This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
n
be
.> since a(~n,n# n' ~n.a{en,n} is a choice sequence, say S, we have
< 2 -n. Since
n ~ n . The correspondence
is equivalent with
X.
is lawllke.
In case the sequencen is lawllke relative to a choice parameter, correspondence is lawllke relative to the same choice oarameter.
the
12.2. A basis
n
for a separable metric space
(n is called discrete if theorem however.
Theorem. Let
<X,p>
find a sub-sequence Proof. Let
b
<X,p>
~nAm(Pn ~ P m V P n
<Pan>n
which is a basis for
be a lawlike function such that
a
by recursion,
and we put
aO : O. We want to achieve
= Pm ))" We can prove the following
be a separable metric space with a basis
Jb(n,m,k).2 -k - p(pn,Pm) J < 2 -k. Now we define
is not necessarily discrete
~
(Po,Pl ) < 2 -2 § al = O, (Po,Pl)
~ 2 -1 § al = 1.
<X,p>
n . Then we can
and which is discrete.
-
66
-
This can be done by stipulating b(O, I, 2) < 2 § al = 0 al = I otherwise. For
Ib(O, 1, 2)2 -2 - 0(po , pl)l < 2 -2 , hence
b(O, 1, 2) < I + 4p(po , pl ) < 2. Conversely,
(b(O, I , 2) >1 4P(PO, Pl ) - 1 1 { 1, i . e . Now suppose aO,..., ac(k - I) ~d(i A cCk-1) § al & k-l). Let that
<Jo,...,Jn >
<J~,..., J~> !
JO''''' Jm-I (a k)
such that
Jk >
I I< m - 1
so
as follows. Take J~
~_ i
such that p(pj~ , pjm ) < 2 -k-1
i ~ m-i o(Pjl , pjm ) { 2-k, then
Jm = Sm "
: 0 & I & m} by d.
Jm = Jm-1
p(Pj~ , pj ) < 2 -k-1
for an I & m-1. Then
Ib(Jm, J~, k+l) 2 -k~l - 0(pj~ , pjm) I < 2 -k-i, therefore b(Jm, J~, k + 1) < 1 + 1, so
Then
suppose
b(Jm, J~, k+l) & I, hence
p(pj~, pjm ) { 2 -k
b(Jm, J~, k + 1) 9 I
for all
This shows that our construction of
for every
i & m-l, hence J~
for
such that
Jm = Jm-1"
I f for every
such
= k~
J~ = Ji
Then we put
Conversely,
Ok
!
Denote Inf {b(Jm , J~, k+l)
Suppose
step, such that
be an ordering of the elements of
n I < n 2 § Jn+nl < Jn+n2
from <Jo''''' Jk >
If there exists an
k th
{aO,..., ac(k-1)}
have been chosen, we want to choose
then
(b k)
implies
2) 9 1 .
be an orderlng of the elements of
{0,..., k} - {Jl,.., Jn }
If
b(O, l s
<Jn+l'''''
implies
P(Po' Pl ) ~ 2"I
to be constructed after the
hi< n 2 * in1 < Jn2 . Let
Construct
0(po , pl ) < 2 -2
satisfies
d
t <m-1. d > 1.
(ak), (~).
i An.
-
67
-
Finally we take ' a(c(k-l) + I) = Jn+l Apparently
for
1 A i
k-n.
c(k) = c(k-1) + (k-n).
Now we prove the discreteness
of
<PaO''''' Pac(k) >
k. Suppose the discreteness
for every
to be proved. From the fact that follows that PJm
in case
a
Jm(Jm ~
<Pan>n
by proving the discreteness
(ak) , (b k)
{Jo''''' Jm-1 })
<Pao''''' Pac(k-l) >
are satisfied at every step, it
is included in
lies apart from every element of
<PaO''''' Pac(k) >
of
of
{J~,..., J~} only PJm-1 }. Therefore
(PJ0 ' ' ' "
is discrete.
Thus we obtain the desired conslusion by ordinary induction. There remains to be shown that <Pan>n is a basis. In view of 12.1.6 m it is sufficient to show <Pan>n to be dense in the space. To see this, we remark that for every
I < k
~(0 & ~ & c(k)) at the
k th
So if we take
either
PI~{Pao'''''
such that
Pac(k) }p or there exists an
~(Pi' PaL ) < 2"k+l
step, there exists a
J~
(for if
such that
i ~ {J~,..., J~}, i ~ k
o(Pj~ , Pl ) < 2 "k+l, m ~ k).
p(p, pn ) < 2 -k-l, then we can effectively find a
~ c(k+2), such that
0(pn , paL) < 2 -k-l, hence
PaL'
o(P, PaL) < 2-ko Q.e.d.
12.3. 12.3.1. Theorem. Every complete sepaFable metric space represented by a spread.
r
=
(IT 1966], 2.2.1)
<X,
can
o>
be
Proof. Letn be a discrete basis for r, and let <s(k, n, m)> m be a lawlike enumeration of all indexes i (possibly with repetitions) such that 0(pn, pi ) < 2 -k. We may suppose s to be chosen such @hat Pn = Pm - * A k A I
(s(k, n, i) = s(k, m, I)).
~hls is not essential but technically convenient.) taking a = Xx.l, and fixing ~ by
~
Now we define
5
=
= Pko
= Pm § r
Then Just as in 11.3, S
represents
r.
kn-l' k> = Ps(n, m, k)
by
-
68
-
12.3.2. Lemma. Let r = 9 p9 be a complete metric space, then the spread representation indicated in the proof of 12.3.1 has the following property: AxgxV~(lim
r
: x ^ A k ( U ( 2 -k, x ) ~ W ~ k ) )
where W n = {y : V S ~ n ( l i m Proof.
Take a certain
9qn~n ~_ 9
such that An(p(qn'
so
x~ X
$~8 = y)}.
and let
y
be such that
p(x, y) 9 2 -k-2. Construct
~n(e(qn , x) 9 2-n-2). Then apparently
qn+l ) ~ P(qn' x) + P(qn+l' x) 9 2 -n)
r
Vk-l> = qk-1 " Clearly,
p(y, qk.1) ~ ~(x, y) + o(x, qk.1) < 2 -k-2 + 2"k-i 9 3 .2"k-2 9 Then we can find An(p(q~,
(q~>n ~
9
y) 9 2-n-2), hence
Then the sequence
9
such that
= y'
~(q~, qk_i ) 9 2 -k.
qi''''' qk-i' qk' qk+i" q k + 2 ' ' "
verified. This sequence is equal to Therefore
lim 9
~$8
9 E S, as is easily
for a suitable
8, with
~k = 9
y&W 9 "'''Vk-l>
Q.e.d.
12.4. 12.4.1. Definitlon. A topological space is a pair species of V such that
(b)
Finite intersections to
7
,
~a
species of sub-
and arbitrary unions of elements of ~
again belong
"
This definition is exactly the same as one of the well known classical definitions. 12.4.2. Definition. f is called a continuous mapping from a topological spaceinto a space if f is a mapping (V)V' such that
The definitions of homeomorphlsm, given accordingly.
bamis of open species, and nei~hbourhood ' are
A point p is said to be a closure point of X ~ V , if (every neighbourhood of p contains a point of X).
AW6 ~(p~W
* Vq(q6X~W))
9
-
X'~the closure of
X
69
-
consists of the species of closure points of
X.
So far, all things look very much the same as in classical topology. But we must be aware of the fact that classically equivalent definitions are not necessarily intuitionistically equivalent. For example, the notion of weak continuity, characterized by:
Axgv,
(r'1[x'] - : f-l[x-])
(the complete original or counterimage of a closed set is always closed) is classically equivalent with, but intuitionistically weaker than continuity (IT 1966], 2.1.8). With a metric space
~
we may associate in the usual way a topological space
Ap6xV,,O
^qev
(p(p, q) 9 c ~ q e X ) .
Or if we put U(r
p)
= {q : oCP,
q) 9 r
then
The r
U(r
p)
constitute a basis for ~ ( ~ ) .
Usually we shall not distinguish between
and
when the meaning
is clear from the context. 12.5. The results mentioned in this subsection may serve as an illustration of the more typical questions of intuitionistic topology. i2.5.1. Definition. A space r = ~ V , ~ > complete, metrizable (i.e. ~ = ~ ( p ) , 9 representable by a finitary spread.
is said to be located compact if complete, for a suitable e)
r is and
12.5.2. Definition. A space r = ( V , ~ , is said to be locall~ compact, if r is complete, metrizable, and if every p ~ V possesses a located compact neighbourhood U(i.e. U is located compact in the relative topology ~' = { X ~ U : X e ~ } induced by ~ ). Closed intervals
[a, b]~__ ~ ,
a ~ b
are examples of located compact spaces;
-
70
-
, the species of (choice-) reals is an example of a locally compact space. Arbitrary pointspecies in a topological space may be defined very nonconstructively; so one feels the need for a subclass of pointspecies for which some extra information is available. The notion of a located pointspecies turns out to be useful: 12.5.3. Definition. Let r = < V j y > X~V is said to be located, if Ap&VAU~
be a topological space.
q'~(p~U § {Vq(q~U,',x) ~, V w 6 ~ C p G W
i2.5.4. Definitlon. A species metrically located if
X~V,
^wr~x = ~
)}).
a metric space, is said to be
p(x, X) = inf (p(x, y) : y ~ X }
is defined for every
x~V.
The significance of the notion of locatedness becomes clear from the following result: 12.5.5. Theorem. Let r = < V , ~ ( p } ~ be a located compact space. Then an inhabited Xm.V is located iff X is metrically located. (Proof e.g. in IT 1988 A], 3.14(a)). So metrical locatedness is a topological notion. In general, this is not the case; e.g. ~ can be metrized in such a way that a certain located pointspecies cannot be proved to be metrically located. But the following theorem holds: i2.5.6. Theorem. To every locally compact space < V , ~ > we can find a metric p such that ~ = ~ ( p ) jcomplete, and every located inhabited species X ~ V is metricall~ located with respect to ~ (the converse is trivial). A proof may be found in IT 1968 A], 4.6. 12.5.7. Definitions.
Let
be a topological space, X ~ V ,
Interior X = ~ t
X = {p : Vr
U(r p ) ~ X } .
X ~ Y ~m AP ~ V ( p ~ Y v p ~ X ) is the analogue of the classical relation X" ~ 12.5.8. Theorem. Let located, X~Y
iff
~V,0>
Y~V.
Int Y.
be a complete separable metric space. Then if
X" ~ I n t
X
Y.
Proof. The proof is contained in IT 1968 B], and uses continuity principles for choice sequences in an essential way.
Is
-
w 13. App!Ication~of
71
-
the c ontinulty principles
and the fan theorem
13.1. We start wlth some applications In its simplest form
A~Vx x(~, x, % , . . . ) Proof-theoretlcally,
Def!nltlon.
mapping from
V
Let
n~___V
wlth a limit
fp.
13.1.2.
Theorem.
r
metric space,
f
Let
~'>
consequences
and let
Is sald to be sequentlally wlth a limit
be a separable
a mapping from
r
.
Is not an essential strengthening
r =, r' =
V'. f
n
for choice sequences
= j y § x(B, x, % , . . . ) )
[Ho,K 1967]) but Its mathematical
Into
verging sequence
§
thls principle
(as follows e.g. from and rather elegant. 13.1.1.
of the continuity principle
p
f
are interesting
be a (lawllke)
continuous
If every con-
Is mapped onto a converging
complete metric space,
Into
of analysls
r'. Then
f
r'
sequence
a separable
Is sequentially
continuous.
