I checked some examples and I always received that skew-symmetric matrix of even dimension has only pure imaginary eigenvalues.

For example:
$\begin{bmatrix} 0 & 2 & 3 & 1 \\ -2 & 0 & 1 & 4 \\ -3 & -1 & 0 & 1 \\ -1 & -4 & -1 & 0 \end{bmatrix}$

Eigenvalues: $( 0.000, 5.406i) ( 0.000,-5.406i) ( 0.000, 1.665i) ( 0.000,-1.665i)$

How can be explained such property?
Additionally why skew-symmetric of even dimension has non-zero determinant in opposition to odd dimensional skew-symmetric matrices ?

(I'm not considering here zero matrices) Interesting is also fact that probably every matrix (of even dimension) can be decomposed into symetrical part which has only real eigenvalues and skew-symmetrical which has only pure imaginary values what makes interesting analogy to complex numbers and their two parts, but I don't know whether there are importatnt consequences of this fact.


It's because $-A^2 = A^T A$ has only real nonnegative eigenvalues: if $v$ is an eigenvector of $A$ with eigenvalue $\lambda$ then $$-\lambda^2 \|v\|^2 = -v^T A^2 v = v^T A^T A v = \|A v \|^2.$$

Skew-symmetric matrices do not have to have nonzero determinant, the zero matrix is a counterexample.

  • $\begingroup$ Let's exclude zero matrices from our problem. It is so special that it could be at the same time formally symmetric and skew-symmetric what gives really no information about its nature. $\endgroup$ – Widawensen Jul 26 '16 at 7:52
  • $\begingroup$ @Widawensen OK but there are plenty of other examples of skew-symmetric matrices with determinant $0$, including in even dimension $\endgroup$ – user356288 Jul 26 '16 at 7:54
  • $\begingroup$ @ user356288, hmm, do you point me some example ? Additionally, such even dimensional skew-symmetric matrices with zero determinant have some special property which other don't have? $\endgroup$ – Widawensen Jul 26 '16 at 8:10
  • $\begingroup$ The equation written by you, dear user356288, provides so much information about the problem, probably it gives also a hint why odd dimensional skew-symmetric matrices must have determinant equal 0. Only for one eigenvalue = 0 in this case we have fulfilled this equation. Maybe we should have two eigenvalues for even dimensions equal 0 in order to have non-invertible matrix ? Is it possible? $\endgroup$ – Widawensen Jul 26 '16 at 8:16
  • $\begingroup$ Ok. I've found counterexample. It has two 0 eigenvalues. $\begin{bmatrix} 0 & 1 & 0 & 0 \\ -1 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 \end{bmatrix}$ It's seems that such non-invertible matrix must have some even dimensional sub-matrix on diagonal as the zero matrix. $\endgroup$ – Widawensen Jul 26 '16 at 8:30

Say A is a skew symmetry matrix
Say $\lambda$ is a real eigenvalue of A, then there exists a real vector $v≠0$ such that $Av=\lambda v$

By skew symmetry we have:
$v^TAv=(v^TAv)^T$ (this holds because the $v^TAv$ is a real and reals are equal to their transpose)
Thus $v^TAv=-v^TAv=0$

Thus $0 = v^TAv=v^T\lambda v = \lambda v^Tv = \lambda |v|^2$
(this last equality holds because v is real)
Thus $0=\lambda |v|^2$
Thus $\lambda=0$ or $v=0$, however $v≠0$ thus $\lambda=0$

Thus the only possible real eigenvalue of A is $0$

Stated differently, all non zero eigenvalues of a skew symmetric matrix have non zero imaginary component


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.