The cohomological Kudla conjecture for unitary Shimura varieties
François Greer and Salim Tayou · 2025
Read The cohomological Kudla conjecture for unitary Shimura varieties by François Greer and Salim Tayou (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-13299.txt.
Introduction
François Greer and Salim Tayou prove a cohomological version of the Kudla conjecture for unitary Shimura varieties, a topic in modern number theory. Their 2025 paper concerns special cycles, geometric objects inside these spaces, and the generating series that packages their cohomology classes. Kudla's program predicts such series are modular forms, a highly structured kind of function.
The authors extend the classic Kudla-Millson series to compactified spaces and prove the extensions are Hermitian modular forms. The proof covers all codimensions up to the middle. They also develop a theory of Hermitian quasi-modular forms, centered on weighted theta functions. A second series, built from cycle closures, is shown to be quasi-modular.
Catalog record
| CREATOR | François Greer and Salim Tayou |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 106 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-13299
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The cohomological Kudla conjecture for unitary Shimura varieties François Greer and Salim Tayou Abstract . We construct natural extensions of the Kudla–Millson generating series of cohomology classes of special cycles in compactified unitary Shimura varieties of signature ( n + 1 , 1 ) (n+1,1) and prove that they are holomorphic Hermitian modular forms. This proves the cohomological version of a conjecture of Kudla and Bruinier–Rosu–Zemel, in all codimensions up to the middle. We also develop the theory of Hermitian quasi-modular forms, with a particular focus on polynomial weighted theta functions, and prove that the generating series of Zariski closures of special cycles is a Hermitian quasi-modular form. 1. Introduction In their seminal work [ KM90 ] , Kudla and Millson introduced a geometric version of the classical theta lift considered by Siegel and Weil. They defined generating series of special cycles on locally symmetric spaces of orthogonal and unitary type and proved that these series are Fourier expansions of modular forms. Since the locally symmetric spaces involved are typically non-compact, it is natural to seek an extension of this modularity to compactifications. In the case where the underlying algebraic group is \mathrm{O}(n,2) or \mathrm{U}(a,b) , the locally symmetric space is a Shimura variety, which is a quasi-projective algebraic variety by [ BB66 ] . Several recent results elucidate how the modularity phenomenon survives upon taking the closures of special cycles. For cycles of codimension one, there are boundary contributions which are either quasi-modular or mixed mock modular forms [ EGT23 ] . In this paper, we generalize this behavior to higher codimension special cycles associated to the unitary groups \mathrm{U}(n+1,1) . The Shimura variety is a complex ball quotient in this case. Our study features a natural class of compactifications of Shimura varieties called toroidal compactifications , first introduced in [ AMRT10 ] . In the case of orthogonal Shimura varieties, the following conjecture was originally posed by Kudla [ Kud04 , Problem 3] and then reformulated more recently by Bruinier-Rosu-Zemel [ HRZ24 ] . Conjecture 1.1 . Let X_{\Gamma}^{\Sigma} be a toroidal compactification of an orthogonal type Shimura variety, resp. a unitary type Shimura variety of signature (n+1,1) . There exist boundary corrections to the generating series of special cycles of codimension g\geq 1 such that the resulting series is a holomorphic Siegel, resp. Hermitian, modular form, valued in \mathrm{CH}^{g}(X_{\Gamma}^{\Sigma}) . The goal of this paper is to prove the cohomological incarnation of Conjecture 1.1 in the unitary case, for special cycles of codimension up to the middle, as well as to provide non-holomorphic completions, as we did in our previous work in the orthogonal case for codimension 1 [ EGT23 ] . 