Well... That's not true I don't think. Example.
Is positive definite.
But that multiplied by any other positive definite matrix (call it B) will result in B, which is positive definite.
... is not positive definite?
I read this in my notes, but I think it might be wrong. I can't think of a counter example. I know that not being able to think of a counter example does not make it true, but I would think that the product of two positive definite matrices would still be positive definite.
Can anyone think of an example of where the product of two positive definite matrices would not be positive definite?
The examples we can give depend on our definition of positive definiteness. Taking the more general definition which allows non-Hermitian matrices (or non-symmetric matrices, in the real case), we say the n x n complex matrix is positive definite if, for any n x 1 complex vector , , where is the real part of and is the conjugate transpose of . We say the n x n real matrix is positive definite if, for any n x 1 real vector , . [1]
Consider the 2 x 2 real matrix
This matrix rotates a vector counter-clockwise by . As long as , will be positive definite. To see this, note that , where is the angle between the vectors and .
Now, suppose . Then is a product of two positive definite matrices. However rotates a vector by . As a result, and will be orthogonal, so . Thus B is not positive definite.
We can even find a negative definite matrix as a product of positive definite matrices. Try .
[1] Weisstein, Eric W. "Positive Definite Matrix." From MathWorld--A Wolfram Web Resource. Positive Definite Matrix -- from Wolfram MathWorld. Accessed 30 July 2010.