If $A$ is a C*-algebra, denote its minimal unitization by $\tilde A$ and its suspension by $SA$, thought of as all continuous $a:[0,1] \to A$ with $a(0)=a(1)=0$. The unitized suspension $\widetilde{SA}$ can then be identified with the algebra of continuous $a:[0,1] \to B$ such that $a(0)=a(1) \in \mathbb{C} \subset B$ where $B$ is either $A$ or $\tilde A$ according as $A$ is already unital or not.

In C*-algebra K-theory, we have the isomorphism $K_1(A) \to K_0(SA)$. This is not the isomorphism generally equated with Bott periodicty, but it is still a significant isomorphism (note: I am thinking of the definition of $K_1$ in terms of unitaries). In the commutative context, this isomorphism comes from a clutching construction for vector bundles. We could formulate it like this:

For $X = (X,x_0)$ a pointed compact Hausdorff space, let $\mathrm{Clutch}_n(X)$ denote the set of homotopy classes of maps $X \to U(n)$ and let $\mathrm{Vect}_n(X)$ denote the set of isomorphism classes of complex vector bundles over $X$ with $n$-dimensional fibre over $x_0$. Then, there is a bijection $$ \mathrm{Clutch}_n(X) \to \mathrm{Vect}_n(SX). $$ Here, $SX$ is the pointed suspension got by forming the cylinder $X \times I$ and collapsing $(\{x_0\} \times I) \cup (X \times \{0,1\})$ to a point. $I$ is the closed unit interval.

Now, we reformulate for the C*-algebraic setting:

For $A$ be a C*-algebra, let $\mathrm{Clutch}_n(A)$ denote the set of homotopy classes of unitaries in $M_n(\tilde A)$. Let $\mathrm{Vect}_n(A)$ denote the set of equivalence classes of projections in $M_\infty(\tilde A)$ whose scalar part is a rank-$n$ projection. Here $M_\infty(\tilde A)$ means the $*$-algebra of infinite matrices with finitely many nonzero entries in $\tilde A$ and equivalence of projections in $M_\infty(\tilde A)$ is given, for example, by homotopies which stay stay inside one of the finite matrix algebras. There is a map $$ \mathrm{Clutch}_n(A) \to \mathrm{Vect}_n(SA) $$ given as follows. Given $u \in U_n(\tilde A)$, choose $w : [0,1] \to U_{2n}(\tilde A)$ such that $w(0) = 1_{2n}$ and $w(1) = \left( \begin{smallmatrix} u & 0 \\ 0 & u^* \end{smallmatrix} \right)$. We send the class of $u$ to the class represented by the projection $p_u = w 1_n w^* \in M_{2n}(\widetilde{SA})$.

The above construction is just the usual one appearing in most introductory books on K-theory, see e.g. Rørdam, Larsen & Laustsen or Wegge-Olsen's books. Standard arguments will show you that the map described above does not depend on choice of $u$ or $w$, and as well that the map is surjective. For example, this is done Theorem 7.3.5 of Wegge-0lsen's book in all gory detail. But, at least in the references I've looked at, it is only shown that, if $p_u \sim p_v$ for $u,v \in \mathrm{Clutch}_n(A)$, then $u \oplus 1_k \sim v \oplus 1_k$ for some $k$, i.e. $u$ and $v$ are stably-homotopic. This is enough to get the K-theory isomorphism, but...

Question: I would like to see an an example where the map described above is not injective.

The fact that you get a bijection in the classical commutative setting is rather nice, and I'm very curious what goes wrong in general.


Your Answer

By clicking "Post Your Answer", you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.