of

( 2 . 1 ) , w e c a l l t h i s ca tegory t h e e f f e c t i v e topos and denote it by

Eff h e r e a f t e r .

Theorem 2 . 1 . Eff is a topos .

Proof: See H J P C19801 f o r d e t a i l s .

W e can ex tend t h e non-standard i n t e r p r e t a t i o n of 91 t o g ive an

account of t h e i n t e r n a l l o g i c of t h e ca tegory Eff. This goes a s f o r

t h e l o g i c of sheaves except f o r obvious modi f ica t ions t o d e a l wi th

t h e f a c t t h a t func t ions are (only) r ep resen ted by func t iona l rela-

t i o n s . A gene ra l account of t h e i n t e r n a l f i r s t - o r d e r l o g i c of

c a t e g o r i e s i s given i n Makkai and Reyes C19771, and accounts of t h e

h ighe r o rde r l o g i c of toposes can be found i n Fourman C19771 and

Boileau-Joyal C19811. A s t h e s e accounts make c l e a r , c a t e g o r i c a l

cons t ruc t ions can be de f ined by means of t h e i n t e r n a l l o g i c . Thus,

n o t on ly can (an ex tens ion o f ) v a l i d i t y i n t h e sense of 9 1 , be used

t o determine what i s t r u e i n Eff, b u t it can a l s o be used t o d e f i n e

c a t e g o r i c a l c o n s t r u c t s . (Now cont inuing t h e i n t e r p l a y , t h e s e c a t e g o r i -

c a l c o n s t r u c t s can then be used t o e s t a b l i s h f u r t h e r f a c t s about what

i s t r u e i n Eff.) W e now g ive some simple examples of t h e l o g i c a l

d e s c r i p t i o n of t h e s t r u c t u r e of Eff. (On a few occasions w e w i l l

need t o quote some more complicated f a c t s of t h e same kind . )

1) A map [GI: (X,=) --> (Y,=) i s monic i f f

G ( x , y ) ~ G ( x ' , y ) --> x = X'

i s v a l i d .

A subob jec t of (X,=) can always be r ep resen ted (though n o t

uniquely) by a canonica l monic of t h e form

Page 9: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos

[ = ' I : (X,=') --> (X,=)

where

I[ x = ' x q = A ( X ) A ux = x l n

f o r some A E Zx s t r i c t and r e l a t i o n a l f o r (X,=). Thus subobjec ts

173

always arise by r e s t r i c t i n g t h e membership p r e d i c a t e whi le (as f a r

as p o s s i b l e ) l eav ing t h e e q u a l i t y a lone .

2 ) Given two maps CG1,CHI: (X,=) --> (Y , =) , t h e i r e q u a l i z e r is

represented by t h e canonica l monic obta ined from t h e s t r ic t and

r e l a t i o n a l

3y .G(x ,y)hH(xfy) E E x .

The cons t ruc t ion of o t h e r f i n i t e l i m i t s i s analogous.

The diagram

W , = ) - [ G " > (Z,=)

[H'I 1 1 [HI V

( X r = ) - > ( Y r = ) [GI

i s a pullback i f f CHIo[G'I = CGIo[H'I, (CG'1,CH'I):W -> Z X X i s a

monic and

G ( x , y ) ~ H ( z , y ) --> 3W.G' ( W , Z ) A H ' (w,x)

is v a l i d . The cond i t ion t h a t o t h e r diagrams g ive f i n i t e l i m i t s can

be expressed s i m i l a r l y i n the l o g i c .

3 ) A map [GI: (X,=) --> (Y,=) is s u r j e c t i v e i f f

Ey -> 3x.G(xry)

i s v a l i d .

A q u o t i e n t can always be r ep resen ted as

[ - I : ( X r = ) --> ( X v - )

where - i s s t r i c t r e l a t i o n a l f o r (X,=) and such t h a t

"- is an equiva lence r e l a t i o n on (X,=)"

i s v a l i d . Thus q u o t i e n t s are a ma t t e r of ex tending t h e e q u a l i t y

r e l a t i o n and l eav ing t h e membership p r e d i c a t e a lone .

Page 10: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

174 J.M.E. HYLAND

W e can now show t h a t any ob jec t ( X I = ) of Eff i s a quo t i en t of a

subobject of an "ordinary set" , j u s t i f y i n g the explanat ion a t t h e

s t a r t of t h i s s e c t i o n . For a s e t X w e l e t AX ( a s i n 5 4 ) be t h e

ob jec t of Eff with underlying set X and (non-standard r ep resen ta t ion

o f ) standard equa l i ty .

Proposit ion 2 . 2 . Any ob jec t ( X , = ) of Eff i s a quo t i en t by = of the

subobject EX of AX obtained from t h e ex i s t ence p red ica t e of ( X , = ) . Proof: Obvious i n view of 1) and 3) above.

- Note. W e have s t a r t e d using open f ace brackets t o ensure r e a d a b i l i t y

( e spec ia l ly i n connection with e q u a l i t y ) , a s promised i n §I. W e

a l s o abuse no ta t ion and w r i t e X f o r ( X , = ) where context makes the

meaning obvious.

5 3 . Some o b j e c t s and maps i n Eff. - W e can e a s i l y descr ibe a terminal ob jec t 1 i n Eff. I n view of

5 2 , 1 is ( I * } , = ) where { * I i s a s i n g l e t o n , and

[ I* = * I = T

Of course any p equ iva len t t o T i n Z { * ' = Z, t h a t i s , any non-empty

p would do a s the value

sec t ions of an a r b i t r a r y o b j e c t (Y,=) of Eff , t h a t is t h e maps from

1 t o (Y,=) . Since { * I i s a s i n g l e t o n , such maps a r e represented by

degenerate func t iona l r e l a t i o n s G E Z y , such t h a t

1 * = * I . W e now c a l c u l a t e t h e g loba l

G ( Y ) A Y = y ' --> G(y')

G(y) --> Ey

G(y)AG(y') --> y = Y '

3y.G(y)

a r e a l l v a l i d .

G(yo) i s non-empty. The r e l a t i o n a l and single-valued condi t ions

imply t h a t

The t o t a l condi t ion t e l l s us t h a t f o r some y , yo say,

G ( Y o ) --> (G(y) <--> yo = y)

and hence ( s ince G(yo) i s non-empty)

Page 11: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 175

G(y) <--> yo = y

a r e v a l i d . C lea r ly i f [I yo = y l l is non-empty, then

Yo = Y <-> Y 1 = Y

i s v a l i d . W e deduce a t once t h e fo l lowing c h a r a c t e r i z a t i o n .

Propos i t ion 3.1. Each map [ G I : 1 -> (Y,=) determines and i s com-

p l e t e l y determined by [y I G(y) non-empty}, which is an equiva lence

c l a s s f o r t h e ( p a r t i a l ) equiva lence r e l a t i o n

" U y = y ' l i s non-empty".

Conversely any such equiva lence class de termines a map from 1 t o

(Y,=).

F i n i t e c o l i m i t s i n Eff a r e ha rd t o g e t used t o because f o r a

s t a r t coproducts a r e odd: t he r e a l i z a b i l i t y i n t e r p r e t a t i o n of d i s -

junc t ion is very r e s t r i c t i v e . I n p a r t i c u l a r , t h e coproduct 2 of 1

with i t s e l f i s n o t t h e obvious o b j e c t A2 wi th s t anda rd e q u a l i t y

(see 54). L e t us look a t maps from A2 t o an a r b i t r a r y o b j e c t IY,=)

of Eff. Suppose G ( i , y ) r e p r e s e n t s such a map (where 2 = { O , l } ) .

Then s i n c e EO = E l = T , t h e t o t a l cond i t ion t e l l s us t h a t t h e r e a r e

yo,yl such t h a t G(O,yo)nG(l,yl) i s non-empty. Arguing a s f o r t h e

te rmina l o b j e c t w e f i n d t h a t

G ( i , y ) <--> yi = y

i s v a l i d . However [GI does n o t correspond simply t o a p a i r o f

equivalence c l a s s e s i n IylEy non-empty?:the union of t h e ex i s t ence of

t h e two equiva lence c l a s s e s must i n t e r s e c t n o n - t r i v i a l l y , and t h i s is

a r e a l r e s t r i c t i o n .

I n f a c t t h e o b j e c t 2 i n Eff can be r ep resen ted as ( 2 , = ) where

EO = 101, E l = {l}, [li = j]l = E i n E j .

(Of course any p,,pl wi th p npl empty would do as t h e va lues EO, E l . )

An argument a s above shows t h a t maps from 2 a r e p a i r s of maps from 1.

Note a l s o t h a t t h e only maps from A2 t o 2 are cons t an t ( t h a t i s ,

f a c t o r through 1). There is an obvious monic from 2 t o A2. I n 516

0

Page 12: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

176 J.M.E. HYLAND

we w i l l show t h a t t h e whole s t r u c t u r e of Eff depends on 2 --> A2

n o t be ing i s o , i n t he sense t h a t t h e topology i n v e r t i n g 2 --> A2

c o l l a p s e s Eff back t o Sets. Since 2 is n o t A2, w e would ha rd ly expec t t h e n a t u r a l number

o b j e c t N i n Eff t o be A m . I n f a c t it is t h e o b j e c t ( I N , =) where

En = i n ] , [ n = m ] = EnnEm.

There are maps 0: 1 --> IN and s : IN-> IN i n Eff r ep resen ted res-

p e c t i v e l y by Go and Gs where

Go(*,n) = IOjnIn) and G,(n,m) = In+l)n{m).

P ropos i t i on 3 . 2 . IN t o g e t h e r w i th 0: 1 --> IN and s: IN-> IN i s a

n a t u r a l number o b j e c t i n Eff . Proof: Suppose t h a t w e are g iven maps a : 1 --> ( X , = ) and

g: ( X , = ) --> ( X , = ) r ep resen ted r e s p e c t i v e l y by Ga E C x and

G E 1'''. W e can d e f i n e r e p r e s e n t a t i v e s Gn f o r gn i n d u c t i v e l y by 4 9

( x , x l ) = ~ X ~ ~ . G ~ ( X , X ~ ~ ) A G ( X ~ * , X ' ) . Gn+l G O ( X , X I ) = U X = x 8 n ,

g 9 9 9 Now w e can d e f i n e a func t ion f : IN-> ( X , = ) r ep resen ted by

Gf (n ,x ) = EnA3x'.Ga(x')AGn(x' ,x). 9

W e c la im t h a t

f i fl (X,=)->(X,=) a

9

commutes. This amounts t o showing t h a t

G (x ) <--> 3n.Go(*,n)AGf(n,x)

and

3x ' .Gf ( n , x ' ) A G ( x ' , x ) <--> 3m.Gs (n,m) A G f (m,x) 9

a r e both v a l i d . These can both be e s t a b l i s h e d by use of e lementary

l o g i c .

It remains t o show t h a t f is unique such t h a t ( * ) commutes. SO

suppose t h a t f ' r ep resen ted by Gf I i s ano the r such map.

l o g i c w e see r e a d i l y t h a t

By use of

Page 13: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 177

G (0 ,x ) <-> Gf I (0 ,x ) f i s v a l i d and t h a t

G f ( n + l , x ) <-> 3x ' .Gf(n ,x ' )AG ( x ' , x )

G f , (n+l,x)<-> 3 x ' . G f , ( n , x ' ) AG (x ' , x ) g

g a r e both v a l i d . But i n t e r m s o f t h i s d a t a w e can d e f i n e , by primi-

t i v e r e c u r s i o n , a p a r t i a l r e c u r s i v e f u n c t i o n uniformly mapping

Gf ( n ,x ) t o G f , ( n , x ) , and t h i s i s enough t o show t h a t f = f ' ,

Remark. S ince q u a n t i f i c a t i o n i n ou r l o g i c (see Fourman-Scott C 1 9 7 9 1

and S c o t t C19791) i nvo lves t h e e x i s t e n c e p r e d i c a t e , w e see a t once

on the b a s i s of ( 3 . 2 ) t h a t t h e r e a l i z a b i l i t y i n t e r p r e t a t i o n co r re s -

ponds t o the l o g i c of t h e n a t u r a l number o b j e c t i n Eff. Corol la ry 3.3. A s en tence of Heyting a r i t h m e t i c is r e c u r s i v e l y

r e a l i z e d i f f it i s t r u e of t h e n a t u r a l number o b j e c t i n Eff. A l l t he s p e c i f i c o b j e c t s w e have looked a t so f a r have been

o b j e c t s (X,=) where [ x = x ' J non-empty impl i e s x = x ' ( i n X).

Indeed a l l t h e o b j e c t s w e cons ide r u n t i l 1 1 4 w i l l be (isomorphic t o )

ones o f t h i s s o r t (see 5 6 f o r a d i scuss ion of what t h e cond i t ion

means). I t is a s w e l l t o have an example of an o b j e c t n o t o f t h i s

form. The most obvious example i s t h e subob jec t c l a s s i f i e r i n Eff ( t h a t i s , t h e o b j e c t of t r u t h v a l u e s ) . A s i n d i c a t e d i n HJP C19801

t h i s is t h e o b j e c t (X ,<->), t h a t i s , t h e se t X = P(JN ) wi th

e q u a l i t y given by t h e non-standard b i - impl i ca t ion . W e leave it a s

an easy e x e r c i s e t o show t h a t (I,<-->) i s n o t isomorphic t o any

o b j e c t (X,=) where I[x = x ' l non-empty impl i e s x = x ' . (Show

f i r s t t h a t X would have t o have j u s t two e lements . ) As f u r t h e r

examples it i s n a t u r a l t o cons ide r (EA,<->) where now e q u a l i t y is

the poin twise b i - imp l i ca t ion . These are e s s e n t i a l l y t h e o b j e c t s ou t

of which w e cons t ruc t ed Eff i n t h e f i r s t p l a c e ; they a r e i n f a c t t h e

power sets of t h e o b j e c t s AX (see 9 4 ) .

