Splitting monoidal stable model categories (Q1004489): Difference between revisions

The main result is a splitting theorem for a stable monoidal category: Let \(\mathcal{C}\) be a stable monoidal category with unit \(S\) for the monoidal product. I.e., \(\mathcal{C}\) is a pointed model category, such that the natural suspension and loop adjunction \((\Sigma,\Omega)\) is an equivalence on the homotopy category Ho \(\mathcal{C}\). And \(\mathcal{C}\) has a closed monoidal product \((\wedge,Hom)\) with unit \(S\) and such that the pushout product axiom holds. \([X,Y]^{\mathcal{C}}\), the set of morphisms in Ho \(\mathcal{C}\), is an abelian group. Let \([S,S]^{\mathcal{C}}\) be the set of maps in Ho \(\mathcal{C}\). Lemma 2.1 states that \([S,S]^{\mathcal{C}}\) is a commutative ring, and \([X,Y]^{\mathcal{C}}\) is then an \([S,S]^{\mathcal{C}}\) module. Let \(a\in[S,S]^{\mathcal{C}}\) be represented by \(a':\hat{c}\hat{f}S\to \hat{c}\hat{f}S\) where \(\hat{c}\) is cofibrant replacement and \(\hat f\) is fibrant replacement. The homotopy colimit of \(\hat{c}\hat{f}S\to^{a'} \hat{c}\hat{f}S \to^{a'}\hat{c}\hat{f}S \cdots\) is denoted \(aS\) and is well defined up to weak equivalence. The similar construction over the map \(a'\wedge Id_X:\hat{c}\hat{f}S\wedge X\to \hat{c}\hat{f}S\wedge X\) defines the object \(aX\) up to weak equivalence. When \(a\) is an idempotent, \(a=e\), the \(lim^1\) term in \(0\to \lim^1[X,Y]^{\mathcal{C}}\to [aX,Y]^{\mathcal{C}}\to\lim [X,Y]^{\mathcal{C}}\to 0\) vanishes, and \(e[X,Y]^{\mathcal{C}}=\lim[X,Y]\) is then isomorphic to \([eX,Y]\). The idempotent \((Id_S-e)\) is denoted \(1-e\) and there is a canonical natural isomorphism \([X,Y]^{\mathcal{C}}\cong e[X,Y]^{\mathcal{C}}\oplus(1-e)[X,Y]^{\mathcal{C}}\) -- well known for modules over a commutative ring. Hence we have \([X,Y]^{\mathcal{C}}\cong[eX,Y]^{\mathcal{C}}\oplus[(1-e)X,Y]^{\mathcal{C}}\) for any \(X,Y\in\mathcal{C}\). This gives a natural isomorphism \(X\to eX\prod (1-e)X\) in Ho\(\mathcal{C}\), and Ho\(\mathcal{C}\) can be written as a product \(e\)Ho\(\mathcal{C}\times(1-e)\)Ho\(\mathcal{C}\), where the objects of \(e\)Ho\(\mathcal{C}\) are the objects of Ho\(\mathcal{C}\) and the morphisms are \(e[X,Y]^{\mathcal{C}}\). The Bousfield localization of \(\mathcal{C}\) at a cofibrant object \(E\) is defined in section 3. A map \(f:X\to Y\) is an \(E\)-equivalence, if \(id_E\wedge f:E\wedge X\to E\wedge Y\) is an equivalence. Consider the category \(\mathcal{C}\) with weak equivalences the \(E\)-equivalences and cofibrations unchanged. If this defines a model structure, denote the model category \(L_E\mathcal{C}\). This is the Bousfield localization of \(\mathcal{C}\) at \(E\). The splitting theorem, Thm.~4.4 states, that provided the Bousfield localizations \(L_{eS}\mathcal{C}\) and \(L_{(1-e)S}\mathcal{C}\) exist, then there is a Quillen equivalence \(\Delta:\mathcal{C}\to L_{eS}\mathcal{C}\times L_{(1-e)S}\mathcal{C}:\prod\) which induces the above splitting of the homotopy category. Moreover, as a corollary, if \(\mathcal{C}\) splits as a product of localizations \(L_E\mathcal{C}\times L_F\mathcal{C}\) at cofibrant objects. Then there are orthogonal idempotents \(e_E\) and \(e_F\) in \([S,S]^{\mathcal{C}}\) such that \(e_E+e_F=1\), \(L_E\mathcal{C}=L_{e_ES}\mathcal{C}\) and \(L_F\mathcal{C}=L_{e_F}\mathcal{C}\). The motivating example is from the author's thesis and is a splitting result of the category of rational \(G\)-equivariant spectra. The result is that splittings of this category correspond to idempotents of the rational Burnside ring.