Let me first define the majorization order (or dominance order) on partitions as $\lambda \succeq \mu$ iff $$\sum _{i=1}^{k}\lambda_i \geq \sum_{i=1}^{k}\mu_i$$ for all $1\le k\le l-1$ and $$\lambda_1+\cdots+\lambda_l=\mu_1+\cdots+\mu_l.$$

While playing around with some variation on Muirhead's inequality which says that $$m_{\lambda}(x_1,x_2,\dots,x_n)\geq m_{\mu}(x_1,x_2,\dots,x_n)$$ for all $x_i\geq 0$ when $\lambda \succeq \mu$ (here the $m_{\lambda}$'s are the monomial symmetric polynomials) I ended up investigating if such an inequality holds for Schur polynomials, which are another famous basis of the ring of symmetric functions. I tried a few special cases, and the following seems to hold, but I don't have a proof or a counterexample. So the question is:

Let $x_1,\dots,x_n \geq 0$ and $\lambda\succeq \mu$, is it always true that $$\frac{s _{\lambda}(x _1,x _2,\dots,x _n)}{s _{\lambda}(1,1,\dots,1)}\geq \frac{s _{\mu}(x _1,x _2,\dots,x _n)}{s _{\mu}(1,1,\dots,1)}?$$

If not, does majorization induce some similarly nice order on values of Schur polynomials?

  • $\begingroup$ Perhaps it is possible to introduce a second set of variables, and look at double Schur functions as well... $\endgroup$ – Per Alexandersson Jul 27 '15 at 0:39

This question is listed as a conjecture (conjecture 7.4 in the section "Open questions") in a recent paper of Cuttler, Greene and Skandera.

  • $\begingroup$ Thank you! I will have to wait till tomorrow to gain access to the paper. About a year ago I had proofs for the corresponding inequalities on elementary, complete or power symmetric functions, and I thought I had proved the inequality in the OP as well, but I had made a few silly mistakes. I guess now I'm less hopeful that it can be fixed (i.e. that it can be approached by elementary methods)... $\endgroup$ – Gjergji Zaimi May 17 '11 at 9:42
  • 1
    $\begingroup$ FYI, there are several online versions on author's pages, e.g. haverford.edu/math/cgreene/papers/smean10-15.pdf - in case you don't care if you get the journal version. $\endgroup$ – Vladimir Dotsenko May 17 '11 at 9:59
  • $\begingroup$ Oh no, that's perfect. It looks like an interesting read! $\endgroup$ – Gjergji Zaimi May 17 '11 at 10:13

This was recently proven by Suvrit Sra, on the arXiv. (Make sure to read version 3.)


This was also subsequently proved by Rachid Ait-Haddou and Marie-Laurence Mazure in 2018 in FOCM. There's also a variant in a preprint by Apoorva Khare and Terence Tao (see latest version) with "majorization" replaced by "weak majorization" and "all $x_i \geq 0$" replaced by "all $x_i \geq 1$".


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.