Scribd
Upload
81 views

An Intuitionistic Version of Cantor's Theorem

This document presents an intuitionistic version of Cantor's theorem within Martin-Löf's Intuitionistic Type Theory. It proves that there is no surjective function from the type of natural numbers N to the type of functions from N to N, written as N→N. The proof assumes the existence of such a function f and constructs a function λx.s(f(x)(x)) that cannot be in the image of f, obtaining a contradiction. More generally, a similar result holds for any type A where a function from A to A has no fixed point.

Uploaded by

Rebecca Webb
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
81 views

An Intuitionistic Version of Cantor's Theorem

This document presents an intuitionistic version of Cantor's theorem within Martin-Löf's Intuitionistic Type Theory. It proves that there is no surjective function from the type of natural numbers N to the type of functions from N to N, written as N→N. The proof assumes the existence of such a function f and constructs a function λx.s(f(x)(x)) that cannot be in the image of f, obtaining a contradiction. More generally, a similar result holds for any type A where a function from A to A has no fixed point.

Uploaded by

Rebecca Webb
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

An intuitionistic version of Cantors theorem

Dario Maguolo, Silvio Valentini


Dipartimento di Matematica Pura ed Applicata
Universit`a di Padova
via G. Belzoni n.7, I35131 Padova, Italy
[email protected]
September 24, 1996
Abstract
An intuitionistic version of Cantors theorem, which shows that there
is no surjective function from the type of the natural numbers N into the
type N N of the functions from N into N , is proved within MartinL
ofs Intuitionistic Type Theory with the universe of the small types.

Mathematics Subject Classification: 03B15, 03B20.


Keywords: Intuitionistic type theory, Cartesian closed categories.

The intuitionistic Cantors theorem

In this work we show that within Martin-L


ofs Intuitionistic Type Theory with
the universe of the small types [ML84, NPS90] (ITT for short in the following)
a version of Cantors theorem holds, which shows that there is no surjective
function from the type of the natural numbers N into the type N N of the
functions from N into N . As the matter of fact a similar result can be stated
for any not-empty type A such that there exists a function from A into A which
has no fixed point, as is the case of the successor function for the type N . In
order to express Cantors theorem within ITT we need the Equality proposition:
let A is a type and a, c A, then by a =A c we mean the Equality proposition
for elements of type A [NPS90].
Theorem 1.1 (ITT Cantors theorem) Let N be the type of the natural
numbers; then
(f N (N N ))(y N N )(x N ) f (x) =N N y
To prove this theorem some lemmas are useful. Indeed we need to obtain a
contradiction from the assumption
(f N (N N ))(y N N )(x N ) f (x) =N N y
1

i.e. from the two assumptions


f N (N N )
and
(y N N )(x N ) f (x) =N N y
By using the basic idea of the classic proof of Cantors theorem, from the first
assumption we can prove x.s(f (x)(x)) N N , where s : (x : N )N is the
successor function, by the following deduction:
[x : N ]1

f N (N N )
[x : N ]1
f (x) : N N
f (x)(x) N
s(f (x)(x)) N
1
x.s(f (x)(x)) N N
We can now use this function in the second assumption in order to obtain
(x N ) f (x) =N N x.s(f (x)(x)). So our problem becomes to obtain a contradiction from the two assumptions x : N and f (x) =N N x.s(f (x)(x)). We
can use these assumptions to prove, by transitivity of the equality proposition,
that f (x)(x) =N s(f (x)(x)) is true since in general if A and B are types and
a =A c and f =AB g then f (a) =B g(c) and obviously (x.s(f (x)(x)))(x) =N
s(f (x)(x)) is true.
We can thus re-state our aim by saying that we have to prove that ITT is not
consistent with the assumption that the successor function has a fixed point. To
prove this result we can transpose a well known categorical arguments within
ITT [L69, HP90]. Let us recall that we can solve the usual recursive definition
of the sum between two natural numbers

n+0=n:N
n + s(x) = s(n + x) : N
by putting n + x Rec(x, n, (u, v) s(v)). Then the following lemma can be
proved by induction.
Lemma 1.2 For any n, x N , n + s(x) =N s(n) + x.
As for the sum, we can solve the recursive equation for the predecessor function

p(0) = 0 : N
p(s(x)) = x : N
by putting p(x) Rec(x, 0, (u, v) u), and then that for the subtraction

n0=n:N
n s(x) = p(n x) : N
by putting n x Rec(x, n, (u, v) p(v)).
2

Lemma 1.3 For any x N , (n N ) (n + x) x =N n.


