14

Does there exist a bivariate polynomial $p \in \Bbb R[x,y]$ that is positive iff $x, y > 0$?

My motivation was originally to state multiple positivity conditions with one expression but now I'm just curious and unable to find the answer via search.

Rodrigo de Azevedo
  • 23,223
Christian Bueno
  • 1,320

1 Answers1

23

No.

Consider such a polynomial $P$.

$P$ must be $0$ on the positive real $x$ and $y$ axes, as it must cross from being positive to negative there. As a result, both $x$ and $y$ divide such a polynomial. Write $P=xyQ$.

Now, $Q$ is positive everywhere except the third quadrant, and so $R(x,y)=-Q(-x,-y)$ is positive only on the first quadrant. As a result, we have a polynomial $R$ that satisfies the same property, but has lower degree than $P$. This process cannot continue infinitely, giving a contradiction.

Carl Schildkraut
  • 37,483
  • I think you have to be a bit more careful. $Q$ is certainly positive in the first quadrant, but in the second, third and fourth quadrants, it might be zero. Importantly, I don't see how we can assert that $Q$ is nonzero everywhere in the third quadrant? – Vincent Aug 11 '20 at 09:35
  • @Vincent You're right that there are some issues with $0$, but the proof works verbatim to show that "there is no nonzero polynomial that is nonnegative on the first quadrant and nonpositive on the other three quadrants." – Carl Schildkraut Aug 11 '20 at 21:02
  • 1
    You're right, thanks :) – Vincent Aug 12 '20 at 02:35