7
$\begingroup$

We say that $T(X) \in \mathbb{Q}[X]$ is a trinomial if there exist $A,B,C \in \mathbb{Q}$ such that $T(X) = AX^n + BX^m + C$ for some $n \geq m \in \mathbb{N}$.

Is it true that for each irreducible $g(X) \in \mathbb{Q}[X]$ there exists some nonzero $h(X) \in \mathbb{Q}[X]$ such that $g(X)h(X)$ is a trinomial?

If no, then what are some necessary/sufficient conditions on $g$ for the existence of such $h$?

Improve this question
$\endgroup$
10
  • 5
    $\begingroup$ All non-zero multiples of $(X-1)^k$ require at least $k+1$ non-zero coefficients. A necessary condition on $g$ is that it does not have a root $\alpha \neq 0$ of multiplicity $> 2$. $\endgroup$ Commented Aug 12, 2015 at 12:17
  • 11
    $\begingroup$ How many real zeros can a trinomial have? $\endgroup$ Commented Aug 12, 2015 at 12:19
  • 2
    $\begingroup$ @VesselinDimitrov thanks! I thought about irreducible polynomials. I will edit accordingly. $\endgroup$ Commented Aug 12, 2015 at 12:27
  • 1
    $\begingroup$ What's the answer over finite fields? $\endgroup$ Commented Aug 12, 2015 at 13:00
  • 2
    $\begingroup$ There are a few other splitting constraints as well; e.g., a trinomial has only a bounded number of $\mathbb{Q}_p$-roots, for every fixed $p$. (Also Khovanskii's theory of fewnomials could be something worth looking up for generalizations of this kind of question.) But it would be quite interesting to show that "almost all" degree-$d$ polynomials in $\mathbb{Z}[X]$ (or in $\mathbb{R}[X]$) require at least $d+1$ non-zero coefficients for all their multiples. $\endgroup$ Commented Aug 12, 2015 at 13:10

1 Answer 1

Reset to default
8
$\begingroup$

To summarize some of the discussion in the comments: a trinomial can have at most four distinct real roots. A random polynomial of degree $d$ (random = all coefficients are iid, there are other models) has $\Omega(\log d)$ real roots. Which means that the probability that a random polynomial divides a trinomial goes to zero (exponentially fast with degree, though this is a little harder to prove).

Improve this answer
$\endgroup$
3
  • 2
    $\begingroup$ Should the notation be $\Omega(\log{d})$, or $\asymp \log{d}$? With $O(\cdot)$ one could infer that only an upper bound is stated, while actually here it is the lower bound that matters. $\endgroup$ Commented Aug 12, 2015 at 16:56
  • 1
    $\begingroup$ @VesselinDimitrov Correct, it is actually asymptotic to constant times log. $\endgroup$ Commented Aug 12, 2015 at 19:00
  • $\begingroup$ @VesselinDimitrov OK, fixed... $\endgroup$ Commented Aug 12, 2015 at 19:10

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.