4
$\begingroup$

I am trying to read Jerome Keisler's book "Model Theory of Infinitary Logic". I got to the part about the undefinability of well-orders, and there I got stuck. For completeness of this post, we have a consistency property as:

enter image description here

By a Model Existence Theorem, if $S$ can be shown to be a consistency property, then each $s \in S$ has a model...

Now, the theorem about undefinability of well-orderings in the book is:

enter image description here

The proof is rather short (only 2 pages), which involves defining a certain $S$ as follows:

enter image description here

So the trick seems to be to show that $S$ as defined above is a consistency property. I understand the proof that $S$ satisfies $(C5)$ ... But for $(C7)$, Keisler only writes:

enter image description here

Could anyone help me understand how come $S$ satisfies $(C7)$ ? Any help would be really really appreciated as I am have been stuck here already for quite some time .... I have been trying out some proofs such as showing that:

$$ (\mathfrak{M},b_1,\dots,b_n) \models (\forall x_1 \dots x_n) (\bigwedge s_0(x_1 \dots x_n d_1 \dots d_n) \rightarrow \neg (d_r = c)) $$

and then deriving a contradiction (i.e. arriving at $\neg (d_r = d_r)$), but this seems to be far from Keisler's approach ..

$\endgroup$

1 Answer 1

5
$\begingroup$

To prove that $S$ has $(C7)$ you need $S$ to satisfy three things,

  1. For all $c,d\in C$, if $(c=d)∈s∈S$ then $s∪\{d=c\}∈S$

This part is obvious from the definition of $S$, as if $(\mathfrak A, b_0,\ldots)$ witness $s\in S$, it will also witness $s∪\{d=c\}∈S$

  1. For all basic terms $t$ and for all $c\in C$, if $(t=c),φ(t)∈s\in S$, then $s\cup \{φ(c)\}∈S$

Same argument from above will show that that holds

  1. For all basic terms $t$, there exists $e\in C$ such that if $s\in S$ implies $s\cup\{t=e\}∈S$

To show that, first note that if $t$ is a basic term that is not of the form $d_r$ for some rational $r$ and $s∈S$ (where $s=s_0∪T\cup \cdots$) with witness $(\mathfrak A, b_0,\ldots)$ then $s_0$ doesn't contain any mention of $t$ and any mention of some $c\in C$ (as it contains only finitely many such $c$), so we can just let $(\mathfrak A', b_0,\ldots)$ be the same as $(\mathfrak A, b_0,\ldots)$ but where we assigne $c$ to be $t$.

So all we need to show is that if $d_r∈D$ and $s\in S$, then $s\cup\{d_r=e\}∈S$ for some $e\in C$. The claim Keisler proves is that if $e$ does not accure in $s_0$, then indeed $s\cup\{d_r=e\}∈S$, which proves that $S$ has $(C7)$


To actually show the last paragram consider $s=s_0\cup T\cup \{d_{r_0}<d_{r_1}<d_{r_2}<\cdots<d_{r_n}\}∈S$ where $r_0<r_1<\cdots<r_n$ are rationals, for now assume $r$ appears in $\{d_{r_0}<d_{r_1}<d_{r_2}<\cdots<d_{r_n}\}$, so we call $r=r_p$ where $0\le p\le n$ and $d_r=d_{r_p}$.

Let $s_1(\bar c, e, d_{r_0},\cdots,d_{r_n})=s_0(\bar c, d_{r_0},\cdots,d_{r_n})\cup \{d_r=e\}$ where $e$ does not appear in $s_0$, and consider $s'=s_1\cup T\cup \{d_{r_0}<d_{r_1}<d_{r_2}<\cdots<d_{r_n}\}$, we want to show that $s'∈ S$.

Let $α<ω_1$ be arbitrary, we want to find a model $\mathfrak B_α$ in $K$ and assignments $b_0,\ldots,b_n$ to $d_{r_0},\ldots,d_{r_n}$ such that $(\mathfrak B_α,b_0,\ldots)⊨ ∃\bar x∃y (\bigwedge s_1(\bar x, y, d_{r_0},\ldots d_{r_n}))$ such that [the confusing condition on the $b_i$s is true].

