4
$\begingroup$

Let $p,q,r$ be integers with $r\ge2$; let $f$ be a polynomial of the form $f(X) = g((X+1)^r)$, which is not a $q$-th power. Let $\omega$ be a $p$-th root of unity.

Prove or disprove that the polynomial $f(X)f(\omega X)\cdots f(\omega^{p-1}X)$ is also not a $q$-th power.

NOTE: The special case $p=q=r$ is Lemma 5 in Samuel White, "The group generated by x->x+1 and x->x^p is free", J. Algebra 118 (1988) 408--422; he says that "the idea is attributed to George Bergman". I wondered how general the statement is.

$\endgroup$
4
  • 2
    $\begingroup$ Where does this question come from? $\endgroup$
    – Igor Rivin
    Sep 1, 2014 at 14:51
  • 1
    $\begingroup$ The special case p=q=r is Lemma 5 in [White, "The group generated by x->x+1 and x->x^p is free", J. Algebra 118 (1988) 408--422]; he says that "the idea is attributed to George Bergman". I wondered how general the statement is. $\endgroup$
    – grok
    Sep 1, 2014 at 15:11
  • 5
    $\begingroup$ Cool! Often, questions starting with "Show that..." are homework, so to avoid misunderstandings, might be better to give some back story (as in your comment). $\endgroup$
    – Igor Rivin
    Sep 1, 2014 at 15:49
  • 4
    $\begingroup$ @grok, it is best to put information asuch as that you gice in the comment in the question body itself :-) $\endgroup$ Sep 1, 2014 at 16:17

1 Answer 1

3
$\begingroup$

A family of counterexamples is defined as follows:

$r=2$,

$p_1\ge 2, \quad p_2\ge 3, \quad d_1 \perp p_1, \quad d_2 \perp p_2$,

$a=\dfrac{\tan(\pi d_1/p_1)-\tan(\pi d_2/p_2)}{\tan(\pi d_1/p_1)+\tan(\pi d_2/p_2)} \qquad b=\dfrac{2\tan(\pi d_1/p_1)\tan(\pi d_2/p_2)}{\tan(\pi d_1/p_1)+\tan(\pi d_2/p_2)}$,

(with $a=1, \;b=2\tan(\pi d_2/p_2)\;$ if $\;p_1=2$)

$g(X)=(X-(a+i b)^2)(X-(a-ib)^2)=X^2-2(a^2-b^2)X+(a^2+b^2)^2$,

$p=\text{lcm}(p_1, p_2), \quad q=2$,

$\prod_{j=0}^{p} f(\omega^j X)=(X^{p}-(a-1+ib)^{p})^2(X^{p}-(-a-1+ib)^{p})^2$.

Here is a picture of the four roots of $f$ in the complex plane:

enter image description here

When $p_1\!=\!p_2$, $\;g$ is a square and the formulas simplify to the original answer:

$r\!=\!2, \; p\!\ge\!3, \; b\!=\!\tan(\frac{\pi}{p}), \; g(X)\!=\!X\!+b^2, \; q\!=\!2,\; \prod\limits_j f(\omega^j X)\!=\!\big(X^p\!-(-1\!+ib)^p\big)^2$.


When $\;a=\pm b\;$ the example simplifies to: $\; r=4, \; g=X+4a^4, \quad$ but the only such cases I know of are for $\;(d_1, p_1, d_2, p_2) \!=\! (3, 8, 1, 8)\;$ and $\;\prod f(\omega^j X)=(X^8-(\frac{1+i}{\sqrt{2}}-1)^8)^2(X^8-(\frac{-1+i}{\sqrt{2}}-1)^8)^2$,$\;\;$ or $(d_1, p_1, d_2, p_2) \!=\! (1, 6, 1, 12)\;$ and $\;\prod f(\omega^j X)=(X^{12}-(\frac{-1+i}{1+\sqrt{3}}-1)^{12})^2(X^{12}-(\frac{1+i}{1+\sqrt{3}}-1)^{12})^2\;$.


Ignoring trivial derivations (i.e. replacing $g$ with products of the "primitive" ones above), I'd be very surprised if any other examples esisted - but a proof of that looks to me extremely hard.

$\endgroup$
2
  • $\begingroup$ There is more case for $r=4$, corresponding to $(d_1, p_1, d_2, p_2)=(7, 12, 1, 6)$, for which $a=b=(\sqrt{3}+1)/2$. $\endgroup$ Jan 8, 2015 at 10:11
  • $\begingroup$ I apologize for bumping up this answer one last time. I wanted to make the attached image clearer. $\endgroup$ Jan 8, 2015 at 19:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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