2
$\begingroup$

My REU partner and I are working on a problem involving iterations of quadratic rational maps over an algebraically closed field $K$ that is complete with respect to a non-trivial non-archimedean absolute value $|\cdot|$. We've reduced it to the problem of determining whether $\phi(x) = k(x + x^{-1})$ is post-critically finite for $0 < |k| < 1$ (that is, determining whether the critical points $\pm 1$ are preperiodic). Our conjecture maintains that $\phi$ is not post-critically finite, but this is where we're stuck.

A simple strategy would be to show that some $n$-periodic point $\gamma$ of $\phi$ strictly attracts a critical point of $\phi$. According to a recent paper (Theorem 1.5, p. 4), if $p$ is the residue characteristic of $K$ and either $p=0$ or $p>2$ (assumptions we're willing to make), then a critical point of $\phi$ will be attracted to the cycle containing an $n$-periodic point $\gamma$ of $\phi$ if the multiplier $m(\phi^n,\gamma)$ is strictly between $0$ and $1$.

Unfortunately, the multipliers of the fixed points $\pm\sqrt{k/(1-k)}$ and $\infty$ are $2k - 1$ and $1/k$, respectively, each of which is at least $1$ in absolute value. The periodic points of minimal period $2$, $\pm\sqrt{-k/(1+k)}$, each have a multiplier of $(2k+1)^2$, which is also at least $1$ in absolute value. We're hesitant to continue this method of checking individual periodic points since the calculations become increasingly cumbersome for minimal period greater than $2$. Does anyone have suggestions for how to either: a) show that a periodic point exists with a multiplier in the desired range, or, more broadly, b) show that $\phi$ is(n't) post-critically finite?

$\endgroup$
  • $\begingroup$ If $k=1/2$, then $\pm 1$ are fixed points. I suspect a typo here? $\endgroup$ – David E Speyer Jul 26 '15 at 23:19
  • $\begingroup$ How is $2k-1$ greater than $1$ in absolute value? Is $k$ negative or complex, or do you mean a non-Archimedean absolute value $\alpha ^ {v(\cdot)}$? $\endgroup$ – Douglas Zare Jul 27 '15 at 2:52
  • $\begingroup$ @DouglasZare nonarchimedean. I state it in the first sentence, which is quite a wordy sentence so you may have missed it. $\endgroup$ – Justin Case Jul 27 '15 at 3:11
  • $\begingroup$ I also missed it. Of course, there are no non-archimedean norms for which $|1/2|<1$. $\endgroup$ – David E Speyer Jul 27 '15 at 3:26
3
$\begingroup$

I have found a counterexample. Take $K = \mathbb{Q}_2$. Let $f(t)$ be the polynomial $2 - t + 6 t^2 + 4 t^3 + 8 t^4$. By looking at the Newton polytope, $f$ has a root in $\mathbb{Q}_2$ of norm $1/2$ (so $2$-adic valuation $1$). Let $k$ be that root.

Then I compute that $f(f(f(f(1))))=f(1)$, where $f(x) = k(x+x^{-1})$. This was found by writing a little Mathematica script to determine the algebraic condition on $k$ which implies that $f^m(k) = f^n(k)$ and looking through the output until I saw one whose Newton polytope (for some $p$) was consistent with $|k|<1$.

I have also found examples with other residue characteristic.

I feel guilty about possibly spoiling the fun of finding this yourself, as this is an REU. But I think there is a lot of fun to be had by (1) learning about Newton polytopes and non-archimedean fields, if you haven't yet and (2) writing a similar script for yourself, and looking for patterns in the output. I suspect there is a lot hiding in that data.

$\endgroup$
  • $\begingroup$ I wasn't familiar with Newton polytopes, so this gives me some interesting topics to look into! $\endgroup$ – Justin Case Jul 27 '15 at 4:11
  • 1
    $\begingroup$ Hensel's lemma is enough in this case: Make the change of variables $t=2u$ and note that the root $u=1$ of $1 - u + 12 u^2 + 16 u^3 + 64 u^4$ lifts to $\mathbb{Q}_2$ by Hensel. But you should learn about Newton polytopes! They make spotting this sort of thing obvious, and they are a lot of fun. $\endgroup$ – David E Speyer Jul 27 '15 at 4:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.