Canonical blow-ups of Lagrangian and Orthogonal Grassmannians

Hanlong Fang, Alex Massarenti, Xian Wu · 2026

Read Canonical blow-ups of Lagrangian and Orthogonal Grassmannians by Hanlong Fang, Alex Massarenti, Xian Wu (2026) 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-2603-00386.txt.

Introduction

Fang, Massarenti, and Wu build families for Hilbert quotients of Lagrangian and orthogonal Grassmannians by blowing up the matching isotropic spaces. The result gives smooth toroidal compactifications of maximal rank matrix spaces, symmetric and skew, with simple normal crossing edges that include complete bilinear forms. Over any algebraically closed field they stay smooth and the quotients match complete bilinear form spaces. In characteristic zero they are weak Fano, rigid, and resolve Landsberg-Manivel rational maps.

Catalog record

CREATORHanlong Fang, Alex Massarenti, Xian Wu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE150968
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.00386

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Canonical blow-ups of Lagrangian and Orthogonal Grassmannians Hanlong Fang, Alex Massarenti, Xian Wu Abstract Let 𝐋𝐆 ⁡ ( V ⊕ V ∗ ) \mathbf{LG}(V\oplus V^{*}) and 𝐎𝐆 + ​ ( V ⊕ V ∗ ) \mathbf{OG}^{+}(V\oplus V^{*}) denote the Lagrangian and orthogonal Grassmannians endowed with the natural 𝔾 m \mathbb{G}_{m} -actions, respectively. Thaddeus proved that over ℂ \mathbb{C} , the Hilbert quotients 𝐋𝐆 ( V ⊕ V ∗ ) / / 𝔾 m \mathbf{LG}(V\oplus V^{*})\!/\!/\mathbb{G}_{m} and 𝐎𝐆 + ( V ⊕ V ∗ ) / / 𝔾 m \mathbf{OG}^{+}(V\oplus V^{*})\!/\!/\mathbb{G}_{m} are isomorphic to the wonderful compactifications of the spaces of symmetric and skew - symmetric matrices of maximal ranks, that is, the spaces of complete quadrics and complete skew-forms, respectively. In this paper, we construct the universal families of these Hilbert quotients by explicitly blowing up the corresponding isotropic Grassmannians, resulting in smooth toroidal compactifications of the spaces of symmetric and skew - symmetric matrices of maximal ranks (before projectivization), which have simple normal crossing boundary divisors and include the spaces of the complete bilinear forms among these divisors. Specifically, we prove that over any algebraically closed field, the universal families of these Hilbert quotients are smooth, and the Hilbert quotients themselves are isomorphic to the spaces of the complete bilinear forms. Over an algebraically closed field of characteristic zero, we prove that the universal families are weak Fano varieties with vanishing higher cohomology groups for their tangent bundles, and are therefore locally rigid. Furthermore, we show that these universal families naturally resolve the Landsberg-Manivel rational maps from projective spaces to isotropic Grassmannians. 1 Introduction The study of the moduli spaces of complete bilinear forms originated from classical 19th-century works in enumerative and projective geometry [ 9 , 23 , 45 ] , and later evolved towards a more geometric perspective in higher dimensions, establishing connections to fields such as combinatorics and algebraic statistics, with key contributions from [ 52 , 46 , 50 , 51 , 31 , 12 , 13 , 49 , 30 , 39 ] , among others. In particular, De Concini and Procesi developed the foundational theory of what later became known as wonderful compactifications, generalizing earlier constructions to all symmetric spaces of adjoint type. More recently, such spaces have been used in constructing other moduli spaces [ 49 , 37 , 10 , 8 , 42 ] . In particular, Thaddeus proved that over \mathbb{C} , the Hilbert quotients of the natural one-dimensional torus actions on Grassmannians, Lagrangian Grassmannians, and orthogonal Grassmannians are isomorphic to the spaces of complete collineations, complete quadrics, and complete skew-forms, respectively. The theory of wonderful compactifications constitutes a cornerstone in the study of homogeneous varieties with various applications in arithmetic geometry and representation theory [ 17 , 47 , 38 , 14 ] . Nevertheless, the