5
$\begingroup$

The paper Asymptotic properties of polynomials and algebraic functions of several variables by Gorin contains the following.

Lemma 3.1. Let $f\in \mathbb R[x_1,\dots,x_n]$. Suppose $f$ has a root in the interior of the unit circle. Then there exists a positive constant $c$ such that $$\|f(x)\|\geq c \cdot d(x,f^{-1}(0))^{\deg f},$$ where $d$ denotes the standard metric on $\mathbb R^n$.

The proof involves a previous theorem which is itself rather involved.

Question. Is there a quick direct proof of the above lemma?

$\endgroup$
  • 1
    $\begingroup$ The statement is weird. The stated bound on $\|f(x)\|$ is invariant under shifts, hence the condition on a root in the unit circle should be redundant – you can always shift the polynomial so that e.g. the origin is a root. If $f$ has any root at all, that is. $\endgroup$ – Emil Jeřábek Jun 11 '19 at 8:22
  • 1
    $\begingroup$ @EmilJeřábek I agree the condition is redundant, but it's part of the statement in the reference, so I figured I should write it down. $\endgroup$ – Arrow Jun 11 '19 at 8:25
  • 4
    $\begingroup$ Sure. I just mentioned it because it raises the possibility that there is a typo in the original paper. $\endgroup$ – Emil Jeřábek Jun 11 '19 at 8:32
  • 1
    $\begingroup$ @EmilJeřábek I guess the evaluation point is supposed to be in the disk (or some other compact) too. The statement is clearly false in the entire space. $\endgroup$ – fedja Jun 11 '19 at 19:57
2
$\begingroup$

The Russian original text clearly states that the exponent is not $\deg f$ but just some $\alpha>0$. For $\alpha=\deg f$ is it not true: a polynomial $f(x,y)=x^{2n}+(x-y^n)^2$ has a unique zero $(0,0)$, but $f(\delta^n,\delta)=\delta^{2n^2}$.

I remember obtaining a cubic upper bound for $\alpha$ in the case of a polynomial in two variables with a single zero (but the estimate holds only in a disk centred at the root, not on the whole plane). I may try to find it if you wish.

| cite | improve this answer | |
$\endgroup$

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.