An introduction to motivic Hall algebras (Q655338): Difference between revisions

The paper under review is an introduction to Joyce's construction of the motivic Hall algebra of coherent sheaves on a variety \(M\). One of the main results of the paper is the definition of a semi-classical integration map from a Poisson subalgebra of the Hall algebra to the ring of functions on a symplectic torus. An application of these results is the proof in the author's paper [J.\ Am.\ Math.\ Soc.\ 24, No.\ 4, 969--998 (2011; Zbl 1234.14039)] that, roughly speaking, reduced Donaldson--Thomas invariants and stable pair invariants coincide when \(M\) is a Calabi--Yau threefold. In Section 2 basic definitions concerning Grothendieck rings of varieties, schemes and algebraic spaces, denoted by \(K(\text{Var}/\mathbb{C})\) etc., are reviewed. Roughly, the Grothendieck ring is the free abelian group on isomorphism classes of the respective category subject to relations, namely that the class of a disjoint union of two objects is the sum of their classes and, furthermore, that the classes of two objects are identified if there exists a geometric bijection between them. The latter condition means that there exists a morphism inducing a bijection between the sets of \(\mathbb{C}\)-valued points. The product is induced by the fibre product. In fact, it can be checked that all three rings coincide. The next section is devoted to Grothendieck rings of algebraic stacks of finite type with affine stabilisers, where the latter 2-category is denoted by \(\text{St}/\mathbb{C}\). The definition is slightly more involved than in the variety case, but it can be nevertheless proved that localising \(K(\text{Var}/\mathbb{C})\) with respect to the general linear groups of various dimensions gives \(K(\text{St}/\mathbb{C})\). The section ends with a discussion of relative Grothendieck groups, that is, we consider stacks over a fixed stack \(S\) and not just over the ground field \(\mathbb{C}\). Section 4 introduces the motivic Hall algebra \(H(\mathcal{A})\). The setup is the following. We consider a smooth projective variety \(M\) and the abelian category \(\mathcal{A}\) of coherent sheaves on \(M\). One can consider \(\mathcal{M}^{(n)}\), the moduli stack of \(n\)-flags of coherent sheaves, that is, the objects over a scheme \(S\) are chain of monomorphisms of coherent sheaves on \(S\times M\). There are also various maps between these stacks, for example, an element in \(\mathcal{M}^{(2)}\) can be identified with a short exact sequence and there are maps \(b: \mathcal{M}^{(2)}\rightarrow \mathcal{M}\) sending the sequence to its middle term and \((a_1,a_2): \mathcal{M}^{(2)}\rightarrow \mathcal{M}\times\mathcal{M}\) which sends the sequence to the product of its outer terms. The motivic Hall algebra is the Grothendieck ring of stacks over \(\mathcal{M}\), \(K(\text{St}/\mathcal{M})\), and the convolution product uses the maps just introduced. Then \(H(\mathcal{A})\) is an associative unital algebra over \(K(\text{St}/\mathbb{C})\). This algebra can also be made graded if one uses the monoid of effective classes in the numerical Grothendieck group of \(M\). The main results of the paper can be roughly summarised as follows. From here on, \(M\) is a smooth Calabi--Yau threefold. Firstly, one can consider the submodule \(H_{\mathrm{reg}}(\mathcal{A})\) of \(H(\mathcal{A})\) consisting of regular elements, that is, the span of classes \([X\rightarrow \mathcal{M}]\), where \(X\) is a variety. This is closed under the convolution product and the quotient \(H_{sc}(\mathcal{A})=H_{\mathrm{reg}}(\mathcal{A})/(\mathbb{L}-1)H_{\mathrm{reg}}(\mathcal{A})\), where \(\mathbb{L}\) is the class of the affine line, is a commutative \(K(\text{Var}/\mathbb{C})\) algebra. There is, furthermore, a Poisson bracket on \(H_{\mathrm{sc}}(\mathcal{A})\). Secondly, for any constructible function \(\mathcal{M}\rightarrow \mathbb{Z}\) there is a unique group homomomorphism from \(H_{sc}(\mathcal{A})\) to an explicitly constructed skew polynomial ring. Under additional assumptions this morphism also respects the algebra and the Poisson structure. Roughly, this map is the ``semi-classical integration map'' from the first paragraph and the target space is the ``ring of functions on a symplectic torus''. The main tool in the proofs of the main results, which are given in Section 7, is a careful analysis of the fibres of the morphism of stacks \((a_1,a_2)\) presented in Section 6. The paper concludes with an appendix which collects some facts on fibre products of stacks.