1

So I'm trying to come up with a proof of the above action being possible or not, using the compactness theorem from logics (a set of first-order sentences has a model if and only if every finite subset of it has a model) on this one but I'm not quite sure how to split up the finite subsets of $\mathbb{N}$ to make the condition true.

(Note: Neighboring here means that for instance $X,Y \subseteq \mathbb{N}$ are neighboring if you can get $X$ by adding an element to Y, i.e. $X=Y \cup \{c\}$ for a c $\not\in Y$ or the other way around. $\{2,41\}$ and $\{0,2,41\}$ would be neighbors but not $\{3,4\}$ and $\{3,5\}$

Note 2: $\dot\cup$ is the disjoint union

Note 3: Referring to all possible subsets of $\mathcal{P}(\mathbb{N})$. @Gae. S. provided a very nice answer using the axiom of choice, but I am looking for a way to prove it with the compactness theorem of logics which is in a certain way similar to the axiom of choice. )

I'd really appreciate some help on this :) Thanks so much!

  • 2
    What are "neighboring sets" ? – Fred May 03 '21 at 10:17
  • Well, the definition of a neighborhood. $X,Y \subseteq \mathbb{N}$ are neighboring if for instance you can get $X$ by adding an element to Y, i.e. $X = Y \cup {c}$ for a c $\not\in Y$ or the other way round. ${5,42}$ and ${0,5,42}$ would be neighboring, but not ${4,5}$ and ${4,6}$. Maybe I got the wrong term for this definition? – Emma Lee May 03 '21 at 10:17
  • Is $\mathcal P (N)$ the same as $\mathcal P (\mathbb N)$? If so, please edit the question to correct that, or explain what $\mathcal P (N)$ actually is. For that matter, also worth explaining that $\mathcal P$ denotes the power set. Also, can you explain the notation $\dot \cup$? – Prime Mover May 03 '21 at 10:17
  • Oh yes I'm sorry, just edited the question! And well, $\dot\cup$ is the definition for the union of two disjoint sets... – Emma Lee May 03 '21 at 10:22
  • Okay, so you need to add that definition to the question itself, it doesn't help much to leave it hidden in the comments. – Prime Mover May 03 '21 at 10:27
  • 1
    ... and of course add that $\dot \cup$ means the disjoint union. Some mathematicians use different notations, and this (for example) is unfamiliar to me, as I use $\sqcup$. – Prime Mover May 03 '21 at 10:37
  • @EmmaLee Can you check to see whether the question actually refers to finite subsets of $\mathcal P (\mathbb N)$? If so, then the answer is as straightforward as the one I presented. Otherwise it gets into properties of infinite sets which are difficult to grasp intuitively, and probably inappropriate for a class in logic. OTOH the compactness theorem is itself pretty advanced, so it may be the case that you do need to consider infinite subsets -- but at this stage it's way over my head. – Prime Mover May 03 '21 at 11:38
  • It refers to all possible subsets of $\mathcal{P}(\mathbb{N})$, but it'd be a start to determine how to split up the finite subsets. I think for that purpose your answer is just right! – Emma Lee May 03 '21 at 11:47
  • May I ask where you encountered this problem? If it comes from a textbook, the context of the material leading up to the exercise would be relevant. – hardmath May 10 '21 at 03:30

3 Answers3

5

This is essentially a variation of the dwarfs-and-hats problem, or at least the same use of the axiom of choice.

Consider the equivalence relation on $\mathcal P(\Bbb N)$ defined by $$X\equiv Y\iff \lvert X\triangle Y\rvert<\aleph_0$$

where $X\triangle Y=(X\setminus Y)\cup(Y\setminus X)$ and $\lvert X\rvert$ is the cardinality of $X$. Consider the quotient set $M=\mathcal P(\Bbb N)/_\equiv$ and a map $f:M\to \mathcal P(\Bbb N)$ assigning to each equivalence class a representative. Id est, a map such that $f([X]_\equiv)\triangle X$ is a finite set for all $X\subseteq \Bbb N$.

