# Positivity of coefficients of a polynomial derived from Schubert polynomials

Let $W=\bigcup_{n=1}^\infty S_n$ be the union of all symmetric groups $S_n$. For an element $w\in W$, denote by $\mathfrak{S}_w$ the Schubert polynomial associated to $w$, and by $\partial_w$ the divided difference operator associated to $w$.

Question: Let $u,v\in W$. Why is $$\partial_u(\mathfrak{S}_u\mathfrak{S}_v)-\mathfrak{S}_v$$ always a positive polynomial, in the sense that all coefficients are non-negative integers?

Preliminary thoughts:

• It is clear that $\partial_u(\mathfrak{S}_u\mathfrak{S}_v)$ is positive.

• One may use the twisted Leipniz rule and induction on the length of $u$ to reduce the problem to an apparently simpler one. Indeed, let $s_i$ be the rightmost factor in a reduced expression of $u$. Then it suffices to show that $\partial_{us_i}(s_i\mathfrak{S}_u\partial_i\mathfrak{S}_v)$ is positive. (Here, $W$ acts on $\mathbb{Z}[x_1,x_2,x_3,\ldots]$ by permuting the variables.)

• I was thinking to realize the polynomials in question as characters of representations, in the spirit of the work of M. Watanabe on KP modules, but had not much success.

Any help would be greatly appreciated. I computed many examples and cannot find a counter example. In fact, I have a stronger "positivity conjecture" which implies this one, and even to this stronger conjecture, I was not able to find neither a counterexample nor a proof.

• Is this expression perhaps positive in Demazure key polynomials, or Demazure atoms? Jul 11, 2017 at 11:44
• Demazure polynomials / atoms might be relevant for the question - thanks - I googled them just now. It raises the question if Demazure polynomials / atoms are positive themselves. Jul 11, 2017 at 13:09
• Both Demazure polynomials and Demazure atoms are positive - there are several characterizations of these (as sum over lattice points in polytopes, as sum over (subset of) semi-standard tableaux, sum over skyline fillings...) Jul 11, 2017 at 19:08
• Why don't you post your stronger conjecture too? Jul 12, 2017 at 12:08
• I was reluctant to post the stronger conjecture because it is less easy to state. Basically, if you expand the product $\partial_u(\mathfrak{S}_u\mathfrak{S}_v)$ via the twisted Leipniz rule and the lexicographically smallest reduced expression of $u$, you get many summands. One of them is $\mathfrak{S}_v$. I expect each of the summands to be a positive polynomial, but I am not completely sure about this, because I have only checked it heuristically for $u=s_\alpha$ - the case I am mostly interested in. Jul 12, 2017 at 15:55