Proof. See IT 1967]. This theorem Is contained In the stronger theorem 13.1.6 given below. 13.1.5. space.
Theorem. If
f
continuous, Proof.
Let
r'
be a separable metric space,
Is a (lawllke) mapping from then
f
([T 1967]).
r'
Into
and let
r"
and
F" f
be a metric
Is sequentially
Is continuous. Let
n be a lawllke sequence of points dense in r' = . Let r" = <W,0'>. Let x be an arbitrary point of r' and suppose llm n = x. Now we want to construct i such that p(pB{n,l},X)
O(pl, The construction From
n
< 2.2 -n
for every
II
(1)
x) ~ 2 -n § Pl ~j
proceeds
as follows.
we obtain effectively
n , llm n
= x, such that
An(p(pyn J x) < 2-n). Let
b(n, m, k)
be a (lawlike)
An~r~k
enumeration
of rationals
Ip(pn , pm ) - b(n, m, k) I < 2 -k.
such that
-
?2
-
We have b(i, v(n+2), n+2) 9 6.2 -n-2 v b(i, v(n§ If
0(pi , x) ~ 2 -n, then
Ip(pi, py(n+2))
P(Pi' Py(n+2) )
n+2) ~ 6.2 -n-2.
5'2"n-2; also
9
- b(i, v(n+2), n+2) I 9 2 "n-2, therefore
b(i, v(n+2), n+2) 9 6.2"n'20 If
b(i, y(n+2), n+2)
9
6.2 "n'2, then
0(pi , pv(n+2))
7.2 "n-2, so
<
P(Pi' x) 9 8.2 -n-2 = 2.2 -n. Therefore we may take for
i
b(J, y(n+2), n+2) 9 6.2"n-2; then tinuous operation from
~)).
Now we define a spread
<~x.i,+>
an enumeration of the
(I)
is satisfied.
for every
Furthermore,
a'@*s
Ps(n,in}"
is obtained by a con-
9
is a fundamental sequence in
for any fixed natural number
A6VnAm(P'(fPs{n+m,
such that
by stipulating
$=
(6
J
P'
converging towards
~ :
6(n+m)}' fx) 9 2-v),
hence with an application of the continuity principle we find, for any n, n' such that A6(~'n'
~n'
=
+
Am(p'(fps(n+m ' ~(n+m)}'
Without restriction we may suppose Ai(p(pi,
2-v)).
<
n' = n. From this it follows that
x) < 2 -n * o'(fPi' fx) < 2-v).
For whenever
P(Pi' x) 9 2 -n, there is a
Now take any
y
n_~
fx)
6',
such that such that
f is sequentially Therefore.
6
with
~'n = ~n,
6{n, 6n} = i.
~(x,y) < 2 -n. Then there is a sequence Am(~(p6m , x) 9 2"n), l~m P6n = y'
continuousj hence for a suitable
k
P'(fP~k'
fy) < 2"v"
p'(fy, fx) $ D'(fp6k, fy) + D'(fp~k, fx) 9 2 "~ + 2 -~ = 2.2 -~ and thus Ay(p(y,
x) 9 2 -n * ~'(fy, fx) < 2.2 -v)
.
q.e.d.
x.
- 73
13.1.4. Theorem. Let Let
r
:
r
be covered by a species
-
be a separable, complete, metrizable space.
O~
of pointspecles; then
r
Is also covered by
the species of interiors of these pointspecies. Proof. Let us use Then we have By 12.3.1
W
hxgV
as a variable for polntspecles. VW(W60~^ x@W).
there is a representation
~x.l,r
for
r.
Hence we also have:
Now we apply (13) of 10.7
and we obtain
A~VnVWAs(~n : ~n § llm r We then use lemma 12.3.2. Take any
x, and find lim r
We have an
n
Hence
W6~
: ~n
U(2 -n, x)_~W, so
This proves
AxEV
such that
: x ^ A k ( U ( 2 -k, x ) ~ W ~ k ) .
and a
As(~n
~
such that
§ lim
~
86W).
x &Int W.
VW(W~^x~Int
13.1.5. Remark~ In case O ~ w a s
W). a subspecies of
N, it would have been sufficient
to use (3) of w 10 instead of (13). We remark that the lawllkeness of ~
and the elements of ~
is essential.
For consider the covering by species of one element; then the interiors do not cover
rl The reason is
that arbitrary points are not lawlike. The theorem may be
generalized however to the case where the elements of O ~ contain choice parameters from a fixed finite collection, From 13.1.4
one also obtains
13.1.6. Theorem. let
r
be a complete, s e p a r a b l % m e t r i c
separable metric space; then any mapping from Progf. Either we may combine 13.1.2
r
into
i : i~N},
r"
r"
be a
is continuous.
and 13.1.3, or we deduce the theorem directly
form 13.1.4, by looking for any fixed natural number (W
space, and let
defined as follows. Let
n
~
at the coverings
be a basis for
r" :; then
- 74 -
W , i = {y : y ~ r ^ P(Pi, fy) <2-~} " Then an application of i3.1.4
readily yields continuity of
f.
i3.2. Intermediate in strength between (i3) of w iO and the correspom44ing stronger form A~ VXa~(~,X a) § VeAn(en ~ 0 §
VxeAs&nY(s,Xa))
we have As VXaY(s,X a) + V a [ A n ( a n
~ 0 §
VX~As6nY(m,X
a))^A~vn(a~n
~ 0)] .
I f we apply this principle instead of the weak form of continuity, one obtains an intuitionistic version of Lindel6fs theorem: 13.2.1 Theorem. Let r =be a separable, complete, metric space, and let r be covered by a species ~ of pointspecies. Then there exists an enumerable subspecies ~ ~ such that {Int W : W G ~ } is a covering for r. Proof. The proof can be given by remarking that if As~q~lim and if a is such that then there is a species
~sG
AsVn(a~n ~ 0), An(an ~ o § We[nJ with parameter n such that
An(an ~ o § W ~ [ n ] & ~ The collection
~=
w
{W~[n]
^A~6n(lim
: an ~ O}
6 n(lim Ct= ~ W)),
Cea6W*[n])).
covers
r; then apply 13.1.4.
13.3. We finish this section with applications of the fan theorem. Very well known is i3.3.1. Theorem ([H 1966], 3.4.3). A real-valued (lawlike) function on [-I, I] is uniformly continuous. Proof. Let us use the spread 11.5.
S 3 =
, f
S3
described in 11.4, with the theorem proved in
is a real-valued lawlike function on
Let ~ be a fixed natural number, and let any fixed v 6 N : A=6a3Vmlf
llm ~
-m.2-vl
S
be the undressed spread
< 2-v
or
A=Vmlf lira r
c= - m.2-vl
I-i, 1].
<2 -v.
a3 Applying the fan theorem yields:
VnA~VmAB(ra3~ n = ra3's n § If llm Cmra38 - m.2-ul
< 2 -v)
a 3. For
-
75
-
or equivalently VnAa6a3VmAsga3(~n Suppose
n
and let
x,y
= ~n § If lim ~es - m.2"vl
to be such that
Aa~a3VmAs~a3(~n
be such that
lim Cea = x, lim r
Ix-yl
= y,
{W~n
: ~a
{[(k-l)
Let
m
3}
2 -n-l,
= ~n * If lim ceB - m.2-vJ
Then we can flnd
~n = ~n, since It follows
W~n = {lim r then
<2 -n-i.
< 2-v).
: ~n = ~n ^ B 6 a
Is exactly (k+l)2-n-1]
a,S@ a 3
from 11.5
< 2-v),
such that
that if we define
3}
the collection : -2 n*l < k < 2 n*l}
.
be such that
A~6as(~n
=
~n
§
If llm ~*~ - m.a-vl
<
2 -v)
then
Ifx If
13.3.2.
- fYl nm
~
=
If
llm
- m.2-~l
~em - f l l m *
If
nm
~sl
,~B
If we adapt the proof of 13.1.4
form of Heine-Borel
for
~I,
form of Helne-Borel
for every
- =.2-~1
compact
Q.e.d.
9
we obtain a proof
1] . More generally, located
< 2-~.1
of the int~Itlonlstlc
we can prove an intultlonlstlc
space.
See
EH 1966]
5.2.2.
-
w 14. Well-orderings.and
76
-
ordinals
14.1. In this section we dSscuss a simplified version of Brouwers theory of well-orderings and ordinals
([Br 1926],
[H 1959]).In Brouwers original definition,
the type of a
well-ordered species may become arbitrarily high, and is in fact itself characterized by an arbitrary well-ordering.
Therefore we shall use a standard coding of well-
orderings by means of species of natural numbers, in order to avoid such conceptual difficulties. 14.1.I. Definition.
Let
XO,... , Xn~__N. We define operations
+, Z, 9 :
X 0 +...e X n = { ~ x : 0 .< i .< n ^ x E X i} ; as a special case we have Xo+
=
{<0>
~
x
x 6 X O}
:
.
We write 9 X
or < x ~ X
for { < x ~ y
: y~X}
.
The infinite sum is defined by X 0 + X I +... = ~X I = { < l > ~ x XI.X 0 = XIX 0 =
{n@m
: I~NAxEX
: n~X OAmEX
I} 9
I} 9
Now we introduce a class of species, WO, by a generalized inductive definition (analogous to the introduction of 14.1.2. Definition.
Every
F@WO
K
is section 9).
(WO = the class (species) of well ordered species)
is constructed by means of the following principles: (a)
(b) (c)
(o}
~
wo.
FO,... , F n ~ W On
§
Fo+~WOAPo
lawllke, A n ( F n ~ W O ) ,
14.1.3. Convention. elements of WO.
+'''+ F n ~ W O "
then
~Fi6WO.
In this section, we reserve the capitals
Corresponding to the generalized inductive definition of 14.1.4.
Principle of proof by ~nduction over
If the following hypotheses
(a), (b), (c)
WO.
are satisfied
W0
F, G, H, J
we have the
for
- 77
(a)
{0}~
-
X,
(b)
FO,... , F n ~ X
(c)
An(F n6x),
§
Fo+6XAFo+...+
F n ~ X,
n lawlike § Z F n ~ X . n
then WO~X. Just as in the case of ordinary induction, we may Justify definition by induction over
WO
14.1.5.
by the principle of proof by induction over Definition.
Let
n = 9
n~m
~-+ Vi(i ~ r ^ i
WO.
~ is an ordering on the natural numbers defined as follows.
, Xr> , m = 9
Ys ~ "
~ s ^x i 9
^Aj
9 i(x i = y~)) v (r , s A A J
~ s(xj = yj)).
This ordering figures in the literature as the so-called B r o u w e r - K l e e n e between
finite sequences
is a refinement n
9
m ~-+Vn'(n'
of natural numbers.
of the partial ordering on sequences defined by ~ OAn~n'
= m).
is a total ordering, hence that
FO+...+ Fn, ZF n
sense.
In the sequB1,
standard ordering ~IF 14.1.6. X ~ Y
Definltlon.
correspond
induces a total ordering on any
(restriction of
If
X, Y
r
i4.1.7.
~
to
are order preserving. WO
Remark
FGWO
refers to the
F).
are two (partially)
of
FGWO.
to the formation of ordered sums in the usual
any t e r m i n o l o g y i n v o l v i n g order on
The following properties
(I)
~
ordered species, we write
to indicate the existence of a one-to-one m a p p i n g
such that
ordering
(In short:
r
r
from
X
onto
Y
is an order isomorphism.)
are easily proved by induction over
WO:
Theorem.
