On the subring of special cycles (Q2136077)

The special cycles on orthogonal Shimura varieties, namely those coming from embeddings of Shimura subvarieties, are the subject of a vast program in Number Theory and Algebraic Geometry, which is named after the author of the current paper. The main theme is that in many settings (and conjecturally in all settings) their images in Chow or cohomology groups are the coefficients of a modular form, the parameters of which are determined by the ambient Shimura variety. Both the Chow and the cohomology groups are the graded pieces of a ring, where the product is essentially defined by intersecting cycles (or cup product on cohomology). Since the cohomology of top degree is equipped, in the compact setting, with a degree map to scalars, this product can be combined with the tautological bundle on the Shimura variety to obtain scalar-valued pairings between different parts of the Chow or cohomology ring. This paper is concerned with investigating these notions in the Hilbert setting, namely for orthogonal Shimura subvarieties that are associated with quadratic spaces over totally real fields. In more detail, let \(V\) be a quadratic space over a totally real number field \(F\) of degree \(d\) over \(\mathbb{Q}\), which has signature \((m,2)\) for real embeddings \(\sigma\) in a set \(\Sigma\) of size \(0 \leq d_{+}<d\) of real places of \(F\) (the strict inequality is required for the Shimura variety to be compact), and signature \((m+2,0)\) in the remaining \(d-d_{+}\) real places of \(F\). To each open compact subgroup \(K\) of the corresponding spin group of \(V\) over the finite adéles one associates a Shimura variety \(S_{K}\) of dimension \(md_{+}\), and for every positive definite matrix \(T\) of size \(n\) over \(F\) corresponds a special cycle, which is a finite sum of Shimura subvarieties of codimension \(nd_{+}\) in \(S_{K}\). Their Chow or cohomology classes form a direct system over the groups \(K\) (ordered by inclusion), yielding a well-defined limit cycle \(Z(T,\varphi)\) in the appropriate limit group, for each Schwartz function \(\varphi\) on \(V\) over the finite adéles. This definition extends to positive semi-definite matrices \(T\) using the appropriate power of the tautological bundle on \(T\), and the modularity theorem (already proved before) is that with this definition, for every \(\varphi\), the generating series over \(T\) of the classes in the limit cohomology group is a Hilbert--Siegel modular form of parallel weight \(\frac{m}{2}+1\). There is a natural product formula for the cycles \(Z(T_{i},\varphi_{i})\), as the sum of cycles associated with the tensor product of the \(\varphi_{i}\)'s, and block matrices having the \(T_{i}\)'s as their diagonal blocks (this translates to a formula for the modular generating series as well). The author then considers the subring of the cohomology ring that is generated by the cycles \(Z(T,\varphi)\) (divided by its nilpotents, to become reduced and behave better with respect to pairings), and defines a pairing on it, that is concentrated on the top degree part of the product. The main theorem of the paper is relating these structure (cycles, product, pairing) for different quadratic spaces over \(F\). More precisely, consider two such spaces, and assume that are isomorphic at each local place of \(F\) (so that they have the same dimension \(m+2\) and their values of \(d_{+}\) must have the same parity). Then the isomorphism between the associated spaces of Schwartz functions on \(V\) over the finite adéles produces an isomorphism between the corresponding reduced rings of special cycles, which is also an isometry. The basic idea is that when the Schwartz functions are completed with an infinite part (involving the Schwartz form for \(\sigma\in\Sigma\) and the Gaussian at the other real places), the generating series of cohomology classes was shown in Reference number 6 (by the same author) to be the formal cohomolgy class of the resulting theta function arising from the Weil representation of the metaplectic double cover of the corresponding symplectic group. The map from the Schwartz functions on \(V\) over the total adéle ring into a principal series representation of the metaplectic group allows one to relate, by the Siegel-Weil formula, the associated Eisenstein series with an integral of said theta function. This allows one to express the pairings of products in terms of such Eisenstein series (this is Theorem A of the paper), and the matching principle, arising from the relation to such principal series representations by Reference number 5 (by the same author yet again), is then used for proving the isometrical isomorphism between the rings arising from two quadratic spaces over \(F\), under the said conditions. Reference number 4 of the author dealt with the case \(d_{+}=1\), and is used as a reference for many details, some of which are omitted from the present paper. An interesting new analogous construction is given for the case \(d_{+}=0\), and the isometrical isomorphism is shown to extend to such cases as well. Note that the inequality \(d_{+}<d\) is assumed throughout, as otherwise boundary components can show up and add complicated contributions. The ring for \(d_{+}=0\) is a truncated polynomial ring in one variable (equivalent to the class of the tautological bundle) over continuous functions on quotients of the finite adélic points of the special orthogonal group, a degree map is again defined (using the top degree coefficient) and with it a pairing, and the special cycles are based on representation numbers (indeed, Fourier coefficients of theta functions). Finally, some local matching conditions are described, a few points are made about comparisons between the rings associated with quadratic spaces that are not isomorphic over all the finite places of \(F\), and this analysis is used to exemplify how the matching principle cannot respect tensor products. The paper is divided into 5 sections. Section 1 is the Introduction, including the definitions of many objects (the special cycles, the resulting rings, and more) and the statements of some of the results. Section 2 considers generating series, theta functions, and Eisenstein series, and Section 3 presents the matching principle and establishes the isometric isomorphism between different reduced rings of special cycles. Section 4 constructs the objects required for defining the ring for \(d_{+}=0\), and extends the isometric isomorphism to this case as well. Finally, Section 5 discusses local representations and local matching conditions.