In my matrix theory studies I recently came across the following problem:

We have the following real symmetric $ n \times n $ matrix $ A=\begin{pmatrix} d_1 & u^T E^\dagger v & u^T \\ v^T E^\dagger u & d_2 & v^T \\ u & v & E \end{pmatrix} $ where $ d_1,d_2 $ are real numbers and $E$ is a square $n-2\times n-2$ matrix, $u,v$ are column vectors of length $n-2$, T denotes transpose as usual and $ \dagger $ denotes generalized inverse (or classic inverse if it is easier to solve) and we know that any principal submatrix of $A$ that does not include entries $ a_{1,2},a_{2,1} $ is positive semidefinite, and we are asked to show $A$ is positive semidefinite.

I thought a direct approach might be difficult, so I though about using the characterization theorem of positive semidefinite matrices to show an easier equivalent condition. I would certainly appreciate all help.


Well, according to what you say if you choose
$$n = 4 $$ $$E = I$$ $$u = v = [1 \quad 1]^T$$ $$d_1 = d_2 = 1$$ You can show that matrix that the principle submatrices by deleting the second row and column are positive definite, however $A$ is not positive definite. The eigenvalues, in that case, are given as $\lambda = [-1,\ -0.2361,\ 1, \ 4.2361]$.

I hope i understood your question correctly.

  • $\begingroup$ I mean principal submatrices that don't include index 2 $\endgroup$ – kroner Oct 17 '16 at 23:09
  • 1
    $\begingroup$ Can you elaborate on what you mean ? $\endgroup$ – Ahmad Bazzi Oct 17 '16 at 23:10
  • $\begingroup$ I mean any principal submatrix of A that does not include entries $ a_{1,2}, a_{2,1}$ $\endgroup$ – kroner Oct 17 '16 at 23:11
  • 1
    $\begingroup$ Aha that is different now.. $\endgroup$ – Ahmad Bazzi Oct 17 '16 at 23:12
  • $\begingroup$ sorry about that $\endgroup$ – kroner Oct 17 '16 at 23:12

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.