Every
F
is decidable.
(II) Every
F
possesses
(III)For all
x~F
a first element.
(FgWO)
is a last element
of
(IV) ~ F A n ~ m ( n ~ F ^ n @ m 6 F AF~aAxAy (~x~FA~y~F
either there is an immediate successor in
F
or
F. § m = 0). This property is more simply expressed as § x = y).
The proofs are left as an exercise to the reader.
-
14.1.8. Theorem For any species
78
-
(Transfinite induction with respect to X~_~F:
41F).
A x ~ F [ A y ~ F ( y ~ x § Xy) § Xx] * A x ~ F ( X x ) . Proof. By i n d u c t i o n over
W0 with r e s p e c t to
F. For
F = {0}
the theorem i s
evident. Suppose now the assertion of the theorem to be valid for From this hypothesis we prove for any n (by ordinary induction)
F1, F2, F3, . . . .
A X 6 F o + . . . + Fn[AY6Fo+...+ Fn(Y ~x -* Xy) * Xx] § A X e F o + , . . + Fn(XX). For
n = 0, this is immediate. Ax6F0+...+
Suppose
Fk+I[~Y6F0+...+
(1) to be proved up to
(1)
n = k, and let
Fk+l(y ~ x § Xy) ~ Xx].
(2)
This implies AXEFo+...§
Fk[~Y~Fo§247
and hence by our ~ d u c t i o n AXs
Fk(YlX
~ Xy) ~ Xx]
hypothesis
Fk(XX).
Now take an arbitrary As a consequence
(3) z ~r
+ I) @ Fk+ I .
of (3) we have
A Y e F o + . . . + Fk+l(Y.~Z * Xy)'~, A y 6 , k
+ 1 , @ F k + l ( y . ~ z -* Xz),
therefore from (2)
Az ~~ F k + l [ ~ y ~ . ~ k + 1 , ~ F k + i ( y ~ z
-, Xy) § Xzl,
and this yields, since the theorem was supposed
to hold for
Fk+i,
Az 6
Fk+l(Xx).
Now if F = ZFi, then any x ~ F the previo-siargument Xx.
belongs to
FO+...+ F k
for some
k, hence by
14.1.9. Theorem. Let F be any well-ordered species, Every strictly descending (w.r.t ~) sequence contains an element outside F.
- 79
-
Proof. Take xx to be: every strictly descending contains an element outside F. Clearly Ax E F~y hence
beginning
for an arbitrary
<Xn, n
with
x0 = x
6 F(y ~ x § Xy) § Xx],
(by the previous
sequence
sequence
theorem)
Ax ~ F(Xx),
with an element
strictly
of
descending
F
i.e.
contains
sequence
every strictly an element
descending
outside
F. Since
x 0 ~ F v x 0 6 F, this proves
our
theorem. 14.2. 14.2.1.
Theorem.
G(Z Fi)= z GFi~ G(Fo+...+F n) = GFo+...+GF n i i
Proof.. immediate. 14.2.2.
Theorem
~ F ~G(.WG ~ WO).
Proof.
by induction
14.2.3.
Defl nition. F[u)
= (x
on
WO
with respect
Fer any
u, v E F
to
F, using the previous
theorem.
we define
: uix^x~F),
Flu,v) = : u i x Iv ^ x E F(V) = (x : x ~ V ^ X ~ F}, where F(v),
u ~ v
FKu,v) , FKu)
or an element For
u ~v,
14.2.4. Let (a)
Is an abbreviation
of
FKu,v)
for
are detachable
subspecies
= FKu) F) F(v).
Definition
§ ~u v
= v. F. For any
of
v, F(v)
is empty
WO.
~u
F = F(u) %2 Flu).
is a mapping
u = <Xo,... , xt, Xt+l,... , Xt+s> v~u
u ~vvu
defined
as follows.
= <0,...,
Xt+l,... ; Xt+s >.
= v
(b) u_~, for some
i .< t
Yi > 0 § Su
yr > =
Yr >"
(c) u ~ < Y o , . . . , Yr'' t < r < t+s, Y0 = ''" = Yt = 0 § ~u
(d) If ~u 9
u ~
Yr>' t+s
''''' Yr > =
Yt' Yt+l - Xt+l''''J
Yr - Xr>'
9 r § Yt' Yt+l - xt+IJ'''s
Yt+s - Xt+s*
Yt+s+l'''''
Yr >"
- 80-
It may be helpful to the reader to verify by representing
a simple finite
a construction
Cu
tree
(see the picture below) what exactly
0
F
by
does.
F = F0 + F I + F2 F 0 = FO, 0 + FO, I
<010~.... <0,0}0>
<0,i~... <0,0,I>
<1,1><2,0>
<2,1>
<2,2>
FO, 0 = FO00 + FO01
<0,1,0>
FO00 = (0}
(The terminal nodes of the tree represent
the elements
of
We use in the remainder of thls section S,T as letters and species Flu), F(u), F [ u , v ) ( u , v 6 F , F E W O ) .
F.) for species
We introduce a mapping 0 from species Into specles;, Is defined If u is the first element of S, then *S = {r x : x ~ S}. Clearly
AS(@S g WO))by induction),
and
... ~ F(nl)
+
of
Definition.
(Equivalently: 14.2.7. (I)
F < G
for any
S.
of
~nl> I b e a strictly
~F[n2,n3)+.. .
F, hence
F(n 1)
Is
~D Vx(x 6 G A F ~ G(x)).
F < G ~D VH(F+H ~ G).
Theorem.
If r is an order-lsomorphlsm and r Is the identity.
(II) AF A G A x , Proof.
WO, and let
@F[nl,n2 ) +
Proof. Since n o < nl,n I Is not the first element inhabited. Otherwise the proof Is routine. 14.2.6.
WO
AS(S ~ 0S).
14.2.5. Theorem. Let F be an arbitrary element increasing sequence of elements of F. Then F = F(ni) v F[nl,n 2) ~ F[n2,n3)
of
between
F
and a subset
y 6 G(F ~ G(x) a F ~ O(y) ~ x = y).
(I). It Is possible
to prove
A x E F(Ay E F(y ~ x § Cy = y) § r
= x).
X~
F, then
X = F
-
For suppose
Cx ~ x A A y
r 1 6 2 1 6 2 = r162 Then
r
and o b t a i n
Let
to the wlth
conflicts
Ax ~ F(r
(I).
for
Deflnltlon.
14.2.9.
x
~ y; h e n c Z x
theorem
F 9 G
r
n
r162 = Cx, h e n c e
~ x. S u p p o s e
would
Therefore
from
from
x ~ Cx.
be an i n f i n i t e
descending
Cx = x. Now we a p p l y
F ~ G(x), G(y)
14.1.8
onto
F ~ G(y) G(x),
we c o n c l u d e
In c o n t r a d i c t i o n
= y.
we d e f i n e :
It f o l l o w s
that
If
thls
F ~ G(u),
definition
then
G-F
~D CG[u).
Is unique.
Theorem. F ~ F',
(II)
F < G § F + (G - F)
(III)
F < G r H.F
(IV)
F 9 G § H.(G
G ~ G' § F . G ~ F'.G'.
Proof.
(I) Is e v i d e n t ,
proved
simultaneously
H.F HF
14.2.10. Then
then
isomorphism
(I)
Hence
or
14.1.9.
F ~ G(y),
F o r any
F r o m the p r e c e d i n g
n, and so
wlth
of an o r d e r
Likewise
Cx = x
: x).
x ~ y, F ~ G(x),
existence
14.2.8.
for e v e r y
and thls
-
g F(y < x + Cy = y); t h e n a p p a r e n t l y
Cx = x. T h e r e f o r e
< r
sequence;
(II).
l.e.
81
< HG,
< H.G. - ~) ~ H . G
(II)
f r o m the
definitions,
(III)
and
(IV)
are
as f o l l o w s .
H(G-F)
lim H n = z H' n n n
- H.F.
follows
+ H.(G-F)
Deflnltlon.
~ G.
~ H.(F
~ HG-HF
Let with
(with
n H6
+ (G-F)) 14.2.8
~ H.G II).
be a l a w l l k e
= HO,
.
sequence
H'n = H n - H n _ I
such that
for
A I ( H I < HI+l).
n 9 1.
14.2.11. Lemma. A n ( H n ~ Jn ) § lim H n ~ lim Jn" n n G . llm H n ~ llm G . H n n n
(II)
Proof. (II)
(I)
is i m m e d i a t e 9
O . llm H n : G(H 0 + (H i - No)
+ (H 2 - H1)
+...)
=
n
G . H 0 + G . ( H 1 - HO)
+...
~ GHO + (GH 1 -
GHo)
§ (GH 2 -
GH1)
+...
= llm n
In v l e w contains
of the n e x t exactly
one
definition or m o r e
we r e m a r k
t h a n one
that
element.
It is d e c i d a b l e
whether
an
F
GHn .
-
14.2.12.
Definltlon.
For any
8 2 -
F, G ~ W0
we define
F G = F exp G
inductively
as follows 9 For
F ~ {0}
we put
FG = F
for any
(a)
F {0}= F .
~b)
F exp (G O +...+ Gn) = F
(c)
F exp Z G I = llm F 0 i
GO FG1
m 9149
Lemma.
lawllke
sequence
Let
F
Gn 9
n
be a strictly
of natural numbers,
A n ( H n < Hn+1).
increasing
and let
n
(In the natural
ordering)
be a lawllke sequence
such
Then
llm H n ~ llm Hk(n) n
.
n
Proof. We remark that If since
we put
Gn Oe.
F
F , {0}
n
14.2.13. that
G. For
F 1 < F 2 < F3, then
F 5 - F I ~ (F 2 - F I) + (F 3 - F2) ,
F i +((F 2 - FI) + (F 3 - F2) ) ~ (F i + (F 2 - F1) ) + (F 3 - F2) ~ F 2 + (F 3 - F2) ~ F 3
Repeated Hk(n+l)
application - Hk(n)
of thls result yields:
~ (Hk(n)+l
- Hk(n))
9 ..+ (Hk(n+l) One verifies
~l(Gi
obtains
Z G i ~ Z Jl" Hence by (4) I i
14.2.14.
Lemma.
sequence
of natural numbers,
(I) If
n
- Hk(n)+1)
+...
(4)
- Hk(n+l)_l )) .
easily that from
Let
+ (Hk(n)+2
~ Jw(i)
+'''+ Jw(i+l)-I )' w(O)
we conclude
be a strictly and let
n
to
llm H ~ llm Hk(n) n n n
increasing
(In the natural
be a lawlike sequence.
~ n ( H n < Hn+1) , F 9 {0}, then llm F exp (H0+...+ Hn) ~ llm F exp (Ho+...+ Hk(n)). n
(II) Let
n
A I ( H I = Gk(i)
+...+ Gk(l+l)_l),
k(O)
= 0. Then
F exp Z G I ~ F exp Z H i . I I (III) Let
~ l < n ( H I = Gk(1)+...+JGk(l+l)_l) , k(O) H n = ~ Gk(n)+j
. Then
F exp
= O, one
= O, and
IZ G I ~ F exp(Ho+...+
Hn).
.
order)
-
83
-
Proof. F 9 (0}, F G > {0}. Hence
(I) For F exp
(H 0 +...+ Hi).
AI(F exp (H 0 +...+ Hi+ 1) >
Then apply 14.2.13.
