Hecke correspondences for smooth moduli spaces of sheaves (Q2141719)

There are deep connections between the representation theory of infinite-dimensional algebras and Hilbert schemes of points on surfaces. Inspired by \textit{L. Göttsche} [Math. Ann. 286, No. 1--3, 193--207 (1990; Zbl 0679.14007)] computation of the dimension of the cohomology groups of the Hilbert schemes, \textit{H. Nakajima} [Ann. Math. (2) 145, No. 2, 379--388 (1997; Zbl 0915.14001)] and \textit{I. Grojnowski} [Math. Res. Lett. 3, No. 2, 275--291 (1996; Zbl 0879.17011)] constructed an action of the Heisenberg algebra on the cohomology \[ \mathbb H_S:=\bigoplus_{n\in \mathbb N} H^*(S^{[n]},\mathbb C)\cong H^*\Bigl(\coprod_{n\in \mathbb N} S^{[n]},\mathbb C\Bigr) \] of all Hilbert schemes of points associated to a fixed smooth quasi-projective surface \(S\). Two questions about possible generalisations were studied quite immediately thereafter: \begin{itemize} \item[1.] Are there analogous actions on other invariants of the Hilbert schemes such as the Grothendieck groups \(\mathbb K_S:=\bigoplus_n K(S^{[n]})\) or the derived categories \(\mathbb D_S:=\bigoplus_n D^b\bigl(\operatorname*{Coh}(S^{[n]})\bigr)\)? \item[2.] Sending a zero-dimensional subscheme to its ideal sheaf gives an isomorphism \(S^{[n]}\cong \mathcal M_S(1,0,n)\) to the moduli space of stable sheaves of rank \(1\), trivial determinant, and \(c_2=n\). Are there similar actions on other moduli spaces of stable sheaves? \end{itemize} Already shortly after the works of Nakajima and Grojnowski, there was an affirmative answer, on the level of cohomology, to the second question for a large class of moduli spaces by \textit{V. Baranovsky} [J. Differ. Geom. 55, No. 2, 193--227 (2000; Zbl 1033.14028)]. Concerning the first question, the first successful constructions of algebra actions on the level of (torus equivariant) Grothendieck groups were given by \textit{B. L. Feigin} and \textit{A. I. Tsymbaliuk} [Kyoto J. Math. 51, No. 4, 831--854 (2011; Zbl 1242.14006)] and \textit{O. Schiffmann} and \textit{E. Vasserot} [Duke Math. J. 162, No. 2, 279--366 (2013; Zbl 1290.19001)] for Hilbert schemes of points on the affine plane. Actually, an even richer structure than an action of a Heisenberg algebra was discovered in these works. Namely, an action of the elliptic Hall algebra, which contains the (deformed) Heisenberg algebra as a subalgebra. The paper under review can be seen as a vast generalisation and strengthening of these results. Under certain assumptions on the smooth polarised surface \((S,H)\) and on the pair \((r,c_1)\in \mathbb N\times H^2(S,\mathbb Z)\) guaranteeing the smoothness of the relevant moduli spaces as well as the existence of universal families, an action of generators of the elliptic Hall algebra on the derived category of the disjoint union of moduli spaces of stable sheaves \[ \mathcal M:=\coprod_{c_2\in H^4(S,\mathbb Z)\cong \mathbb Z} \mathcal M_{(S,H)}(r,c_1,c_2) \] is constructed. In particular, this gives an action of the elliptic Hall algebra, hence of its Heisenberg subalgebras, on the Grothendieck groups. The construction of the functors on the derived category uses iterated Fourier-Mukai transforms along natural line bundles on incidence schemes of the form \begin{align*} \mathfrak Z_1&=\Bigl\{(\mathcal F_0, \mathcal F_1)\in \mathcal M\times \mathcal M\mid \mathcal F_0\subset \mathcal F_1\,,\, \mathcal F_1/\mathcal F_0\cong \mathcal O_x \text{ for some \(x\in S\)}\Bigr\}\,, \\ \mathfrak Z_2^\bullet&=\Bigl\{(\mathcal F_0, \mathcal F_1, \mathcal F_2)\in \mathcal M\times \mathcal M\times \mathcal M\mid \mathcal F_0\subset \mathcal F_1\subset \mathcal F_2\,,\, \mathcal F_1/\mathcal F_0\cong \mathcal O_x\cong \mathcal F_1/\mathcal F_0 \text{ for some \(x\in S\)}\Bigr\}\,. \end{align*} As part of the proof of the main results, interesting geometric properties of these two, and some other related, incidence schemes are established, such as being local complete intersections, in some cases even smooth, together with concrete descriptions as subvarieties of projective bundles. Actions of Heisenberg algebras on the derived categories of Hilbert schemes of points on surfaces were constructed in [\textit{S. Cautis} and \textit{A. Licata}, Duke Math. J. 161, No. 13, 2469--2547 (2012; Zbl 1263.14020); \textit{A. Krug}, Math. Z. 288, No. 1--2, 11--22 (2018; Zbl 1453.14025)] using the derived McKay correspondence. However, even when restricted to the special case of Hilbert schemes of points and to Heisenberg algebras, the paper under review still gives something interesting new. The reason is that the constructions in the two aforementioned papers are quite far away from Nakajima's original construction. They construct operators corresponding to another set of generators of the Heisenberg algebra (with rational coefficients). Objects in the images of all these operators can have support on the whole Hilbert scheme. In contrast, cohomology classes in the images of the higher Nakajima operators are supported only in loci parametrising highly non-reduced zero-dimensional subschemes. The construction in the paper under review is much closer to the original Nakajima operators. In particular, when writing the operators as a single Fourier-Mukai transform, their kernel objects (a concrete description of them would be desirable but seems very difficult to achieve) are supported exactly on the Nakajima correspondences. The paper is very well-written. But, unavoidably, there are still some technically difficult parts which might require the reader to work really hard to follow. Hence, I would like to also recommend to the interested reader the survey article [\textit{A. Neguţ}, Lect. Notes Math. 2248, 53--81 (2019; Zbl 1439.14048)] of the same author for a gentle introduction into the topic, which I found very helpful.