Abstract. In this paper we describe the Goodwillie tower of the stable homotopy of a space of maps from a finite-dimensional complex to a highly enough connected space. One way to view it is as a partial generalization of some wellknown results on stable splittings of mapping spaces in terms of configuration spaces.
IntroductionIt has been known for a while (see [1] for a survey article and a list of references) that given a parallelizable, compact m-dimensional manifold M with a nonempty boundary, and given a connected, pointed space Z, there is a configuration space model for the space of unbased maps Map(M, S m Z), which stably splits. More precisely, there is a weak equivalence:where C(M, ∂M ; n) stands for the space of n-tuples of distinct points in M , where all n-tuples whose intersection with ∂M is not empty have been identified to a point. There is an analogous splitting for the space of based maps. A closely related result is the stable splitting of spaces of the form Ω m Σ m X. This later splitting is sometimes refered to as the Snaith splitting (at least in the case m = ∞), and we refer to (0.1) as Snaith-type splitting.It is, therefore, natural to ask if for a based space K, that is not a manifold, but, say, a finite CW-complex, anything can be said about the functor X → QMap * (K, X) (where Map * (K, X) stands for the space of based maps from K to X). One does not expect this functor to split, in general, but it is still reasonable to try to approximate it by more elementary functors in a way that would give the splitting above in the case when K = M and X = S m Z. It turns out that this question (in fact a generalization of it) can be answered positively within the framework of the theory referred to as calculus of functors, which had been developed by T. Goodwillie in [4], [5], [6]. Since [6] has not been published yet, we present a brief outline of the theory of "Taylor towers" in the appendix. In what follows we will freely use notation from there.Consider again the identity (0.1). Observe that the factorsReceived by the editors July 21, 1994 and, in revised form, February 4, 1997. 1991 Mathematics Subject Classification. Primary 55P99. are homogeneous functors (of Z). Therefore, the functorsare excisive of degree n. Moreover, it is, in fact, true that the (weak) mapis (n + 1)k-connected if the space Z is k-connected. Therefore, by universality, the right hand side of (0.1) is nothing but the Taylor tower of the functor Z → QMap(M, S m Z), where Z ∈ T * . One can think of it as the "Maclaurin tower" of QMap(M, S m Z). Hence the generalization that we seek in this paper is the one of identifying the Taylor tower of the functor QMap * (K, Y ), where K is a finite CWcomplex and Y ∈ T X . More concretely, we want to answer the following question: given a space Y containing X as a retract, how can we describe QMap * (K, Y ) as a sequence of extensions of QMap * (K, X) by homogeneous functors? Let us discuss the case X = * (Maclaurin tower) first. We use implicitly the fact that the differentials of ...