Page 14: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

178 J.M.E. HYLAND

5 4 . The inc lus ion of t h e category of sets i n the e f f e c t i v e topos.

In t h e l a s t s e c t i o n w e s a w glimpses of a func to r A from t h e

category Sets of sets t o Eff, t h e e f f e c t i v e topos.

Def in i t i on . For a se t X, de f ine AX t o be ( X , = A x ) l where

is t h e n a t u r a l i n t e r p r e t a t i o n of t he e q u a l i t y i n Sets. For a map

f : X -> Y i n Sets, def ine A f : AX -> AY t o be the map represented

by the func t iona l r e l a t i o n ,

This d e f i n i t i o n can be made whenever w e cons t ruc t a topos from a

t r i p o s (see P i t t s [ 1 9 8 l l ) , and w e always have o w f i r s t r e s u l t .

Proposi t ion 4 . 1 . A : Sets -> Eff is a Cartesian func to r ( t h a t i s ,

func to r preserving f i n i t e l i m i t s ) . Proof: F u n c t o r i a l i t y i s obvious. That A i s Cartesian follows

e a s i l y from t h e way f i n i t e l i m i t s a r e def ined l o g i c a l l y i n 53. For

d e t a i l s see H J P C19801 o r P i t t s C19811.

The next r e s u l t i s a general f e a t u r e of r e a l i z a b i l i t y toposes .

- Propcsi t ion 4 . 2 . A : Sets -> Eff i s f u l l and f a i t h f u l .

Proof: Suppose t h a t f , g : X --> Y i n Sets, and t h a t

= y n <--> u g ( x ) = y~

i s va1i.d. Then I [ f (x ) = y l = T i f f f g ( x ) = yl] = T whence

f ( x ) = y i f f g ( x ) = y: t hus f = g. This shows t h a t A i s f a i t h f u l .

To show t h a t A i s f u l l , l e t G E Z x x y be a func t iona l r e l a t i o n from

AX t o A Y . The r e l a t i o n a l and s t r i c t cond i t ions axe au tomat i ca l ly

s a t i s f i e d , t h e s ingle-valued condi t ion implies t h a t f o r given x

t h e r e is a t most one y with G(x,y) non-empty ( c o n s t r u c t i v e l y , in-

h a b i t e d ) , and t h e t o t a l condi t ion imples t h a t t h e r e is a t least one

such y f o r given x. Thus w e have g: X -> Y such t h a t G(x,y) i s

non-empty i f f g(x) = y . Then c l e a r l y

~ ( x , y ) --> u g w = yn

Page 15: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos

i s v a l i d : an index f o r t h e i d e n t i t y realizes it. But t h e t o t a l

condi t ion becomes

E x -> G ( x , g ( x ) )

i s v a l i d , whence

I g ( x ) = y n -> G

i s v a l i d . Thus G r e p r e s e n t s t h e map Ag.

Remark. ndr6 Joya l has po in t ed o u t t h a

179

X I Y )

This shows t h a t A is f u l l .

A i s analogous t o t h e

Yoneda embedding: it i s Car t e s i an , f u l l and f a i t h f u l , and (so) pre-

s e rves exponen t i a t ion . But I do n o t understand t h e f o r c e of t h i s

analogy.

The main r e s u l t o f t h i s s e c t i o n i s ano the r gene ra l f e a t u r e of

r e a l i z a b i l i t y toposes . Reca l l t h a t i n (3.1) w e showed i n e f f e c t

t h a t t h e g l o b a l s e c t i o n f u n c t o r on Eff i s n a t u r a l l y isomorphic t o

r : Eff --> Sets def ined by

(i) r ( X , = ) = {XIEX i s non-empty]/_ where x - x ' i f [Ix = x i ] '

i s non-empty;

(ii) i f G i s a f u n c t i o n a l r e l a t i o n from ( X , = ) t o (Y,=)

r e p r e s e n t i n g g , t hen

T(g) ([XI) = {ylG(x ,y) i s non-empty} where [XI denotes

t h e equiva lence c l a s s o f x .

r i s a conc re t e ve r s ion of t h e g loba l s e c t i o n f u n c t o r , w i th which

w e can work, even c o n s t r u c t i v e l y : s t a r t i n g from an a r b i t r a r y base

topos E , T ( X , = ) s t i l l makes sense a s t h e i n t e r p r e t a t i o n i n E of

" the set o f maps from 1 t o ( X I = ) " . (Of course , "non-empty" must be

r ep laced by " inhab i t ed" ) . Theorem 4.3. A is t h e d i r e c t image f u n c t o r of a geometric morphism,

whose i n v e r s e image f u n c t o r i s r . Proof: The g l o b a l s e c t i o n f u n c t o r i s always Car t e s i an : a l t e r n a t i v e l y ,

r a s de f ined is C a r t e s i a n by t h e l o g i c a l c o n s t r u c t i o n of f i n i t e

l i m i t s de sc r ibed i n 53. So w e concen t r a t e on the a d j o i n t n e s s .

Page 16: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

180 J.M.E. HYLAND

W e d e f i n e t h e u n i t o f t h e ad junc t ion ny : Y --> ATY f o r (Y,=) i n Eff by t h e f u n c t i o n a l r e l a t i o n

(y , [y ' l ) - -> U{Eyly E Cy ' l}={yy ' i f y - y ' , o the rwise .

Now l e t G be a f u n c t i o n a l r e l a t i o n from (Y,=) t o AX. The t o t a l and

s ingle-va lued c o n d i t i o n s imply t h a t i f Ey i s non-empty, t hen t h e r e

i s a unique x E X wi th G(y ,x ) non-empty. The r e l a t i o n a l cond i t ion

impl i e s t h a t i f y = y ' ] is non-empty, then w e g e t t h e same x

f o r y ' a s f o r y . Thus w e have a wel l -def ined map g: rY --> X . By

l o g i c , t h e composite A ( 9 ) o r l y i s r e p r e s e n t e d by

~ ( y , x ) = U I E ~ A ug(cy11) = x n ~y E iy11 E ~ Y I .

By t h e s t r i c t c o n d i t i o n

G(y,X) --> EY

i s v a l i d ; s o s i n c e G(y ,x) i s non-empty i f f x = q(Cy1) and s i n c e

c l e a r l y

Ey --> H ( y , g ( c y l ) )

is v a l i d , w e deduce t h a t

G(y ,x) --> H(y,x)

i s v a l i d . S ince both G and H a r e f u n c t i o n a l r e l a t i o n s , t h i s shows

( a s remarked i n 5 2 ) t h a t t h e y r e p r e s e n t t h e same f u n c t i o n , and w e

have our f a c t o r i z a t i o n

[GI = A(q) o n y .

I t remains t o show t h a t q i s unique wi th t h i s p r o p e r t y . But i f

9 ' : TY --> X is such t h a t

G ( Y , X ) <--> U{EYA u g i ( c y i i ) = x n Iy E c y ' i E TY}

i s v a l i d , t h e n G ( y , g ' ( C y l ) ) i s non-empty, so t h a t g ' = q. This

completes t h e p roof .

Remark. I t is an easy c o r o l l a r y of t h e proof of ( 4 . 3 ) t h a t maps

( Y , = ) --> A X i n Eff have a s imple canon ica l r e p r e s e n t a t i v e . L e t

g: TY --> X cor respond under t h e ad junc t ion t o a map IY,=) -> AX.

Then t h i s l a t t e r map i s r e p r e s e n t e d by t h e f u n c t i o n a l r e l a t i o n

Page 17: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

T h e effective topos 181

(y ,x) --> UiEylg ( [y l ) = XI = { y : W e can now i n d i c a t e how category theo ry may be appl ied t o s tudy

r e a l i z a b i l i t y . (4 - .2 ) and ( 4 . 3 ) t oge the r say t h a t A : Sets --> Eff

i s an inc lus ion of toposes (see Johnstone C19771) so t h a t Sets i s

j-sheaves on Eff f o r a s u i t a b l e topology j . W e g ive an i d e n t i f i -

cat ion of j , which depends on the f a c t t h a t Sets has c l a s s i c a l l o g i c ,

Proposi t ion 4 . 4 . The topology j such t h a t E f f . Sets i s t h e double

negation topology

Proof: In t h e f i r s t p l ace , Sets is dense i n Eff s ince A

preserves t h e i n i t i a l o b j e c t (see (8.1) ) ; so j is a t most 71 ( t h e

g r e a t e s t dense topology) . But Sets i s boolean, and from t h i s

it follows t h a t j must be T-, .

-J

W e can now desc r ibe what the use of "classical o b j e c t s " i n

in tu i t i on i sm amounts t o i n our context : s i n c e they are defined by

l i b e r a l use of 7-,, they are when i n t e r p r e t e d i n Eff, t h e o b j e c t s i n

the image of A . Thus A ( = ) should be regarded a s t h e world of

c l a s s i c a l mathematics w i th in Eff.

15. Basic f a c t s from t h e l o g i c of sheaves.

While the material presented i n this s e c t i o n is i m p l i c i t i n t h e

topos t h e o r e t i c l i t e r a t u r e , it can n o t be found i n the form w e r e -

quire . With Grothendieck toposes one has t y p i c a l l y a subtopos E

of a topos E which one understands ( E i s usua l ly a func to r category)

and one r e q u i r e s r e s u l t s which enable one t o d i scuss E i n terms of

E and t h e topology j . For us however t h e s i t u a t i o n i s d i f f e r e n t .

It is the topos E ( t h a t is Sets) which w e understand and w e wish t o

obtain information about E ( t h a t i s Eff) i n terms of E and j .

j

j

j

j W e p r e sen t t h e material i n the following general context . E is

a topos with a topology j , E . i s t h e f u l l subcategory of E cons i s t ing

of j-sheaves and L: E -> E j i s t h e s h e a f i f i c a t i o n func to r l e f t J

Page 18: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

182 J.M.E. HYLAND

a d j o i n t t o the inc lus ion E -> E . W e g ive t h e b a s i c d e f i n i t i o n s

i n a number of u se fu l equ iva len t forms which a r e i m p l i c i t e i t h e r i n

Johnstone C19771 o r i n Fourman-Scott 119791.

Def in i t i on . An ob jec t F of E i s j -separated i f f any of t he following

equ iva len t cond i t ions is s a t i s f i e d :

(i) f o r any j-dense monic m: Y' >--> Y and maps f , g : Y --> F

with fm = gm, w e have f = g;

(ii) t h e u n i t nF : F --> L(F) of t h e adjunct ion i s monic;

(iii) E I= V f , f ' E ~ . j ( f = f ' ) --> ( f = f ' ) .

A subobject (monic) A >--Z E of an o b j e c t E of E i s j - c losed

i f f any one of t h e following equ iva len t cond i t ions i s s a t i s f i e d :

(i) i f a: E --> R c l a s s i f i e s A >--> E , then j a = a ;

(ii) the commutative square A --> L ( A ) i s a pul lback;

E --> L ( E j

(iii) E I= V e . j ( e E A) --> e E A .

Of t hese d i f f e r e n t formulat ions, (i) i s the t r a d i t i o n a l category

t h e o r e t i c one, (ii) i s p a r t i c u l a r l y use fu l f o r understanding Eff and (iii) i s t h e l o g i c a l formulation ( t r e a t i n g j a s a p ropos i t i ona l

ope ra to r ) . I t i s obvious from t h e d e f i n i t i o n s t h a t F is j -separated i f f

t h e e q u a l i t y on F i s j -c losed, and t h a t a subobject of a j -separated

o b j e c t i s i t s e l f j -separated. W e c o l l e c t some f u r t h e r f o l k l o r i c

f a c t s about t hese not ions i n the nex t theorem.

Theorem 5 . l . ( a ) I f E and F are j -separated, then SO i s EXF. A l s o

n E x F : ExF --> L(ExF) = L(E)xL(F) i s n E X n F .

( b ) I f F i s j -separated, then so i s FE f o r any E . A l s o

t he composite of nFE: FE --5 L(FE) with t h e n a t u r a l map

L ( F ~ ) %> L ( F ) L ( E ) followed by t h e isomorphism L(F)nE: L ( F ) L ( E )

-> L ( F ) E is the monic q F : FE --> L ( F ) E , and t h e eva lua t ion map E

Page 19: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 183

E F X E --> F i s obtained by f a c t o r i n g evo(aon Xn ) through qF . FE

( c ) If c >--> F i s j -c losed and a: E --> F then Cr*(C) >--> E

i s j-closed. A l s o L ( a * ( C ) ) = L ( a ) * ( L ( C ) ) .

( d ) I f A >--> E and B >--> E are j-closed then SO is

AAB >--> E. A l s o L(AAB) = L ( A ) A L ( B ) .

(e) I f B 2-> E is j -c losed then so is ( A --> B ) >--> E f o r

any A >--> E. Also L ( A --> B) = L ( A ) --> L ( B ) .

( f ) I f A >--> E i s j-closed and a: E --> F then Va .A >--> F i s

j-closed. A l s o L ( V a . A ) = t l L ( a ) . L ( A ) .

(9) I f R >--z ExE i s a j-closed equivalence r e l a t i o n on E l

then the q u o t i e n t E/R i s j -separated. Also the image ( o r s u r j e c t i v e

monic) f a c t o r i z a t i o n of

E- > L ( E ) - > L(E) /L (R) nE

i s

E- > E/R --> L ( E / R ) = L ( E ) / ~ ( ~ ) . n E/R

Proof: A l l t r i v i a l by t h e l o g i c of j -operators (sketched a t t h e end

of Fourman-Scott C19791) . Category t h e o r e t i c proofs are ( impl i c i t )

i n Johnstone C19771.

L e t us now exp la in why w e a r e i n t e r e s t e d i n c losed subobjects .

Our understanding of Grothendieck toposes rests on the f a c t t h a t

inverse image func to r s preserve coherent l o g i c ( t h a t is A , v , ~ ) . But

t h e inc lus ion of Sets i n Eff i s i n the wrong d i r e c t i o n i f W e wish t o

see some of t h e l o g i c of Sets preserved i n Eff. In general a d i r e c t

image func to r preserves l i t t l e , bu t w e can g e t r a t h e r s t rong r e s u l t s ,

when dea l ing with inc lus ions E --> E l by r e s t r i c t i n g a t t e n t i o n t o

j -c losed subob jec t s . This is s i g n i f i c a n t because a j -c losed subobject

A >--> E "agrees with i t s meaning i n E " i n t h e sense t h a t

nE*(LA) = A .

j

j

(This i s vers ion (ii) of the d e f i n i t i o n . )

Page 20: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

184 J.M.E. HYLAND

Given an i n t e r p r e t a t i o n of t h e atomic formulae of a f i r s t

order language i n E w e g e t

(i) an i n t e r p r e t a t i o n U $ 1 of an a r b i t r a r y formula i n E , and (ii)

by applying L an i n t e r p r e t a t i o n of t h e atomic formula i n E and

hence an i n t e r p r e t a t i o n [ $ I of an a r b i t r a r y formula i n E .

Clear ly i f II $ 1 i s a subobject of E , then B $ i s a subobject of

L ( E ) . W e a r e i n t e r e s t e d i n when I[ $ ]I "agrees with t h e in t e rp re -

t a t i o n II $ ]I i n E " i n t h e sense t h a t

j

3 '

j

j j n E * ( 1 4 n j) = u $ n .

The r e l e v a n t d e f i n i t i o n i s of a form fami l i a r from T r o e l s t r a C19731.

Def in i t i on . I n a f i r s t o rde r language, t h e negative formulae ( o r

formulae i n the negat ive fragment) a r e those b u i l t up from atomic

formulae using A , -> ,V.

Theorem 5 . 2 . I f an i n t e r p r e t a t i o n of a f i r s t order language i n E

i n t e r p r e t s t h e atomic formulae a s j-closed subobjects and $ i s a

negativeformula with I $ 1 >-> E , then

r l E * ( v ~ n j ) = ~ $ 1 .

Proof: Induction on the complexity of $ using (5.1) (c) (d) (e) and

( f ) .

Remarks 1) W e can only have e q u a l i t y f o r j-separated o b j e c t s .

2) As $ n i s j -c losed, nE*( I$ 1 j ) = f $ ]I i s equ iva len t

to n + n j = L ( u $ D ) .

3) The r e s u l t i s j u s t a consequence of the " j - in t e rp re t a t iod '

of t he l o g i c of E. For negative formulae w e a r e reading it no t a s a

p re sc r ip t ion f o r de r iv ing the l o g i c of E from t h a t of E , bu t a s t h e

statement t h a t the l o g i c of E agrees with t h a t of E j

j '

16 Separated o b j e c t s and closed subobjects i n Eff. In t h i s s e c t i o n , we desc r ibe what (5.1) means f o r t he p a r t i c u l a r

case when E i s Eff and E

topology. ( I n f a c t w e do n o t use t h i s l a s t f a c t , so t h a t t he

i s Sets so t h a t j i s t h e double negation j

Page 21: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 185

mater ia l r e l a t i v i z e s t o an a r b i t r a r y base topos E . ) . W e w i l l say

t h a t an ob jec t of Eff is separated when it is j -separated and t h a t

a subobject of an ob jec t is closed when it i s j -c losed.

Proposit ion 6 . 1 . An ob jec t of Eff is separated i f f it is isomorph

t o one of t he form ( X , = ) where fi x = x ' 1 non-empty implies x = x '

Proof: By version (ii) of t h e d e f i n i t i o n of j -separated, w e see

t h a t i f an ob jec t i s sepa ra t ed , it is a subobject of some A X . But

J

C

any canonical monic i n t o A X i s of t he r equ i r ed form. Conversely any

ob jec t of t he required form is a subobject of a AX Cthe obvious map

is monic), and subobjects of separated o b j e c t s a r e separated.

Defini t ion. An ob jec t ( X , = ) of Eff, where 1[ x = x ' i non-empty

implies x = x', i s a canonical ly separated ob jec t of Eff.

(Such an ob jec t i s completely determined by t h e values [Ix E X i f o r

each x i n X, and i s e s s e n t i a l l y ( t h a t i s , modulo t r i v i a l coding)

given a s a canonical monic i n t o A X . )

Proposit ion 6 . 2 . I f ( X , = ) and (Y,=) a r e canonical ly separated, then

so i s t h e usual product ( X x Y , = ) where

u ( x ,y ) = ( x l , y ~ ) n = IIX = x l n A uy = y l n . Proof: Immediate from (5.1) ( a ) and the d e f i n i t i o n of t he product of

maps i n the l o g i c .

The case of funct ion spaces i s more complex than t h a t of products.

Since t h e general desc r ip t ion of a funct ion space (see HJP C19801 i s

too clumsy, w e must use (5.1) (b) t o cons t ruc t a s u i t a b l e represen-

t a t i o n .

Proposit ion 6 . 3 . L e t ( Y , = ) and (Z,=) be o b j e c t s of Eff with (Z,=)

canonically separated.

taken t o be t h e canonical ly separated o b j e c t (rZrY,=) where ( taking

r ( z , = ) 5 2)

Then the funct ion space (Z,=) ("=) may be

I n = II V ~ ~ Y . ~ ( C ~ I ) = ~ I ( c Y n n = n~ II Ey-->a(Cyl)=a' ( c y i ) i l y c ~ i

and where t h e evaluat ion map is represented by t h e func t iona l r e l a t i o n

Page 22: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

186 J.M.E. HYLAND

I [ E ~ A EY A a ( c y 1 ) = z n . Proof:

i n t he l o g i c by

(5.1) (b) g ives us a monic from (Z,=) ("=) t o rZrY def ined

uvY Y . ~ C C ~ I ) z n which i s equ iva len t t o t h e formulae given a s CZ,=) i s canon ica l ly

sepa ra t ed . The r e p r e s e n t a t i o n of t h e eva lua t ion map fo l lows from

t h e d e f i n i t i o n i n t h e l o g i c of t h e map descr ibed i n (5 .1) (b ) by

elementary l o g i c .

Remark. I f f o r every y i n Y , Ey i s non-empty (and w e may a s w e l l

d i s r e g a r d t h e o t h e r s ) , then t h e fo l lowing a l t e r n a t i v e r e p r e s e n t a t i o n

of t h e func t ion space is canon ica l ly sepa ra t ed : IZy ,=) where

a = a l n = n{ I I ~ = ~ ~ - - > ~ ( ~ ) = ~ ~ ( ~ ~ ) n l y , y l E ~ ~

and where t h e eva lua t ion map i s r ep resen ted a s above. (We g e t t h i s

a l t e r n a t i v e r e p r e s e n t a t i o n by cons ide r ing t h e obvious map from

r Z ( y l = ) t o Z E Y l where EY i s t h e canonica l subobjec t of AY of which

( Y , = ) is a q u o t i e n t , )

Ea is non-empty, w e can cont inue t h i s process and o b t a i n a simple

d e s c r i p t i o n of i t e r a t e d func t ion spaces of s epa ra t ed o b j e c t s . W e

cons ider t h i s f u r t h e r i n 5 5 7 and 11.

Then i f w e d i s r ega rd those a i n Z y such t h a t

W e nex t cons ide r c losed subob jec t s i n Eff. Propos i t i on 6 . 4 . A subob jec t of an o b j e c t (X,=) of Eff i s c losed

i f f it i s r ep resen ted by a canonica l monic determined by A E Cn of

f o r s o m e A 5 I ' ( X , = ) .

( I t does no harm t o l e t A denote t h e subse t of r ( X , = ) , t h e canonica l

monic as de f ined and t h e c losed subob jec t which it r e p r e s e n t s ) .

Proof: By ve r s ion (ii) of t h e d e f i n i t i o n of j - c losed , a c losed sub-

o b j e c t of (X,=) must be of t h e form rlx (AA) f o r some A TIXI=) . -1

Page 23: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 187

But what w e have desc r ibed i s e a s i l y seen t o be e q u i v a l e n t t o t h e

d e f i n i t i o n of nx (AA) i n t h e l o g i c .

Def in i t i on .

-1

A monic of form (X,=') - [="> (X,=) where

[x=xD, i f [XI E A , I x = ' x ' ] = U { I [x=x ' JJ IlxleA} = { I , otherwise,

f o r some A 5 T ( X , = ) i s a canon ica l c losed monic. ( ( 6 . 4 ) shows

e s s e n t i a l l y t h a t t h e c losed subob jec t s a r e j u s t those r ep resen ted

by canon ica l c losed monics) .

Remark. On many occas ions it i s more n a t u r a l t o d i s r e g a r d i n (X,=')

the x which a r e n o t i n A. W e s h a l l s u i t terminology t o need and

r e f e r t o t h i s mod i f i ca t ion a l s o a s a canon ica l c losed monic. Note

t h a t t h e no t ion becomes p a r t i c u l a r l y simple i n case (X,=) i s

canon ica l ly s e p a r a t e d , a s t hen w e may t a k e A 5 X ( t ak ing

r ( X , = ) 5 X a g a i n ) .

We now say what (5 .1) ( c ) , ( d ) , ( f ) mean f o r t he e f f e c t i v e topos .

P ropos i t i on 6 . 5 . L e t A >-> (X,=) and B >-> (X,=) be subob jec t s

of (x,=), C >--> (Y,=) a subob jec t of (Y,=) and [GI: (X,=) --> (Y,=)

a map i n Eff. I f C i s a canon ica l c losed monic (de f ined from C 5 r ( Y , = ) ) , then

[ G l - l ( C ) i s t h e canon ica l c l o s e d monic de f ined from

( r ( G ) - l ( C ) = {Cxl l{ylG(x ,y) non-empty) E C).

I f A , B are canon ica l c l o s e d monics (de f ined from A , B 5 r ( X , = ) ) , then

AAB i s t h e canonica l c losed monic de f ined from AnB. I f B i s a

canonica l c losed monic (de f ined from B 5 r ( X , = ) ) , then A -> B i s

t h e canon ica l c l o s e d monic de f ined from

TA -> B = {Cxll i f [ X I E rA then [XI E B},

and VCG1.B is t h e canon ica l c losed monic de f ined from

V r ( C G 1 ) .B = {Cyll i f G(x,y) non-empty then [-XI E B}.

Proof: (5 .1) t e l l s us t h a t t h e r e l e v a n t subob jec t s are c losed and

t h a t w e g e t a r e p r e s e n t a t i o n by apply ing T, doing t h e r equ i r ed

c o n s t r u c t i o n i n Sets, and t a k i n g t h e cor responding canon ica l c losed

monic.

Page 24: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

188 J.M.E. HYLAND

Remark. The cons t ruc t ions desc r ibed i n 6 . 5 are p a r t i c u l a r l y simple

i n case t h e o b j e c t s ( X , = ) and (Y,=) are canon ica l ly sepa ra t ed .

F i n a l l y w e cons ide r t h e meaning of (5 .1) (9) f o r t h e e f f e c t i v e

topos. I t g ives a converse t o t h e obvious remark t h a t i f ( X I = ) i s

canon ica l ly sepa ra t ed ] t hen t h e e q u a l i t y ( o r d iagonal ) i n

( X l = ) x ( X l = ) i s t h e canon ica l c losed monic de f ined by t h e d iagonal i n

xxx. Propos i t ion 6 . 6 .

l ence r e l a t i o n on ( X , = ) i n Eff . Then t h e q u o t i e n t ( X , - ) i s i s o -

morphic ( i n t h e obvious way) t o t h e canon ica l ly sepa ra t ed o b j e c t

( r ( X I - ) , z ) where

1 ~ x l ~ C x l l J

Suppose t h a t - E C x x x r e p r e s e n t s a c losed equiva-

= U { I[ x ' -x l ' J Ix' E [ X I and xl' E [ x l l } .

Proof: The composite ( X , = ) - > ( X , - ) -> A r ( X l - ) i s represen-

t ed by [ - I rl ( X I - )

H(x,Cxll) = U { x-xl') Ixl' E I x l l ) .

By (5 .1) (9) w e r e q u i r e t h e image f a c t o r i z a t i o n of [HI, and what w e

have is a s t anda rd d e f i n i t i o n of t h i s f a c t o r i z a t i o n i n the l o g i c .

57. The e f f e c t i v e o b j e c t s .

Since Sets i s inc luded i n Eff, Eff con ta ins c l a s s i c a l mathema-

t i c s so much of it i s n o t p a r t i c u l a r l y " e f f e c t i v e " . I n t h i s s e c t i o n

w e cons ider o b j e c t s whose c l o s e r e l a t i o n t o t h e a p p l i c a t i v e s t r u c -

ture (IN , .) ensu res t h a t ope ra t ions on them a r e genuinely "ef fec t ive" .

I n l a t e r s e c t i o n s w e w i l l show t h a t t h e o b j e c t s of a n a l y s i s i n Eff a r e ( q u i t e familiar) o b j e c t s of t h i s k ind .

Def in i t i on . An o b j e c t ( X , = ) i s ( s t r i c t l y ) e f f e c t i v e i f f

(i) 1 x E X J is non-empty each x E X ,

(ii) U x E X J n f x ' E X J non-empty impl i e s x = x ' ,

and (iii) U x = x ' l = II x E X I n Ux' E X I . (Occasionally w e may d e s c r i b e an o b j e c t as e f f e c t i v e when it i s

isomorphic t o one of t h e above form. It w i l l be obvious when t h i s

l oose sense is meant.)

Page 25: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 189

Clearly e f f e c t i v e o b j e c t s a r e (canonical ly) separated, and w e

can e a s i l y show t h a t they sha re the c losu re p rope r t i e s of separated

objects . Proposition 7 .1 . ( a ) I f ( X , = ) and (Y,=) a r e e f f e c t i v e , then so i s

t h e i r product.

(b) I f (Z,=) i s effect ' ive , then so i s t h e funct ion

space (z,=) (',=) f o r any (Y,=) i n ~ f f .

(c) A subobject of an e f f e c t i v e ob jec t is

e f f ec t ive .

