I want to compare a standard model with a non-standard model.
Call the standard model M and the non-standard model M'.
> As I noted in [1], M has a element which in M satifies the
> definition of aleph_1. In [1] I didn't name that element, but now let
> us call it w1.
>
> I neglected to spell out in [1] that M also has an element which
> in M satisfies the usual ZF definition of omega. Call this element w.
M has w and w1, M' has w' and w1'.
> ZFC proves there is no bijection from omega to omega_1. So M will
> satisfy the corresponding sentence about w and w1.
M proves there is no bijection from w to w1.
M' proves there is no bijection from w' to w1'.
> This sentence talks about bijections, which are functions, which are
> formalized in ZFC as sets of Kuratowski ordered pairs. So we expand
> out all those definitions, including the definition of Kuratoswki
> ordered pairs. That becomes a definition in just primitive relation
> symbols = and epsilon, and usual f.o. logic over those relation
> symbols.
Sets don't exist "inside the model" because PA doesn't have
axioms for sets. We can define a weak set theory "inside the model".
> To see how that sentence evaluates in M, we replace all occurences
> of epsilon symbol by E.
I assume E somehow encodes order type.
Given two sets, E and E', could we tell which
one is the epsilon function for the standard model?
Does a standard model have to have order type omega?
> So we get a complicated statement in M, mentioning E and all
> f.o. quantifiers from the definition become quantiofication over the
> universe of M.
>
> In this sense we get the statment that there is no bijection between
> w and w1 in M, ie "omega" and "omega_1".
Our weak set theory "inside the model" doesn't have
an axiom of infinity, so the model can't "prove" w and w1
even exist.
Are you saying ZFC proves the set that is a
bijection between w and w1 (as defined in ZFC)
can't exist inside the model?
> In [1] I neglected to mention "omega" and the bijection in question
> was between "omega" and "omega_1".
>
> But stepping back up to ZFC above M, there is another sentence true
> in ZFC language about M, not in M's internal language, which is a
> reasonable formalization of the statement that in some sense there is
> a bijection between w and w1.
Does "in some sense" mean we can construct a set of
ordered pairs, (wi, w1i). If so, why can't we encode
this set using our weak set theory inside the model?
> Namely w is an object in M, a model of set theory. So w has
> certain "members" in the sense of that model, namely those element in
> the universe of M which are in the E relation to w.
>
> Since M was a countable model, this set of E members of w is a
> countable set.
I don't think PA has any models, let alone a
countable non-standard model.
Infinite Sets Can't Exist
Posted: Feb 19, 2010 1:45 PM
http://www.mathforum.com/kb/message.jspa?messageID=6987273&tstart=0
I show how, given a bijection between A and B,
we can construct a bijection between A/B and B/A.
This construction always works when A and B are finite.
If A is infinite and B is a proper subset of A,
the method constructs a bijection between a
non-empty set and the empty set.
What Does a Non-Standard Model of PA Look Like in ZFC?
RussellE Nov 2, 6:10 pm
http://groups.google.com/group/sci.logic/browse_thread/thread/9c6a302796385add
I show how we can use the powerset of a non-standard
number to define an "uncountable" number of
different Z chains in the model.
> Similarly, we can in ZFC over M form the set of all those elements
> of the M universe which are in E relation to w1, ie E members of
> w1.
ZFC proves "uncountable in the model" is really countable in ZFC.
> These 2 sets are reasonable formalizations in ZFC over M of what
> those "sets" from inside M represent.
>
> But w1 only has according to ZFC countably many such actual E
> members, since ZFC knows M was only a countable model.
Because Compactness is a theorem of ZFC.
> So ZFC sees both those defined sets based on w and w1 are actually
> countable and so have a bijection.
Can ZFC "construct" this bijection?
If so, we can encode the bijection "in the model".
The model can't prove this is a bijection
but the bijection would exist in the model.
> That is what [1] was trying to write about, though it sure skipped
> details.
>
> In [2] you wrote, respopnding to [1]:
>
> >I don't understand this part at all.
>
> Yes, I had skipped over too much in [1]. I hope I have explained it
> better now.
>
> You continued:
>
> >Let B = {0,1,2,...,b-1} where b is non-standard.
> >Let P be the powerset of B.
> >In the model, the cardinality of P equals
> >some non-standard natural number.
> >Let e = |P|. In the model, there is
> >a bijection between P and E where
> >E = {0,1,2,...,e-1}.
>
> Yes, I agree.
Where P is the powerset of B "in the model".
Since B is an infinite set in ZFC,
ZFC thinks the powerset of B is P2,
an uncountable set.
> >In ZFC, P is an uncountable set.
> >Who is right? Is the model correct
> >and the cardinality of P is "countable"
> >(equal to some natural number).
> >Or is ZFC correct in believing P
> >is uncountable.
>
> I thought we were talking about countable nonstandard PA models in
> this thread.
We are assuming there are such models.
> The PA model would interpret hereitarily finite sets over integers
> as I wrote in [1]. That interprets the epsilon symbol into symbols
> =, S, +, *. Then those symbols are interpeted into realtions ond
> functions on the PA model.
>
> If we want to recast those as sets in ZFC meta-theory over the PA
> model we would extract sets, similar to how I discussed over a ZFC
> model M as above.
>
> If we do that we don't get back ZFC full powerset over the extracted
> version of B = {0, ... , b-1}.
Huh? A natural number is a natural number.
Let M be a standard model and let B={0,1,2}.
ZFC says the powerset of B is infinite?
The standard model only "extracts" a finite
subset of "true" powerset of B?
In PA, the cardinality of the poweset of a
natural number is equal to a natural number.
> Instead we just get the PA models weak version of powerset, recoded
> back into ZFC. In general a subset of the true ZFC pwerset.
Our "weak" set theory inside the model can't
define the entire powerset of a natural number?
> And if the PA model was countable, it will indeed be a proper subset
> since it will be only a countable set in ZFC.
>
> So ZFC will take that P to be countable.
We are talking about two different sets here.
We have P which is what our weak set theory
inside the model says is the powerset of B.
ZFC thinks B is an infinite set and the
powerset of B is P2, an uncountable set.
The model and ZFC agree there is a bijection between P and E.
The model thinks P is the powerset of B.
ZFC thinks B and E have a bijection.
ZFC proves there is a bijection between B and P.
We can encode the bijection between B and P
inside the model.
> Above with Skolem's "paradox" was non-existence statements about
> bijections need not lift from the model to above.
So, this isn't really Skolem's paradox.
ZFC and the model agree there is a bijection
between P and E.
> Ie, you are the model and I am ZFC over the model. You are
> nearsighted, and I can see farther than you.
I don't know why you think "the model" is smaller than ZFC.
Yes, we are only allowing the model to have a
weak form of set theory. Even so, sets ZFC thinks
are "infinite" are "finite" in the model.
If we allowed the model to have its own
non-standard version of ZFC, this non-standard
ZFC would laugh at our puny standard ZFC.
Imagine a set theory that thinks {0,1,2}
is an infinite set.
> When you see things, I see them too, and they are really there.
>
> When you can't see something and you say there are no such things,
> maybe there really are none as you said, or else maybe there are and
> I can see them, and you just couldn't because you didn't see far
> enough.
>
> Anyway, the issue here was never about the bijection between E and
> P. The PA model see it, and as your question suggests ZFC over the
> model sees it.
This is like the great and powerful Oz saying
"Pay no attention to that man behind the curtain!"
http://www.imdb.com/title/tt0032138/quotes
> The bijection in question is not that, but instead the bijection
> between B = {0, ... , b-1} and B's powerset P.
ZFC proves there is a bijection between P and E.
ZFC proves there is a bijection between B and E.
ZFC proves there is a bijection between B and P.
We don't even have to talk about powersets.
Proving there is a bijection between B and E
is enough to get a contradiction.
B = {0,1,2,...,b-1}
E = {0,1,2,...,e-1}
e > b
B is a proper subset of E.
In ZFC, B and E are infinite sets of finite ordinals.
Finite ordinals can be ordered by set membership.
(I mean set membership as defined in ZFC,
not the model.)
Assume we order B and E by set membership.
We can now define a bijection between B and E.
This bijection is a set of ordered pairs, (b,e),
where b and e are finite ordinals.
This bijection can be encoded as a set
"in the model".
There exists a bijection between B and E
"in the model" even if the model can't
prove such a bijection exists.