current framework fails for homogeneous spaces whose automorphism groups possess a non-trivial center. The development of a more general framework is still an enticing open question, while a major step forward has been taken by Kausz [ 27 ] in 2000, who constructed a novel modular compactification for the general linear group. A simple uniform picture was recently established in [ 18 ] to incorporate the Kausz compactifications and the spaces of complete collineations. In particular, it was shown that over any algebraically closed field, the Kausz-type compactifications are the universal families for the Hilbert quotients of the \mathbb{G}_{m} -action on Grassmannians, which aligns with Thaddeus’s results. Investigating birational maps from projective spaces to smooth prime Fano varieties has been significantly advanced by the study of special Cremona transformations with a smooth irreducible base scheme [ 53 , 11 , 16 , 24 , 44 , 1 , 21 ] . Landsberg and Manivel [ 33 ] constructed natural birational maps from projective spaces onto minuscule varieties, for which the base scheme may be singular. Zak [ 53 ] , and Fu and Hwang [ 21 ] factored the rational maps associated with complex quadrics into a blow-up and a blow-down. Ding [ 15 ] further generalized to any compact Hermitian symmetric spaces, based on the blow-up centers constructed via the Białynicki-Birula decomposition (first appeared in [ 20 ] ) and the theory of varieties of minimal rational tangents [ 25 ] . It was shown in [ 18 ] that the Landsberg–Manivel maps from projective spaces onto Grassmannians can naturally factor through the Kausz-type compactifications over {\rm Spec\,}\mathbb{Z} . The focus of this paper is to study the universal families of the two remaining types of Hilbert quotients that arise in Thaddeus’s work. We explicitly construct the universal families by generalizing the blow - up technique of [ 18 ] . Using the concrete description, we are able to examine their differential and birational geometric properties. Let us be more precise. Let V be a free \mathbb{Z} -module of rank n . The module V\oplus V^{*} possesses a canonical symplectic form \omega defined by \omega((v_{1},\alpha_{1}),(v_{2},\alpha_{2})):=\alpha_{2}(v_{1})-\alpha_{1}(v_{2}) . The Lagrangian Grassmannian \mathbf{LG}(V\oplus V^{*}) is a smooth, projective scheme over \mathbb{Z} which represents the functor assigning to each commutative ring R the set \left\{L\subset(V\oplus V^{*})\otimes_{\mathbb{Z}}R\left|\,\begin{array}[]{l}L\text{ is a}\text{ rank }n\text{ direct summand of }(V\oplus V^{*})\otimes_{\mathbb{Z}}R,\\ \omega_{R}(u,w)=0\text{ for all }u,w\in L\end{array}\right.\right\}. The Plücker embedding of the Grassmannian induces an embedding e_{{\bf LG}}:\mathbf{LG}(V\oplus V^{*})\xhookrightarrow{\,\,\,\,\,\,\,\,\,\,\,}\mathbb{P}\big(\bigwedge\nolimits^{n}(V\oplus V^{*})\big). (1) The \mathbb{G}_{m} -action on V\oplus V^{*} defined by t\cdot(v,\alpha)\mapsto(v,t\alpha),\ \ \,\,(v,\alpha)\in V\oplus V^{*}, (2) induces a \mathbb{G}_{m} -action on \mathbf{LG}(V\oplus V^{*}) . Define a subgroup {\rm GL}(V^{*})\hookrightarrow{\rm Sp}(V\oplus V^{*},\omega) by \begin{split}A&\longmapsto\left(\begin{matrix}A^{-T}&0\\ 0&A\end{matrix}\right)\end{split}. Noticing that {\rm GL}(V^{*})\times\mathbb{G}_{m} stabilizes each summand in the decomposition \bigwedge\nolimits^{n}(V\oplus V^{*})=\bigoplus\limits_{k=0}^{n}\bigwedge\nolimits^{k}V\otimes\bigwedge\nolimits^{n-k}V^{*}, we can define a {\rm GL}(V^{*})\times\mathbb{G}_{m} -equivariant rational map \mathcal{KL}(V):\mathbf{LG}(V\oplus V^{*})\mathrel{\mathchoice{\mathrel{}\mathrel{\mathop{\dabar@\dabar@\dabar@\dabar@}\limits^{\hbox{$\scriptstyle\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,$}}}\mathrel{\mathchar 0\symAMSa 04B}{}}{\mathrel{}\mathrel{\mathop{\dabar@\dabar@\dabar@\dabar@}\limits^{\hbox{$\scriptstyle\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,$}}}\mathrel{\mathchar 0\symAMSa 04B}{}}{\mathrel{}\mathrel{\mathop{\dabar@\dabar@\dabar@\dabar@\dabar@\dabar@}\limits^{\hbox{$\scriptscriptstyle\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,$}}}\mathrel{\mathchar 