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

Let $A$ be HPD. Denote by $D$ the diagonal matrix obtained by observing the diagonal elements of A, i.e. $D = \operatorname{diag}(a_{11},a_{22},\ldots,a_{nn})$.

I would like to show that if the Jacobi method for linear systems converges for the system $Ax = b$, then $2D-A$ must also be HPD.

The inverse is also true; Let $A = D - E - E^{*}$ be the standard decomposition of A to diagonal matrix, upper triangular matrix and lower triangular matrix. The Jacobi matrix can be expressed as $J = D^{-1}(E+E^{*})$. If $(\lambda,v)$ is an eigenpair of $J$, then there holds $v^{*}(E+E^{*})v = \lambda v^{*}Dv$, and we can extract $\lambda$ since $D$ must be HPD, and find a condition so that $|\lambda| < 1$. Then, we need to show the condition holds using the fact that $v^{*}Av* > 0$ and $v^{*}(2D-A)v > 0$, which is rather simple to do.

However, I cannot seem to find a concrete method to prove the original statement. I have came to a conclusion the Jacobi method converges for both $Ax=b$ if and only if it converges for $(2D-A)x = b$, but I do not know how to conclude that $v^{*}(2D-A)v > 0$ for non-zero vectors $v$.

share|cite|improve this question
Welcome to Math.SE! – Belgi Jun 1 '12 at 17:22

After some careful thought I was able to prove the claim.

Assume the Jacobi method converges for the linear system $Ax = b$. Then, $|\lambda(J)| < 1$. As before, if $(v, \lambda)$ is an eigenpair of $J$, I can write $v^{*}Dv > 2|Re(v^{*}Ev)|$. From there, one can easily show that all eigenvalues of $2D - A$ are positive. $2D - A$ is hermitian so it has only real eigenvalues, thus it suffices to show none of them can be zero or negative, or it will contradict $v^{*}Dv > 2|Re(v^{*}Ev)|$. Since $2D - A$ is hermitian with only positive eigenvalues, it must be HPD.

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.