2
$\begingroup$

This question is coming from a section of Rordam's "K-theory for C*-algebras", section 9.4.2.

Let $H$ be an infinite dimensional separable Hilbert space, and let $T$ be a Fredholm operator in $B(H)$. Let $T=S|T|$ be the polar decomposition of $T$, where $S$ is a partial isometry and $|T|=(T^*T)^{1/2}$.

For $t\in[0,1]$, define the path $R_t=S(t|T|+(1-t)I)$. If $\pi$ is the projection of $B(H)$ into the Calkin algebra (modulo compact operators), then $\pi(R_t)=\pi(S)(t|\pi(T)|+(1-t)I)$. The book claims then for each $t$, this element is invertible in the Calkin algebra and hence $R_t$ is Fredholm.

Since $T$ is Fredholm, then $\pi(T)$ is invertible and hence $\pi(S)$ is unitary, so really I just need to see that $(t|\pi(T)|+(1-t)I)$ is invertible. Now $|\pi(T)|$ and $I$ are both invertible in the Calkin algebra, and so I'm basically creating the straight line between two invertible elements, but I don't immediately see why everything on that path must also be invertible.

$\endgroup$
2
$\begingroup$

It follows from this fact:

If you have a positive invertible operator, T, in a unital $C^*$-algebra, then the path $(1-t)T+tI$ consists of invertible (positive) operators.
Just apply the functional calculus to the element $T$, whose spectrum is contained in $[\epsilon, \infty)$ for some $\epsilon>0$. The function $(1-t)x+t1$ is strictly positive in this segment (greater than epsilon), and therefore invertible.

$\endgroup$

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.