0
$\begingroup$

From the description of positive semidefinite matrix, I found that determinant of positive semidefinite is multiplication of its eigenvalues. Can anyone show me how to derive this property?

$\endgroup$
  • $\begingroup$ Hint: Eigenvalue decomposition $\endgroup$ – user251257 Apr 4 '16 at 23:39
3
$\begingroup$

This holds for all matrices. Suppose that $\mathbf M \in \mathbb{F}^{n,n}$ is a matrix. Its eigenvalues $(\lambda_k)_{k=0}^{n-1}$ are roots of the characteristic polynomial $$ \det (\lambda \mathbf I - \mathbf M) = c \prod_{k=0}^{n-1} (\lambda_k - \lambda) $$ Expanding along the diagonal shows that the leading coefficient is $c = 1$. Therefore you can set $\lambda = 0$ to obtain the desired result.

$\endgroup$
  • $\begingroup$ Thank you for information. Can you explain me again the rationale behind setting c = 1 & λ=0? $\endgroup$ – pippp Apr 11 '16 at 11:15
  • $\begingroup$ @pippp Every polynomial can be factored, and the leading coefficient can vary. In this case the leading coefficient $c$ is $1$. We can set $\lambda = 0$ since $\lambda$ is an arbitrary variable. $\endgroup$ – Henricus V. Apr 11 '16 at 13:19
1
$\begingroup$

The determinant of any matrix $M$ is the constant term of the characteristic polynomial $\det(M - \lambda I)$, and thus the product of the roots of that polynomial, which are the eigenvalues (counted by algebraic multiplicity).

$\endgroup$
  • $\begingroup$ As long as the characteristic polynomial factorizes... $\endgroup$ – user251257 Apr 5 '16 at 10:05
  • $\begingroup$ Over a splitting field, it does. The OP mentioned "positive semidefinite", so in this case the field can be $\mathbb R$. $\endgroup$ – Robert Israel Apr 5 '16 at 15:18
  • $\begingroup$ yes. I think it is a fact worth to mention. $\endgroup$ – user251257 Apr 5 '16 at 16:39

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.