0
$\begingroup$

This might be really dumb question but I've just started dealing with such matrices. I would like to know why $$\begin{bmatrix} 0 & 1 \\ 1 & x \end{bmatrix}$$ cannot be a positive semidefinite matrix.

Also, given the condition $$ \begin{bmatrix} 0 & x & y\\ x & z & p\\ y & p & 1-x \end{bmatrix} \succeq 0$$

how do we get following matrix $$\begin{bmatrix} 0 & 0 & 0\\ 0 & z & p\\ 0 & p & 1\\ \end{bmatrix}$$

Any help would be much appreciated.

$\endgroup$
0
$\begingroup$

Have you heard of Sylvester's Condition? So, in a nutshell, what it says is that you start looking at (top left) 1x1 2x2 3x3 ... nxn blocks of matrices all have positive determinants.

As a corollary, a matrix with negative determinant can never be positive definite. In your example, no matter what x is, the determinant is always -1 which by Sylvester's Condition (Note : It is IFF) is not positive definite.

EDIT :

Now that you clarified in the comments what you mean by "get". It makes more sense.

Think about it, if the entire top row wasn't zero: Apply the Sylvester's Criterion.

The first 1x1 block is just 0 which is OK (but the matrix can't be positive definite, at most semidefinite) The next block is 2x2 which has determinant $-x^2$. As long as you choose a real value of $x$, the criterion won't hold and you don't even need to go ahead if $x$ wasn't zero. However since we want a positive definite matrix, let's set $x=0$ and continue. If you carry out the same determinant criterion on the entire matrix you'll realize that $y$ needs to be zero as well (Assuming $z\geq0$. Else the determinant will be negative. (The determinant is $-zy^2$)

$\endgroup$
  • $\begingroup$ That means that if I implement the condition on the matrix to be positive semidefinite then it would yield the matrix that follows with zeros in top row. $\endgroup$ – mat_123 May 8 '13 at 4:48
  • $\begingroup$ I think by your argument of Sylvester's Condition all the top row entries should be zeros. $\endgroup$ – mat_123 May 8 '13 at 4:51
  • $\begingroup$ @mat_123, edited. $\endgroup$ – Inquest May 8 '13 at 4:55

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.