5
$\begingroup$

Let $P_{d, n}$ be the space of polynomial maps $\mathbb{R}^n\to \mathbb{R}$ of degree at most $d$.

Is the subset $S\subset P_{d, n}$ of nowhere negative polynomials semialgebraic?

$\endgroup$
3
  • $\begingroup$ I am not an expert in this, but it seems to me that this paper is likely to lead to a positive answer: dmle.icmat.es/revistas/detalle.php?numero=4502 . $\endgroup$ Commented May 2, 2021 at 11:12
  • $\begingroup$ In case the link gets outdated, here is the full info on the article I cite: Cukierman, Fernando. "Positive polynomials and hyperdeterminants." Collectanea Mathematica 58.3 (2007): 279-289. $\endgroup$ Commented May 2, 2021 at 11:13
  • $\begingroup$ Yes, this is very well known, it follows immediately for example from the Tarski–Seidenberg theorem. $\endgroup$
    – Arno Fehm
    Commented May 2, 2021 at 11:25

2 Answers 2

10
$\begingroup$

As I said in the comments this is very well known: $S$ is the complement of the projection of the semialgebraic set $\{(f,a)\in P_{d,n}\times\mathbb{R}^n:f(a)<0\}$, hence semialgebraic by the Tarski-Seidenberg theorem.

$\endgroup$
5
$\begingroup$

The answer is yes.

Recall first that the sums-of-squares polynomials of degree at most $D$ and in $n$ variables form a spectrahedron, and in particular a semialgebraic set.

Now if $f \in P_{d,n}$ is a positive¹ polynomial, then by Hilbert's 17th problem and Artin's solution there are sums-of-squares polynomials $g$ and $h$ such that $f = \frac{g}{h}$. It is possible to bound the relevant degrees of $g$ and $h$ in terms of $n$ and $d$. (I'm not sure what the best currently known degree bounds are, but the linked 2014 paper gives a tower of five exponentials!)

Thus the $f \in P_{d,n}$ are characterized in terms of the equations $hf = g$ and $h \neq 0$ for sums-of-squares polynomials $g$ and $h$ with a given number of unknown coefficients. Hence $P_{d,n}$ is the projection of a semialgebraic set, and by Tarski's theorem therefore semialgebraic itself.

¹ "Positive polynomial" is the standard term for nowhere negative polynomial.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .