Moderators: kiore, Blip, The_Metatron
Definitely the latter. ω specifically refers to the first infinite ordinal, and ω + 1 to the second, ω + 2 to the third and so on. The first infinite ordinal is the natural numbers.cateye wrote:Which brings up another subtlety which I'd like to have cleared up before we really start getting into this: Do we, by ω mean the set that is postulated to exist by the axiom of infinity (as in ZF or NBG) or do we mean the minimal set that is closed under the successor operation, ie. the intersection of all sets that are closed under the successor operation?
cateye wrote:This thread is a spawn of the "Riemann Hypothesis" thread, which led to a discussion on model theory. Before we start getting onto it, I propose that we get our terminology clear:
"True" and "False" should be used semantically, ie. referring to whether a formula is semantically true or false in a specific model of a theory, whereas "provable" and "refutable" should be used syntactically, ie. referring to whether a specific formula can be deduced in some theory.
Also "theory" should be defined: I propse the definition that a (deductive) theory is a set of sentences in a formal language which is closed under logical consequence.
Furthermore I propose to distinguish between a model and an interpretation of a theory - an Interpretation shall be called model iff. any sentence of the theory is semantically true in that interpretation.
For starters, because this is spawned from the RH thread, and in order to ensure a passionate discussion, I will throw this question out:
Are hypercomplex numbers a model of complex analysis? What models of complex analysis are there and how many of them?
What do you mean by "complex analysis"? Complex analysis is first of all a branch of mathematics (normally performed within the framework of some theory like ZFC), not a formal theory.cateye wrote:Are hypercomplex numbers a model of complex analysis?
There are infinitely many models of "complex analysis" (assuming by "complex analysis" you mean some first-order theory with an infinite model), at least one for each infinite cardinality.What models of complex analysis are there and how many of them?
Meaning? In set theory, ω is defined as the set of natural numbers (finite ordinals).VazScep wrote:So I'd like to formulate the question as: can you recover the hypercomplex numbers by considering a model of set theory in which ω is not isomorphic to the set of natural numbers?
Preno wrote:What do you mean by "complex analysis"? Complex analysis is first of all a branch of mathematics (normally performed within the framework of some theory like ZFC), not a formal theory.cateye wrote:Are hypercomplex numbers a model of complex analysis?
Since complex analysis includes complex numbers, functions and expressions like "every derivative (of some function)", if you want to come up with some theory tailored specifically to complex analysis, you'd presumably have to use many-sorted logic (i.e. some form of type theory).There are infinitely many models of "complex analysis" (assuming by "complex analysis" you mean some first-order theory with an infinite model), at least one for each infinite cardinality.What models of complex analysis are there and how many of them?Meaning? In set theory, ω is defined as the set of natural numbers (finite ordinals).VazScep wrote:So I'd like to formulate the question as: can you recover the hypercomplex numbers by considering a model of set theory in which ω is not isomorphic to the set of natural numbers?
We're leaving out type theory for now (though you might guess from my avatar that I am very much fond of it!). We want the completeness theorem. But that's okay. We have first-order set theory in which functions and quantification over functions and properties is just quantification over sets.Preno wrote:Since complex analysis includes complex numbers, functions and expressions like "every derivative (of some function)", if you want to come up with some theory tailored specifically to complex analysis, you'd presumably have to use many-sorted logic (i.e. some form of type theory).
Right. But if you embed PA in ZFC using the finite ordinals and then construct the Godel sentence of ZFC, you're forced to conclude that there are non-standard models. I'd need to think about this a little more carefully to be absolutely sure, but in these non-standard models, the interpretation of "ω", by which I mean the appropriate constant symbol definable in ZFC, would not be isomorphic to ω, by which I mean the set of natural numbers.Meaning? In set theory, ω is defined as the set of natural numbers (finite ordinals).VazScep wrote:So I'd like to formulate the question as: can you recover the hypercomplex numbers by considering a model of set theory in which ω is not isomorphic to the set of natural numbers?
cateye wrote:Yeah, I think it's confusing to just jump into this thread out of context. You should have a quick glance at the Riemann Hypothesis thread to get an idea of what we're up to. Just saying..
Welcome to the forum, btw!
Preno wrote:cateye wrote:Yeah, I think it's confusing to just jump into this thread out of context. You should have a quick glance at the Riemann Hypothesis thread to get an idea of what we're up to. Just saying..
Okay, I just the last couple of pages, but tbh I still don't understand what complex analysis (as a formal theory) is supposed to be, or what the distinction between ω and the set of natural numbers is supposed to be.Welcome to the forum, btw!
Thanks.
Preno wrote:I still don't understand ... what the distinction between ω and the set of natural numbers is supposed to be.
cateye wrote:The (more general) question we're trying to figure out, is if we can assume to have the usual complex numbers (and everything that's proven about them) in any model of set theory or not, or if in some models we necessarily end up having hypercomplex numbers.
cateye wrote:The (more general) question we're trying to figure out, is if we can assume to have the usual complex numbers (and everything that's proven about them) in any model of set theory or not, or if in some models we necessarily end up having hypercomplex numbers. Ashley brought up the idea that in nonstandard models constructing the rational numbers from ω in the usual way (equivalence classes of pairs, etc. ) will lead to hyperrational and therefore eventually to hypercomplex numbers, which would have implications on the decidability of the RH. At least thats my (limited) understanding.
ashley wrote:The natural numbers are the real natural numbers that Platonists would say really exist. ω is a thing in set theory that tries to formalise them, but since set theories are first-order theories they can only guarantee that ω satisfies all the right first-order formulae of the language of arithmetic, which is not the same as being isomorphic to the naturals. We are contemplating models of set theory in which ω is not actually the natural numbers.
ashley wrote:Didn't you mention Skolem's paradox in the other thread? There the whole model is countable so clearly its C can't be the usual C.
ashley wrote:
Didn't you mention Skolem's paradox in the other thread? There the whole model is countable so clearly its C can't be the usual C. I can sketch a proof that there are models with complex numbers whose absolute value is bigger than any of the usual natural numbers if you like.
ashley wrote:
I've thought of a good example of a bad argument analogous to the one that was being made in the other thread and has a more obviously wrong conclusion.
Take a Turing machine and an input. Formalise the notion of Turing machine in ZF. Either we can prove in ZF that the machine halts or we can't. If we can't, then there must be a model of ZF where it does halt. But if it halts we can show it halts by running it. So it must be decidable. So we can always decide whether a Turing machine will halt.
And this fails for quite a similar reason. When we formalised Turing machines we will have done something along the lines of constructing for each machine T and input i a function fT,i : ω -> ω that we can prove is always primitive recursive and has 0 in its image iff the machine halts on the input. It would have to be something like that. Then we took a model in which 0 is in the image of fT,i. But this model must have had non-standard naturals in ω, so that even though 0 is in the image you would never find it. So it wasn't true to say that if it halts according to some model, we can decide whether the actual Turing machine will halt.
Come to think of it, we can probably use the Weierstrass factorization theorem to construct a meromorphic function that has zeros off the line Re(z)=1/2 iff the machine halts.
Preno wrote:cateye wrote:The (more general) question we're trying to figure out, is if we can assume to have the usual complex numbers (and everything that's proven about them) in any model of set theory or not, or if in some models we necessarily end up having hypercomplex numbers. Ashley brought up the idea that in nonstandard models constructing the rational numbers from ω in the usual way (equivalence classes of pairs, etc. ) will lead to hyperrational and therefore eventually to hypercomplex numbers, which would have implications on the decidability of the RH. At least thats my (limited) understanding.
Well, that depends on what you mean by "the usual complex numbers". Yes, in any model of ZF(C), one can construct the complex numbers as ordered pairs of reals, reals as the Dedekind completion of the rationals and rationals as ordered pairs of naturals, but the question is how you're equating the complex numbers of different models. Yes, everything that can be proven about the complex numbers in ZF(C) is true in any model of ZF(C). No, the complex numbers in one model don't have to be isomorphic to the complex numbers in another model, even if the models are elementarily equivalent. Yes, if ω (defined as the intersection of all inductive sets, i.e. sets which along with any set x also contain its successor x+{x}) contains non-standard members in a given model of ZF(C) - i.e. members not designated by any term - then naturally, the complex numbers will also contain non-standard members (i.e. members which aren't ordered pairs of the Dedekind completion of ordered pairs of standard naturals). I doubt this has any relevance to the RH, though, as any sum or product which involves a non-standard natural is necessarily non-standard (although I haven't been following that argument in detail).
cateye wrote:We have established a pretty airtight case in the other thread that if the RH is false in the "standard" model, then it can be proven false in that model (ie. always assuming we're talking about the usual complex numbers).
Preno wrote:cateye wrote:We have established a pretty airtight case in the other thread that if the RH is false in the "standard" model, then it can be proven false in that model (ie. always assuming we're talking about the usual complex numbers).
If you have, that's pretty curious, since the phrase "proven false in a model" is meaningless.
cateye wrote:Preno wrote:cateye wrote:We have established a pretty airtight case in the other thread that if the RH is false in the "standard" model, then it can be proven false in that model (ie. always assuming we're talking about the usual complex numbers).
If you have, that's pretty curious, since the phrase "proven false in a model" is meaningless.
I was being sloppy in terminology here: if it is false there exists a counterexample, ie. a non-trivial zero outside the Im(z)=1/2 line that we can find in finite steps.
Preno wrote:cateye wrote:Preno wrote:cateye wrote:We have established a pretty airtight case in the other thread that if the RH is false in the "standard" model, then it can be proven false in that model (ie. always assuming we're talking about the usual complex numbers).
If you have, that's pretty curious, since the phrase "proven false in a model" is meaningless.
I was being sloppy in terminology here: if it is false there exists a counterexample, ie. a non-trivial zero outside the Im(z)=1/2 line that we can find in finite steps.
What do you mean by being "found in finite steps"?
I skimmed the thread again, and I can't find the argument to that effect. The book of de Sautoy's which I just - let's use a neutral term - procured also contains merely the bare assertion that "if it is false then there is a zero off the critical line which we can use to prove it is false". Yeah, no. If the Goldbach hypothesis is false, then it's decidable, since we can find a term n referring to the counter-example. But no such thing is true of the real or complex numbers (not to mention that our notion of what it means to be true of the real numbers is much less clear than our notion of what it means to be true of the natural numbers).
Users viewing this topic: No registered users and 1 guest