(II) F exp Z H i = lim F exp (H 0 +...+ H i ) = limF exp(G 0 +...+ Gk(n+1)_l) I i i lim F exp (G O +...+ Gi) (by (I)) ~ F exp Z Gn. i n (III) F exp (H 0 +...+ Hn) F exp
= F exp H 0 . . . . .
(G O +...+ Gk(n)_1)
F exp H n =
F exp Z Gk(n)+j
=
J F exp
(G O +...+ Gk(n)_l)
lim F exp (G O + . . . +
lim P exp (Gk(n)
J
Gk(n)+j)
(14.2.11,
+...+ Gk(n)+j)
II)
~ l l m P exp (G O + . . . + J
J
= F exp
Gj)
Z Gj (by (I)).
J 14.2.15.
Definition,
H
is called a refinement
of
F
(notation
F = H) if
+...+ H n (H = Z Hi) ' H i = r i = 0r... , n i (i ~ N resp.), F = S O U . . . u S n (F = n ~ S n resp,)~
H
= H0
AIAJAx 6 SlAY E Sj (i < J § x ~y). 14.2.16.
Remark.
The relation
14.2.17.
Lemma. When
=
is evidently not symmetrical.
G = H, then
F G ~ F H.
Proof. We prove this lemma by induction
over
distinguish
step.
four cases in the induction
Case I. Let Let
Hi = r i
Si =
for
to
G. We
i ~ n, S O U . . . u S n = F. Ai AJAx 6 s i A y ~ sj (l<J ~ x~y).
TO,... , T k
such that
<xi>@Ta(i) v . . . a(O)
v < x i + y i > ~ Ta(i+l)_l , i ~ n,
= O, Yi = a(i+l)-
G I = Tb(i) u . . . U T b ( i + l ) _ l CTi = Ji
for
<1> 9 eTa(i)+1 Gi = Jb(i)
a(1)-l, A l ( a ( i ) < a ( l + l ) ) .
, I ~ m, b(O)
0 ~ i ~ k. Then
~(<xi>eTa(i)U<x
Also
with respect
G = G O +...+ Gn, G = H 0 +...+ H m .
Now we can find
Take
WO
+'''+ Jb(i+l)-l'
b(l+1)).
Hi = r i =
i + 1> 9 T a ( i ) + l U . . . ) ... = eTa(i)
= O, A l ( b ( i ) <
=9 C T a ( i ) ~
+...+ eTa(i+1)_1
i < m, b(O)
= O.
= Ja(i)
+'''+ Ja(i+l)-I
'
-
Now, using our induction F G = F GO . . . . .
hypothesis
84
-
with respect
F Gn ~ F J0 . . . . .
to
F Jk ~ F HO
GO,... , Gn:
.....
F Hm = F H.
~s~_~. G = G O +...+ Gn, G - z H i . Dy a reasoning analogous the previous case, we can find iJi, i E N such that Gi = Ja(o) +'''+ Ja(l+l)-1'
i < n, a(O)
to the argument
= O, Al(a(1)
in
~ a(l+l)),
Gn = ~ Ja(n)+k Hi = Jb(i) Then,
+'''+ Jb(i+i)-l'
with the use of the induction F G = FGo ... F Gn F exp (Ja(o)
Case Ill.
9 b(i+l)).
F exp Ja(n)-I
+'''+ Ja(n)-I ) limk F exp
(Ja(n)
Ja(n)+k ) ~ lim F exp
F exp Z Ja(n)+k k +'''+ Ja(n)+k) ~
(JO+...+
G = Z Gi, G = H 0 +...+ H n = H. Now we can find i
Jk )
Jl, i ~ N, such that
Gi = Ja(i)
+'''+ Ja(i+l)-l'
i ~ N, a(O)
= O, A l ( a ( i )
< a(i+l))
HI = Jb(1)
+'''+ Jb(i+1)-l'
i < n, b(0)
= O, A i ( b ( i )
< b(i+l))
"
Then F G = F H Is proved the previous case.
from the induction
hypothesis
G = z Gi, G = Z H i = H. Now we can find i I GI = Ja(i) +'''+ Ja(i+l)-l'
a(O)
in the same manner as in
Ji' i ~ N, such that
= O, i ~ N, Ai(a(i)
HI = Jb(1) +'''§ Jb(l+l)-l' b(O) = O, i 6
N, Al(b(~)
< a(i+l))
< b(•
ete.etc.
14.2.18. Theorem. Proof.
for
G ~ H §
By induction
Then to every
~
14.2.14).
Hn = ~ Jb(n)+k
Case IV.
= O, Ai(b(1)
hypothesis:
F exp Ja(o)''"
lim F exp (Ja(o)+...+ k
(14.2.11 ( I I ) ,
i ~ N, b(O)
Gi
over
FG ~
WO
F H. wlth respect
we can find a
i < n, Jn : CH[Vn)'
Jl ~ Gi
Ji'JO for
to
G. Let
= cH(Vo)' i ~ n.
G = GO+...+
Ji = CH[vi-1'
vi)
Gn .
- 85
FGo
... F
Gn
~
FJo
So
FG =
If
G = Z Gi, we can find I
Jo = ~H(Vo)'
... F
Jn
FH
~
-
(by the previous
Ji' i ~ N, with
Ji ~ Gi
lemma,
for every
since
H = Jo+...+
i,
Vl) and H a z Jl" Therefore FG = I lim F GO +'''+ Gn = l~m F Jo +. ..+ Jn = F exp z Ji ~ F exp H n i
~y
Jl = ~H[Vl-1'
the previous
14.2.19.
lemma).
Theorem.
G ~ H § G F ~ H F.
Proof. by induction i4.2.20.
over
WO
with respect
to
F.
Th.eorem. F exp GH ~ (F exp G) exp H.
Proof. We apply induction
over
WO
with respect
to
H.
F exp G. {0} = F exp G = (F exp G) exp {0}. Now let F exp G(Ho+...+ Hn)
= F exp
H = HO+...+ H n. ~nen
(GH 0 +...+ GH n) H0
F exp GH 0 . . . . . Let
H = z H i . Then i
F exp GH n ~ (F exp G)
Hn ... (F exp G)
(F exp G) exp H.
F exp G(z H i ) = F exp r GH i = i i
l~m F exp (GH 0 +...+ GH n) ~ l~m(F exp G) exp H0
(H O +...+ H n)
Hn
l~m (F exp G)
... (F exp G)
= (F exp G) exp H.
14.5. 14.3.i. Definit%on. Ax 6 G (F x ~ W0)
We define
by induction
~ Z{F x
an ordered over
{F x : x g G}
with respect
G:
(a)
G = {0}
(b)
G = G O +...+ G n § Z{F x : x 6 G} = Z(F,o, t y
(C)
G = Z GI§
Z{F x
: x G G}
WO
sum of a species
= F0 +
: y ~ GO} +'''+ Z { F , n , g y : y 6 Gn}. : x 6
G}
= Z
I 14.3.2. Suppose
Lemma.
(Z(F,z,m
z Let
to
with
F -- G, and let
r
Y
Gz})
be an o r d e r - i s o m o r p h i s m
Ax ~ F (H x = H'r ), then E{H x : x g F} ~ 2{H'r
: y E
: x ~ G~,
9
from
F onto
H.
Jn ).
-
Proof.
As in 14.2.17,
86
we apply induction
-
over
WO
with respect
have to distinguish four cases. Let us treat as a typical case Then we can find Ji such that
F = Z Fi, G = Z G i. i i
+'''+ Ja(i+i)-1'
a(O)
= O, i ~ N, ~ i ( a ( i )
< a(i+l)),
Gi = Jb(i)
+'''+
b(O)
= O, i ~ N, Ai(b(i)
9 b(i+i)).
Jb(i+l)-i'
= Zz Z{H~ y : y 6 F z} ~ Z{Z{H z z
where
is such that if
Hz
,Y
Ja(z)
+'''+ Ja(z+l)-I
The coneluslon Hence
F, and we
Fi = Ja(i)
Z{H x : x ~ ~
and
to
~y. r
for every
then follows
AzAvAy (o ~ v ~ a(z+i)-a(z)
AzAuAy 14.3.3.
-I
: y ~ G z} = z{H'y
(O~u~b(z+i)-b(z)-1 Theorem.
Let
^
Completely
14.3.4.
Theorem.
Proof.
by induction
~z,y
: y ~ Ja(z)}+...+
G m H, and let
between
~
Fz
y) = H < z > e y ) -
such that
r
§
H' < z , u > ~ y
Z{~a(z+i).i ' y : y 6 Ja(z+l)-i H'y
to be such that
be an o r d e r - l s o m o r p h l s m
to the derivation
between
G
and H;
: x E H}.
of 14.2.18
and {H x : x 6 G}
from 14.2.17.
be given such that
Z{F x : x ~ G} = Z{H x : x 6_G}.
over
W0
with respect to
G.
14.4. 14.4.1.
Definition.
species
of
Ordinals
WO. The ordinal
therefore
defined
Ord ({0})
= 1, Ord ({
The theorems
as
are the equivalence m = Ord
classes
(F) associated
of order-isomorphic
with a species
F ~ WO
is
{G : G * F}. : i ~ N})
= ~ .
of 14.2 permit us to define
}}=
~b(z)+u~y)~
:
Z{F x : x G G} = Z{Jcx
{F x : x ~ G}
Ax 6 G (F x ~ Hx) , then
Az AY(Hz,r
))
y ~ Ja(z)+v § ~z0
^ y E Jb(z)+u
analogous Let
z, then
: y E G}, if we take
suppose ~ x ~ G (F x = Jcx); then Proof.
is an order-lsomorphlsm
: y 6 Jz}}, with
: Y 6 Jz }} ~ ~{Z{~a(z),y
z{H~z > @ y
+'''+ Ja(z+l)-i
from the induction hypothesis.
Z{H x : x 6 F} ~ Z{Z{~z,y z
~{Z(~z,y
y)
'Y : y 6 Ja(z)
a few arithmetical
operations
on ordinals.
-
14.4.2. Definition.
87
-
Ord(F + ~ } )
= Ord(F)
Ord(F + G)
= 0rd(F) + Ord(S).
Ord(F G)
= Ord(F) 0rd(G)
Ord(FO)
= 0rd(F).
Some simple identities
+ I.
of ordinal arlthmetlc
proved for w e l l - o r d e r e d
= Ord(F) exp Ord(G).
Ord(O). follow readily
from the theorems
species:
Ca + B) + v = ~ + (s + y).
8+y (8)y
14.4.3.
If we define
~ < 8
8 y9 = ~8y =
by
VY(~
+ Y 9 s) (recall that we have excluded
as an ordinal for reasons of technical convenience)
then it may be concluded from
14.3 that we can define well ordered sums of ordinals unambiguously: 14.4.4.
It is a consequence
ordered for any given species
of 14.1.5,
14.2.7
0
(II), that
z{aS:
{~ : s < 8}
8 < y}.
is totally
B . But on the other hand we have no hope of p r o v i n g the
of all ordinals to be totally ordered,
as is illustrated by the following
(weak) counterexample. Let
a
be an arbltrary lawllke function of
Define the sequence
(N)N.
n : an = 0 § F n = {0}, an ~ 0 § F n = { : 1 6 N}. Then
we have no general m e t h o d to decide
Ord
(~ F n) = ~ v m <
Ord (~ F n)
since this would imply Aa(An(an
= O) v Vn (an ~ 0)).
For a concrete illustration we only have to m a k e a depend on the decimals
of
in the usual way. We shall refrain here from a further development reader is referred to
of ordinal arlthmetlc~
bhe
[Br 1926].
14.5. 14.5.1. and
It will be clear that there is much similarity in b e h a v l o u r between
WO. By a quite general t h e o r e m one obtains immediately that
definable in terms of
K. (For a statement of the theorem,
WO
K
is explicitly
see section 15.)
-
Here we shall demonstrate are explicitly and induction
definable
B8
-
by a more straightforward in terms of each other,
scheme for
K, WO
respectively
argument that
K
and that the closure
are provable
and
WO
conditions
from these explicit
definitions. For an easy comparison, elements
of
WO
we study
instead of
W, the species
WO
of characteristic
functions
of
itself.
If we put A W (P,a) ~D
[a = ~x.(iZx) I v [aO = 0 ^ Ay
~x.a(y,
x)@
P]v
v [aO = O A VzAy((y ~ z ^ ~x.a(9 9 x) e P) v (y 9 z * A ~ a(y m x) = O))] then the introduction by
of
Aw(W,a)
W
K
inductive definition
is expressed
+ Wa, +
I%.5.2. Theorem.
by a generalized
Qa]
+
is explicitly
definable
in terms of
W
such that
AK(K,a) + Ka, Aa[AK(Q,a) become provable Proof.
+ Qa] § [K __.Q~
(by induction
If we replace
Vz[a = Ax.(z+i)],
W).
a = Ax.(1~x)
we o b t a i n
If we define a species
over
in the definition
of
W* by
a E We+-+ Vb @ W Vc(a = kn.bn, sg h(b,n),(ch(b,n) where m'(m'
AW(Q,a)
~(Q,a).
sg x = i A (i z x), h(b,n)
= O
= m + i
otherwise.
Then one proves easily
Aa(A w, CW',a) + w'a) + + Qa] + The proof is left to the reader.
if
Q].
bm ~ O, m o
+ i)) n,
by
-
Now
K
-
is explicitly definable by a g K,-* a g W * A AK(K,a)
Is readily proved.
AnVy
a(n @ y) ~ o.
§ Ka
Suppose now
Aa[AK(Q,a) Take
89
(4)
§ Qa].
QOa ~D AnVy a(n t y) ~ O § Qa.
If a : ~x.z+l, then Now let
AK(Q,a) , hence
Qa.
aO = o ^ Ay(AnVz a(y e n 9 z) ~ O § ~m.a(y @ m) ~ Q), then aO : O ^ AnVz(a(n @ z) ~ o) § ~y ~m.a(y m m) ~ Q, hence ~nVz(a(n • z) ~ O) § (aO = O ^ therefore by (4) AnVz a(n m z) ~ O § a ~ Q, so
Ay
~m.a(y 9 m) ~ Q),
a g QO.
If aO : O ^ V z Ay[(y < z ^ ~x.a(y ~ x) ~ QO) v (y > z ^ Ax a(y ~ x) = o)] we have trivially
QOa, since
AnVy a(n e y) ~ o
clearly does not hold.
Hence we have shown that (4) § Aa[&./~(QO,a) and therefore
We~_QO
-~ QOa] i.e. Aa(a ~ W~§ a 6 QO).
Hence Aa(a E We^ AnVy a(n 9 y) ~ o § a 6 Q O ^ so
AnVy a(n , y) ~ o)
K ~___Q.
14.5.3. Theorem.
W
is explicitly definable in terms of
K, such that
Aa [Aw(W,a): § Wa] Aa[Aw(Q,a) becomes provable Proof. Let
V
§ Qa S §
(by induction
~__Q] over
K).
denote a class of spread directions
satisfying
-
Va
=-D aO ~ 0 A An(an
90
-
~ o § lax a(n ~ x) ~ o v
v V z { A x ~ z(a(n ~ x) ~ o) ^ Ax 9 z(a(n ~ x) : o) 1~). we define
W~:
b ~ W ~ -D V a 6 V
Ve An(bn
~ 0 ,-~V~(m < o n ^ sg(am).em
: bn ~ 0)).
We prove easily
AW*( w" ,a) ~ Wa. Now suppose Aa[~-(Q,a)
§
Qa].
(5)
We apply induction with respect to e to prove Wa~__Q. Let us introduce y(epa) (e g K, a E ~ v a -- ~x.O) for the unique b
function
such that An(bn
~ O .-~Vm(m
~o n ^ bn = sg(am).em
Now we want to prove by induction
over
~ O)).
K
AeAa g V (~(e,a) ~ Q). If e = Ax.z+1, Let now
then
w(e,a)
= Ax.z+l,
hence
u
g Q.
eO : 0 ^ AxAc 6 v (~(~n.e(x ~ n),c) 6 Q) and let
(6)
a g V. We remark that a ~ V § Ax(~n.a(x ~ n) 6 v) v Vz(Ax
< z(knoa(X ~ n ) g V ) ^
lm
Ax 9 z(~na(x e n )
= o)).
Furthermore ax ~ 0 A e O Suppose
first
= 0 § An.w(eja)(x
Ax(kn.a(x
Ax(ln.v(e,a)(x u and thus
@ n) 6 V ) .
= w(An.e(x ~ n), a(x e n)).
Then from
46):
9 n) ~ Q )
47)
5
= 0
from (5) and
, n)
(7) (which implies
Now let Vz(Ax ( z(kn.a(x @ n ) ~ Then from (6): VzCAx < z(kn~165
V)
~(Qbu
A AX > z(~n a(x 9 n) = 0)).
9 n) E Q) a A x
>z ~n(u
and (e,a)
we conclude
: O,
therefore Aw~(Q,~(e,a)) , hence with (5) Q~(e,a). Thus we have proved AeAa ~ V(w(e,a) s Q), i.e. Wm_~Q.
9 n) = 0))
to
Q~ (eja).
-
w 15. Species revisited|the
91
-
role qf the comprehension
prlnciple
15.1. In the IntultionIstic the fundamental
theory of species,
question of classical
we are confronted with the analogue
axiomatic
set theory:
said to exist? Or in terms of a theory of constructions: structions? Let
X
The problem is illustrated
be a certain given species.
there is nothing problematic Let
~
be a language
prehension principle
(Y
elements
(Yx~
is accepted Xn
for any
x, y
as a well-deflned
of
may be expressed by a schema: y
we accept the universal
(1)
for subspecies
principle ~
of
relative
x.)
to
X
clearly depends
is a first order language with variables
X, the resulting predicative
seems to us to be quite unobjectionable.
version of the comprehension
The strength 15.1.1.
of full comprehension
Theorem
arithmetic
(see e.g.
[Kr
~
AX,VX,
for
Let us call this weak comprehension.
contains
quantifiers
is illustrated
1968 AS). Suppose
with the language extended by variables
and quantifiers and N~
on the
principle
Evidently we can build a ramified hierarchy by repeated weak comprehension, in classical ramified analysis. or ful ! comprehension, where the other extreme possibility.
object,
X. The com-
F(x))
of a comprehension
principle.
n ~ N.
for elements
not containing
a variable
power of .C . When
of
, X
F(x) of #~
VYAx
expressive
X
containing variables
For every formula closure of
The strength
Once
relative to ~
which species may be
which notions are con-
by the role of the comprehension
in accepting
of
and the comprehension
as
AY, VY, represents
by ~
to consist X, Y, ...
of Intultionistic for subspecies
of N,
axiom relative to thls language
If ~ + denotes the corresponding classical system, then the G6del translation of w 3 extends to ~ (taking (AX F(X))- = AX F'(X), (%/X F(X))- = n A X ~ F - ( X ) , and preserves validity. Proof. We only have to verify that the translation for F Is a consequence this Is straightforward. Therefore
~+
of the comprehension
is consistent
if ~
of the comprehension
scheme in ~
is consistent.
applied to
scheme F-
and
(I)
-
92
-
15.2. Inductive definitions
like those of
represent examples of intermediate If we accept full comprehension,
is Justified
K
in w 9, and the definition of
WO
in
PA
satisfying
w
14
forms of comprehension.
then the introduction of a predicate
Ax(A(PA,
x) § PA x )
Ax(A(Q,
x) § Qx) § PAC__ Q
(2)
)
(classically as well as Intuitionistically)
whenever
A
satisfies
the condition of monotonlcity: A(P, x) ^ P ~ PA
(3)
P' § A(P', x).
Is~ said to be introduced by a generalized inductive definition
Justification is given by remarking that
PAy Since
PA
--AX[Ax (A(X. x) + Xx)
PA
(4)
over species, we might,
if we think
Just as well accept full comprehension outright.
But if we impose more stringent requirements for the introduction of
may be defined by
Xy ] .
(4) requires universal quantification
this Justification satisfactory,
(g.i.d). The
on
A, sometimes better Justifications
can be given. For example, the introduction of
K
(introduced and discussed in section 9) is essentially Justified by observing that and
AK(P , a) § Pa K
expresses closure of
under certain simple operations;
is then viewed as the species such that
using these simple closure properties e g K
P
e ~ K
only. Moreover,
iff this can be proved it is to be remarked that any
may be said (in a sense) to codify itself a standard proof of
Likewise we may Justify the introduction of to be explicitly definable in terms of
e g K.
WO. In fact, we have even shown
WO
K.
Once we have accepted
K, a quite general class of g.i.d.'s also becomes acceptable,
since the species
required to exist by the g.i.d, may also be defined
PA
plicitly in terms of
K. We have the following result:
15.1.2. Theorem.
IDK ~
w I0, and let such that talning
A P
Let
A(P, a)
be an extension of the system
be any formula of
is constructed by means of
IDK ~
IDK
as described in
with a single predicate
^,v, Ax~x,Va
~1,b2,...,
xl,x2,..,
yS), t
letter
from formulae not con-
and formulae of the form P(~y.t
ex-
a term of
IDK ~.
P,
-
Then we can explicitly
define
a)
Aa(A(PA, Aa[A(Q, for any
Q
§
(in
IDK r
-
a predicate
PA
such that
PAa)P
a) § Qa] §
in the language
93
of
AaKPAa § Qa]
IDK e.
Proof. We shall not present a full proof here;
for more details
see
[Kr, T]. The
essential idea is that for A(P, a) of the form described above, a ~ PA must have a standard "cut-free" proof, which may be codified by a well-founded tree, hence by a function Let for example
of
K.
A(P, a)
be of the form
~JbAx(RCa, b, x) v PC[a, b, I ] ) (r
b, x] : ky.t
Then we take'as PA a
[a, b, x, y]
our explicit
~D VeVcVd
§ d TM *= r
for a suitable term
t).
definition:
{d ~ : a ^ eO = 0 ^ AmAy TM, c m, y])
((e(m e y) = O §
^ (e(m m y) ~ 0 ^ em # 0 ~ R(d m, cTM, Y))}
where d TM : ~x.d {m, x}, c TM : ~x.c {m, x}.
15.3. A typical application definition
union of all connected not find
of the full comprehension
of a component
of a point Y~__X
YI' Y2' closed in
p
such that
Y, such that
We shall present here a typical
principle
in a topological p ~ Y. ( Y ~ YI ~ Y 2
(and essential)
X
is given by the space
<X,~>
is connected if we can-
= Y' Yi ~ Y2 = ~)"
application
of the predieatlve
prehension principle, which clearly shows the role of the comprehension as a creator of new objects. 15.3.1. We write X~IY if there are (lawlike) mappings r r such that for some Z C_r r162 (r162 = {r : y~ r Let X ~ 2 Y mean: there Is a (lawlike) mapping may be properly included in X.) 15.3.2.
Theorem.
X~2Y
§ X~IY
r
as the
such that
(IT 1967 A], lemma 2.4).