Now if $\mathfrak A_α$ with $a_0,\ldots, a_n$ is witness for $(\mathfrak A_α,a_0,\ldots)⊨ ∃\bar x (\bigwedge s_0(\bar x, d_{r_0},\ldots d_{r_n}))$, then note that by letting $\mathfrak B_\alpha=\mathfrak A_α$, and $a_i=b_i$ we are finish (witness with $y$ being $b_p$)

All we left to do is to show why we can assume $r=r_p$ for some $p$.

Assume $r$ is not in there, so let $p$ be such that $r_p<r<r_{p+1}$ (the cases where $r<r_0$ or $r>r_n$ are pretty much identical so I won't talk about them).

As before we shall open the definition, let $α<ω_1$ be arbitrary, we want to find a model $\mathfrak B_α$ in $K$ and assignments $b_0,\ldots,b_n$ to $d_{r_0},\ldots,d_{r_n}$ as well as an assignment $b$ for $d_r$ such that $(\mathfrak B_α,b_0,\ldots,b_n,b)⊨ ∃\bar x(\bigwedge s_0(\bar x, d_{r_0},\ldots d_{r_n}))$ (note here that $s_0$ does not depends on $d_r$) such that $$α≤b_0, b_0+α≤b_1,\cdots, b_p+α≤b, b+α≤b_{p+1},\cdots b_{n-1}+α≤b_n$$.

For this, consider $β=α+α<ω_1$ and the model $\mathfrak A_{β}$ and the assignments $a_0,\ldots, a_n$ to $d_{r_0},\ldots,d_{r_n}$ such that $(\mathfrak A_α,a_0,\ldots,a_n)⊨ ∃\bar x(\bigwedge s_0(\bar x, d_{r_0},\ldots d_{r_n}))$ such that $$β≤a_0, a_0+β≤a_1,\cdots, a_p+β≤a_{p+1},\cdots a_{n-1}+β≤a_n$$ (Such $\mathfrak A_β, a_i$ exists because $s\in S$)

Note that we also have we also have $$α≤a_0, a_0+α≤a_1,\cdots, a_p+α≤a_p+α, (a_p+α)+α≤a_{p+1},\cdots a_{n-1}+α≤a_n$$

So we finish the proof by letting $\mathfrak B_α=\mathfrak A_β$, $b_i=a_i$ and $b=b_p+α$.

$\endgroup$
5
  • $\begingroup$ thanks a lot, could you please expound a bit on how Keisler gets to show that if $e$ does not occur on $s_0$ then indeed we have $d_r=e$ ? I do not understand how the $\beta$ and $\alpha$ in his proof comes into play $\endgroup$ Commented Jul 25 at 22:02
  • 3
    $\begingroup$ @LinkL looking at the proof, Keisler couldn't have wrote it in a more messy way even if he tried... I will add to my answer a more detailed explanation $\endgroup$ Commented Jul 25 at 22:06
  • 1
    $\begingroup$ @LinkL I added a detailed explanation, I hope it is clear from it $\endgroup$ Commented Jul 25 at 22:40
  • $\begingroup$ one last clarification, you wrote "Now if $\mathfrak A_α$ with $a_0,\ldots, a_n$ is witness for $(\mathfrak A_α,a_0,\ldots)⊨ ∃\bar x (\bigwedge s_0(\bar x, d_{r_0},\ldots d_{r_n}))$, then note that by letting $\mathfrak B_\alpha=\mathfrak A_α$, and $a_i=b_i$ we are finish (witness with $y$ being $b_p$)" ... ah just to clarify this bit .... am i right that we can say its finished because $e$ does not appear in $s_0$ so that we can have $(d_r=e) \wedge \bigwedge s_0$ ? thank you ! $\endgroup$ Commented Jul 25 at 23:46
  • 2
    $\begingroup$ @LinkL Yes, if we write $\bigwedge s_1$ explicitly we want to get $(\mathfrak B_α,b_0,\ldots)⊨ ∃\bar x∃y (d_r=y\land \bigwedge s_0(\bar x, y, d_{r_0},\ldots d_{r_n}))$ but because $e$ does not appear in $s_0$ we get $(\mathfrak B_α,b_0,\ldots)⊨ ∃\bar x∃y (d_r=y\land \bigwedge s_0(\bar x, d_{r_0},\ldots d_{r_n}))$, which obviously holds for our choice of $\mathfrak B_\alpha$ and $b_i$ $\endgroup$ Commented Jul 26 at 0:22

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.