Stele IV · 1986 · with K. Uhlenbeck

唐纳森–乌伦贝克–丘定理Donaldson–Uhlenbeck–Yau

代数里的"稳定",等于分析里的"最优联络"Algebraic stability equals the existence of an optimal connection
原文Paper · Comm. Pure Appl. Math. 39 (1986) 阅读约 20 分钟~20 min read
TL;DR

在一个复几何空间上,有一种叫"全纯向量丛"的对象:在每一点上挂一个复向量空间,并且挂得很光滑。代数几何学家用一个纯代数的条件来判断它好不好,这个条件叫斜率稳定。规范场论和微分几何学家则问另一个问题:能不能给它配一个最"平衡"的度量,即厄米–爱因斯坦度量?

乌伦贝克与丘成桐在 1986 年证明:在任意紧凯勒流形上,这两件事是等价的。唐纳森此前对代数曲面、此后对射影流形也证明了这一点,所以这个定理以三人的名字命名。

On a complex manifold there are objects called holomorphic vector bundles — a complex vector space attached smoothly to every point. Algebraic geometers judge whether one is "good" by a purely algebraic test: slope stability. Gauge theorists and differential geometers ask a different question: can it carry the most "balanced" metric, a Hermitian–Einstein metric?

In 1986 Uhlenbeck and Yau proved that on every compact Kähler manifold, these two are equivalent. Donaldson had proved it for algebraic surfaces before, and for projective manifolds after — hence the three names.

1965
Narasimhan–Seshadri:曲线情形
Narasimhan–Seshadri: curves
1985
Donaldson:代数曲面
Donaldson: algebraic surfaces
1986
Uhlenbeck–Yau:任意紧凯勒流形
Uhlenbeck–Yau: all compact Kähler manifolds
1987
Donaldson:射影流形,另一种方法
Donaldson: projective, another method
Chapter 1

向量丛、联络与曲率Bundles, connections, curvature

想象地球表面的每一点上都插着一面小旗,可以朝任意切方向摆动。所有这些"可能的方向"拼在一起,就是一个向量丛。要比较相邻两点上的向量,需要一条规则,叫联络。沿一个小圈平移一圈回来,向量会转过一个角度,这就是曲率 $F$。

Imagine a little flag planted at every point on the Earth's surface, free to swing in any tangent direction. All those possible directions together form a vector bundle. To compare vectors at neighbouring points you need a rule — a connection. Transport a vector around a small loop and it comes back rotated: that is curvature $F$.

物理学家对这个画面很熟悉:电磁场就是一个 $U(1)$ 丛上的联络,场强就是曲率;描述强相互作用和弱相互作用的杨–米尔斯理论用的是非交换的丛。

Physicists know this picture well: electromagnetism is a connection on a $U(1)$ bundle, with field strength as curvature; Yang–Mills theories of the strong and weak forces use non-abelian bundles.

在复流形上,全纯向量丛 $E$ 上选定一个厄米度量 $h$,就自动确定了一个与复结构相容的联络(陈联络),以及它的曲率 $F_h$。问题变成:在所有 $h$ 里,有没有一个"最好"的?

On a complex manifold, choosing a Hermitian metric $h$ on a holomorphic vector bundle $E$ automatically determines a compatible connection (the Chern connection) and its curvature $F_h$. The question becomes: among all $h$, is there a "best" one?

Chapter 2

两种"好"Two notions of "good"

分析的"好":厄米–爱因斯坦度量Analytic: the Hermitian–Einstein metric

设底空间是紧凯勒流形 $(M,\omega)$。用 $\omega$ 对曲率做"迹"(记作 $\Lambda$),要求结果是单位矩阵的常数倍:

Let the base be a compact Kähler manifold $(M,\omega)$. Contract the curvature with $\omega$ (written $\Lambda$) and require the result to be a constant multiple of the identity:

$$\sqrt{-1}\,\Lambda_\omega F_h=\lambda\cdot\mathrm{Id}_E .$$

这就是厄米–爱因斯坦方程,也叫厄米–杨–米尔斯方程。常数 $\lambda$ 由拓扑决定。满足它的联络是杨–米尔斯泛函的极小值点,在物理上对应"能量最低"的规范场。

This is the Hermitian–Einstein equation (also called Hermitian–Yang–Mills). The constant $\lambda$ is fixed by topology. Solutions minimize the Yang–Mills functional — physically, the lowest-energy gauge fields.

代数的"好":斜率稳定Algebraic: slope stability

定义丛的度数 $\deg E=\int_M c_1(E)\wedge\omega^{n-1}$ 和斜率 $\mu(E)=\deg E/\operatorname{rk}E$。称 $E$ 是稳定的,如果它的每个真子层 $\mathcal F$($0<\operatorname{rk}\mathcal F<\operatorname{rk}E$)都满足

Define the degree $\deg E=\int_M c_1(E)\wedge\omega^{n-1}$ and slope $\mu(E)=\deg E/\operatorname{rk}E$. $E$ is stable if every proper coherent subsheaf $\mathcal F$ ($0<\operatorname{rk}\mathcal F<\operatorname{rk}E$) satisfies

$$\mu(\mathcal F)<\mu(E).$$

若 $E$ 是若干斜率相同的稳定丛的直和,则称为多稳定。

$E$ is polystable if it is a direct sum of stable bundles of the same slope.

◈

打个比方:一个团队是"稳定"的,如果没有哪个小团体的平均战斗力高于全队平均。只要存在这样的小团体,团队就有分裂的倾向。

An analogy: a team is "stable" if no sub-group has a higher average strength than the whole team. If such a sub-group exists, the team tends to split.

