Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Let $S$ be any positive semi-definite symmetric matrix (Hermitian psd matrices work as well). The Hadamard inequality is that $$\det S\le\prod_{i=1}^n s_{ii}.$$ My question is whether there are some other upper bounds of $\det S$ in terms of a partial sum $$\sum_{\sigma\in F}\epsilon(\sigma)\prod_{i=1}^n s_{i\sigma(i)},$$ where $F$ is some subset of ${\frak S}_n$. The Hadamard inequality corresponds to the case $F=({\rm id})$.

More precisely, I am interested in the validity of $$\det S\le\sum_{\epsilon(\sigma)=1}\prod_{i=1}^n s_{i\sigma(i)}.\qquad\qquad(1)$$

This latter inequality is true at least when $n\le4$, essentially because the difference $$\sum_{\epsilon(\sigma)=-1}\prod_{i=1}^n s_{i\sigma(i)}$$ contains enough many non-negative terms (the half if $n=4$, all of them if $n=2$ or $3$). This reason (which can be combined to the Cauchy-Schwarz inequality) fails if $n\ge5$, and therefore I suspect that the inequality (1) could be false if $n\ge5$.

Edit. Felix Goldberg's answer tells us that inequality (1) is true. More generally, Schur proved that if $G$ is a subgroup of ${\frak S}$, and $\chi$ is a character of $G$, then $$\chi(e)\det S\le\sum_{\sigma\in G}\chi(\sigma)\prod_{i=1}^n s_{i\sigma(i)}=:d_\chi(S).$$ This gives (1) when $G={\frak A}_n$ and $\chi={\bf 1}$.

This leads me raising a second question:

If $G$ is a subgroup of ${\frak S}_n$, let $K_G$ be the set of class functions $f$ over $G$ with the property that, for every positive definite symmetric matrix $S$, one has $$f(e)\det S\le\sum_{\sigma\in G}\chi(\sigma)\prod_{i=1}^n s_{i\sigma(i)}=:d_\chi(S).$$ Clearly, $K_G$ is a convex cone, which contains all the characters. Is it equal to the convex cone spanned by the ireeducible characters of $G$ ?

share|improve this question
2  
It may be trivial, but neverhteless: there exist many different inequalities of this kind (for the sets $F$ different from the one you need). Actually, if you split $[n]=\sqcup_i N_i$ and denote by $S_i$ the principal minor on the set of indices $N_i$, then $\det S\leq\prod_i \det S_i$. This is true because of the interpretation of $S$ as the Gram matrix of some vectors, and of $\det S_i$ as the squares of the respective volumes. –  Ilya Bogdanov Aug 28 '12 at 11:13
    
@Ilya. Shame on me, I should have think to it. This is block-wise Hadamard inequality, and its proof is essentially the same as that for the entrywise inequality. –  Denis Serre Aug 28 '12 at 11:34
add comment

1 Answer 1

up vote 8 down vote accepted

If I interpret correctly the terms and notations, this seems to be a result of Schur. See p.4 here.

share|improve this answer
    
Nice! Thanks a lot. –  Denis Serre Aug 28 '12 at 11:49
    
Nice survey. May be you can add it as an answer to MO question. Open problems everyone can understand. –  Alexander Chervov Oct 2 '12 at 16:13
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.