Then, define the two sets by $$A=\{X\in\mathcal P(\Bbb N)\,:\, \lvert X\triangle f([X]_\equiv)\rvert\text{ is odd}\}\\ B=\{X\in\mathcal P(\Bbb N)\,:\, \lvert X\triangle f([X]_\equiv)\rvert\text{ is even}\}$$

  • I bet this answer is magnificent but there are a lot of definitons I don't understand here that are way too advanced for me :( This was actually a question from my logic class, that's why I'd wanted to try the compactness theorem. – Emma Lee May 03 '21 at 10:49
  • I know this solution and I don't know the compactness theorem, so I cannot help you with that. –  May 03 '21 at 10:49
  • No worries, thank you for your help though!! – Emma Lee May 03 '21 at 10:53
  • To see that $\equiv$ is a equivalence relation Henno Brandmsma's answer to this post might be helpful – Peter Melech May 03 '21 at 11:02
2

Let $p_S$ be a propositional variable for each $S \subset \mathbb N$.

Take for axioms the uncountably many sentences:

$$ p_S \not\equiv p_T $$

whenever $S,T \subset \mathbb N$ are neighboring, i.e. their symmetric difference is a singleton set.

We show that any finite subset $\mathbf S$ of those sentences has a model (and so is consistent).

Consider the graph $\mathcal G$ whose vertices are the finitely many sets corresponding to any of the sentential variables $p_S$ appearing in $\mathbf S$, and whose edges are between pairs of neighboring sets.

Lemma The graph $\mathcal G$ is bipartite.

Proof It suffices to show that $\mathcal G$ has no cycles of odd length.

Let $S_0 \sim S_1 \sim \ldots \sim S_n$ be a path in $\mathcal G$.

Since each edge in a cycle signifies the addition or removal of an element from one set to obtain the next set, any two nodes in a path will have a finite symmetric difference. By induction these symmetric differences satisfy:

$$ | S_i \Delta S_j | \equiv |i-j| \bmod 2 $$

Applying this to a cycle of length $n$, where necessarily $S_0 = S_n$, we see that the cycle length is even. QED

Finally we interpret $\mathcal G$ as the model for the sentences $\mathbf S$. Let all the sentential variables $p_S$ for sets $S$ in one part of $\mathcal G$ be true and those variables for sets in the other part be false. Since the axioms merely require that two propositional variables which correspond to neighboring subsets of $\mathbb N$ have opposing truth values, all the axioms in $\mathbf S$ are satisfied in this model.

It follows by the compactness theorem for the propositional calculus that there exists a model, and therefore a consistent interpretation, of the full set of (uncountably many) axioms. This interpretation bipartitions $\mathcal P(\mathbb N)$ into those subsets $S$ of $\mathbb N$ corresponding to true values of $p_S$ and those corresponding to false values, just as desired.

hardmath
  • 37,715
0

The way I understand it, two sets are neighbors if one has $n$ elements and the other has $n + 1$ elements.

You could put all the sets with an odd number of elements into $A$, and all the sets with an even number of elements into $B$.

Prime Mover
  • 5,214
  • 2
  • 16
  • 32
  • $\mathcal{P}(\mathbb{N})$ does not consist only of finite sets though? – Peter Melech May 03 '21 at 10:38
  • Good call ... hadn't thought about that. Can you use transfinite induction at this stage? I would presume that at the point at which the question has been asked, the student would have been introduced to it. – Prime Mover May 03 '21 at 10:40
  • I have no idea what a transfinite induction is xD – Emma Lee May 03 '21 at 10:41
  • @EmmaLee okay no worries, mark my answer down then, sorry, it just does not do the job. – Prime Mover May 03 '21 at 10:43
  • nooo its okay dont worry :) thanks for your help anyway! – Emma Lee May 03 '21 at 10:49
  • Considering the level of mathematics at which OP is working, and that this is an exercise in logic, this answer may in fact be adequate for its purpose. If I had not been introduced to the delights of the infinite, and especially if the question discussed finite subsets of $\mathcal P (\mathbb N)$, if I were @EmmaLee I would probably present this as my solution to the exercise. – Prime Mover May 03 '21 at 11:36