0\symAMSa 04B}{}}{\mathrel{}\mathrel{\mathop{\dabar@\dabar@\dabar@\dabar@\dabar@\dabar@\dabar@\dabar@}\limits^{\hbox{$\scriptstyle\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,$}}}\mathrel{\mathchar 0\symAMSa 04B}{}}}\mathbb{P}\big(\bigwedge\nolimits^{n}(V\oplus V^{*})\big)\times\prod\limits_{k=0}^{n}\mathbb{P}\big(\bigwedge\nolimits^{k}V\otimes\bigwedge\nolimits^{n-k}V^{*}\big). (3) Denote by \mathcal{TL}(V) the closure of the image of \mathcal{KL}(V) , and \mathcal{ML}(V) the image of \mathcal{TL}(V) under the natural projection \mathbb{P}\big(\bigwedge\nolimits^{n}(V\oplus V^{*})\big)\times\prod\limits_{k=0}^{n}\mathbb{P}\big(\bigwedge\nolimits^{k}V\otimes\bigwedge\nolimits^{n-k}V^{*}\big)\longrightarrow\prod\limits_{k=0}^{n}\mathbb{P}\big(\bigwedge\nolimits^{k}V\otimes\bigwedge\nolimits^{n-k}V^{*}\big). (4) Similarly, the orthogonal Grassmannian \mathbf{OG}(V\oplus V^{*}) is a smooth, projective scheme over \mathbb{Z} which represents the functor assigning to each commutative ring R the set \left\{L\subset(V\oplus V^{*})\otimes_{\mathbb{Z}}R\left|\,\begin{array}[]{l}L\text{ is a}\text{ rank }n\text{ direct summand of }(V\oplus V^{*})\otimes_{\mathbb{Z}}R,\\ Q_{R}|_{L}=0\end{array}\right.\right\}. Here V\cong\mathbb{Z}^{n} with n\geq 2 , and V\oplus V^{*} possesses a a canonical quadratic form Q defined by Q(v,\alpha):=\alpha(v) . We denote by \mathbf{OG}^{+}(V\oplus V^{*}) the connected component of \mathbf{OG}(V\oplus V^{*}) that contains the maximal isotropic submodule V\oplus\{0\}\subset V\oplus V^{*} . The first canonical embedding of \mathbf{OG}^{+}(V\oplus V^{*}) can be given as follows (see [ 28 ] for details). Let C(V\oplus V^{*},Q) be the Clifford algebra with a \mathbb{Z} -linear map i:V\oplus V^{*}\rightarrow C(V\oplus V^{*},Q) such that [i(x)]^{2}=Q(x) . The spinor representation \rho:C(V\oplus V^{*},Q)\rightarrow{\rm End}_{\mathbb{Z}}(\bigwedge V^{*}) defined by \rho(v,\alpha)=l_{\alpha}+\iota_{v} splits into half-spinor representations \bigwedge^{\rm even}V^{*} , \bigwedge^{\rm odd}V^{*} , where l_{\alpha} is the left multiplication and i_{v} is the derivation. A spinor \phi\in\bigwedge^{\rm even}V^{*} is pure if its annihilator L_{\phi}:=\{x\in V\oplus V^{*}:x\cdot\phi=0\} is a maximal isotropic submodule. Each L\in\mathbf{OG}^{+}(V\oplus V^{*}) corresponds to a unique pure spinor line [\phi_{L}] . we have the spinor embedding e_{{\bf OG^{+}}}:\mathbf{OG}^{+}(V\oplus V^{*})\xhookrightarrow{\,\,\,\,\,\,\,\,\,\,\,}\mathbb{P}\big(\bigwedge\nolimits^{\rm even}V^{*}\big). (5) Similarly, we can define a {\rm GL}(V^{*})\times\mathbb{G}_{m} -equivariant rational map \mathcal{KO}(V):\mathbf{OG}^{+}(V\oplus V^{*})\mathrel{\mathchoice{\mathrel{}\mathrel{\mathop{\dabar@\dabar@\dabar@\dabar@}\limits^{\hbox{$\scriptstyle\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,$}}}\mathrel{\mathchar 0\symAMSa 04B}{}}{\mathrel{}\mathrel{\mathop{\dabar@\dabar@\dabar@\dabar@}\limits^{\hbox{$\scriptstyle\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,$}}}\mathrel{\mathchar 0\symAMSa 04B}{}}{\mathrel{}\mathrel{\mathop{\dabar@\dabar@\dabar@\dabar@\dabar@\dabar@}\limits^{\hbox{$\scriptscriptstyle\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,$}}}\mathrel{\mathchar 0\symAMSa 04B}{}}{\mathrel{}\mathrel{\mathop{\dabar@\dabar@\dabar@\dabar@\dabar@\dabar@\dabar@\dabar@}\limits^{\hbox{$\scriptstyle\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,$}}}\mathrel{\mathchar 0\symAMSa 04B}{}}}\mathbb{P}\big(\bigoplus\limits_{k=0}^{[n/2]}\bigwedge\nolimits^{2k}V^{*}\big)\times\prod\limits_{k=0}^{[n/2]}\mathbb{P}\big(\bigwedge\nolimits^{2k}V^{*}\big). (6) Denote by \mathcal{TO}(V) the closure of the image of \mathcal{KO}(V) , and \mathcal{MO}(V) the image of \mathcal{TO}(V) under the natural projection \mathbb{P}\big(\bigoplus\limits_{k=0}^{[n/2]}\bigwedge\nolimits^{2k}V^{*}\big)\times\prod\limits_{k=0}^{[n/2]}\mathbb{P}\big(\bigwedge\nolimits^{2k}V^{*}\big)\longrightarrow\prod\limits_{k=0}^{[n/2]}\mathbb{P}\big(\bigwedge\nolimits^{2k}V^{*}\big). (7) When there is

Read the complete text