1.1. Main results Let k be an imaginary quadratic field with ring of integers \mathcal{O}_{k} and let (V,h) be a Hermitian vector space over k of signature (n+1,1) . Let L\subset V be an \mathcal{O}_{k} -lattice, i.e., h is \mathcal{O}_{k} -valued on L , and L\otimes_{\mathcal{O}_{k}}k=V . One can attach to this data a unitary Shimura variety X_{\Gamma} whose construction is recalled in Section 3 . Let 1\leq g\leq n+1 , \underline{\nu}\in(L^{\vee}/L)^{g} and let N\in\mathrm{Herm}_{g}(k)_{\geq 0} be a Hermitian semi-positive matrix. For each (\underline{\nu},N) , there is a special cycle {\mathcal{Z}}(\underline{\nu},N)\hookrightarrow X_{\Gamma} of codimension g . Let \rho_{L,g}:\mathrm{U}(g,g)(\mathbb{Z})\rightarrow\mathrm{Aut}(\mathbb{C}[(L^{\vee}/L)^{g}]) be the genus g Weil representation associated to the Hermitian lattice (L,h) , where L^{\vee} denotes dual lattice of L with respect to the quadratic form \mathrm{Tr}_{L/K}(h) . Consider the generating series: \Phi_{L}^{g}(\tau)=\sum_{\underset{\underline{\nu}\in(L^{\vee}/L)^{g}}{N\in\mathrm{Herm}_{g}(k)_{\geq 0}}}[{\mathcal{Z}}(\underline{\nu},N)]q^{N}\mathfrak{e}_{\underline{\nu}}\in\mathrm{CH}^{g}(X_{\Gamma})\otimes\mathbb{C}[(L^{\vee}/L)^{g}]\llbracket q\rrbracket\penalty\ , where \tau\in{\mathcal{H}}_{g} is an element of the Hermitian upper half-space and q^{N}=e^{2i\pi\mathrm{Tr}(\tau N)} . There is a unique toroidal compactification X_{\Gamma}\subset X_{\Gamma}^{\mathrm{tor}} , and we denote by \overline{{\mathcal{Z}}}(\underline{\nu},N) the Zariski closure of the special cycle {\mathcal{Z}}(\underline{\nu},N) in X^{\mathrm{tor}}_{\Sigma} . Our first main result is the following. Theorem 1.2 . Assume that g\leq\frac{n}{2} . There exists a correction \widetilde{{\mathcal{Z}}}(\underline{\nu},N)\subset X^{\mathrm{tor}}_{\Sigma} to each \overline{{\mathcal{Z}}}(\underline{\nu},N) , by a cycle supported in the boundary X^{\mathrm{tor}}_{\Sigma}\setminus X_{\Sigma} , such that the generating series: \sum_{\underset{\underline{\nu}\in(L^{\vee}/L)^{g}}{N\in\mathrm{Herm}_{g}(k)_{\geq 0}}}[\widetilde{{\mathcal{Z}}}(\underline{\nu},N)]q^{N}\mathfrak{e}_{\underline{\nu}}\in\mathrm{H}^{2g}\left(X^{\mathrm{tor}}_{\Gamma}\right)\otimes\mathbb{C}[(L^{\vee}/L)^{g}]\llbracket q\rrbracket\penalty\ . is a Hermitian modular form of weight n+2 and representation \rho_{L,g} with respect to \mathrm{U}(g,g)(\mathbb{Z}) . In particular, this resolves Conjecture 1.1 in cohomology in the case of \mathrm{U}(n+1,1) , for g up to the middle codimension. Our second result provides non-holomorphic completions of the series of Zariski closures \overline{{\mathcal{Z}}}(\underline{\nu},N) in the spirit of [ EGT23 ] , which realizes these objects as cycle-valued Hermitian quasi-modular forms. Theorem 1.3 . The generating series of cycle classes for the Zariski closures \overline{{\mathcal{Z}}}(\underline{\nu},N) is a Hermitian quasi-modular form. That is, it admits a non-holomorphic completion which transforms like a Hermitian modular form of weight n+2 and representation \rho_{L,g} with respect to \mathrm{U}(g,g)(\mathbb{Z}) . An analogous statement was obtained for special divisors in the orthogonal case in [ EGT23 ] . To our knowledge, Theorem 1.3 is the first instance of Hermitian quasi-modular forms appearing in the context of special cycles. To give a more explicit form to the corrections and completions in Theorems 1.2 and 1.3 , we need to recall the structure of the toroidal compactification X_{\Gamma}^{\mathrm{tor}} . For each equivalence class of primitive isotropic line \mathfrak{J}\subset L under the action of \Gamma , we have an associated boundary