1
$\begingroup$

Consider a real number $x > 0$, a real symmetric matrix ${\bf A} = {\bf A}^\mathrm{T}$, and the imaginary unit $i$. I want to show that the real part of $$(x{\bf I} + i{\bf A})^{-1}$$ is positive definite.

It's probably a one-liner with the right approach but I just can't find it. The matrix inversion lemma (Woodbury) didn't help me. I'm confident that the statement is true for a physical reason and after checking it numerically for a million random ${\bf A}$ and tiny $x$-values.

A more general form of the statement (also backed up numerically) is: the real part of $$({\bf D} + i{\bf A})^{-1}$$ is positive definite where ${\bf D} = \mathrm{diag}_n(x_n)$ with all $x_n > 0$.

$\endgroup$
1
$\begingroup$

HINT: For the first party: $B\colon =(x I + i A)^{-1}$ is normal, and all its eigenvalues have real part $>0$. The real part of $B$ it will have eigenvalues the real part of the eigenvalues of $B$, so it will be positive definite.

But in general, if $M$ is invertible and the real part of $M$ is positive, then the real part of $M^{-1}$ is positive. To see this, use $$\langle M v, v \rangle + \langle v, M v\rangle > 0$$ and take $w \colon = M^{-1} v$. This takes care of the second part too.

$\endgroup$

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.