(d) A q u o t i e n t of an e f f e c t i v e o b j e c t by a c losed

equivalence r e l a t i o n i s e f f e c t i v e .

Proof: ( a ) i s t r i v i a l : look a t ( 6 . 2 ) .

(b) follows by in spec t ion of (6.3). I f w e res t r ic t t o those

a E r Z r Y with Ea

(i) , (ii) and (iii) above.

non-empty, then w e g e t an ob jec t s a t i s f y i n g

(c) r equ i r e s more work. L e t ( X I = ) be s t r i c t l y e f f e c t i v e and

l e t (XI=') --> ( X , = ) be a canonical monic with

[ x ='XI II = R ( x ) A Il x=x'n for some s t r i c t r e l a t i o n a l R E c . Write x E' X f o r x ='x and put X ' = Ix E XI Ux E ' X 1 i s non-empty}.

Since [ x E ' x n n [x' € ' X I non-empty implies Ilx E x D n [ x u E x n non-empty which implies x = x', w e g e t a s t r i c t l y e f f e c t i v e ob jec t

(XI,=) with [ x E X ' l = I[ x E ' X 1 . I t is isomorphic t o (XI=')

because

X

n ( ( u x E x'I n Ux' E xi] 1 <--> I [ x ='x'II )

is non-empty . (d) follows from ( 6 . 6 ) . I f - i s a c losed equivalence

r e l a t i o n on (X,=) which is s t r i c t l y e f f e c t i v e , then [x-xi n [xl-x'D

non-empty imples x E X I n I[ X'E X 1 non-empty which implies x = x'.

I t follows t h a t ( r ( X I = ) ,%) i s s t r i c t l y e f f e c t i v e .

The f u l l subcategory of Eff whose o b j e c t s a r e t h e e f f e c t i v e

ones has a concrete r ep resen ta t ion f a m i l i a r t o log ic i ans i n

Page 26: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

190 J.M.E. HYLAND

connection with t h e e f f e c t i v e ope ra t ions . Take p a r t i a l equivalence

r e l a t i o n s on IN ( t h a t i s equivalence r e l a t i o n s on t h e i r f i e l d s )

R,S, ... and w r i t e m / R = {CnlR(n E F i e l d ( R ) } f o r t h e set of

equivalence c l a s s e s of R . L e t a m x F: R -> S be a map F: N/R->lN/S

such t h a t t h e r e is f E IN with

F(CnlR) = [ f ( n ) l S

f o r a l l n E F i e l d ( R ) . C lea r ly w e have a category.

Each p a r t i a l equivalence r e l a t i o n R gives r ise t o a s t r i c t l y

e f f e c t i v e o b j e c t ( B / R , = ) of Eff where E ( l n l R ) = Cnl,.

F: R --> S gives r ise t o a map (IN/R ,=) -> (B/S ,=) represented

A map

by

F(CnlR,[mlS) = U{CnlRh[mlSIF([nlR) = [mlsl ,

and so w e have a func to r i n t o Eff which i s c l e a r l y f a i t h f u l and i s

f u l l by applying g loba l s e c t i o n s t o ( 7 . 1 ) ( b ) . Clear ly any s t r i c t l y

e f f e c t i v e o b j e c t i s isomorphic t o one obtained from a p a r t i a l

equivalence r e l a t i o n .

category of p a r t i a l equivalence r e l a t i o n . I t i s given by

L e t us desc r ibe t h e funct ion space SR i n the

R eS f i f f nRm implies e ( n ) S f ( m ) . A moment's thought shows t h a t t h i s corresponds t o the p r e s c r i p t i o n

f o r f ind ing t h e space of func t ions from ( B / R ,=) t o (N/S ,=) given

by ( 7 . 1 ) (b) . This i s a use fu l way t o think of t h e m a t e r i a l i n

§ § l o and 11. ( I n f a c t t h e embedding of t h e p a r t i a l equivalence

r e l a t i o n s i n Eff prese rves t h e l o c a l Cartesian closed s t r u c t u r e of

t h e former category. )

One p a r t i c u l a r e f f e c t i v e o b j e c t i s c ry ing o u t f o r a t t e n t i o n :

t h a t corresponding t o t h e e q u a l i t y r e l a t i o n on N . This i s t h e

object . I N = (IN ,=) where

[In = m l l = inlnIm}.

As w e noted i n 13, t h i s i s t h e n a t u r a l number o b j e c t ; w e consider some

of i t s p r o p e r t i e s i n l a te r sec t ions . F i r s t however, w e w i l l use it

t o give a c h a r a c t e r i z a t i o n of e f f e c t i v e o b j e c t s . Reca l l t h a t any

Page 27: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 191

ob jec t ( X , = ) i s a q u o t i e n t of a subobject of AX. For e f f e c t i v e

ob jec t s w e can r ep lace AX by ( I N ,=) . Proposi t ion 7 . 2 . Every e f f e c t i v e o b j e c t i s a q u o t i e n t by a closed

equivalence r e l a t i o n of a c losed subobject of (IN ,=) . Proof: I f ( X , = ) corresponds a s above t o the p a r t i a l equivalence

r e l a t i o n R on I N , then the closed subobject of LIN ,=) i s t h a t

determined by Fld(R) 5 I N and t h e closed equivalence r e l a t i o n - i s

given by

That t h e r e s u l t i n g q u o t i e n t of a subobject of ( IN ,=) gives rise t o

the same R is immediate i n view of ( 6 . 4 ) and ( 6 . 6 ) .

Now w e can s t a t e our c h a r a c t e r i z a t i o n theorem.

Theorem 7.3. The following cond i t ions on a o b j e c t X of Eff a r e

equivalent :

( i) X i s isomorphic t o a s t r i c t l y e f f e c t i v e o b j e c t ;

(ii) X i s a c losed q u o t i e n t of a c losed subobject of (IN ,=) :

(iii) X i s a c losed q u o t i e n t of a subobject of ( I N , = I .

Proof: (i) implies (ii) i s (7 .2 ) , (ii) implies (iii) is t r i v i a l

and (iii) implies (i) follows from ( 7 . 1 ) ( c ) and ( a ) .

Remark. Since (m,=) is t h e n a t u r a l number o b j e c t , w e have shown

t h a t t h e e f f e c t i v e o b j e c t s are those subnumerable i n a c e r t a i n way.

However t h e e q u a l i t y on an e f f e c t i v e o b j e c t must be c losed (as it

is a sepa ra t ed o b j e c t ) and t h e r e a r e quo t i en t s o f (m ,=) by

equivalence r e l a t i o n s which are emphat ical ly no t c losed. (The

reader w i l l know where t o look a f t e r reading t h e next s ec t ion ! )

So the e f f e c t i v e o b j e c t s a r e a proper subc la s s of t h e quo t i en t s of

decidable o b j e c t s r e c e n t l y s tud ied by Peter Johnstone i n a general

context.

Page 28: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

192 J.M.E. HYLAND

18. Markov's p r i n c i p l e and almost negat ive formulae.

In t h i s s ec t ion w e see how the general r e s u l t of (5.2) can be

extended i n the case of t h e topos Eff and (double negation) topology

j with Effj = Sets. Lemma 8.1. A : Sets -> Eff preserves t h e i n i t i a l ob jec t . Thus

lI is always a c losed subobject , and hence decidable subobjects

a r e closed.

Proof: T r i v i a l category theory.

Lemma 8.2. Markov's p r i n c i p l e

VR L P ( N ) . ( V n . R ( n ) v ~ R ( n ) h ~ l 3 n ( R ( n ) -> 3nRCn))

holds i n Eff. Proof: A s t h e a r i t h m e t i c a l s ta tements holding i n Eff a r e those

r e a l i z e d i n the o r i g i n a l sense of Kleene (see 13) t h i s is t h e

s tandard argument ( T r o e l s t r a C19731). Note t h a t w e do n o t need t o

know about P ( N ) 1

Lemma 8 . 3 . I f R >-> N x X is a decidable subobject i n E X , then

3n.R(n,x) >-> X i s closed and r ( 3 n . R ( n f x ) ) = 3 n . r ( R ( n f x ) ) .

Proof: This amounts t o --3n.R(n,x) 5 3n.R(n,x) which follows by

( 8 . 2 ) .

Remark. Though ( 8 . 2 ) depends on Markov's p r i n c i p l e i n w, and so

does n o t r e l a t i v i z e t o an a r b i t r a r y topos, ( 8 . 3 ) does r e l a t i v i z e :

we w i l l always have j (3n.R(n,x)) 5 3n.R(n,x) .

(8.1) and ( 8 . 3 ) suggest t h a t w e extend the c l a s s of negat ive

formulae.

Defini t ion. A formula is c a l l e d almost negat ive i f f it i s b u i l t up

from atomic formulae using A , ->,V,i, and sequences of 3n appl ied

t o decidable formulae ( t y p i c a l l y equat ions between numerical-valued

terms) . W e now give our extension of (5 .2) .

Theorem 8 . 4 . I f t he atomic formulae of a f i r s t o rde r language a r e

Page 29: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 193

i n t e r p r e t e d as c losed subobjects i n Eff and @ i s almost negat ive

with [ @ 1 >-> E, then

n E * ( u + n j ) = t @ n . Proof: As f o r (5.2) using (8.1) and (8.3) as w e l l .

The fo rce of (8.4) is t h a t , f o r @ almost negat ive @ i s t r u e

i n Eff i f f t h e corresponding i n t e r p r e t a t i o n of @ i n Sets i s t r u e :

t h a t i s , t h e meaning of @ i n Eff "agrees with" i t s c l a s s i c a l meaning.

(8.4) i s a vers ion of 3 .2 .11 ( i ) and (ii) of T r o e l s t r a 119731;

w e could o b t a i n a more p roof - theo re t i c ve r s ion by s e l a t i v i z i n g t o

the f r e e topos (with n a t u r a l number o b j e c t ) . For a language which

can "express i t s own r e a l i z a b i l i t y " we could obviously ob ta in

vers ions of 3.2.12 and 3.2.13 of Tsoe l s t r a 119731. For t h e sake of

completeness w e give a ve r s ion of 3.6.5 of T r o e l s t r a C19731.

Defini t ion ( c f . Hyland C19771) PR(a.n.1 is t h e l e a s t class C of

formulae such t h a t

(i) C con ta ins a l l atomic formulae;

(ii) C i s closed under ~ , v , V , 3 ;

(iii) i f @ is almost negat ive (more gene ra l ly almost negat ive

preceded by e x i s t e n t i a l q u a n t i f i e r s ) and $ is i n C, then ( @ -> $1

i s i n C.

Proposi t ion 8.5. In the s i t u a t i o n of ( 8 . 4 1 , i f @ is i n PR(a.n)

with I@ 1 >-> E , then

n + n 5 nE*( I@ n j ) .

Proof: By induct ion on the complexity of @. (Note t h a t v and + 'are

ca l cu la t ed d i f f e r e n t l y i n E from t h e way they are i n E).

Remark. For a general sheaf subtopos E of E w e have

5 nE*( fi @ 1 . ) f o r a l l @ i n PR(j-closed) .

j

j [ @ I So i f atomics are

3 i n t e r p r e t e d as j -c losed, t hen w e g e t t h e r e s u l t f o r a l l @ i n

PR(negative) . The f o r c e of (8.5) i s t h a t , f o r @ i n PR(a.n) , i f @ is t r u e i n

Page 30: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

194 J.M.E. HYLAND

- E f f , then t h e cor responding i n t e r p r e t a t i o n of @ i n Sets i s t r u e .

This g ives r ise t o a conse rva t ive ex tens ion r e s u l t (when r e l a t i v i z e d )

a s i n T r o e l s t r a C19731 53.6.

§ 9 Choice p r i n c i p l e s and t h e r e a l numbers.

I n t h i s s e c t i o n w e make a s t a r t towards showing t h a t a n a l y s i s

i n Eff is j u s t c o n s t r u c t i v e r e c u r s i v e a n a l y s i s . (We a l r eady have

Markov's p r i n c i p l e ( 8 . 2 ) . ) W e do t h i s i n two s t e p s . F i r s t w e show

t h a t w e have t h e choice p r i n c i p l e s t o ensu re t h a t t h e Dedekind

r e a l s ( t h e r i g h t r e a l s i n a topos) a r e Cauchy (see Fourman-Hyland

C19791 and a l s o Fourman-Grayson t h i s volume). Then w e use t h e re-

s u l t s of 57 t o show t h a t t h e Cauchy r e a l s i n Eff can be i d e n t i f i e d

wi th a f a m i l i a r s t r i c t l y e f f e c t i v e o b j e c t used i n c o n s t r u c t i v e

r e c u r s i v e a n a l y s i s .

F i r s t we need t o know what t h e space of func t ions from IN t o

an a r b i t r a r y ( X , = ) looks l i k e i n Eff. A s s t a t e d i n HJP C19801, by

l o g i c a l c o n s i d e r a t i o n s it is

(ZIN xx,=)

where

[ G = H I = EGAn{G(n,x) <-> H(n,x) In E TN , x d X I ,

wi th EG t h e non-standard va lue of "G i s a f u n c t i o n a l r e l a t i o n " .

Suppose now t h a t e realizes Vn E IN ,3x E ( X I = ) . @ ( n , x ) . Then

f o r every n , e ( n ) E UIEXA [ @ ( n , x ) ] Ix E X I . For each n p i ck xn

such t h a t e ( n ) E E X ~ A [ @ ( n , x n ) l . S e t G(n ,x) = EnA [ x = x n 1 . Now (uniformly i n e) w e can f i n d numbers r e a l i z i n g EG and

Vn.3x .G(nIx)h$(n ,x) : G i s r e l a t i o n a l , s t r i c t and s ing leva lued i n a

s t anda rd way from i t s d e f i n i t i o n : A n . < n , n l ( e ( n ) ) > r e a l i z e s G i s

t o t a l ; Xn.<a ( e ( n ) ) , < < n , ~ l ( e , n ) > , a 2 ( e , n ) > > r e a l i z e s

Vn.3x .G(x ,n)h@(xln) . Thus (uni formly i n e ) w e have a number

r e a l i z i n g

1

Page 31: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 195

3g: Ri-> ( X , = ) .Vn .$ (n ,g (n ) )

and so we have proved t h e fol lowing r e s u l t .

