0
$\begingroup$

Let $\alpha$ be a positive definite automorphism of an inner product space $V$. Is $\alpha^{-1}$ necessarily positive definite?

I know the answer is true for invertible positive definite matrices, but I tried to prove this problem for the general case. However, still, I feel that I'm missing something.

Since $\alpha$ is positive definite, it's self-adjoint and for all nonzero $v\in V$, we have $\langle\alpha(v),v\rangle> 0.$

Let $v\in V$, so $$\langle \alpha^{-1}(v),v\rangle=\langle \alpha^{-1}(v),\alpha\alpha^{-1}(v)\rangle=\langle \alpha\alpha^{-1}(v),\alpha^{-1}(v) \rangle > 0$$

Now for each $v\in V$, we got $\langle \alpha^{-1}(v),v\rangle>0$, but still I need to show $\alpha^{-1}$ is self-adjoint. For finitely generated inner space, I can show that it's true.

Let $v, \:w\in V$. Then,

$$\langle v,w\rangle=\langle \alpha^{-1}\alpha(v),w\rangle=\langle\alpha(v), (\alpha^{-1})^*(v)\rangle=\langle v,\alpha^*(\alpha^{-1})^*(v)\rangle, $$ so $(\alpha^{-1})^*=(\alpha^*)^{-1}. $ Moreover, $$\langle\alpha^{-1}(v),w\rangle=\langle v,(\alpha^{-1})^*(w)\rangle=\langle v,(\alpha^*)^{-1}(w)\rangle=\langle v,\alpha^{-1}(w)\rangle.$$ Hence $\alpha^{-1}$ is self-adjoint.

But how about when $V$ is not finitely generated?

$\endgroup$
1
$\begingroup$

Your argument is correct (though you should mention that you are assuming $v\neq 0$, which implies $\alpha^{-1}(v)\neq 0$ in order to get your inequality). You still need to show that $\alpha^{-1}$ is self-adjoint, though. You can do so by an argument very similar to the one you used.

$\endgroup$
  • $\begingroup$ I'm not sure if I can prove it, for $V=\mathbb{C}\:\:or\:\:\mathbb{R}$, my argument is true, but how about when $V$ is infinitely generated inner product space and not equal $\mathbb{C}\:\:or\:\:\mathbb{R}$. And plus in general we don't have $(\alpha^{-1})^* =(\alpha^*)^{-1}$ $\endgroup$ – Parisina Dec 30 '16 at 6:16
  • 1
    $\begingroup$ Just prove directly that $\langle \alpha^{-1}(v),w\rangle=\langle v,\alpha^{-1}(w)\rangle$. The proof is nearly identical to your proof that $\langle \alpha^{-1}(v),v\rangle>0$. $\endgroup$ – Eric Wofsey Dec 30 '16 at 22:22
  • 1
    $\begingroup$ $\langle \alpha^{-1}(v),w\rangle=\langle \alpha^{-1}(v),\alpha\alpha^{-1}(w)\rangle=\langle\alpha\alpha^{-1}(v),\alpha^{-1}(w)\rangle=\langle v,\alpha^{-1}(w)\rangle.$ $\endgroup$ – Parisina Dec 31 '16 at 0:56
  • $\begingroup$ You know why I got confused,because of this example: Let $V=\mathbb{R}^{(\infty)}$ with inner product $\langle [a_0,a_1,\dots],[b_0,b_1,\dots\rangle=\sum_{i=0}^{\infty}a_ib_i$, Let $\alpha\in End(V)$ be given by $\alpha: [a_0,a_1,dots]\to [0,a_1,a_2,\dots]$. Then $\alpha^*$ exists and is given by $\alpha^*:[a_0,a_1,\dots]\to [a_1,a_2,\dots].$ Here, $\alpha$ is monomorphism but not $\alpha^*.$ $\endgroup$ – Parisina Dec 31 '16 at 1:04
  • $\begingroup$ But, in this example $\alpha$ isn't an automorphism. Thanks! $\endgroup$ – Parisina Dec 31 '16 at 1:07

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.