divisor {\mathcal{B}}_{\mathfrak{J}} which is isomorphic to E_{M} , where M=\mathfrak{J}^{\bot}/\mathfrak{J} , and E=\mathbb{C}/{\mathcal{O}}_{k} is a CM elliptic curve. Let \iota_{\mathfrak{J}}:{\mathcal{B}}_{\mathfrak{J}}\hookrightarrow X_{\Gamma}^{\mathrm{tor}} denote the inclusion and let \iota_{{\mathfrak{J}},*}:H^{*}({\mathcal{B}}_{\mathfrak{J}},\mathbb{Q})\rightarrow H^{*+2}(X_{\Gamma}^{\mathrm{tor}},\mathbb{Q})\penalty\ , denote the Gysin morphism. The divisor {\mathcal{B}}_{\mathfrak{J}} is polarized by an ample line bundle \mathcal{L} isomorphic to the conormal bundle of {\mathcal{B}}_{\mathfrak{J}} in X_{\Gamma}^{\mathrm{tor}} , so there is a natural Lefschetz decomposition on the cohomology of {\mathcal{B}}_{\mathfrak{J}} , where \mathcal{L} denotes also the associated Lefschetz operator: \mathrm{H}^{2g}({\mathcal{B}}_{\mathfrak{J}},\mathbb{Q})\simeq\bigoplus_{0\leq\ell\leq g}{\mathcal{L}}^{g-\ell}\mathrm{H}^{2\ell}_{\mathrm{prim}}(\mathcal{B}_{\mathfrak{J}},\mathbb{Q})\penalty\ . As this is a decomposition of Hodge structures, we get an induced decomposition at the level of Hodge classes \mathrm{Hdg}^{g}(-)=\mathrm{H}^{g,g}(-,\mathbb{Q}) : \mathrm{Hdg}^{g}({\mathcal{B}}_{\mathfrak{J}})\simeq\bigoplus_{0\leq\ell\leq g}{\mathcal{L}}^{g-\ell}\mathrm{Hdg}^{\ell}_{\mathrm{prim}}(\mathcal{B}_{\mathfrak{J}})\penalty\ . For each 0\leq\ell\leq\frac{n}{2} , consider a basis (W_{i}^{\ell})_{i\in I_{\ell}} of the primitive cohomology \mathrm{Hdg}^{\ell,\ell}_{\mathrm{prim}}(\mathcal{B}_{\mathfrak{J}},\mathbb{Q}) . As the Hodge conjecture holds for {\mathcal{B}}_{\mathfrak{J}} , the W_{i}^{\ell} are classes of algebraic cycles; see 3.6 . To each primitive cycle W^{\ell}_{i} , we associate a harmonic polynomial 1 1 1 Polynomial in the real and imaginary parts of the \lambda_{i} ’s. P^{\ell}_{i} in (\lambda_{1},\ldots,\lambda_{\ell})\in(L_{\mathbb{R}})^{\ell} that satisfies the following equivariance property for A\in\mathrm{GL}_{\ell}(\mathbb{C}) : P\left((\lambda_{1},\ldots,\lambda_{\ell})\cdot A\right)=|\det(A)|^{2}\cdot P(\lambda_{1},\ldots,\lambda_{\ell})\penalty\ . The space of all polynomials that satisfy this homogeneity condition is in fact a finite-dimensional vector space denoted by {\mathcal{F}}_{n,\ell} , and the direct sum {\mathcal{F}}_{n,\bullet}=\bigoplus_{\ell=0}^{n}{\mathcal{F}}_{n,\ell} admits an \mathfrak{sl}_{2} -action introduced in Section 2.2 , the primitive elements of which are exactly the harmonic polynomials, i.e., the span of the P_{i}^{\ell} . Let (E,F,H) be the corresponding \mathfrak{sl}_{2} triple. Our main theorem is the following. Theorem 1.4 . Let g\leq\frac{n}{2} , let \underline{\nu}\in(L^{\vee}/L)^{g} , and let N\in\mathrm{Herm}_{g}(k)_{\geq 0} be a Hermitian semi-positive matrix. The correction of the cycle {\mathcal{Z}}(\underline{\nu},N) is trivial if \uparrow_{L}^{M}(\underline{\nu})=0 2 2 2 The arrow operator between Weil representations. and otherwise, it is given by the formula: [\widetilde{{\mathcal{Z}}}(\underline{\nu},N)]=[{\mathcal{Z}}(\underline{\nu},N)]+\sum_{{\mathfrak{J}}/\sim}\sum_{\underset{i\in I_{\ell}}{0\leq\ell\leq g-1}}\frac{r_{\mathfrak{J}}}{d_{k}}P^{\ell}_{i}(\underline{\nu},N)\,\mathfrak{\iota}_{{\mathfrak{J}},*}[W^{\ell}_{i}\cup\mathcal{L}^{g-\ell-1}]\penalty\ , where by definition P^{\ell}_{i}(\underline{\nu},N):=\sum_{\begin{subarray}{c}(\lambda_{1},\ldots,\lambda_{g})\in(M^{\vee})^{g},\\ {\mathrm{Gram}}(\underline{\lambda})=N,\\ [\underline{\lambda}]=\uparrow^{M}_{L}(\underline{\nu})\\ \end{subarray}}(E^{g-\ell}P^{\ell,\ell}_{i})(\lambda_{1},\ldots,\lambda_{g})\penalty\ , E is the raising operator on the \mathfrak{sl}_{2} -module {\mathcal{F}}_{n,\bullet} , -d_{k} is the discriminant of k , and r_{\mathfrak{J}} is a certain rational number associated to {\mathfrak{J}} , see Section 3.3 . Let \vartheta_{M,P} denotes the weighted theta series of the homogeneous