Proposit ion 9 . 1 . A C ( I N , X ) , t he axiom of choice from t h e n a t u r a l

numbers t o an a r b i t r a r y type X , holds i n Eff. Remark. W e used AC ( IN , X ) i n sets i n t h e above proof . But i f ( X , = )

i s e f f e c t i v e , then no use of a choice p r i n c i p l e i n the base topos

i s needed (compare ( 7 . 1 ) (b) : i n t h i s case t h e argument i s contained

within T r o e l s t r a 119731 3.2 15.

By a s i m i l a r proof ( l e f t t o t he r eade r ) w e a l s o have t h e s t ronge r

r e s u l t .

Proposit ion 9 . 2 . D C ( X ) , t h e axiom of dependent choices on an

a r b i t r a r y type X , holds i n Eff. Remark. Again D C ( X ) i s used i n the proof , but i s no t needed f o r

e f f e c t i v e o b j e c t s X.

A C ( W , I N ) is enough t o show t h a t t h e Cauchy and Dedekind r e a l s

a r e the s a m e . To g e t an e x p l i c i t r ep resen ta t ion of R as a

s t r i c t l y e f f e c t i v e o b j e c t , w e use t h e Cauchy sequence d e f i n i t i o n .

Lemma 9.3. The i n t e g e r s Z and r a t i o n a l s Q i n Eff can be taken a s

s t r i c t l y e f f e c t i v e o b j e c t s a,=) and (Q,=) where f o r x i n Z o r Q ,

Ex = # x) where # x is an elementary code f o r X.

Proof: They are obtained successively from (IN ,=) by taking closed

(decidable) q u o t i e n t s of c losed (decidable) subobjects o f products:

so the r e s u l t fol lows from the p r e s c r i p t i o n s involved i n (7.1) ( a ) ,

( c ) [easy case of c losed subob jec t s l and ( a ) . Lemma 9 . 4 . The space of maps from IN t o Q i n Eff i s t h e s t r i c t l y

e f f e c t i v e o b j e c t (Qm ,=) where

Qm = t h e r e c u r s i v e func t ions from IN t o Q

and [[ c1 E QIN! = { e j e ( n ) = # a ( n ) 1 , t h e set of ind ices f o r a .

Page 32: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

196 J.M.E. HYLAND

Proof: This i s t h e p re sc r ip t ion i m p l i c i t i n ( 7 . 1 ) (b) . Since w e

have enough choice t o show t h a t any reasonable not ions of Cauchy

sequence give the same r e a l s i n Eff w e de f ine C S , t h e c o l l e c t i o n of

( r e s t r i c t e d ) Cauchy sequences by

Thus d e f i n i t i o n i s i n the negat ive fragment and so s ince < i s

decidable on the r a t i o n a l s and hence by (8.1) c losed, de f ines a

closed subobject of Qm i n Eff. it.

Lemma 9.5. The space of Cauchy sequences i n Eff is t h e s t r i c t l y

e f f e c t i v e ob jec t ( C S , = ) where CS i s t h e set of r ecu r s ive Cauchy

sequences and Ur E C S ] I i s t h e set of i nd ices f o r r .

Proof: By the discussion above.

To ob ta in the r e a l s IR, w e t ake the quo t i en t of CS by t h e equiva-

lence r e l a t i o n

I n view of (8.4) w e can i d e n t i f y

1 r - s i f f vn.lrn-snl < - p - 3 ' (This choice of d e f i n i t i o n gives one p l en ty of "elbow room".)

Proposit ion 9 . 6 . The space IR of r e a l s i n Eff i s t h e s t r i c t l y

e f f e c t i v e ob jec t (IR ,=) where

IR = t he r ecu r s ive r e a l s ( t h a t i s r e a l s with r ecu r s ive

Cauchy sequences converging t o them)

and Ux E IR ]I = t h e set of i nd ices f o r Cauchy sequences converging

t o x.

Proof: As before - def ines a c losed equivalence r e l a t i o n so this

i s by the p re sc r ip t ion of ( 7 . 1 ) ( a ) . W e have shown t h a t t h e r e a l s i n Eff a r e represented j u s t a s

they are i n ( cons t ruc t ive ) r ecu r s ive ana lys i s . Of course, as t hey

t o o a r e def ined i n the negative fragment, t h e operat ions of

add i t ion , m u l t i p l i c a t i o n and so f o r t h a r e what they should be. To

do s e r i o u s ana lys i s however w e need t o consider funct ions which w e

do i n the next few sec t ions .

Page 33: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 197

Remark. I t is seldom e f f i c i e n t t o g r ind t h i n g s o u t i n models f o r

cons t ruc t ive ana lys i s : where possible one should use the axiomatic

point of view. Consider f o r example t h e quest ion of t he fundamental

theorem of a lgebra i n Eff . This theorem is proved i n Bishop C19671.

One way of reading Bishop's cons t ruc t ive mathematics (though no t

the intended one!) is t o regard it a s formalized i n an i n t u i t i o n i s t i c

type theory with extensional e q u a l i t y and using ( D C ) . Hence i n view

of ( 9 . 2 ) the fundamental theorem of a lgebra i s t r u e i n Eff ( a s it i s

i n o the r r e a l i z a b i l i t y t o p o s e s ) . I n view of t he obvious represen-

t a t i o n of C i n Eff derived from ( 9 . 6 ) , and t h e f a c t t h a t

"al, ... ,an a re the r o o t s of zn+an-lzn-l+. . .+ao = 0"

de f ines a c losed subobject ( i n C2n) w e can i n t e r p r e t t h i s f a c t a s

follows. There i s an e f f e c t i v e process t ak ing ind ices f o r the

r ecu r s ive complex c o e f f i c i e n t s of a monic polynomial of degree n

over t he r ecu r s ive complex numbers t o i n d i c e s f o r the r ecu r s ive

roots . I t is n o t t r i v i a l t h a t a r ecu r s ive polynomial has recursive

roo t s and any n a t u r a l proof would s e e m t o e s t a b l i s h the s t ronger

r e s u l t and a s such would have t h e form of an a b s t r a c t proof using

(DC) - 510. E f f e c t i v i t y and Church's Thesis.

I t i s t i m e t o give substance t o the claim made i n 17 t h a t

operat ions on e f f e c t i v e o b j e c t s a r e " e f f e c t i v e " . W e f i r s t consider

t he s p e c i a l case of Church's Thesis.

Lemma 10.1. The space of maps from IN t o IN i n Eff i s t h e s t r i c t l y

e f f e c t i v e o b j e c t ( I N r n ,=) where

ININ = t h e r ecu r s ive funct ions from IN t o IN , and 6 a E I N I N ] = { e l e ( n ) = a ( n ) 1 , t h e set of i nd ices f o r a

Proof: This i s the p r e s c r i p t i o n i m p l i c i t i n (7.1) (b) . Proposit ion 1 0 . 2 . "Church's Thesis" t h a t a l l funct ions a r e r ecu r s ive ,

Page 34: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

198 J.M.E. HYLAND

~ c i E mIN .3e .Vn .3y(T(e ,n ,y )~U(y) = c i (n ) )

ho lds i n Eff. CT i s Xleene ' s T-predicate and U h i s ou tpu t f u n c t i o n . ]

Proof: I n view of (3.3) elementary r ecu r s ion theory can be developed

i n Eff as i n T r o e l s t r a E19731. So by ( 8 . 4 ) Vn. 3y . (T(e ,n ,y )AU(y)=

a ( n ) )

real izes "Church's Thes is" .

Remark. Church's Thes is a s t r a d i t i o n a l l y formulated i n Heyt ing ' s

Ar i thmet ic (see T r o e l s t r a C19731) is a n amalgam of ou r "Church's

Thes is" and AC(IN ,IN).

i n Eff agrees wi th i t s meaning i n Sets. Then Xe.<e,<e,e>>

W e can hope t o g e n e r a l i z e ( 1 0 . 2 ) t o a l l e f f e c t i v e o b j e c t s i n

view of (7.3) which states t h a t t hey can i n a c e r t a i n way be sub-

numerated (by t h e codes f o r t h e i r e l emen t s ) .

Lemma 10.3. I f (Z,=) i s s t r i c t l y e f f e c t i v e and ( Y , = ) is a r b i t r a r y

i n Eff , t hen t h e space of maps from (Y,=) t o (Z ,= ) i n Eff i s t h e

s t r i c t l y e f f e c t i v e o b j e c t ( Z r Y , = ) where

Z r y = t h e " r ecu r s ive" maps from T Y t o Z ( t h a t i s , t h e maps

wi th i n d i c e s ) ,

and ci E Z Ty = { e l e ( n ) E Eci(y) f o r a l l n E Ey}, t h e set of

i n d i c e s f o r a.

Proof: This i s t h e p r e s c r i p t i o n i m p l i c i t i n (7 .1 ) ( b ) .

Here then is a g e n e r a l i z a t i o n of (10 .2 ) .

Propos i t ion 1 0 . 4 . L e t (Y,=) <- B >-> IN r e p r e s e n t t h e e f f e c t i v e

o b j e c t (Y,=) as a q u o t i e n t of a c losed subob jec t of I N , and l e t

( X , = ) <s A >--> IN r e p r e s e n t (X,=) as a q u o t i e n t of a c losed

subobjec t of I N . Then a "genera l ized Church's Thes is"

sY

X

Yci E YX.3e.Va E A . 3 z ( T ( e , a , a ) h a ( S X ( a ) ) = Sy(U(z)))

holds i n Eff . (One can u s e f u l l y compare t h i s r e s u l t wi th t h e t r ea tmen t of t h e

extended Church's Thes is i n T r o e l s t r a C19731.)

Page 35: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 199

Proof: The cond i t ions given ensu re t h a t a ( S x ( a ) ) = Sy(U(z)) i n t e r -

p r e t s as a c losed subob jec t . (Note t h a t U(y) E B i s i m p l i c i t , s o w e

need B c losed . ) S ince

3 z ( T ( e , a , z ) A a ( S X ( a ) ) = Sy(U(z)))

is e q u i v a l e n t t o

3~.T(e,a,z)AVz.(T(e,a,z) --> a ( S , ( a ) ) = Sy(U(z)),

it a l s o i n t e r p r e t s a s a c losed subob jec t . It remains t o determine

e from an index f o r a .

t ak ing any a E A t o an e lement of ESX(a ) :

t o EaSX(a ) ; t he cond i t ion t h a t Sy i s on to p rov ides a map from t h i s

t o some b E B wi th Sy(b) = a S x ( a ) .

which can c l e a r l y be chosen e f f e c t i v e l y i n the index f o r a .

The t o t a l cond i t ion f o r Sx g ives a map

an index f o r a maps t h i s

e i s an index f o r t h i s composite

In p a r t i c u l a r , w e can see t h a t when e f f e c t i v e o b j e c t s a r e

p re sen ted ( v i a p a r t i a l equiva lence r e l a t i o n s ) a s c losed q u o t i e n t s

of c losed subob jec t s of I N , t h e n maps between them a r e e f f e c t i v e i n

the i n d i c e s (and t h i s ho lds i n Eff). This i s t y p i c a l l y t h e s i t u a t i o n

i n c o n s t r u c t i v e r e c u r s i v e a n a l y s i s .

111. The e f f e c t i v e ope ra t ions .

In t h i s s e c t i o n w e use ( 1 0 . 4 ) a s t h e induc t ion s t e p t o show

t h a t t h e s t a t emen t t h a t t h e f i n i t e t ypes ove r IN a r e t h e he red i -

t a r i l y ( e x t e n s i o n a l ) e f f e c t i v e o p e r a t i o n s ho lds i n Eff . A s s u m e f o r n o t a t i o n a l purposes a c o l l e c t i o n of type symbols

genera ted from 0 by x ( f o r p roduc t s ) and --> ( f o r func t ion s p a c e s ) .

The f i n i t e t ypes over t h e n a t u r a l numbers ( m u la a t ype symbol) a r e

de f ined i n d u c t i v e l y by INo = I N ,

= m ' x m IN U X T u 7'

INu+= = (rnT ) 0 . El

The h e r e d i t a r i l y e f f e c t i v e ope ra t ions ( H E O u l u a t ype symbol)

(see Kreisel 119591 and T s o e l s t r a C19731) may be de f ined by f i r s t

d e f i n i n g a c o l l e c t i o n ( R u I a a type symbol) of p a r t i a l equiva lence

Page 36: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

200 JM.E. HYLAND

r e l a t i o n s induc t ive ly by

nROm i f f n = m,

nRaxTm i f f r1 (n) Raml ( m ) and r 2 (n )RTr2 (m) ,

eRa+.rf iff i f nRam then e ( n ) , f (m) are def ined and

e ( n ) R T f ( m ) .

W e can then regard HEOa as t h e equivalence classes IN/Ra .

of t h e d i scuss ion i n 9 7 , w e can equa l ly regard HEOa a s b u i l t up

( toge the r with ind ices f o r i t s elements) from t h e n a t u r a l number

o b j e c t i n Eff, by t ak ing t h e usual products and funct ion spaces as

i n (10.3) . Thus i n Eff No = (HEOaI=) where

In view

Ex = t he i n d i c e s f o r x,

and where nRam i f f n,m a r e i n d i c e s f o r t h e same x E HEOg. Then w e

can regard HEOa as t h e g loba l s e c t i o n s of t h e f i n i t e t ypes over JN

i n Eff .

These d e f i n i t i o n s a l l r e l a t i v i z e and our next r e s u l t s tates

t h a t Eff "knows t h a t i t s f i n i t e t ypes a r e t h e e f f e c t i v e ope ra t ions" .

Theorem -11.1.

t h a t t h e products and func t ion spaces correspond.

