Every real matrix $A$ can be decomposed into symmetric and skew-symmetric part. Symmetric matrix has only real numbers as its eigenvalues (including $0$) and skew-symmetric matrix has only imaginary values (also including $0$).

  • Could we infer from separate calculations of eigenvalues for symmetric and skew-symmetric about eigenvalues for matrix $A$ ?
  • If so then can the same be said about eigenvectors calculated separately for these two parts of matrix and their relevance for eigenvectors of the whole matrix $A$?

Edit (after 1 day)

If the answer for the questions above is too difficult to obtain in a general case maybe it would be possible to answer for a particular case:

  • why in the case of orthogonal matrices $R$ we can write down: $EGV(sym(R))+EGV(sk(R))= EGV(R)$ where $EGV$ is here a vector obtained from respectively ordered values of eigenvalues for symmetric part of $R$, skew-sym. part of $R$ and full matrix $R$.

for 3-D rotation matrix we have ( it's hard to believe that it is just a coincidence)

$EGV(sym(R)) =\begin{bmatrix} \cos(\theta)\\ \cos(\theta)\\ 1 \end{bmatrix} , EGV(sk(R)) \begin{bmatrix} i\sin(\theta)\\ -i\sin(\theta)\\ 0 \end{bmatrix}, EGV(R)=\begin{bmatrix} \cos(\theta)+i\sin(\theta)\\ \cos(\theta)-i\sin(\theta)\\ 1 \end{bmatrix}$

Of course it would be better if it were proved starting from general properties of orthogonal matrix without calculating exact values of eigenvalues as I showed in the example above.

  • $\begingroup$ I am afraid that there are no relations between the eigenvalues for symm and skewsymm and eigenvalues for $A$ $\endgroup$ Nov 8 '16 at 9:16
  • $\begingroup$ More generally, knowing the eigenvalues of matrices $A$ and $B$ does not provide information on the eigenvalues of $aA+bB$. $\endgroup$
    – Jean Marie
    Nov 8 '16 at 9:20
  • 2
    $\begingroup$ @AlessandroBlasetti Indeed, not much can be said, but some loose relations do exist, e.g. the trace of $A$ must be the sum of traces of the two decomposed parts. $\endgroup$
    – user1551
    Nov 8 '16 at 9:22
  • 1
    $\begingroup$ You are right, it may happen that we have particular properties for this decomposition. Sometimes something can be said about singular values instead of eigenvalues (in this respect, for example, I am almost certain that the symmetrical part of a matrix $A$ is the closest symmetrical matrix to $A$ in the sense of Frobenius norm). $\endgroup$
    – Jean Marie
    Nov 8 '16 at 9:28
  • 1
    $\begingroup$ have a look at (math.stackexchange.com/q/1941701) $\endgroup$
    – Jean Marie
    Nov 8 '16 at 9:31

I try to give a partial answer. As @JeanMarie said in the comments there is no relationship between the eigenvalues of two matrices, $A$ and $B$, and some linear combination $aA+bB$.

Since $0$ is an eigenvalue of both the symmetric part of $A$ and the anty-symmetric part, if $\ker(A+A^T)\cap\ker(A-A^T)\neq\emptyset$, we can easily prove that that also $A$ is not invertible.

  • $\begingroup$ Ok. Maybe it is a very limited conclusion but gives however some information about one eigenvalue for A. Conclusions for more number of eigenvalues can be sometimes obtained from sym, and sk-sym, part of a matrix A when we have some additional information about the matrix as it is - for example - in the case of 2-d and 3-d orthogonal matrices where sym. and sk-sym. parts give clues for eigenvalues. $\endgroup$
    – Widawensen
    Nov 8 '16 at 16:35

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.