[x]
r
principle
bi-uniqu%
: Y.
com-
(dom
-
Proof.
Let
r
We p u t
for
any
,
,
,
,
= Y, dora r C
X, where
x 6 Dom r
We define
-
dora r = {x
:
Vy(r
= y)}.
x 6 X:
Yx = {y : y g Dom r ^ x E D o m If
94
then
r
Yx
on {u
r A r
is inhabited,
: V z ( z ~ Yy)}
#'(u
= # x.
~'[(u
: Vz(z E vy)}]
= Cy},
and conversely.
by p u t t i n g
Clearly
There remains
to be proved that
$'(u
= $'(u
§ u (We use 15.3.3.
u
= r162
Remark.
is bi-unique. $(x)
= $(y)
: u
= {y : y g Dora r A r X ~2 Y
: X ~ X ^ y ~
The proof uses the comprehension application
r §
The hypothesis
{~:xjy:.
= Y.
Y ^ y
requires
for
the existence
x ~ Dora r of
X, Ym and
= r
principle
of weak comprehension,
= r
see e.g.
relative to these species. IT 1967 A], theorem 4.3.
For another
-
95
-
16. Brouwer's theory of the creative subject i6.1. In a number of papers published after 1945 (e.g. [Br 1949~,
[Br 1948~, p. 1246,
[Br 1948 A],
[Br 1949 A]), Brouwer introduced the Idea of the creative subject
(or the idealized mathematician).
Thls concept gave rlse to much discussion and
It is likely to do so for some period of tlme to come. A systematic and coherent theory has not yet been developed,
so I am restricted to presenting a few fragments.
The central idea Is that of an idealized mathematician Jectivlstic viewpoint of Intuitlonlsm, to obtain the required idealization,
~conslstent wlth the sub-
we may think of ourselves;
or even better,
we may think of ourselves as we should like
to be), who performs hls mathematical activities in a certain order (you may think of the order given by time). The process of bls mathematical activity proceeds in discrete stages. Therefore we introduce a basic notion:
~m
A
to be read as: "the creative subject has a proof of
A
"the creative subject has evidence for
m II .
We suppose
~m A
~ --mA At stage for
m
A
at stage
to be a decidable relation,
v ~m
at stage
m"
or better
l.e.
A.
we know If we have evidence for
A
or If we do not have evidence
A.
Clearly (Vm ~ m A) § A. "If we have evidence for
A
(2) at stage
m, then we can find a proof of
A".
In order to simplify the interpretation we also suppose
(3)
( l---mA) A (n > m) § ( l--nA) . "The evidence at stage
m
Is also contained In all following stages".
If we boldly identify the provable assertions with the assertions
for which we
can obtain evidence at a certain stage, we also have A *-~ Vm( ~ m A)
(4)
- 96
-
or in combination with (2)
Vm( ~ - m A) ~-* A
(5)
.
If we want to be more cautious, we may satisfy ourselves with the following assertion instead of (4): (6)
A § -1-1Vm( ~--mA) . (6) may be read as follows: there is a proof of
"I am completely free in making deductions. Hence if
A, it is absurd that I would be able to prove that I will
never find a proof of If we want to assert
A
(at no stage will have evidence for
A)".
(6), without asserting (4), this means that we do not want
to identify all possible constructive proofs with the collection of proofs whose existence becomes evident to me at a certain time (stage). Let us call in the sequel the theory based on (6) instead of (4) the "weak theory", and the theory based on (4) the "strong theory". 16.2. In the existing literature, most of the deductions are based on the weak theory. In developing consequences from the weak or the strong theory, I shall try to be cautious and hence proceed more or less axiomatically, in order to show what is actually used in certain deductions. In the weak system, we can derive the following scheme (called Kripke's scheme in the literature) :
Vx[(Ax(xx
- o) + ~
In the strong system, Vx(Vx(xx
~A)
^ (Vx(•
~ O)
(7)
+ A)].
(7) can be strengthened to
(8)
# o) +~ A).
This is seen by defining
A)
x
relative to a given assertion
A
by:
:I ]
(n ~--nA) § xn = O. If we have a definite prescription involving the actions of the creative subject (by means of a relation like
~ n A) for determining the values of a sequence, we
speak of an empirical sequence (as is done e.g. in [M 1968]).
-
97
-
Our idea of lawlike sequence does not exclude empirical
sequences,
at least not
as lon E as we are willing to consider reference to our own course of activity by means of
l"-n a s " d e f i n i t e " .
It is clear however, stricter,
that e.g. primitive
more objective
recursive
functions
sense~ their values are independent
are lawlike in a of future decisions
about the order in which we want to make deductions. Let us call a sequence which is given by a complete description a mathematical
([M 1968])
If a sequence
~
or absqlutely
is defined by a complete description
We shall return to the distinction
between empirical
is really evident the use of
~-n
§
since it is conceAvable
(even if
~n'
in 16.8. A
VaAx A(x, ax)
A
that
sequences
from a proof of
AxVy A(x, y)
(A
we ought to restrict
(This can be done, I believe,
from the species
AxVy A(x, y)
itself does not refer to
Va~x A(x, ax) with a mathematiaal, not Involving
...,
or
in section(~ i.e. for an
only if we do not exclude empirical
In order to conclude
Xl, X2,
mathematical
for lawllke objects onl~
A(x, y)
of lawlike sequences,
~
and mathematical
It seems to me that the axiom of choice as discussed
AxYy
from sequences
we shall call
X2 , . . . .
Xl ,
containing free variables
I n
lawlike sequence.
without reference to the creative subject, lawlike in
not involving
is proved with
~-n ) . not involving ourselves
consistently
~ n ) to
to arguments
throughout
the
preceding sections.) For lawllke sequences,
we obtain from (7)
AaVb~(Ax(ax If we use
= 0)*-* Vy(by
= 0))
(9)
(8) instead, we obtain the stronger form: AaVb(Ax(ax
This is seen by applying
= 0)*-~ Vy(by
= 0))
"weak" counterexamples
turn out to be essentially
collapse,
equivalent.
(1o)
.
(7) and (8) respectively
In virtue of (9), various types of unsolvable tlonlstlc
.
to the formula
problems
A
-D Ax(ax = 0 ) .
that are used in intui-
i.e. some of these classes of problems
-
A few of these types by the following
of problems,
formulae,
98
-
some of which we have met before,
which express
assertions
are represented
which we have no hope of p r o v i n g
intuitionistically: Ab(Vx(bx
: O)v-IVx(bx
Ab(-7-1Vx(bx
: O) § V x ( b x
AD(-)Vx(Dx
: 0))
= O)v~IVx(bx
AaAb[-7(Vx(ax
(11)
: 0))
= 0))
= 0) ^ V x ( b x
(11) and (13) are restricted
forms
the principle of testability Markov's principle.
respectively.
~a~b(Ay~(Vx Vc~y[(Vx
b{x,y}
A
a{x,y} = O ^ V x
a{x,y}=
of the excluded third and
(12) is an intuitionistic
analogue
of
to : O)
b{x,y} = O) §
o § cy = o) ~ (~x b { x , y } =
0 + cy ~ 0)]).
the analgon of the assertion:
set is recursive,
the second one expresses
not containing non-lawlike V b ( A 4-~ Vx(bx
Since the members
(14)
: O) ^ (cy ~ 0 §
of disJolng r.e. sets can be separated For an
.
to
The first formula expresses enumerable
: 0))]
of the principle
(11) Is equivalent
AbVc~y~cy : o § and (14) is equivalent
(13)
= o)) §
("lVx(ax = O)v"lVx(bx
Intuitionistically,
(12)
the analogon
by a recursive variables,
every recurslvely of: every pair
set.
(8) simplifies
to
= 0)).
(15)
of the pairs ~IA
§ A, ~ A
v A
and (A^B) possess
equal strength
§ (~A v ~ B ) , ~ A v ~ A
as axiom schemes when added to intuitionistic
see easily that as a consequence lawllke parameters (11) *--'* ( 1 2 ) , (13) It is worthwhile using
of (15),
only; likewise ~ (14).
knowing however,
(9) only, without
further
(11) implies
(13) implies
that
~ Av~IA
(11) *-~ (12),
reference
Av
~A
logic, we
for all
for such
A
with
A. Therefore
(15) 4-+ (14) can be proved
to the creative
subject.
-
Prqof,
(i) (11) ~ (12) is immediate,
Suppose
(12), and take any
99
-
even without
(9).
b. According to (9) we can find a
-*1(-IVy(by
: 0)~9
V'y(cy = 0))
b
such that
.
So -rVy(by = O) ~-~ -~-lVy(cy = 0). Since
-11(Vy(by = O) v-IVy(by : 0)), it follows that -7~(Vy(by = o)v,-iVz(cz
P v-~'IQ §
-I"1(P v Q)
: 0))
.
is a t h e o r e m of intuitionistic p r o p o s i t i o n a l
nn(VyCby
: 0))v
VzCCcz
logic, hence
= o)),
so ~Vx(bx.cx Applying
: 0).
(12), we conclude to
Vx(bx.cx
: 0), hence
Vx(bx
= O) v V x ( c x
: o),
which is equivalent to Vx(bx (ii). Suppose
= O)v-/Wx(bx
(13). For
~Vx(ax
= o)~
a, b
= 0). sueM that
conversely, Take any
~Vx(ax suppose
b, and let "~(Ax(bx
Then and
~
Ax(bx ~Ax(bx
= o)^Vx(bx
= 0)) we conclude to
Vx(bx = O)
-iVx(ax = O)~-~-iVx(bx Then by 413)
1(Vx(ax
: 0).
= o)v~Vx(ax
= o).
(14). a
be such that
~ O) ~
Vy(ay
= 0)).
# O)*-~-~Vy(ay = o)) ~ O)*-*nVy(ay
= 0).
From (14) and the fact that "~(Vx(bx = O ) a V y ( a y = 0 ) ) i t follows that -~Vx(ax = O) v - I V x ( b x = O) therefore we may conclude to I V x ( b x = O ) v ~ w V x ( b x = O).
Now we shall proceed with somewhat more i n t e r e s t i n g theorems.
-
iO0
-
16.3. Theorem.
We can prove in the weak system (in fact, using
~VbAaVxAz[Vy
a(z,y}
= 0 *-~Vh b{X,{Z,U}}:
This result has been called a refutation sequences
(16)
O] .
of "Church's
thesis",
but since empiric~
are rather far removed from the idea of "mechanically
functions,
it is perhaps better to describe
Myhill proposed Proof.
(9) only):
computable"
it as a non-enumerability
result,
as
([M 1967]).
Suppose AaVxAz{Vy
a(z,y}
= o ~Vu
b{x,
(17)
(z,u}}: O} .
We remark that ~ V u b{x,{x,u}}:
0 ~Au(1
Now we can find (by (9)) a
: b{x,(x,u}}: c
~ n ( A u bKx,{x,u}}# In virtue of our hypothesis
O) *-~u(b{x,{x,u}}~
O) .
such that
O ~Vv
(18)
c{x,v} : o) .
(17), there exists an
Ax(Vv c{x,v} = o ~ V w
b{Xo,{X,W}}:
X0
such that
(19)
O .
So we obtain
InVw
b{Xo,{Xo,W}}:
O~Vv
+-~Au
b{Xo,{Xo,V}}~
0
~-~Au b{Xo,{Xo,V}}~
0
~Vw contradiction.
C{Xo,V} = O
b{Xo,{Xo,W}}=
This disproves
(from (19))
(from (18))
O :
(17).
16.4. Theorem
([M 1967]).
AxVx' B(X,X'),
There exists au extensional
but for no continuous
functional
AxB(x,rx). Proof.
Apply Kripke's
scheme
(7) to
AX -DVx~y 9 x(xy = 0).
predicate r
of type
B
such that ((N)N)CN)N
-
101
-
Then we obtain
AxVx'[('~VxAy
= o))^~
> x(xy = o).,-,-Ax(x'x
(20)
^(Vxx'x / o-. Vx4y 9 O(x~' : o))] Let us denote the part of (20) within the square brackets by suppose that for some continuous
B(X, X ') , and
r
AxB(x,rx). In case Since
(tX)X # O, it follows that r
is a continuous
VxAy 9 x ( x y
funct&onal,
= o).
there exists an initial segment
~y
of
•
such that
Ax"C~y
= ~"y § (rx)x = (rx")x)
and therefore
Ax"(~y
= ~"y ~ VxAz
9 x(•
= o)).
This is obviously false; we only have to take for ~"y = ~y, X"(y+z) Hence
(r•
= 0
the function such that
for all z.
leads to a contradiction;
the zero functional. tradiction,
= I
•
therefore A • 2 1 5
As a consequence A x n V x A y
= O, i.e.
r
is
9 x(xy = o), which is a plain con-
since the zero function provides a counterexample.
16.5. 16.5.1. Theorem.
-TAX(n~x(•
Proof. We apply (7) to Vx[(nnVx(xx Take for Since
A•
= O) + ~x(• -DVx(•
~nA6
we find a
holds x
(AS
Vx(xx
= O)
:
= O) § AX')].
B.
is an application
such that ~ n V x ( x x
would hold, then
= O)v ~Vx(x'x
= O) * ~ I A x ' ) ^ V x ( x x
x' any choice sequence
= O))([Kr 1967], p. 160).
of the principle
= 0), hence if
= 0), hence
AS. A8A8
Ax(~Vx(xx
of the excluded third) = o) § Vx(•
is contradictory
= o))
(compare 9.10 (I));
so the assertion of the theorem follows. 16.5.2.
Remark.
At first sight,
in section 9; but actually,
A x ~ 06 ( ~ V x ( • where
~
16.5.1
is Just weaker than the result 9.10
(V)
we have proved more, we have shown:
= o) ~ V x ( •
is the class of sequences
(but not necessarily extensional)
= o))
obtained from choice sequences by a lawllke
operation.
-
I02
-
16.6. Brouwer
gave
number
x
takes
([Br 1948 ~ ) a s i m p l e
such that
an a s s e r t i o n
know whether
x # O, w h i l e A
wA
or
w h i c h has not b e e n ~A),
and d e f i n e s
-I ~ n ( ~ A v - ~ n A )
§
rn
~m(IAvnnA)^
Then would
one v e r i f i e s imply
use
subject,
further
reference
xb
denote
a lawlike
equivalence
b
~
n(bm
(12) is e q u i v a l e n t
implies
bn
generator
:
^ n 9 m § r n = 2 -m.
or s t r o n g
a proof
another
system
of
x ~ O; but
x ~ 0
x ~ O.
(12)
relative
formulation
a real number
to the t h e o r y
of this
generator
with
§
rn
result
of the
without
defined w.r.t.
= 0
< m(bn'
= 0)) § r n = 2 - m
A b ( x b ~ 0 § x b ~ 0); but on the
x ~
Ax(x
Then
0 §247
o) ~ (11)
)
other hand
(12)
.
= 0 v x 4 0). F o r
let for a n y
/~n(r n - 2 -n < x < r n 9 2-n),
Therefore
other
~ o.
examples
(II)
xr
o) ~ A x ( x
of t h e o r e m s
A x(x ~ o §
implies
Ax(x ~
0 v x = 0).
x
= o vx~
to w h i c h
we can
o)
.
find w e a k
same k i n d :
Ay~z(y
be s u c h t h a t
n - 2 - n 9 0 v r n + 2 -n < O) ~ and i f we put
for l a w l l k e
Ax(x
xn
< r n - 2 " n > n 6 x m < r n + 2 - n > n 6 x.
= 0 4-~ r n . 2 -n 9 0 v r n + 2 -n < O, we see t h a t
Some
(that is, we d o not
by a real number
(ll) and
~ 0 ^ An'
~ o ~ x~
in t u r n
Anlx-rnl<2-n-i. Hence
up t i l l n o w
he
(II), t h e r e f o r e
Ax(x (II)
real
subject.
with
~ O)
(m & n A bm then
of a l a w l i k e
x @ O. To do this,
by
~Vm
implies
of
to o b t a i n
a real number
of
= 0
in the w e a k
to the c r e a t i v e
function
tested
"~m_l(-1Avn-lA)
that
in o r d e r
counterexample
x
so we d o n o t h a v e
of the
creative
Let
easily
~Av~A;
We can m a k e
(weak)
we h a v e n o p r o o f
x } o vx
# o) j
~ o a z ~ 0 4 Vx(yx
+ z : o))
.
counterexamples
of the
-
103
-
If we sufficiently enlarge our notion of real number, we can give m a t h e m a t i c a l disproofs by means of theorems
Ax(x
like 16.5.1
, e.g.
o)
o § x
is provable if the class of reals is sufficiently all reals obtained from real number generators enumeration of the rationals,
• 6~
large (roughly speaking:
n,
n
take
being a standard
as indicated in remark 16.5.2).
~6.7. 16.7.1.
Lemma.
mappings
r
Let
a, b
be two lawlike sequences
satisfying
~CRange
(a)~
~
Range
(b), r
Then there exists a bl-unique ~ERange (Or if
~
(a)]
= Range
~
(b)] ~ R a n g e
defined on Range
generated by
a'
such that
is defined likewise
Now we define
~
from
as follows.
such that
Lemma.
Consider
Now we put
We then use the existence Let
a lawlike sequence Range
X a
of
(y+l) th "new" number
= ax ^ ax ~ (aO,..., az} ^
y
different e l e m e n t s ) V ( a x
{r
ax, and let 0),...,
r
to demonstrate
AzVx(~ax
= bz).
such that
let us suppose
is further defined by: = aO
~ - m z 6 X § a{m,z}
= z.
Now one verifies easily that on account of Vm
-mX
y
~ax = bz.
~-m z 6 X § a(m,z}
x e x
= O)
@(ax)} must contain at least
(a) = X.
aO = O. a
= aO A y
a'x = y. Then we can find
be an i n h a b i t e d species of natural numbers.
Proof. For sake of simplicity, Then we put
(b).)
b.
b'z = y, because
different elements.
16.7.2.
is the
(6) = Range
a. Or explicitly:
{aO,...,a~} contains
bz
(a) such that
(a) only, Range
a'x = y ,-~ ax
a'x = y ,-~Vz < x(a(z+1)
a
(a).
(b) .
is taken to be defined on Range
Proof. We define
b'
such that there exist bl-unlque
x)
(an application of (6)) we have Range (a) = X.
0 6 X,~-oO ~ X.
Then there is
-
I04
-
Remark. The proof as described above proceeds directly from the axioms for the creative subject, but we could have obtained the result also from the specialization of (8) to species:
AxVa[Vy(ay hence
Ax[Vy(b{x,y}
= o) *-~ x 6 x I;
= 0 ~-* x ~ X]
for a certain
b~ and since
X
is inhabited,
we may construct in the usual way a constructive function which enumerates (x : V y b(x,y} : 0}. 16.7.3. Theorem. Let
X, Y
be hi-unique mappings, 6 (X)Y
be inhabited species of natural numbers, and let
r 6 (X)Y, ~ 6(Y)X.
Then there exists a hi-unique mapping
such that ~ [XI= Y (Intuitionistic
Prqqf. Immediate by 16.7.1, 16.7.4. Remark.
r
analogue of Cantor-Bernstein).
16.7.2.
The requirement that
X, Y
are inhabited can be omitted; the
proof is slightly more involved in that case. We would have had no hope of obtaining this result if we had required mathematical in
r
~
to be
This is demonstrated by the following counterexample:
X = N, Y = (2n+1 : n E N } v ( m
take
: ~VXHm(X)} , cn = 2n + I, ~n = n.
It is Instructive to compare 16.7.3 with the following theorem in the literature, which can be proved without reference to the creative subject: 16.7.5. Theorem. let
r
that
Let
X, Y
be inhabited subspecies of the natural numbers, and
be bi-unique mappings of ~X] is detachable in
mapping Proof.
~
from
X
onto
X
into
Y, ~ d e t a c h a b l e
Y, Y in
into
X
respectively,
such
X. Then there exists a bi-unique
Y.
(formulated with "recursive" instead of "lawlike")
[D, M 1960] theorem 13(b) or [Rog 1967] w T.4, theorem VI. As it stands,
16.?.5 le
Just weaker than 16.7.4. However, the existing proofs of
16.7.5 contain more information than is contained in 16.?.5, since they do not refer to the creative subject. Hence we may strengthen 16.7.5 to: 16.7.6. Theorem. Under the conditions from
X
onto
Y, ~ mathematical in
of 16.7.5, there exists a bi-unique mapping r
The formulation of many results in the preceding sections may be strengthened in the same manner. Our intuitive insight
that there is an essential difference between empirical
and mathematical sequences, the creative subject.
can be expressed to a certain extent in the theory of
-
lO5
-
Take the following property as an example. Let associated with a definite assertion variables),
aA
~-x~A Let
amath
A (i.e.
aA A
be an empirical sequence does not contain non-lawlike
defined by
§ aAx = 1,
~ x A § aAx = 2, ~ x ( A V ~ A )
§ aAx = 0.
be a variable for mathematical sequences.
Then (~x(AV~A)) expresses
§
: a A)
a difference between mathem&tlcal and empirical sequences.
16.8. Let me finish with the indication of some points in the theory of the creative subject in need of further clarification. creative subject has evidence to assert
Intentionally, A
downrightly "the creative subject proves
at stage A
at stage
I used the phrase "the
m" in preference to saying m".
The second formulation suggests the creative subject as making his conclusions by one, so we might ask ourselves if we could not assert that with a sufficient "refinement" in the distinction of the stages, the creative subject draws one conclusion at a time. This supposition can be expressed axiomatically by introducing the constant
A (m)
(denoting the new conclusion of stage
m) by the
statement: ~---mA(m) ^ An(n
9 m §
{(~-m B ^ An(n
9 m §
A(m))
^
§ (B *-* A (m))}.
In some respects this looks plausible; however, in case we accept this axiom, ~ m ~ X Ax would oblige us beforehand to make in the future all deductions AO, A1, A2, A3,... explicitly, since
~X AX ~
~xVm~---mAX ,
and this seems to be a rather unnatural restraint on the future activity of the creative subject. But even worse: our new axiom leads to a downright contradiction. For since it is natural to assume that we know when a conclusion has the form "a
is a lawlike sequence" we have: A (m)
is a conclusion of the form "a
A (m)
is a conclusion of another kind.
is a lawlike sequence" or
one
-
106
Then it is possible for us to enumerate the let
A (bx)'"
-
A (m)
of the form "a
is lawlike";
is a lawlike sequence".
be the x th conclusion of this form, stating "a x
Then AxVa(A (bx)
~ "a
is a lawllke sequence")
and so we conclude to the existence of b'{x, y} : ax(Y) Intuitively
c = ~x.b'{x,x} + 1
able to indicate a
z ~ N A (bz)
a
b'
such that
.
