Let $K\ne \mathbb{Q}$ be a number field, let $\alpha\in \mathcal{O}_K$ and let $f(X)\in \mathcal{O}_K[X]$. Denote the Mahler measure by $M$.

Is there any known result about the comparison of the values $M(\alpha)$ and $M(f(\alpha))$?

  • 5
    $\begingroup$ I don't know, but it would be good to get some trivial cases out of the way. E.g., if $f$ is a constant polynomial, then clearly $M(f(\alpha))$ has nothing to do with $M(\alpha)$. $\endgroup$ – Gerry Myerson Apr 15 '19 at 5:27
  • 2
    $\begingroup$ In which sense do you want to compare the values? Do you mean inequalities or something else? $\endgroup$ – François Brunault Apr 20 '19 at 8:51
  • 1
    $\begingroup$ @FrançoisBrunault Yes, I mean inequalities. For example, do constants $C, \, n$ (depending on $K$ and $f$, but not on $\alpha$) exist such that $M(f(\alpha))\le C M(\alpha)^n$? A bound in the other direction is probably way more problematic. $\endgroup$ – Maurizio Moreschi Apr 20 '19 at 12:43

This is all standard stuff about height functions. More generally, if we use absolute heights, then for any $f(x)\in\overline{\mathbb{Q}}[x]$ of degree $n$ there are constants $C_1(f)>0$ and $C_2(f)>0$ so that for any $\alpha\in\overline{\mathbb{Q}}$, $$ C_1(f)H(\alpha)^n \le H\bigl(f(\alpha)\bigr) \le C_2(f)H(\alpha)^n.$$ For $C_2(f)$ it's easy to get an explicit formula in terms of the coefficients of $f$ using the triangle inequality, as in Moreschi's answer. For $C_1(f)$, one can get an explicit formula that depends on the coefficients and the resultant of $f$. See for example Lang's Fundamentals of Diophantine Geometry.

| cite | improve this answer | |

Let $H(\alpha)$ denote the absolute multiplicative height of $\alpha$, which is a number in $[1,\infty)$, and $M(\alpha) = H(\alpha)^d$, where $d$ is the degree of $\alpha$. Basic properties of the height include:

  • $H(\alpha \beta) \leq H(\alpha)H(\beta)$, for all algebraic numbers $\alpha$ and $\beta$,
  • $H(\alpha_1 + \cdots + \alpha_n) \leq n \prod_{i=1}^n H(\alpha_i)$.

So if $f(x) = c_n + c_{n-1}x + \cdots + c_0x^n,$ we have $$H(f(\alpha)) \leq (n+1)\prod_{i=0}^n H(c_i) H(\alpha)^i \leq (n+1) \left[\max_{0 \leq i \leq n} H(c_i)\right]H(\alpha)^{n(n+1)/2}.$$ Let $d$ be the degree of $\alpha$, $m = n(n+1)/2,$ and let $c = \max_{0 \leq i \leq n} H(c_i)$. Raising both sides to the $d$-th power gives $$H(f(\alpha))^d \leq ((n+1)c)^d M(\alpha)^m.$$ Now $f(\alpha) \in \mathbb{Q}(\alpha)$, so the degree of $f(\alpha)$ is less than or equal to that of $\alpha$, and therefore $M((f(\alpha))$ is at most equal to the left-hand-side of the above inequality. [I had made a comment similar to this, but it contained an error, and so I deleted it.]

Note that $d$ is at most $[K:\mathbb{Q}]$, and so this upper bound is of the form $CM(\alpha)^m$, where $m$ depends only on $f$ and $C$ depends on $K$ and $f$.

As for the reverse direction, I don't think any such inequality can exist, because for an arbitrary $\alpha$ (i.e. of arbitrarily large Mahler measure), we may find a polynomial such that $f(\alpha) = 1$, so $f(\alpha)$ has Mahler measure as small as possible.

| cite | improve this answer | |
  • $\begingroup$ The bound you prove can actually be improved to $H(f(\alpha))\le C H(\alpha)^n$ (which is what I was aiming for) under the assumption of my question. I post the proof of this as an answer, because it is too long for a comment. $\endgroup$ – Maurizio Moreschi May 6 '19 at 13:20
  • $\begingroup$ There is a lower bound $H(f(\alpha))\ge CH(\alpha)^n$, but note that the implied constant will depend on $f$, as it will for the upper bound. $\endgroup$ – Joe Silverman May 6 '19 at 19:09

This the proof of the claim in my comment to Bobby Grizzard's answer.

Since, by definition $$ H(\alpha):=\prod_{v\in \mathfrak{S}_K} \max\{1,|\alpha|_v\}\quad (\alpha\in K), $$ one has $$ H(\alpha)=\prod_{v\in \mathfrak{S}_{K,\infty}} \max\{1,|\alpha|_v\}\quad \forall\alpha\in \mathcal{O}_{K},$$ where $\mathfrak{S}_K$, $\mathfrak{S}_{K,\infty}$ denote the set of all places of $K$ and the set of Archimedean places of $K$ respectively.

Now, let $\alpha\in \mathcal{O}_{K}$ and $f(X)=a_n X^n+\dots+a_0\in \mathcal{O}_{K}[X]$. Then $f(\alpha)\in \mathcal{O}_K$.

Moreover, for any $v\in \mathfrak{S}_{K,\infty}$ one has $$ |f(\alpha)|_{v}\le f_v(|\alpha|_v),\qquad f_v(X):=|a_n|_{v} X^n+\dots+|a_0|_{v}\in \mathbb{R}_{\ge 0}[X].$$

It follows that $$ |f(\alpha)|_{v}\le (|a_n|_{v}+\dots +|a_0|_{v}) \big(\max\{1,|\alpha|_v\}\big)^n $$ and thus $$ \begin{split} H(f(\alpha))&=\prod_{v\in \mathfrak{S}_{K,\infty}} \max\{1,|f(\alpha)|_{v}\} \\ &\le \prod_{v\in \mathfrak{S}_{K,\infty}} \max\big\{1,(|a_n|_{v}+\dots +|a_0|_{v}) \big(\max\{1,|\alpha|_v\}\big)^n\big\} \\ & \le \prod_{v\in \mathfrak{S}_{K,\infty}} \max\{1,|a_n|_{v}+\dots +|a_0|_{v}\} \big(\max\{1,|\alpha|_v\}\big)^n \\ &= C_f H(\alpha)^n, \end{split} $$ where $$ C_f:=\prod_{v\in \mathfrak{S}_{K,\infty}} \max\{1,|a_n|_{v}+\dots +|a_0|_{v}\}.$$

| cite | improve this answer | |

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.