The maximal destabilizers for Chow and K-stability

YI YAO · 2025

Read The maximal destabilizers for Chow and K-stability by YI YAO (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-2511-16401.txt.

Introduction

Yi Yao works on quantizing stability in Kähler geometry, following Donaldson's insight that constant scalar curvature metrics correspond to balanced norms on spaces of sections. The setting here is the unstable case, where a polarized manifold admits no such metric and the question becomes what destabilizes it most strongly.

For large embeddings, a projective embedding of an unstable manifold is Chow-unstable, with a unique maximal destabilizing filtration on its section space. K-stability has its own maximal destabilizer as well, the steepest descent direction of the K-energy. Yao uses non-Archimedean pluripotential theory of Boucksom and Jonsson, plus stated assumptions. He lays out a route to prove the maximal K-destabilizer is quantized by the maximal Chow-destabilizers, linking the two notions of worst instability.

Catalog record

CREATORYI YAO
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE139 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-16401

Browse the full library · About the Tower of Babel

Full text

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

The maximal destabilizers for Chow and K-stability YI YAO Abstract . Donaldson showed that the constant scalar curvature Kähler metrics can be quantized by the balanced Hermitian norms on the spaces of global sections. We explore an analogous problem in the unstable situation. For a K-unstable manifold ( X , L ) (X,L) , its projective embedding via | k ​ L | \left|kL\right| will be Chow-unstable when k k is sufficiently large and divisible. There is a unique filtration on H 0 ​ ( X , k ​ L ) \mathrm{H}^{0}(X,kL) , that corresponds to the maximal destabilizer for Chow-stability of the embedded variety. On the other hand, there is a maximal destabilizer for K-stability after the work of Xia and Li, which corresponds to the steepest descent direction of K-energy. Based on Boucksom–Jonsson’s non-Archimedean pluripotential theory and some idealistic assumptions, we provide a route to show that maximal K-destabilizers are quantized by the maximal Chow-destabilizers. 1. Introduction A central problem in Kähler geometry is the existence of canonical metrics in a given Kähler class. Let (X,L) be a polarized Kähler manifold, the Yau–Tian–Donaldson conjecture predicts that the existence of constant scalar curvature Kähler (cscK) metrics in c_{1}(L) is equivalent to the K-polystability (or some enhanced version) of (X,L) . During the last decade, this conjecture has been verified for the Fano case (X,-K_{X}) by various approaches, such as [ 20 , 73 , 3 ] . Following the variational approach of [ 3 ] , there has been big progress for the general polarization, see [ 52 , 12 , 51 ] . The quantization method plays a key role in the study of canonical metrics problems. The main idea is that the infinite-dimensional space \mathcal{H}(L) of Kähler potentials can be approximated (quantized) by the Riemannian symmetric space \mathcal{H}(R_{k}) of Hermitian norms on R_{k}\coloneqq\mathrm{H}^{0}(X,kL) , as the Planck’s constant \hbar\sim k^{-1} goes to zero. A cornerstone is the Tian–Yau–Zelditch’s asymptotic expansion of the Bergman kernel. There are many studies involving quantization in this area. We mention one work that motivates our study. The seminar work of Donaldson [ 33 , 37 ] shows that if (X,L) admits a cscK metric \omega and the automorphism group is discrete, then there exists balanced norms on R_{k} for k\gg 1 . Moreover, the sequence of Fubini–Study metrics induced by balanced norms will converge to \omega , see Theorem 30 for a review. In a nutshell, balanced norms quantize cscK metrics. By [ 81 , 56 ] , the existence of balanced norms on R_{k} is equivalent to the Chow-polystability of the embedding image X\hookrightarrow\mathbb{P}R_{k}^{*} . Hence in this case, the existence of cscK metrics can imply the asymptotic Chow-stability. Conversely, Ross–Thomas [ 67 ] showed that the asymptotic Chow-semistability can imply the K-semistability of (X,L) . In a rough sense, one can say that Chow-stability quantizes K-stability. In this paper, we explore the parallel story on the unstable side. If (X,L) is K-unstable, by Theorem 29 there exists r,N\in\mathbb{N}_{+} such that the embedding image X\hookrightarrow\mathbb{P}R_{kr}^{*} is Chow-unstable for k>N . Replacing L by rL , we can assume r=1 . In GIT, after the work of Kempf [ 48 ] and Rousseau [ 68 ] , one can consider the maximal destabilizing 1-parameter subgroups to an unstable point, which minimizes the normalized Hilbert–Mumford weight and is unique up to conjugations. Hence for Chow-unstable subvariety X\hookrightarrow\mathbb{P}R_{k}^{*} , we obtain a set of 1-PS in \mathrm{SL}(R_{k}) which are conjugate to each other. To eliminate the conjugacy ambiguity, we consider the filtration (i.e. NA norm) on R_{k} induced by these 1-PS, that is unique up to scaling. We call this NA norm on R_{k} the maximal Chow-destabilizer for X\hookrightarrow\mathbb{P}R_{k}^{*} . Comparing with the picture [ 33 ] in the stable case, we ask: How do this sequence of maximal Chow-destabilizers relate to the maximal K-destabilizer? The maximal K-destabilizer was first studied by Székelyhidi [ 69 ] in the toric setting, where it is given by a concave function \Theta (optimal test-configuration) on the moment polytope P that minimizes the Donaldson–Futaki invariant. Assuming that the Calabi flow exists for all time, Székelyhidi showed that the scalar curvature under the flow converges to \Theta . Moreover, if \Theta is piecewise linear, it induces a polyhedral decomposition of P , such that each piece is relatively K-semistable in a certain sense. These support a conjecture of Donaldson [ 34 , 35 ] , which predicts that when the extremal metrics are absent, Calabi flow will break up the manifold into pieces, and each piece admits a complete extremal metric or collapses. Also see [ 70 ] for the case of ruled surfaces. In the general case, maximal K-destabilizer is defined in the work of Xia [ 77 ] . The K-energy \mathcal{M} can be extended to a convex functional over the space \mathcal{E}^{2}(L) of L^{2} -finite energy metrics on L , which is a Hadamard space. By considering the associated gradient flow, Xia showed that in the geodesically unstable case, modulo the asymptotic relations, there exists a unique geodesic ray \ell^{\mathrm{K}} in \mathcal{E}^{2}(L) along which \mathcal{M} achieves the minimal limit slope, namely it minimizes the radial K-energy \mathcal{M}^{\mathrm{rad}} ( 6.3 ). We call \ell^{\mathrm{K}} the maximal K-destabilizing ray . By the construction in [ 3 ] , \ell^{\mathrm{K}} induces a non-Archimedean metric \phi^{\mathrm{K}} on the Berkovich analytification L^{\mathrm{an}} of L (see § 5.3 ). By [ 52 ] , \ell^{\mathrm{K}} is maximal in the sense of [ 3 ] , so it can be recovered from the NA data \phi^{\mathrm{K}} . We call \phi^{\mathrm{K}} the maximal K-destabilizer for (X,L) , which is unique up to scaling, see Definition 39 for details. In the toric setting, it can be reduced into Székelyhidi’s optimal test-configuration, see Example 43 . In order to connect the maximal Chow-destabilizers to the maximal K-destabilizer, first we need to characterize them in a unified way, namely as the “steepest descent direction” for some energy functionals. Instead of being a 1-PS that minimizes the Chow-weight, the maximal Chow-destabilizer can also be represented by the norm geodesic ray along which the Kempf–Ness functional achieves the minimal limit slope. Maximal Chow-destabilizers Let X\subset\mathbb{P}V^{*} be a projective variety and \mathcal{H}(V) be the space of Hermitian norms on V . By works [ 81 , 56 , 62 , 60 ] , the Kempf–Ness functional for the Chow-stability of X can be identified with the following functional on \mathcal{H}(V) , \mathbf{M}_{X}(H)\coloneqq\mathcal{E}_{X}\left(\mathbf{FS}(H),\mathbf{FS}(H_{r})\right)+\frac{1}{\dim V}\log\left(\frac{\det H}{\det H_{r}}\right), where \mathbf{FS}(H) is the Fubini–Study metric on \mathcal{O}(1)|_{X} induced by H , \mathcal{E}_{X} is the Monge–Ampère energy, and H_{r} is a reference norm. Chow-weights are obtained by taking the limit slope of \mathbf{M}_{X} along the norm geodesic rays. Consider the asymptotic cone \hat{\mathcal{H}}(V) of \mathcal{H}(V) , which is a Hadamard space consisting of geodesic rays in \mathcal{H}(V) modulo the asymptotic relations. By [ 15 ] , a norm geodesic ray H=(H_{t})_{t\geq 0} induces a NA norm \chi_{H} on V . This gives an isometry \hat{\mathcal{H}}(V)\xrightarrow{\sim}\mathcal{N}(V) , the target is the space of NA norms on V . We define the Chow-weight M_{X}:\mathcal{N}(V)\rightarrow\mathbb{R} as follows (1.1) M_{X}(\chi)\coloneqq\lim_{t\rightarrow\infty}\frac{1}{t}\mathbf{M}_{X}(H_{t}), where (H_{t}) is any norm geodesic ray such that \chi_{H}=\chi . It is well-defined since \mathbf{M}_{X} is convex and

Read the complete text