Why should one want to go further, beyond small stable categories? Let us give two reasons, which we will see resolved by Efimov’s theory later in this chapter.
K-theory of analytic spaces
Suppose we want to develop a theory of algebraic K-theory for analytic spaces, such as formal schemes, rigid analytic varieties, or complex analytic spaces, with good descent properties and a Grothendieck–Riemann–Roch theorem. Then the naive approach of taking \(K(\Perf (-))\) does not work. The basic reason is that in analytic geometry, restriction to an open subset does not give rise to a localization in the categorical sense: for instance, if \(U \subseteq X\) is a rational open subset of an affinoid adic space, the restriction functor on perfect complexes is in general not a Verdier quotient. Hence the Thomason–Trobaugh argument sketched above breaks down at its very first step.
The K-theory of the compact operators
Let \(\Hh \) be a separable infinite-dimensional complex Hilbert space, let \(B = B(\Hh )\) be the C\(^*\)-algebra of bounded operators, and let \(\Kk = \Kk (\Hh ) \subseteq B\) be the closed two-sided ideal of compact operators. Since \(\Kk \) is a non-unital ring, its algebraic K-theory is defined using excision: \[ K(\Kk ) := \fib \big (K(B) \to K(B/\Kk )\big ). \] A theorem of Suslin–Wodzicki [Suslin and Wodzicki (1992)] shows that this is independent of the choice of \(B\): we have \(K(\Kk ) \simeq \fib (K(R) \to K(R/\Kk ))\) for every unital ring \(R\) containing \(\Kk \) as a two-sided ideal. They also proved Karoubi’s conjecture, which in particular implies that \[ K(\Kk ) \simeq \KU \] is the periodic complex topological K-theory spectrum. This is a beautiful statement, but it raises a question: K-theory is supposed to be an invariant of categories, so where is the category whose K-theory is \(K(\Kk )\)? We will answer this question in Example 1.14 below.