Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am trying to find properties or constraints on a $(p \times n) $matrix $U$ such that upon left multiplying a real symmetric square positive semi-definite matrix $V$ with $U$ the resulting matrix $W$ of dimensions $(p \times n) $ is still real positive semi-definite matrix. In other words i am trying to establish that there is $U$ such that $$W = U V$$ where $W$ and $V$ are real positive semi-definite matrices.

If $W=V$ then obviously $U=I$ fits the equation.

I have a suspicion that this should be possible as by definition of real symmetric positive semi-definite matrix, $V$ should have non-negative eigenvalues, which implies that it should have a right inverse which I suspect would also be positive semi-definite matrix, so if $W$ (some positive semi-definite matrix) is multiplied with the right inverse we should be able to compute $U$.

Any help would be much appreciated.

edit I suspect if my reasoning about computing $U$ above is correct then it should also be positive semi-definite matrix. So perhaps a way to construct $U$ may be by applying Gram-Schmidt process and finding new eigenvectors for W.

share|cite|improve this question
There is something I am missing here. If $p\neq n$, the product $UV$ is not square, so it certainly cannot be symmetric positive definite. – Michael Grant Apr 13 '13 at 22:47
@MichaelGrant I believe a non-square matrix can be symmetric semi-positive definite. – Comic Book Guy Apr 13 '13 at 22:49
@MichaelGrant I am looking to try find an appropriate example for my claim. I think covariance matrices fit this category. – Comic Book Guy Apr 13 '13 at 22:52
No, a non-square matrix cannot be symmetric positive definite. Either property separately implies the matrix is square. – Michael Grant Apr 13 '13 at 22:56
@MichaelGrant I could be wrong but i would like to clarify here. Firstly I am interested in semi-Positive definite matrices not strictly positive definite. Secondly it is well known condition that covariance matrices must be semi-positive definite matrices. As they are allowed to be non square i can only conclude from this that it is possible to have non-square semi-positive definite matrices. If i am in correct in my understanding then please elaborate why this is not possible. – Comic Book Guy Apr 13 '13 at 23:01
up vote 1 down vote accepted

I am unsure how we should properly resolve this question.

As we have discussed at length in the comments, only square matrices can be positive semidefinite. Therefore, if $p\neq n$, there is no way that the product matrix $W=UV$ can be positive semidefinite, because it will also be non-square.

For grins, let's assume that $p=n$. Under what conditions is $W=UV$ positive semidefinite? This requires that $x^HUVx\geq 0$ for all complex vectors $x$. Alternatively, this is true if and only if $UV+VU^H=Q$ where $Q=W+W^H$ has nonnegative eigenvalues (and is therefore PSD itself). If $U$, $V$ are real, then you can relax the Hermitian transposes to real tranposes, and consider only reall vectors $x$.

Now for one special case, I know the answer. If $V$ is positive definite---i.e., not just PSD but nonsingular---and we require $W$ to be positive definite as well, then the answer follows from Lyapunov's theorem applied to linear systems:

  • The eigenvalues of $U$ must have positive real part.

What about the more relaxed cases? That is, what if $V$ is only positive semidefinite? What if $W$ is only required to be positive semidefinite? I am afraid I do not know. I'm sure people who study Lyapunov's theorem for linear systems in some depth know...

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.