Proof. By induction on x. If x = 0 then (n + 0) 0 =N n + 0 =N n; let
us now suppose that (n N ) (n + x) x =N n, then (n + s(x)) s(x) =N
p((n + s(x)) x) =N p((s(n) + x) x) =N p(s(n)) =N n.
2
We can apply this lemma to the case n = 0 and obtain the following corollary.
Corollary 1.4 For any x N , x x =N 0.
Proof. Immediate, since 0 + x =N x holds for each x N .

Now we conclude our proof. Let us write to mean the fixed point of the
successor function, i.e. =N s(); then the following lemma holds.
Lemma 1.5 For any x N , x =N .
Proof. Again a proof by induction on x. If x = 0 then 0 =N and,
supposing x =N , we obtain s(x) =N p(x) =N p() =N p(s()) =N
.
2
So we proved that =N 0 by corollary 1.4 and also that =N
by lemma 1.5; hence 0 =N =N s(). Finally we reach a contradiction.
Theorem 1.6 For any x N , (0 =N s(x))
Proof. By an elimination rule for the type N , from the assumption y : N , we
obtain Rec(y, , (u, v) ) U0 , where U0 is the universe of the small types,
is the empty type and is the one-element type. Now let us assume that x N
and that 0 =N s(x) is true, then Rec(0, , (u, v) ) =U0 Rec(s(x), , (u, v) )
since in general if A and B are types and a =A c is true and b(x) B [x : A] then
b(a) =B b(c) is true. Hence, by transitivity of the equality proposition, =U0
since =U0 Rec(0, , (u, v) ) and Rec(s(x), , (u, v) ) =U0 . Then,
because of one of the properties of the equality proposition for the elements of
the type U0 , is inhabited since is and hence, by discharging the assumption
0 =N s(x), we obtain that (0 =N s(x)) is true.
2
Thus the proof of theorem 1.1 is finished since we have obtained the contradiction we were looking for. Anyhow we stress on the fact that a similar result
holds for any type A such that there exists a function from A into A with no
fixed point. In fact, in this hypothesis, we can prove that there exists a function
g from A (A A) into A A which supplies, for any function h from A
into A A, a function g(h) A A which is not in the image of h.
Theorem 1.7 Let A be a type; then
(f A A)(x A) (f (x) =A x)
(g (A (A A)) (A A))
(h A (A A))
(x A) (g(h) =AA h(x))
3

The proof of this theorem is similar to the first part of the proof of theorem
1.1. In fact we only have to use the function f A A, instead of the successor
function, to construct the function g k.y.f (k(y)(y)) (A (A A))
(A A) such that, for any h A (A A) and any x A, allows to prove
h(x)(x) =A f (h(x)(x)), which is contrary to the assumption that the function
f has no fixed point.

References
[L69]

Lawvere, F.W., Diagonal arguments and cartesian closed categories,


in Category Theory, Homology Theory and their Applications II,
Lecture Notes in Mathematics, n.92, Springer 1969, pp. 134-145.

[HP90]

Huwig, H., A. Poigne, A note on inconsistencies caused by fixpoints


in a cartesian closed category, Theoret. Comput. Sci., 75 (1990), pp.
101-112.

[ML84] Martin-L
of, P., Intuitionistic Type Theory, notes by G. Sambin of a
series of lectures given in Padua, Bibliopolis, Naples, 1984
[NPS90] Nordstr
om, B., K. Peterson, J. Smith, Programming in Martin-L
of s
Type Theory, An introduction, Clarendon Press, Oxford, 1990

You might also like