In noncommutative geometry when one want to go to the differentiable level, one is forced to work with algebras which are no longer $C^*$. It is nice if we don't loose much information by the replacement of $C^*$-algebras by more general algebra. One possible choice is the so called pre $C^*$-algebra which is by definition a subalgebra $\mathcal{A}$ of a $C^*$-algebra $A$ of which is complete with respect to some locally convex topology finer that the topology of $A$ and is closed under the holomorphic functional calculus. Suppose that we have such $\mathcal{A}$ which is moreover a Fréchet algebra and assume that $\mathcal{A}$ is dense in $A$ (in the norm topology of $A$). The claim is that:

The inclusion $i:\mathcal{A} \to A$ induces an isomorphism in $K_0$ groups.

I found the proof of this fact in the book "Elements of Noncommutative Geometry" by Várilly, Gracia-Bondía and Figueroa. There are two moments which are not clear for me:
1. The first is that authors state that it is enough to prove this fact for unital algebras. I believe that it is indeed the case however I'm not sure whether there are no technical problems.
2. The second thing is the end of the proof: let us recall that $K$-theory may be defined in terms of idempotents (not projections, i.e. our idempotents need not to be self-adjoint) in matrix algebras $Q_n(A)=\{e \in M_n(A): e^2=e\}$ (the same for $\mathcal{A}$: one checks that matrix algebras for $\mathcal{A}$ are again Fréchet pre $C^*$-algebras). Of this the quotient space is made by equivalence relation where $e \sim f$ iff there is some invertible matrix $g$ (possibly bigger that $e$ and $f$) such that $(e \oplus 0)g=g(f \oplus 0)$. At the end of the proof authors arrived to the conclusion that the inclusions $Q_n(\mathcal{A} \to Q_n(A)$ are all homotopy equivalences and they claim that this finishes the proof provided we are able to prove that two idempotents which are homotopic are equivalent in the above sense. I don't quite see two things:

Why it is enough to prove this to conclude that the inclusion induces iso on $K_0$ groups?

And also

How to prove that being homotopic implies being equivalent (in the context of general idempotents and invertibles)?


1) First question: unital algebras are sufficient.

If $i:K_0(\tilde {\mathcal A}) \rightarrow K_0(\tilde A)$ is an isomorphism, and since $id:K_0(C) \rightarrow K_0(C)$ is an isomorphism, you make a diagram chase by doubling the exact sequence $$0 \rightarrow K_0(A) \rightarrow K_0(\tilde A) \rightarrow^\rho K_0(C) \rightarrow 0$$ (second case with ${\mathcal A}$) and connect both lines with $i$ to see that $K_0({\mathcal A}) \rightarrow K_0(A)$ is an isomorphism as well. Note that $K_0(A)$ is the kernel of $\rho$ by definition!

3) "How to prove that being homotopic implies being equivalent (in the context of general idempotents and invertibles)?"

Note that ${\mathcal A}$ is a subalgebra of $A$, so you have all the notions from $C^*$-algebras like projections in ${\mathcal A}$ as well. I think the elementary standard proofs from $C^*$-algebras work in ${\mathcal A}$ as well. I think you must go by yourselves through the proofs.

See Wegge-Olsen Proposition 5.2.10 & 5.2.12 that homotopy-equivalence and unitary equivalence (or invertible equivalence (Proposition 4.2.6)) is the same in matrix algebras.

See Wegge-Olsen 5.B: Every idempotent is homotopic to a projection.

2) "Why it is enough to prove this to conclude that the inclusion induces iso on $K_0$ groups?"

By item 2) we may equivalently switch from invertible equivalence of idempotents to homotopy equivalence of idempotents in $Q$.

If $i$ is a homotopy equivalence on the level $i:Q({\mathcal A}) \rightarrow Q(A)$ then there exists a continuous inverse $k$ such that $i\circ k \sim id$ and $k \circ i \sim id$.

$i$ on the level $K_0$ is injective:

Suppose $$i(e) - i(f)=0$$ for $e,f \in Q({\mathcal A})$. Then $i(e) \oplus m \sim i(f) \oplus m$ (homotopy is $\sim$; Grothendieck group) for some $m \in Q(A)$. Note $m \sim i \circ k (m)$. Hence $$i(e) \oplus i(k(m)) \sim i(f) \oplus i(k(m)),$$ or $$i(e \oplus k(m)) \sim i(f \oplus k(m)).$$ Since $k$ is continuous, it respects homotopy (pathes)!. Thus $$e \oplus k(m) \sim k(i(e \oplus k(m))) \sim k(i(f \oplus k(m))) \sim f \oplus k(m).$$ Hence $$e - f = 0.$$

Surjectivity of $i$ is similar.


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.