2s a lawlike sequence, but then we ought to be
such that
~ "c
is a lawlike sequence"
which implies: Ax(b'{x,x} which is contradictory,
+ I : b'{z,x})
Clearly, the axiom of "one-concluslon at a time" makes
the highly Impredlcative
character of the theory of the creative subject acutely
paradoxical. Therefore we must leave open the possibility of infinitely many assertions
A
for which ~ A ,
at
stage
and hence the cautious formulation "having evidence for
m". Example:
if we have proved
for any special number
AxAx
at stage
m, we also have
A ~mAX
x.
But since we also required ~-m A v 1 ~ m A , we are left with the quite genuine problem of finding a satisfactory interpretation of "evidence for
A
at stage
m". We
might try something like ['trivlally deducible from the available data at stage in the same manner as
An
is trivially deducible from
are left with the problem of a general interpretation
m",
AxAx, but once again we of "trivially deducible"
which ensures decidability at any stage. Another possibility which is suggested by the "paradox" is, that the primary cause of our trouble is not so much in the assumption of the "one-conclusion-at-atime" axiom, as well as in the unrestricted possibilities
for self-reference
implicit in our axioms for the creative subject. Specifically: llke ~--nA
and~-m(~nA)
reflection
= mathematical
activities)
do belong to different
although assertions
"levels of self-reflectlon"
(self-
consideration of the course of our own mathematical
we did not distinguish between them in this respect.
-
107
-
So the following possible alternative for a theory of the creative subject suggests Itself. To each mathematical assertion and construction we suppose a level (of selfreflection)
to be assigned. We restrict ourselves to assertions
and constructions
of finite level; the unions of all assertions and constructions of finite levels form the assertions and constructions of level m . Assertions which may be understood o r constructions which can be carried out without reference to ~--n are said to belong to level zero. Assertions which are described using ~--nA of level
p, are sald to belong to level
relative to ~ n A
for
A
of level
p
for
A
of level
p+l. Likewise,
p
and constructions
constructions
are sald to be of level
defined
p+l.
We make further the following basic general assumption: "When an assertion of level we use
a (p)
p
has a proof, It has a proof of level
to denote a constructive function of level
p". Now, If
p, we can assert
(instead
of ( l o ) ) Aa (p) Vb(P+l)(Ax(ax So If we use get at level
a, b
: O) ~ V y ( b y
= O)).
as variables for constructive
AaVb(Ax(ax
= O)+-~Vy(by
functions of level ~, we still
= O)).
On the other hand, our paradox cannot be derived anymore. I feel this approach deserves further investigation.
Another problem we dld not
touch upon, is the combination of the idea of incomplete
(choice-) objects In
connection wlth the theory of the creative subject. For example, how are the "stages" of the theory of lawless and choice sequences to be related with the "stages" In the theory of the creative subject? In conclusion we might say that the theory of the creative subject Is provocative, attractive,
and dangerous; It represents the extreme consequences
subJectlvlsm;
of intultlonlstlc
undoubtedly it deserves further study, precisely for this reason.
16.9. References.
An easily accessible exposition of Brouwer's counterexamples
the creative subject ls to be found In [H 19661, 8.8.1
involving
(pp. 115-I17). Krelsel
([Kr 1967]p. 160) introduced the r e l a t l o n ~ m (there appearing as Z ~ m ) . A further discussion In connection wlth analysis is to be found In [M 1967] and [M 1968].
-
lO8
-
w This list only contains items referred to in the paper. For a more complete bibliography concerning intuitionism, Br 1920
L.E.J. B R O ~ E R ,
see
~
1955],
~
1966].
Besitzt Jede reelle Zahl eine Dezimalbruchentwickelung?
Proc. Akad. Amsterdam 23, pp. 955 - 965. Br 1924
--, Zur BegrGndung der intuitionistischen Mathematische
Br 1925
Mathematik I.
Annalen 93, PP. 244 - 258.
--, Zur BegrGndung der intuitionistischen Mathematik II. Mathematische Annalen 95, pP. 453 - 473.
Br 1926
--, Zur BegrGndung der intuitionistischen
Mathematik III.
Mathematische Annalen 96, pp. 451 - 489. Br 1926A
--, Ober Defintionsbereiche
yon Funktionen.
Mathematische Annalen 9~, PP. 60 - 76. Br 1927
--, Virtuelle Ordnung und unerweiterbare
Ordnung.
Journal fGr reine und angewandte Mathematik I~7 , pp. 255 - 258. Br 1942
--, Zum freien Werden von Mengen und Funktionen. Proc. Akad. Amsterdam 45, pp. 322 - 323 = Indagationes
Br 1942A
Math. ~, pp. 107-108.
--, Beweis dass der Begriff der Menge h6herer Ordnung nicht als Grundbegriff der intuitionistischen
Mathematik in Betracht kommt.
Proc. Akad. Amsterdam 4_~5, pp. 791 - 793 = Indagationes Math. ~, pp. 274-276. Br 1948
--, Consciousness,
philosophy and mathematics.
Proc. X th international Congress Philosophy
(Amsterdam 1948),
pp. 1255 - 1249. Br 1948A
--, Essentieel negatieve eigenschappen. Proc. Akad. Amsterdam 51, PP. 963 - 965 = Indagationes Math. 1~0, pp. 322-324.
Br
1949
--, De non-equivalentie van de constructieve en de negatieve orderelatie in het continuum. Proc. Akad. Amsterdam 52, pp. 122 - 125 = Indagationes Math. 1_~I, pp. 37-40.
-
Br 1949A
--~ Contradlctoriteit
109
-
der elementalre
meetkunde.
Proc. Akad. Amsterdam 52, pp. 315 - 316 = Indagationes Br 1952
--, Historical
background,
principles
and methods
Math. 11, pp. 89-90.
of intultlonlsm.
South African Journal of Science 49, pp. 139 - 146. D, M 1960
J.C.E. DEKKER and J. MYHILL, Berkeley and Los Angeles
F 1936
H. FREUDENTHALp
Recurslve
equivalence
types.
1960.
Zur intuitionistlschen
Deutung logischer Formeln.
Compositlo Math. ~, pp. 112 - 116. G 1958
K. GODEL, b%er eine bisher noch nlcht benutzte
Erwelterung
des flnlten
Standpunktes. Dlalectlca Gn 1968
12, pp. 280 - 287.
N. GOODMAN,
Intuitionistic
Ph.D. thesis,
arithmetic
Gn
--j A theory of constructions
H 1955
A. HEYTING,
Les fondements
de la demonstration. H 1959
--, Inflnitistlc
--, Intultlonlsm,
equivalent
to arithmetic.
des mathgmatlques.
Parls-Louvaln
methods
Proc. Symp. Warszawa H 1966
as a theory of constructions.
Stanford University. To appear.
Intuitlonnlsme.
Th6orle
1955.
from a finltlst point of view.
1959, Warszawa
an introduction.
1959, PP. 185 - 192. Second, revised edition.
Amsterdam 1966. Ho 1968
W.A. HOWARD~
Functional
Composltlo Math. Ho,K 1966
interpretation
of bar induction by bar recurslon.
20, pp. 107 - 124.
W.A. HOWARD and G. KREISEL,
Transfinite
induction and bar induction
of types zero and one, and the role of contlnulty J. of Symbolic K 1952
S.C. KLEENE, Amsterdam,
K,V 1965
analysis.
Logic 51, pp. 525 - 358.
Introduction
Gronlngen,
to metamathematlcs.
New York, Toronto 1952.
S.C. KLEENE and R.E. VESLEY, Amsterdam 1965.
In Intultlonlstlc
Foundations
of Intultlonlstlc
mathematics.
-
Kr 1958
G. KREISEL,
and the topological
proofs.
J. of Symbolic
Logic 23, pp. 369 - 388.
--, Elementary
completeness
note on negations J. of Symbolic Kr 19 62
-
A remark on free choice sequences
completeness
Kr 1958A
110
of intultionlstic
logic wlth a
Logic 23, pp. 31T - 330.
--, On weak completeness J. of Symbolic
properties
of prenex formulae.
of intultlonistic
predicate
logic.
Logic 2?, pp. 139 - 158.
Kr 1965
--, Mathematical logic, pp. 95 - 195 in: Lectures on modern mathematics, vol. ~, ed. T.L. Saaty, New York 1965.
Kr 19 67
--, Informal rigour and completeness Problems
Kr 1968
in the philosophy
--, Lawless sequences Compositio Math.
Kr 1968A
--, Functions,
of mathematics,
ordinals, congress
species, pp. 145 - 159 in: for logics methodology
ed. B. van Rootselaar
G. KREISEL and A.S. TROELSTRA, analysis (in preparation).
M 1967
J. MYHILL,
science,
of
A formal spstem for intuitionistlc
of intuitionlstic
analysis.
35, pp. 280 - 297.
--, Formal systems of intuitionistic 3 rd intern,
and philosophy
and J.F. Staal, Amsterdam 1968.
Notes towards a formalization
Logique et Analyse
Proc.
Amsterdam 1967.
of natural numbers.
Kr,T
M 1968
ed. I. Lakatos,
20, pp. 222 - 248.
Proc. 3 rd intern, science,
proofs, pp. 138 - 186 In:
congress
analysis
l, pp. 161 - 178 in:
for logic, methodology
ed. B. van Rootselaar
and JeF. Staal,
and philosophy
of
Amsterdam 1968.
N 1966
T. NAGASHIMA, An extension of the Cralg-SchGtte interpolation theorem. Annals of the Japan Association for Philosophy of Science ~, pp. 12 - 18.
S 1962
K. SCHOTTE,
Der Interpolationssatz
Mathematische
Annalen ~ ,
pp.
192
der Intultlonlstischen -
200.
Pr~dikatenloglk.
-
Sp 1962
111
-
C. SPECTOR, Provably recurslve functionals
of analysis;
proof of analysis by an extension of principles intultlonlstlc mathematics,
pp. 1 - 27 in:
Recurslve function theory, Proc. Symp. Pure Mathematics
T 1966
A.S. TROELSTRA,
T 1967
--, Intultlonlstlc
Intultlonlstlc
general topology.
(1962).
Thesis. Amsterdam 1966.
continuity.
Nleuw Archlef voor Wlskunde T 1967A
a consistency
formulated in current
(3) 15, pp. 2 - 6.
--, Finite and infinite In lntultlonlstlc mathematics. Composltlo Math. 18, pp. 94 - 116.
T 1968
--, The theory of choice sequences~ pp. 201 - 223 In: Proc. 3 rd intern, congress for logic, methodology and philosophy
of
science, ed. B. van Rootselaar and J.F. Staalm Amsterdam 1968. T 1968A
--, One-point
compactlflcatlons
of Intultlonlstlc
locally compact spaces.
Fundamenta Mathematlcae 57, pp. 75 - 93. T 1968B
--, The use of "Brouwer's principle"
in Intultlonistlc topology w
pp. 289 - 298 In: Contrlbutlons to Mathematical Rog 1967
Hartley ROGERS Jr., Theory of recurslve functions computability,
R 1960
Amsterdam 1968.
an4 effective
New York etc. 1967.
B. van ROOTSELAAR,
0~ Intudtlonistlc difference relations.
Proc. Akad. Amsterdam A 63 R 1963
Logic. ed. by K. SchGtte,
= Indagationes Math. 22, pp. 316 - 322.
--, Corrections to the paper "On Intultionlstlc difference relations". Proc. Akad. Amsterdam A 66 = Indagatlones Math. 25, pp. 132 - 133.
Offsetdruck: Julius Behz, Weinheim/Bergstr.