Problem:
Suppose we have two real, symmetric and positive definite square matrices $A$ and $B$, i.e., $$A_{ij}, B_{ij}\in \mathbb{R}$$ $$A^T=A$$ $$B^T=B$$ $$x^TAx>0 \forall x$$ $$x^TBx>0 \forall x$$
If $A \ge B$, i.e., $A-B$ is semi-postive definite ($x^T(A-B)x \ge0,\,\forall x $), then is the statement that $A^{-1} \le B^{-1}$ true, i.e. $A^{-1}-B^{-1}\ge0$ ($x^T(B^{-1}-A^{-1})x\ge0,\,\forall x $)
Remarks:
Obviously if $A$ and $B$ can be diagonalized simultaneously with the same similarity transformation, then the statement is true. What about the general case? I tried some numerical examples, it seems the statement is true. But I don't know how prove it.
I would appreciate if anyone can give a proof or point out any reference that has the solution of the above problem.
Thanks in advance!