A classification of Taylor towers of functors of spaces and spectra (Q488968)

The authors of this paper continue, and in a sense complete, their investigation of the Taylor towers for functors of spaces and spectra begun in [\textit{G. Arone} and \textit{M. Ching}, Operads and chain rules for the calculus of functors. Paris: Société Mathématique de France (SMF) (2011; Zbl 1239.55004)]. Namely, given a weak equivalence-preserving functor \(F:\text C\to\text D\) where \(\text C\) and \(\text D\) are ``nice'' categories (but usually taken to be based spaces or spectra), one has the standard Taylor tower supplied by homotopy calculus of functors. The \(n\)th stage of this tower, \(P_nF\), is in general a difficult object. However, for functors of based spaces or spectra the \(n\)th homogeneous layer, \(D_nF\), defined as the homotopy fiber of the map \(P_nF\to P_{n-1}F\), was described by Goodwillie. For example, for functors between based spaces, we have \[ D_nF(X)\simeq \Omega^{\infty}\left(\partial_nF\wedge (\Sigma^\infty X)^{\wedge n} \right)_{h\Sigma_n}. \] Here \(\partial_nF\), the \textit{\(n\)th derivative of \(F\)}, is a spectrum with the action of \(\Sigma_n\), and \(D_nF\) is therefore completely characterized by this spectrum. For functors of spectra, the statement is similar. A longstanding issue has been the one of recovering the Taylor stages \(P_nF\) from the derivatives \(D_nF\). The main result of the paper says that this can be done via a cobar construction: For functors between categories \(\text C\) and \(\text D\), where each of those is either the category of based spaces or spectra, \[ P_nF\simeq \text{cobar}(\Phi, \partial_*\Phi, \partial_{\leq n}F). \] Here \(\Phi\) is the adjoint to the functor \(\partial_*\) that assigns to a pair of objects in \((\text C, \text D)\) the category of modules over the operad \(\partial_*I\), where \(I\) is the identity functor (the authors showed in the earlier paper mentioned above that \(\partial_*F\) is a module over \(\partial_*I\)), or the category of symmetric sequences (this depends on what \(\text C\) and \(\text D\) are). It then follows that the composition \(\partial_*\Phi\) is a comonad on one of the target categories of \(\Phi\) and that \(\partial_*F\) is a coalgebra over \(\partial_*\Phi\); thus \(\partial_*\) can be regarded as a functor from the category od functors between \(\text C\) and \(\text D\) to the category of \(\partial_*\Phi\)-coalgebras. Lastly, \(\partial_{\leq n}F\) denotes the \(n\)th truncation of the sequence of derivatives of \(F\). Further, if the Taylor tower converges, then it immediately follows that \(F\simeq \text{cobar}(\Phi, \partial_*\Phi, \partial_{*}F)\). The authors also observe that every \(\partial_*\Phi\)-coalgebra arises as the derivatives of some functor and deduce that there is an equivalence of (pointed, finitary) \(N\)-excisive functors and \(N\)-truncated \(\partial_*\Phi\)-coalgebras. As the authors note, the statements above would be more useful if an explicit description of the category of coalgebras over \(\partial_*\Phi\) were available. They initiate the study of this category in the case of functors with values in spectra, where they show how the \(\partial_*\Phi\)-coalgebra structure is encoded by certain transformations between functors \(K_r\) which assign to a \(\Sigma_n\)-spectrum \(A_n\) the \(\Sigma_r\)-spectrum \(\partial_r(X\mapsto (A_n\wedge X^{\wedge n})^{h\Sigma_n})\). For functors with values in based spaces, an explicit characterization of a coalgebra over \(\partial_*\Phi\) appears to be much more difficult. In the course of proving these important theorems, the authors devise tools of independent interest. For example, they develop the homotopy theory for coalgebras over a comonad. They also set up a variant of homotopic descent theory in the course of dealing with the technical issue of having to replace the comonad \(\partial_*\Phi\) with a ``better'' version where every object is cofibrant.