Proof: The R a t s are def ined by negat ive formulae, and so by 18.4)

i n t e r p r e t i n Eff as c losed p a r t i a l equivalence r e l a t i o n s agreeing

with t h e i r meaning i n Sets. I f we c a l c u l a t e t he equivalence

c l a s s e s i n the obvious way using (7.1) (d) w e j u s t g e t (HEOg,=)

t h a t i s INa i n Eff.

as it ought. ( I f t h i s i s t o o a b s t r a c t , t h e reader can use a

l abor ious induct ion, with ( 1 0 . 4 ) dea l ing with t h e main induct ion

s t e p . )

Remark. Something q u i t e deep i s going on behind (11.1) which is

connected with i t e r a t i o n s of t h e e f f e c t i v e topos cons t ruc t ion a:

s tud ied i n B i t t s C19811. I t i s i n connection with t h e e f f e c t i v e

ob jec t s t h a t w e can g e t a general expression of t h e idempotency o f

For each a , INa = HEOa holds i n Eff, i n such a way

Clear ly t h e rest of t h e s t r u c t u r e corresponds

Page 37: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 20 1

of r e a l i z a b i l i t y (see T r o e l s t r a (1973) 3.2.16).

512 . Sequent ia l con t inu i ty .

From ( 8 . 2 ) , ( 9 . 2 ) , ( 1 0 . 4 ) and t h e discussion i n 19, it should be

c l e a r t h a t a n a l y s i s i n Eff i s j u s t cons t ruc t ive r e c u r s i v e a n a l y s i s .

So w e have t h e usual con t inu i ty r e s u l t s which are ve r s ions of t h e

Kreisel-Lacombe-Shoenfield theorem.

Theorem 1 2 . 1 . "Brouwer's Theorem" t h a t every map from IR t o IR i s

continuous holds i n Eff.

Proof: The r eade r w i l l have t o do t h i s himself (along t h e l i n e s of

( 1 2 . 4 ) below) o r else f i n d (as I have f a i l e d t o do) a readable

account from t h e Russian school .

( 1 2 . 1 ) is only moderately spec tacu la r . Recursive maps on the

r ecu r s ive r e a l s , while n o t t h e r e s t r i c t i o n of continuous funct ions

on the ( c l a s s i c a l ) reals (see ( 1 3 . 4 ) ) , are continuous on t h e i r

domain. (This i s s t a t e d as Exercise 15.35 i n Rogers 119671.) So

we j u s t need e f f e c t i v i t y t o g e t ( 1 2 . 1 ) . By passing t o higher types

we g e t a more i n t e r e s t i n g phenomenon: we g e t e f f e c t i v e maps, which

are n o t continuous on t h e i r e f f e c t i v e domain, b u t which a r e s t i l l

continuous from t h e p o i n t of view of Eff. W e consider t h e h e r e d i t a r i l y e f f e c t i v e operat ions. By (11.1) ,

i n Eff t hese are j u s t t h e f i n i t e types.

much of t h e material can be developed f o r an a r b i t r a r y "type

s t r u c t u r e " over IN .) W e de f ine a not ion of sequence cwvergence on

each HEOu i nduc t ive ly a s follows:

(The r eade r w i l l see t h a t

on HEOO = I N , xn -> x i f f 3k,Vn 2 k.xn = x;

on HEOuxT = HEOuxHEOT, (xn,yn) -> (x ,y) i f f xn ->xand yn-->y;

on H E O ~ + ~ = ( H E O ~ ) HE', , f ->f i f f xn->x implies f n (xn) ->f (x) . n W e say t h a t a funct ion f E HEOO+T i s continuous i f f f preserves

sequence convergence.

Page 38: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

202 JM.E. HYLAND

Remark. The meaning of t h e s e d e f i n i t i o n s i n Eff does not a g r e e wi th

t h e meaning i n Sets. L e t us i n i t i a l l y res t r ic t a t t e n t i o n t o t h e h e r e d i t a r i l y e f f e c -

t i v e o p e r a t i o n s of pure t y p e (HEOklk a pure type symbol) where each

k+ l denotes (k -> 0 ) . For f n , f i n HEOk+l, w e s a y t h a t p E HEOk+l

i s a modulus f o r f n --> f i f f

Vx E HEOk.Vn 2 p ( x ) . f n ( x ) = f ( x ) .

( W e do n o t assume h e r e t h a t f n --> f i n HEOk+l: t h i s i s f a l s e i n

S e t s , though t r u e i n Eff . )

Lemma 1 2 . 2 . ( I n Eff.) A s s u m e f u n c t i o n s i n HEOk+l axe cont inuous .

I f 1.1 i s a modulus f o r f n -> f i n HEOk+l, t h e n f n -> f .

Proof:

f o r a l l n t k , p ( x n ) = ~ ( x ) = k ' s a y . A s f i s cont inuous t h e r e is

k" such t h a t f o r a l l n L k ' , f (x,) = f ( x ) . n 2 m a x ( k , k ' , k " ) ,

L e t xn --> x. S ince 1.1 i s cont inuous , t h e r e i s a k such t h a t

Then f o r a l l

f n (xn) = f ( x n ) = f ( x ) .

Remark. Th i s argument i s e n t i r e l y e lementary and h a s u s e f u l

a p p l i c a t i o n t o a v a r i e t y of type s t r u c t u r e s i n a v a r i e t y o f t oposes .

Lemma 12.3. A s s u m e a l l f u n c t i o n s i n HEOk a r e cont inuous .

I f f n -> f i n HEOk+l, t hen t h e r e i s a modulus 1.1 fox f n --> f .

Proof:

so w e can deduce f n ( x ) -> f ( x ) , t h a t i s

Vx3kVm 2 k f m ( x ) = f (x ) .

( I n Eff.)

The sequence wi th c o n s t a n t va lue x converges t o x i n HEOk,

By b a s i c a r i t h m e t i c choose k minimal f o r each x . This g i v e s us a

func t ion p : HEOk -> IN which by (11.1) o r ( 1 0 . 4 ) i s i n HEOk+l.

Remark. This argument depends on e f f e c t i v i t y i n Eff. Again t h e r e

a r e many u s e f u l v e r s i o n s of it.

Lemma 1 2 . 4 .

F E HEOk+2, t h e n t h e r e i s an r such t h a t

Vn t r F ( f n ) = F ( f ) .

( I n Eff.) I f p is a modulus f o r f n --> f i n HEOk+l and

Page 39: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective top- 203

Proof:

t o f i n d an r such t h a t Vn t r. F ( f n ) = F ( f ) . Following Gandy, use

t h e second r ecu r s ion theorem t o d e f i n e an index b ' by,

From i n d i c e s bn,m,b,c f o r f n l p r f , F r e s p e c t i v e l y w e wish

b ( a ) i f m ( a ) < l e a s t y.y shows c ( b ) = c ( b ' ) C = yo s a y ] , b ' ( a ) = { b n ( a ) f o r n l e a s t t yo wi th c ( b n ) # c ( b ) , o therwise .

l y shows c ( b ) = c ( b ' ) i f f T(c,b,~l(y))AT(c,b',~~(y))~U(~l(y)) =

U ( T 2 ( Y ) ) .I

W e see e a s i l y t h a t yo e x i s t s and t h a t Vn t yo.c(bn)

Markov's p r i n c i p l e ) .

Remark. This e s s e n t i a l l y i s Gandy's proof of t h e Kreisel-lacombe-

Shoenf ie ld theorem. S ince both Vn t k F ( f n ) = F ( f ) and p is a

modulus f o r f n -> f a r e i n t e r p r e t e d a s c losed subob jec t s i n Eff, it makes no d i f f e r e n c e whether w e do ( 1 2 . 4 ) e x t e r n a l l y i n Sets o r

i n t e r n a l l y i n Eff.

Theorem 12 .5 . (i) I n Eff it holds f o r a l l pure types r+l t h a t

f n -> f i n HEOr+l i f f t h e r e i s a modulus p f o r f n -> f i n HEOr+l,

and t h a t a l l m e m b e r s of HEOr+l are cont inuous .

= c ( b ) (us ing

Thus yo is c l e a r l y what w e want.

(ii) I n Eff it holds f o r any types a , ~ , t h a t a l l

members of HEOa+T a r e cont inuous .

Proof: (i) fol lows by induc t ion us ing ( 1 2 . 2 ) , (12.3) and ( 1 2 . 4 ) .

(ii) fo l lows by ex tens ion us ing Car t e s i an c losedness of t h e here-

d i t a r i l y e f f e c t i v e o p e r a t i o n s and o f t h e continuous f u n c t i o n a l s ( i n

the s e q u e n t i a l v e r s i o n , see Hyland C1979 1 ) i

The r e a d e r should compare (12.5) w i th t h e example of Gandy

(see Gandy-Hyland C19771 of a type 3 e f f e c t i v e o p e r a t i o n n o t con-

t inuous on t h e type 2 e f f e c t i v e o p e r a t i o n s ) . Con t inu i ty has a q u i t e

d i f f e r e n t meaning i n t e r n a l l y i n Eff . Remark. The f i n i t e t ypes over R i n Eff co inc ide n o t on ly wi th t h e

h e r e d i t a r i l y e f f e c t i v e o p e r a t i o n s i n Eff , b u t a l s o wi th t h e sequen-

t i a l l y continuous f u n c t i o n a l s i n Eff. The use of t h e modulus was

in t roduced o r i g i n a l l y i n t h e con tex t of r e c u r s i o n theo ry on the

Page 40: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

204 J.M.E. HYLAND

( sequen t i a l ly ) continuous f u n c t i o n a l s by Stan Wainer.

513. F a i l u r e of compactness.

A s i s w e l l known, t h e r e are decidable subse t s R of 2 'B, t he

set of f i n i t e b ina ry sequences such t h a t

(i) any r ecu r s ive a E 2m extends some u E R ,

(ii) t h e r e are a E 2 B which extend no u E R (so t h a t no f i n i t e

S 2 R w i l l s a t i s f y (i)). This has an immediate consequence f o r

E f f .

Proposi t ion 13.1. I n Eff t h e r e is a decidable subobject R of 2

such t h a t

- B

(i) any a extends some u E R ,

(ii) f o r any k, t h e r e is an a which extends no u E R of

length 6 k.

Thus i n Eff t h e r e i s a decidable cover of 2 N , Cantor space, by

b a s i c clopen sets, with no f i n i t e subcover.

Proof: EITHER (i) and (ii) a r e almost negat ive and hold i n sets of

t h e r ecu r s ive r e a l s ,

OR immediate from Church's Thesis .

Corol lary 13.2.

F: 2m-> I N .

Proof: S e t F ( a ) = least l eng th of u E R with a extending u.

I n Eff t h e r e i s a continuous but unbounded func t ion

(13.1) shows t h a t t h e Fan Theorem f a i l s a s badly as p o s s i b l e i n

- Eff . This is why w e g e t (13.2) . There a r e Grothendieck toposes i n

which (13.1) holds without t he s t i p u l a t i o n t h a t R is decidable . I n

t h e known exanples, a l l continuous func t ions from 2 m t o JN are

uniformly continuous and so bounded. It is no t known whether t h e r e

are Grothendieck toposes i n which (13.1) ho lds .

The t r a d i t i o n a l way t o ob ta in r e s u l t s analogous t o (13.1) and

(13.2) f o r t h e reals i s t o use " s ingu la r coverings" as s t u d i e d i n

Zaslavski i -Cei t in C19621. (Of course one can set up (13.1) and

(13.2) i n an analogous fashion.)

Page 41: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 205

Proposi t ion 13.3. I n Eff, t h e r e i s a sequence of r a t i o n a l i n t e r -

v a l s covering JR, bu t of a r b i t r a r i l y small measure.

Proof: E s s e n t i a l l y a diagonal enumeration, see Zaslavski i -Cei t in

C19621. The proof i s a l s o sketched i n Rogers C19671 Exercises

15.36, without cons ide ra t ions of e f f e c t i v i t y . By t h e cond i t ions , t o

be s a t i s f i e d by t h e sequence of r a t i o n a l i n t e r v a l s , can be expressed

as almost negat ive formulae, so by (8.4) t h i s does no t matter.

Corol lary 13.4. In Eff t h e r e i s a continuous funct ion from IR t o IR

which i s unbounded on some c losed bounded i n t e r v a l , and so i n

p a r t i c u l a r i s no t uniformly continuous on some closed bounded i n t e r v a l

Proof: Same re fe rences a s f o r (13 .3 ) .

Remark. The r e s u l t s of t h i s s e c t i o n can a l l be regarded a s proved

i n t e r n a l l y i n Eff , t h a t i s , they fol low from t h e e f f e c t i v i t y w e

e s t a b l i s h e d i n 510.

Though w e know Grothendieck toposes i n which Il? f a i l s t o be

l o c a l l y compact (Fourman-Hyland C19791), i n a l l known examples, t h e

t y p i c a l consequences of l o c a l compactness f o r a n a l y s i s s t i l l hold.

Ce r t a in ly continuous func t ions on bounded closed i n t e r v a l s are

uniformly continuous. So the e f f e c t i v e topos opens up p o s s i b i l i t i e s

unknown amongst Grothendieck toposes. Fu r the r examples can be

found i n Zaslavskiy-Celtin C19621.

814. Quo t i en t s of c l a s s i c a l o b j e c t s , and power o b j e c t s .

It is a f a m i l i a r f e a t u r e of i n t u i t i o n i s t i c mathematics t h a t

c o l l e c t i o n s of sets ( spec ie s ) can appear f a r more amorphous than

c o l l e c t i o n s of func t ions . W e have seen i n Eff t h a t t h e o b j e c t of

funct ions between "well-behaved".objects is i t s e l f "well-behaved''

(16.3) and ( 7 . 1 ) (b) ) . W e have seen t h i s good behaviour i n o t h e r

con tex t s (Moschovakis C19731, S c o t t C19701, are t h e e a r l y r e fe ren -

ces), and it can be made t h e b a s i s f o r n i c e p roof - theo re t i c r e s u l t s .

