-2
$\begingroup$

sum of positive definite matrix $A+B $is positive definite. I want to look at the spectrum of $C=A+B$

can we say the ith largest eigenvalue of $C$ is no less than the ith largest eigenvalue of $A$ i.e. $B$ as positive definite matrix, has contribution to the growth of spectrum?

remark:

sorry my question may be too stupid, actually it is originally from

sum of matrix and its spectrum

and

Sum of Gaussian matched by Brownian Motion?

very appreciated for the help!

if "sum of matrix and its spectrum" is solved, "Sum of Gaussian matched by Brownian Motion?" would be solved.

$\endgroup$
2
  • 4
    $\begingroup$ Doesn't this follow from Weyl's inequalities? $\endgroup$ Commented Oct 15, 2018 at 13:43
  • $\begingroup$ yes, but pls take a look at updated questions. Thanks! $\endgroup$
    – jason
    Commented Oct 15, 2018 at 21:08

1 Answer 1

1
$\begingroup$

Yes, this follows from the Maximin characterization of eigenvalues of symmetric matrices. If $A$ is an $n\times n$ symmetric matrix, form the Rayleigh ratio $R(x)=(x,Ax)/(x,x),$ where $(.,.)$ is the standard dot product. Then the $k$-th eigenvalue is $$\lambda_k=\max_a\min_{ax=0} R(x)$$ where the inner minimum is under the condition that we impose $k-1$ linear restrictions ($a$ is a $(k-1)\times n$ matrix) on $x$, and the outer maximum is over all possible restrictions.

If you add to $A$ a positive-definite matrix, the Rayleigh ratio evidently increases, from which the result follows.

Ref. R. Gantmacher and M. Krein, Oscillation matrices... AMS 2002, or any textbook on classical mechanics, for example V. Arnold, Mathematical methods of classical mechanics.

$\endgroup$
3
  • $\begingroup$ Cool, I did not know this. $\endgroup$
    – Nik Weaver
    Commented Oct 15, 2018 at 14:17
  • $\begingroup$ great! your answer is helpful! $\endgroup$
    – jason
    Commented Oct 15, 2018 at 18:06
  • $\begingroup$ sorry, my question may be too naive, but pls take a look at the updated questions. it is originally from this one "Sum of Gaussian matched by Brownian Motion?", however, if "sum of matrix and its spectrum" is solved, "Sum of Gaussian matched by Brownian Motion?" would be solved. $\endgroup$
    – jason
    Commented Oct 15, 2018 at 21:11

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .