3
$\begingroup$

For $A \in \mathbb{R}^{n\times n}$, I am wondering if the following inequality holds, $$\text{tr}(A)^2 - \text{tr}(A^2) \leq ||A||_*^2 - ||A||_F^2$$ This is equivalent to the following inequality on the elementary symmetric polynomial, $$\sum_{i<j}\lambda_i\lambda_j \leq \sum_{i<j} \sigma_i \sigma_j$$ where $\lambda_i$ and $\sigma_i$ denote the eigenvalues and singular values of $A$.

This question arises originally from a discussion here. I am still looking for a proof or counterexample to this inequality.

$\endgroup$
7
  • $\begingroup$ Are you assuming $A$ to have positive, or at least real eigenvalues? $\endgroup$ Commented Feb 20, 2019 at 2:54
  • $\begingroup$ @darijgrinberg No. $A$ is any real valued square matrix. $\endgroup$ Commented Feb 20, 2019 at 2:55
  • $\begingroup$ Oh, I forgot that the symmetric polynomials are real anyway :) $\endgroup$ Commented Feb 20, 2019 at 2:55
  • 1
    $\begingroup$ I think your inequality is just the inequality from math.stackexchange.com/questions/381808/… , but applied to the compound matrix $\wedge^2 A$ (of size $\dbinom{n}{2} \times \dbinom{n}{2}$) instead of $A$. Can you check? $\endgroup$ Commented Feb 20, 2019 at 2:58
  • $\begingroup$ What is $\|A\|_*$ ? $\endgroup$ Commented Feb 20, 2019 at 16:15

1 Answer 1

3
$\begingroup$

As suggested by Darij in the comments to the OP, the compound matrix $\wedge^2 A$ has eigenvalues $\lambda_i \lambda_j$ and singular values $\sigma_i \sigma_j$. This result can be found, for example in Theorem 2.16 of

Zhan, Xingzhi, Matrix theory, Graduate Studies in Mathematics 147. Providence, RI: American Mathematical Society (AMS) (ISBN 978-0-8218-9491-0/hbk). x, 253 p. (2013). ZBL1275.15001.

see also. Applying the inequality from this post proves the desired inequality.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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