However, when the subobject c l a s s i f i e r is i t s e l f complicated, t h e

Page 42: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

206 J.M.E. HYLAND

power set of however simple a ( t o some e x t e n t i nhab i t ed ) o b j e c t w i l l

be complex. I t i s t i m e t o look a t such o b j e c t s i n Eff.

A s w e mentioned i n 13, t he subobject c l a s s i f i e r R i n Eff can be

taken a s (Z,<-->) where <--> i s t h e r e a l i z a b i l i t y bi- implicat ion on

C = P ( W ) . W e may and so do think of t h e members of C a s e x i s t i n g

"global ly" . C lea r ly then (Z,<-->) i s a q u o t i e n t of A C . This w i l l

mean t h a t w e can ob ta in maps t o R i n Eff from s u i t a b l e maps t o C i n

S e t s .

Lemma 1 4 . 1 . Suppose A Y --> (Y,=) i s a s u r j e c t i o n . Then a map

f : X --> Y induces a map f : ( X , = ) --> (Y,=) i n Eff such t h a t

-

-

4 J ( X , = ) - > (Y,=) commutes

f

i f f x = x ' --> f ( x ) = f ( x ' ) i s v a l i d . (That i s , i f f f preserves t h e

e q u a l i t y r e l a t i o n . ) Under these circumstances f i s represented by

t h e func t iona l r e l a t i o n EXA [ I f ( x ) = y 1 . Proof: By a rou t ine use of l o g i c .

In the case of t he s u r j e c t i o n A C --> R , every map a r i s e s a s i n

( 1 4 . 1 ) .

Proposi t ion 1 4 . 2 . Any map from (X,=) t o R = (Z,<-->) i n Eff is

a s defined i n ( 1 4 . 1 ) f o r an f : X --> Z such t h a t both

-

-

(i) x = x ' -> ( f ( x ) <--> f ( x ' ) )

and (ii) f ( x ) --> Ex

a r e v a l i d .

Remark. Given f : X -> C with (i) v a l i d , one can e a s i l y de f ine

g: X --> Z with both (i) and (ii) v a l i d , and such t h a t z = s. S e t

g ( x ) = ExAf (x) . Proof: Since maps from (X,=) t o R a r e i n b i j e c t i v e correspondence

with maps 1 --> P(X,=)), ( 1 4 . 2 ) i s immediate from t h e d e s c r i p t i o n

of the power set i n ( 2 . 1 2 ) of H J P C19801. A reader who f i n d s t h a t

Page 43: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 207

proof unpa la t ab le , can t a k e a r e p r e s e n t a t i v e G(x ,p) f o r a map ( X , = )

t o f2, se t f (x ) = 1 vq. (vp(G(x ,p) hp --> q ) -> q ]I , and check t h a t

G(x,p) <--> Exh ( f (x) <--> p) is v a l i d : s i n c e (i) i s v a l i d f o r f , t h e

remark above a p p l i e s t o g i v e (ii) f o r g ( x ) = Exhf (x) . (Su i t ab ly

r e l a t i v i z e d , t h i s i s a proof of ( 2 . 1 2 ) of HJP L19801.)

From ( 1 4 . 2 ) w e see t h a t i f ( X , = ) i s sepa ra t ed then any map

( X , = ) t o R f a c t o r s through q: A X --> R . Our next r e s u l t g ives t h i s

wi th in Eff. Propos i t ion 14.3.

f o r any sepa ra t ed (X,=) i n Eff.

Proof:

Then q ( x ' ' ) : A ( Z

The map q( ' '=): A C ( x r = ) --> R (',=) i s a s u r j e c t i o n

From (6 .3) we see t h a t AZ"") i s (isomorphic t o ) A ( Z r X ) .

r x ) --> P ( ( X , = ) ) i s r ep resen ted by

H(f,R) = ERhn{R(x) <-> ExAf([xl) Ix E X I .

But w e can t ake I'X 5 X (assuming ( X , = ) c anon ica l ly sepa ra t ed ) and so

by s e t t i n g f t o be t h e r e s t r i c t i o n of R t o I 'X , we see a t once t h a t

ER --> 3f .H(f , R )

is v a l i d , so t h a t [HI i s s u r j e c t i v e .

§15. The Uniformity P r i n c i p l e .

F i r s t a gene ra l un i formi ty p r i n c i p l e f o r Eff. Propos i t ion 15.1. L e t AX -> ( X , = ) be a s u r j e c t i o n and l e t ( Y , = ) be

an e f f e c t i v e o b j e c t . Then

V t J l V X E ( X , = ) .3y E (Y ,=)$J (x ,y ) --> 3y E (Y,=) .vx E (X ,=)$J (x ,y ) l

holds i n Eff.

Proof: Take (Y,=) s t r i c t e f f e c t i v e and cons ide r f i r s t t h e case when

( X , = ) i s AX. L e t e E [Vx.i 'y.$J(x,y) 1 . Then 0 E Ex each x i n X , so

b = v l ( e ( 0 ) ) E Ey f o r some y i n Y , unique as ( Y , = ) i s s t r i c t e f f e c -

t i v e ; and c = v 2 ( e ( 0 ) ) i s i n 1 $ ( x , y ) 1 . But t hen i f d = Xn.c, w e

f i n d t h a t he .<b ,d> r e a l i z e s t h e formula i n square b r a c k e t s . (There

i s no dependence on E + . ) The r e s u l t f o r a q u o t i e n t of AX is an

immediate consequence of t h e s p e c i a l case.

Page 44: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

208 J.M.E. HYLAND

W e have an immediate c o r o l l a r y .

Corol lary 15.2. The "Uniformity P r i n c i p l e "

\d$"dX E P ( I N ) .3n E IN $ (X,n) --> 3n E IN .VX E P(IN ) $ ( x , n I l

holds i n Eff. Proof: IN is an e f f e c t i v e o b j e c t and by (14.3) P ( R ) i s a q u o t i e n t

of A ( Z m ) .

The uniformity p r i n c i p l e is an extreme form of choice p r inc ip l e :

t h e choice func t ion is cons t an t because t h e domain i s amorphous while

t h e range i s well-behaved. Conditions on both t h e range and the

domain a r e necessary. Obviously t h e r e are non-constant func t ions

from IN t o N. As regards cond i t ions on t h e range, t h e r eade r may

l i k e t o show t h a t t he q u o t i e n t map from A E t o 62 does n o t s p l i t .

916. j -operators : f o r c i n g 2 --> A2 t o be i s o .

I n a topos, j -operators a r e maps j : 62 --> 62 s a t i s f y i n g

P 5 j ( p ) o r equ iva len t ly p --> q i j ( p ) --> j ( q )

j (phq) = j ( p ) h j ( q ) T 5 j ( T )

j ( j (p) ) = j (p) j ( j ( p ) ) 5 j ( p ) .

Of course t h e r e i s a l s o an i n t e r n a l o b j e c t of j -operators , a subobject n

of 62" which w e can desc r ibe i n Eff as fol lows.

Proposi t ion 1 6 . 1 . (i) The o b j e c t 62' i n Eff can be taken as ( E x , = )

where

Bf = n = t v p . f ( p ) <--> gip) 1

(ii) The o b j e c t o f j -operators i n Eff i s t h e subobject

of ( E x , = ) represented by t h e canonical monic de€ined by e i t h e r of t h e

above ways of giving t h e not ion of j -ope ra to r . A l t e r n a t i v e l y i t i s

(J,=) where J is t h e set of j -ope ra to r s and where

t j = k l = E j h [Vp . j (p ) <--> k ( p ) J

with Ej = I[j i s a j -operator ]I . Proof: (i) follows from ( 1 4 . 2 ) i n t he manner of (14.3) and (.ii) i s

then immediate.

Page 45: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 209

Remark. A s explained i n Johnstone [ 1 9 7 7 ] j -ope ra to r s correspond t o

topologies and so t o subtoposes. I t i s known t h a t t h e l a t t i c e of

j -operators under pointwise 2 i s a complete Heyting a lgeb ra

( i n t e r n a l l y ) . The r eade r should refer t o Fourman-Scott [ 1 9 7 9 ] f o r

an e x p l i c i t c o n s t r u c t i v e t reatment . I t i s perhaps worth commenting

f u r t h e r on t h e o rde r r e l a t i o n . W e have

B j 6 k j = EjAEkA n Vp.j(p) --> k ( p ) J

def ining t h e appropr i a t e subobject i n Eff. I f w e are looking a t

ex te rna l j -operators , then j 6 k i f f

Vp.j (p) --> k ( p )

i s v a l i d . F i n a l l y note t h a t i f j ( i ) is non-empty then j is t h e

degenerate topology which c o l l a p s e s t h e topos . L e t us look again a t t h e double negation topology. ( W e do no t

bother with a cons t ruc t ive ve r s ion . )

i f p non-empty, ( 7 7 ) p = U I T ~ ~ i s non-empty) = {I: Clearly then w e have the fol lowing lemma.

Lemma 1 6 . 2 . For any j , (-,-,) 5 j i f f n { j ( p ) / p non-empty} i s non-

empty . Proof: T r i v i a l .

W e now consider how t o fo rce monics t o be i s o . L e t a subobject

of (X,=) be given by a canonical monic A and de f ine a map R t o by

@ (p) = 112x (x,=) . A ( X ) --> p n . A

Clea r ly i f j f o rces A >--> (X,=) t o be i s o , t hen

@ A ( j (p) ) --> j (p)

i s v a l i d . This gives us a way t o desc r ibe t h e l e a s t j -operator

fo rc ing A >--> ( X I = ) t o be i s o .

Proposi t ion 16 .3 .

f o rc ing A >-> (X,=) t o be i s o , i s

In the above s i t u a t i o n , jA , t h e l e a s t j -operator

j A ( p ) = Uvq. CC@,q -> q ) h ( p --> q ) --> q ) ll .

Page 46: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

210 J.M.E. HYLAND

Proof: Obvious, a s i n t h e l o g i c t h i s s ays

l A ( P ) = A {Sl@,q 5 qAp 5 q l

where A i s t aken i n t e r n a l l y i n a. I t is easy t o check t h a t ( a s

s (9 (p ) -> @,(q)) i s v a l i d ) j, i s a j - o p e r a t o r . A ( -> 9)

W e now show t h a t f o r c i n g 2 >-> A2 t o be is0 c o l l a p s e s Eff t o

S e t s .

P ropos i t i on 1 6 . 4 . The l e a s t j - o p e r a t o r f o r c i n g 2 >-> A2 t o be is0

is ( 17).

Proof: L e t j be t h e l e a s t j -ope ra to r f o r c i n g 2 >-> A2, ob ta ined a s

i n (16.3) from @ : -> a . H e r e

@ ( p ) = {01+p u I l 1 - t ~ = I e l e ( 0 ) E p o r e (1 ) E PI. C l e a r l y it s u f f i c e s t o show ( 7 7 ) s j , t h a t i s by ( 1 6 . 2 )

n { j (p) Ip non-empty] i s non-empty.

nI j I n 1 ) In E 1 i s non-empty: f o r i f a i s i n

I n f a c t it i s enough t o show t h a t

VP,q. ( p -> q ) -> ( j ( p ) --> j (9) and x i s i n n { j { n I ) In E N 1 ,

t hen ( a ( X n . n ) ) x is i n n { j ( p ) lp non-empty].

Now t a k e b i n 1Vp.p -> j ( p ) ] , c i n " d p . j ( j ( p ) ) --> j ( p ) J , and

t a k e as 2 >-> A2 i s j-dense d i n j ( { O } ) n j ( { l l ) . Note t h a t

e = Xx.c( ( a x ) d ) i s i n l[Vp.@ ( j ( p ) ) -> j ( p ) 1 . Define u s i n g t h e

second r e c u r s i o n theorem an index f by

( f k ) ( 0 ) = b ( k )

( f k ) (1) = U ( l e a s t y . T ( e , S l ( f , k + l ) , y ) ) . 1

Now by a s t anda rd k i n d of argument, w e can show t h a t

1 S l ( f , k ) ( 0 ) = ( f k ) ( 0 ) = b ( k ) E j ( I k 1 )

and S l ( f , k ) 1 (1) = ( f k ) (1) = e ( S l ( f , k + l ) ) 1

1 a r e a l l de f ined , and t h e n w e see t h a t S l ( f ,O) is i n @ ( j ( { n I ) ) f o r

a l l n and so e ( S l ( f , O ) ) i s i n n { j ( { n } ) In E m l as r e q u i r e d . 1

5 1 7 . j - o p e r a t o r s and d e c i d a b i l i t y .

( 1 6 . 4 ) appea r s t o restrict t h e j - o p e r a t o r s i n Eff, b u t i n f a c t

w e can show t h a t t hey have a r i c h s t r u c t u r e . Apparently it w a s

Page 47: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 21 1

Powell who f i r s t r e a l i z e d t h a t t h e r e i s a connection between not ions

of degree and t h e fo rc ing of d e c i d a b i l i t y i n r e c u r s i v e r e a l i z a b i l i t y .

We con ten t ou r se lves wi th a p r e c i s e s t a t emen t and a ske tch of a

proof.

F i r s t w e g i v e a lemma of Andy P i t t s which s i m p l i f i e s t h e pre-

s e n t a t i o n of t h e proof .

Lemma 1 7 . 1 . I n t h e s i t u a t i o n of (16.3) j, can a l t e r n a t i v e l y be de-

f ined by

@,*(p) = n{q 5 I p ~ { * l 5 q and 0,Cq) 5 qlr

where * is an index f o r t h e empty p a r t i a l func t ion , so long as Ex

non-empty impl ies A(x) non-empty.

Proof: L e t us drop t h e s u b s c r i p t A. Note t h a t $ prese rves inc lu -

s ion . Hence because

P --> (PA{*)) 5 $ ( P ) --> @(PA{*))

i s v a l i d , w e can deduce t h a t

@ ( P ) 5 @ * ( P )

is v a l i d . Also w e have

@ ( @ * ( P ) ) 5 n { @ ( q ) IpAI*l 5 q and @ ( q ) 5 ql

- c n{qjpAt*} 5 q and @ ( q ) 5 q l = @ * ( P I ,

so t h a t

@ ( $ * ( P ) ) 5 @*(P)

is v a l i d , r a t h e r t r i v i a l l y . Thus by t h e d e f i n i t i o n of j i n (16.3)

w e have j s @ * i n Eff, and it remains t o show t h a t $ * s j .

