In Higson and Roe's Analytic K-homology, for a unital $C*$-algebra $A$, the definitions of K-theory and K-homology have quite a similar flavor.

Roughly, the group $K_0(A)$ is given by the Grothendieck group of homotopy classes of matrix algebra projections over $A$.

On the other hand, $K^0(A)$ is the Grothendieck group of homotopy classes of even Fredholm modules over $A$ (or more correctly unitary classes of Fredholm modules).

Now in contrast to the $K_0(A)$ case, every element of $K^0(A)$ contains a representative Fredholm module. (For the inverse of the class of $(H,\rho,F)$, just take the class of $(H^{\text{op}},\rho,-F)$, where op denotes the opposite grading.

Does this means that taking the Grothendieck is not strictly necessarily to define $K^0(A)$? Could one just as well define it as the monoid of homotopy classes of Fredholm modules, and then prove that it was a group?

If this is true, then is there any deeper philosophical reason why $K_0$ requires us to introduce inverses, while $K^0$ does not?

  • 2
    $\begingroup$ One can define K^0 using Fredholm operators with dualizable (co)kernels, this will produce a counterpart to the cited definition of K_0, and the resulting model also has inverses defined in a similar way. The very point of Fredholm stuff is to allow inverses in the model by considering formal differences of kernels and cokernels. $\endgroup$ Mar 13 '19 at 2:28
  • 1
    $\begingroup$ The model of Karoubi of K theory involving gradings (as in the definition of K homology cycles) makes non necessary the introduction of formal inverses. See his introductory book on the subject. $\endgroup$ Mar 13 '19 at 12:03

KK-theory provides one way to eliminate the asymmetry in the definitions of K-theory and K-homology. A cycle in $KK(A, B)$ is a triple $(H, \rho, F)$ where $H$ is a (adjectives) Hilbert $B$-module, $\rho$ is a (adjectives) representation of $A$ on $H$, and $F$ is a (adjectives) bounded operator on $H$ such that $[F, \rho(a)]$, $(F^2 - 1)\rho(a)$, and $(F - F^*)\rho(a)$ are $B$-compact for all $a \in A$. The relations are analogous to the relations in K-homology, and in particular KK-cycles form an abelian group without the need to invert anything.

If $B = \mathbb{C}$ then $H$ is an ordinary Hilbert space and the KK-cycles are exactly the same thing as Fredholm modules over $A$. It is a fact that $KK(\mathbb{C}, B)$ is isomorphic to the K-theory of $B$ for all (adjectives) C*-algebras $B$, so this gives a model of K-theory of the sort you're looking for.


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.