W e can t ake a E [Vp,p -> j (p ) J and s i n c e @ Cj Cp)) S j (p) i n Eff, w e can t ake b E UVx E (X,=). ( A ( x ) -> jCp)) --> jCp) 1 . NOW

de f ine an index e by t h e second r ecu r s ion theorem a s fo l lows .

Cons i de r

t h a t (i)

and (ii)

a ( n ) , i f x = -a,*> e ( x ) = 'b(m)) (Xy.e(z(y))), i f x = <m,z>, z # *. f o r any p , t h e set SCp) = I x l e ( x ) E j ( p ) } . W e see e a s i l y

PAC*) 2 S ( P ) ,

$ ( S ( P ) ) 5 S ( P ) ,

Page 48: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

212 J.M.E. HYLAND

so w e can deduce t h a t $ * ( p ) 5 S ( p ) . Thus c l e a r l y e r e a l i z e s

Vp.@*(p) 5 j (p) , and t h i s completes the proof.

Now f o r A 5 I N , l e t D (n) = {<O,n>ln E A}u{<l,n>ln 4 A} so t h a t A

DA r ep resen t s canonical ly t h e subobject

AV i A >-> m , the " d e c i d a b i l i t y of A" . W e w r i t e

$,(PI = U3n E IN .DA(n) --> p n , and kA f o r the l e a s t j -operator generated by $,, t h a t i s , the l e a s t

j -operator fo rc ing A t o be decidable .

$A* i n Eff. Theorem 1 7 . 2 . (External v e r s i o n ) .

reducible t o B.

Proof:

(holds i n E f f ) . Suppose t h a t e ( n ) E JIB (DA(n)) f o r each n. W e wish t o show how t o

compute A from B , t h a t is how t o determine DA(n) from a knowledge

of D ( m ) f o r f i n i t e l y many m.

e i t h e r x is of form <y,*> and w e e a s i l y see t h a t y must be DA(n) so

w e a r e home,

- o r x is of form < m , e l > say i n which case el E D B ( m ) --> $B* (DA(n) )

so w e take x1 t o be e l ( < O , m > ) o r e l ( < l , m > ) a s appropriate ,

x1 E JIB* (DA(n)) , and r epea t t h i s process.

qB*, t h i s terminates i n a f i n i t e number of s t e p s giving DA(n) a s

required.

Suppose conversely t h a t A i s Turing reducible t o B v i a an index f .

Define using t h e second recursion theorem e Cn,y) where y i s (a code

f o r ) a f i n i t e set of numbers of form <m,O> o r < m , l > as follows.

By ( 1 7 . 1 ) kA i s equal t o

kA 5 kB i n Eff i f f A i s Turing

Note f i r s t t h a t kA 5 kg i n Eff i f f Vn.$,*(DA(n)) is v a l i d

*

* L e t x = e ( n ) E QB (DA(n) ; then B

From t h e d e f i n i t i o n of

<k ,*> , i f t h e r e i s a computation {f}Y(n) = k (using only information i n y ) ,

e ( n f y ) = 1 < m , q > , i f t h e computation { f l Y ( n ) asks f o r a value n o t i n y f and g is an index for < m , i > -> e ( n , y u I < m , i > l ) . i

Page 49: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 213

I t is easy t o see t h a t f o r a l l n , and f o r y information t r u e of B,

e ( n , y ) is def ined and i n @,*(D,(n)).

set, w e have f o r a l l n ,

In p a r t i c u l a r f o r y t h e empty

e (n,y) i n @,* (DA(n) ) . Thus Vn.QB* (DA(n)) i s v a l i d .

This completes the proof .

Remark. In f a c t t h e r e i s a proof of the implicat ion from r i g h t t o

l e f t along t h e following l i n e s : i f j fo rces B decidable , then the

statement t h a t A i s reducible t o B and t h a t t he computation is always

defined, a r e almost negat ive f o r E f f . and so hold i n E f f . . hence i n

E f f . A i s decidable , t h a t is j fo rces A decidable. -1

-1 I -1 '

W e cannot f i n d a crude i n t e r n a l vers ion of (17 .2 ) i n view of

Goodman C19781. However t h e proof of ( 1 7 . 2 ) is e f f e c t i v e , so w e can

ge t something o u t of it.

with A i s i n t e r n a l l y defined i n Eff. "A Turing reducible t o B": w e mean the n a t u r a l not ion of computabili ty

r e l a t i v e t o ( p a r t i a l ) c h a r a c t e r i s t i c funct ions. W e ob ta in a r e s u l t

by r e s t r i c t i n g a t t e n t i o n t o c losed subsets of I N , t h a t i s t o

P ( I N ) = {A 5 IN IVn. 9 - n B A --> n E A ] .

Proposi t ion 17.3. The s ta tement

Clear ly t h e funct ion which a s soc ia t e s kA

W e must say what w e mean by

VA,B E F ( I N ) . (A Turing reducible t o B) <--> k A ' kg

holds i n Eff. Proof: By the e f f e c t i v i t y of t he proof of ( 7 . 2 ) .

118. General remarks on the e f f e c t i v e topos.

The pleasing f e a t u r e of t h e e f f e c t i v e topos i s t h a t i n i t ,

ideas about e f f e c t i v i t y i n mathematics,seem t o have t h e i r na tu ra l

home. W e mention the two main examples.

1) Constructive real ana lys i s . W e have t r i e d t o i n d i c a t e t h a t t h i s

is what a n a l y s i s i n Eff i s i n essence i n 118-13. It is worth not ing

how t h e r e a l i z a b i l i t y l o g i c makes d i s t i n c t i o n s f o r us. Consider t he

Page 50: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

214 J.M.E. HYLAND

examples (Kreisel 119591 t h a t t h e in t e rmed ia t e va lue theorem holds

c l a s s i c a l l y bu t n o t e f f e c t i v e l y f o r r e c u r s i v e (cont inuous) func t ions

on t h e r e c u r s i v e r e a l s . I n Eff ,

vf E lRIR .f (O)<oAf (1)>0 -> 3 X E ( 0 , l ) f ( X ) = 0

i s f a l s e whi le

vf E ~ l R . f ( 0 ) < O A f ( l ) > O --> -1 -13x E ( O , l ) f ( X ) = 0

i s t r u e ( a s it is t r u e i n Sets and e q u i v a l e n t t o a nega t ive formula) .

2 ) E f f e c t i v e a lgeb ra . W e have n o t d i scussed t h i s a t a l l , bu t i t

seems worth po in t ing o u t t h a t t h e d e f i n i t i o n s have a n a t u r a l meaning

i n Eff. A r e c u r s i v e l y p re sen ted f i e l d (see Metakides-Nerode C19791

i s an enumerable (dec idab le ) f i e l d i n Eff. I t has a s p l i t t i n g

a lgor i thm i f f i r r e d u c i b i l i t y of polynomials is dec idable i n Eff. Thus t h e e f f e c t i v e con ten t o f a r e c u r s i v e l y presented s t r u c t u x e

corresponds t o p r o p e r t i e s of it which hold i n Eff. This sugges t s

t h a t p o s i t i v e r e s u l t s i n e f f e c t i v e a lgeb ra should be e s t a b l i s h e d by

proving r e s u l t s i n c o n s t r u c t i v e l o g i c from axioms which ho ld i n Eff, and i n t e r p r e t i n g t h e r e s u l t s i n Eff. That i s , one should use t h e

axiomatic method. Of course , nega t ive r e s u l t s ob ta ined i n e f f e c t i v e

a lgeb ra can be i n t e r p r e t e d i n Eff t o g ive independence r e s u l t s .

What w e l a c k , above a l l , i n our t r ea tmen t of t h e e f f e c t i v e topos,

i s any real informat ion about ax iomat iza t ion analogous t o t h e r e s u l t s

ob ta ined i n T r o e l s t r a 119731 ax iomat iz ing r e a l i z a b i l i t y over both

Heyting and Peano a r i t h m e t i c . Of course , one would expec t t o look a t

t h e e f f e c t i v e topos def ined over a topos o t h e r than Sets (say over

t h e f r e e topos wi th n a t u r a l number o b j e c t ) t o g e t a r e s u l t co r re s -

ponding p r e c i s e l y t o an ax iomat iza t ion . But a l l I wish t o p o i n t ou t

is t h a t ( d e s p i t e t h e sugges t ive work of P i t t s C19811 on i t e r a t i o n )

w e have no good informat ion i n t h i s area. W e can n o t p rope r ly be

s a i d to understand r e a l i z a b i l i t y u n t i l w e do.

Page 51: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

The effective topos 215

REFERENCES

M . J . Beeson C19771, P r i n c i p l e s of continuous choice , Ann. Math. Log. - 1 2 , 249-322.

E. Bishop C19671, Foundations of Cons t ruc t ive Analys is (McGraw-Hill).

A . Boileau & A. Joya l 119811, La log ique des topos , J.S.L. El 6-16.

D. van Dalen C19781, An i n t e r p r e t a t i o n of i n t u i t i o n i s t i c a n a l y s i s , Ann. Math. Log. 13, 1-43.

Logic (ed. Barwise) (North-Holland) . M.P. Fourman C19771, The l o g i c of t o p o i , i n Handbook of Mathematical

M.P. Fourman & R . J . Grayson, t h i s volume, Formal spaces .

M.P. Fourman & J .M.E. Hyland 119791, Sheaf models f o r a n a l y s i s , i n Appl ica t ions of Sheaves (Spr inger Lec ture Notes 753) .

M.P. FQUrman & D . S . S c o t t C19791, Notes on Sheaves and Logic, i n Appl ica t ions of Sheaves (Spr inger Lec ture Notes 753) .

R.O. Gandy & J .M.E . Hyland C19771, Computable and r e c u r s i v e l y countable func t ions of h ighe r t ype , i n Logic Colloquium '76 (ed. Gandy, Hyland) (North-Holland) .

a r i t h m e t i c , J.S.L. 43, 497-501.

J.M.E. Hyland C19771, Aspects of c o n s t r u c t i v i t y i n mathematics, i n Logic Colloquium '76 led. Gandy, Hyland) (North-Holland) .

N.D. Goodman C1978 1, The non-cons t ruc t ive con ten t of sen tences of

J .M.E. Hyland 119791, F i l t e r spaces and continuous f u n c t i o n a l s , Ann. Math. Log. 16, 101-143.

Tr ipos theo ry , Math. Proc. Carnb. P h i l . SOC. E, 205-232. J .M.E. Hyland, P .T. Johnstone & A.M. P i t t s C19801 CHJP C198011,

P.T. Johnstone C19771, Topos Theory (Academic P res s ) . S .C. Kleene C 1945 1 , On t h e i n t e r p r e t a t i o n of i n t u i t i o n i s t i c number

theory , J.S.L. El 109-124 .

Cambridge Summer School i n Mathematical Logic (Spr inger Lec ture Notes 337) .

S.C. Kleene C19731, R e a l i z a b i l i t y : A Ret rospec t ive Survey, i n

S.C. Kleene & R.E. Vesley C19651, The foundat ions of i n t u i t i o n i s t i c mathematics (North-Holland) .

G. Kreisel C19591, I n t e r p r e t a t i o n of a n a l y s i s by means of cons t ruc- t i v e f u n c t i o n a l s of f i n i t e type , i n Cons t ruc t iv i ty i n Mathe- mat ics (ed . Heyting) (North-Holland) .

M. Makkai & G.E. Reyes 119771, F i r s t o r d e r c a t e g o r i c a l l o g i c

G. Metakides & A . Nerode C19791, E f f e c t i v e con ten t of f i e l d theory,

(Spr inger Lec ture Notes 6 1 1 ) .

Ann. Math. Log. 17, 289-320.

Page 52: [Studies in Logic and the Foundations of Mathematics] The L. E. J. Brouwer Centenary Symposium, Proceedings of the Conference held in Noordwijkerhout Volume 110 || The Effective Topos

216 JM.E. HYLAND

J . R . Moschovakis C19731, A t o p o l o g i c a l i n t e r p r e t a t i o n of second o rde r i n t u i t i o n i s t i c a r i t h m e t i c , Comp. Math. 26, 261-276.

A.M. P i t t s C19811, The theo ry of t r i p o s e s , Cambridge Ph.D. Thes is .

H . Rogers j n r . C19671, Theory of computabi l i ty ( M c G r a w - H i l l )

D.S. S c o t t C19701, Extending t h e i n t u i t i o n i s t i c a n a l y s i s 11, (ed. Xino, Myhi l l , Vesley)

r e c u r s i v e func t ions and e f f e c t i v e

topo log ica l i n t e r p r e t a t i o n t o i n I n t u i t i o n i s m and Proof Theory North-Holland) .

D.S. S c o t t C19791, I d e n t i t y and e x i s t e n c e i n i n t u i t i o n i s t i c l o g i c , i n Appl ica t ion of Sheaves (Spr inger Lec ture Notes 753) .

W.W. T a i t C19751, A r e a l i z a b i l i t y i n t e r p r e t a t i o n of t h e theo ry of spec ie s i n Logic Colloquium (Spr inger Lec ture Notes 453) .

A.S. T r o e l s t r a C19731, Metamathematical I n v e s t i g a t i o n of I n t u i t i o n - i s t i c Ar i thmet ic and Analys is (Spr inger Lec ture Notes 344).

I . D . Zaslavski? & G.S. Cextin C19621, On s i n g u l a r cover ings and p r o p e r t i e s of c o n s t r u c t i v e func t ions connected wi th them, Trudy hat: I n s t . S tek lov 67, 458-502 ( t r a n s l a t e d i n AMS Trans l a t ions - ( 2 ) =, 41-89 ( 1 9 7 1 ) ) .