Chapter 3

定理The theorem

唐纳森–乌伦贝克–丘定理Donaldson–Uhlenbeck–Yau Theorem

设 $E$ 是紧凯勒流形 $(M,\omega)$ 上的全纯向量丛。则 $E$ 上存在厄米–爱因斯坦度量,当且仅当 $E$ 是多稳定的。若 $E$ 稳定,这样的度量在相差一个正常数倍的意义下唯一。

Let $E$ be a holomorphic vector bundle over a compact Kähler manifold $(M,\omega)$. Then $E$ admits a Hermitian–Einstein metric if and only if $E$ is polystable. If $E$ is stable, the metric is unique up to a positive constant.

"存在厄米–爱因斯坦度量 ⇒ 多稳定"这一方向比较容易,由小林昭七和 Lübke 证明。困难的是反方向:从代数的稳定性出发,解出一个非线性偏微分方程。这种"代数稳定 ⇔ 存在典范度量"的对应,现在统称为 Kobayashi–Hitchin 对应。

The direction "Hermitian–Einstein ⇒ polystable" is easier (Kobayashi, Lübke). The hard direction is the converse: starting from algebraic stability, solve a nonlinear PDE. Such "algebraic stability ⇔ canonical metric" correspondences are now called Kobayashi–Hitchin correspondences.

Chapter 4

乌伦贝克–丘的证明The Uhlenbeck–Yau proof

核心想法是一个二择一:要么方程可解,要么解的失败会自动"吐出"一个破坏稳定性的子层。

The central idea is a dichotomy: either the equation is solvable, or its failure automatically "spits out" a destabilizing subsheaf.

第一步:扰动方程Step 1: a perturbed equation

固定一个背景度量 $h_0$,把未知度量写成 $h=h_0e^{s}$。对 $\varepsilon>0$,考虑扰动方程

Fix a background metric $h_0$ and write the unknown as $h=h_0e^{s}$. For $\varepsilon>0$ consider the perturbed equation

$$\sqrt{-1}\,\Lambda F_{h_\varepsilon}-\lambda\,\mathrm{Id}=-\varepsilon\,\log\!\big(h_0^{-1}h_\varepsilon\big).$$

右边的项提供了额外的"恢复力",使得对每个 $\varepsilon>0$ 都能用连续性方法解出 $h_\varepsilon$。原方程对应 $\varepsilon\to0$ 的极限。

The right-hand term provides an extra "restoring force", so for every $\varepsilon>0$ the continuity method produces a solution $h_\varepsilon$. The original equation is the limit $\varepsilon\to0$.

第二步:要么有界,要么爆破Step 2: bounded, or blowing up

如果 $\|\log(h_0^{-1}h_\varepsilon)\|$ 在 $\varepsilon\to0$ 时保持有界,椭圆估计就能让 $h_\varepsilon$ 收敛到厄米–爱因斯坦度量,证明结束。

If $\|\log(h_0^{-1}h_\varepsilon)\|$ stays bounded as $\varepsilon\to0$, elliptic estimates make $h_\varepsilon$ converge to a Hermitian–Einstein metric, and we are done.

如果它趋于无穷,就把 $s_\varepsilon=\log(h_0^{-1}h_\varepsilon)$ 除以它的大小重新归一化。在弱极限里,可以从中提取出一个算子 $\pi$,满足

If it blows up, renormalize $s_\varepsilon=\log(h_0^{-1}h_\varepsilon)$ by its size. In a weak limit one extracts an operator $\pi$ with

$$\pi=\pi^*=\pi^2,\qquad(\mathrm{Id}-\pi)\,\bar\partial\pi=0,\qquad\pi\in L^2_1 .$$

这是一个弱全纯子丛:一个只在索伯列夫意义下光滑的投影,但它"想要"是全纯子丛。

This is a weakly holomorphic subbundle: a projection that is smooth only in a Sobolev sense, yet "wants" to be a holomorphic subbundle.

第三步:正则性定理(论文的心脏)Step 3: the regularity theorem — the heart of the paper

乌伦贝克与丘证明:任何弱全纯子丛,在一个余维至少为 2 的解析集之外都是光滑的,并且定义了 $E$ 的一个凝聚子层 $\mathcal F$。再通过陈–韦伊公式计算,可以发现 $\mu(\mathcal F)\ge\mu(E)$,与稳定性矛盾。所以爆破不可能发生。

Uhlenbeck and Yau prove: every weakly holomorphic subbundle is smooth outside an analytic set of codimension at least 2 and defines a coherent subsheaf $\mathcal F\subset E$. A Chern–Weil computation then shows $\mu(\mathcal F)\ge\mu(E)$, contradicting stability. So blow-up cannot occur.

Why Kähler, not just projective

唐纳森 1987 年的方法依赖射影嵌入,即需要底空间是代数的。乌伦贝克–丘的正则性定理是纯解析的,因此对一般的紧凯勒流形也成立,包括那些不是代数簇的。

Donaldson's 1987 method uses a projective embedding, i.e. needs an algebraic base. The Uhlenbeck–Yau regularity theorem is purely analytic, so it works on every compact Kähler manifold, including non-algebraic ones.

Chapter 5

影响Legacy

  • 陈类不等式:稳定丛自动满足 Bogomolov–Lübke 不等式 $\int\big(2r\,c_2-(r-1)c_1^2\big)\wedge\omega^{n-2}\ge0$,这为构造或排除向量丛提供了约束。
  • Chern class inequality: stable bundles automatically satisfy the Bogomolov–Lübke inequality $\int\big(2r\,c_2-(r-1)c_1^2\big)\wedge\omega^{n-2}\ge0$, constraining which bundles can exist.
  • 弦论:杂化弦在卡拉比–丘流形上的紧化,需要满足厄米–杨–米尔斯方程的规范丛。DUY 定理把这个物理要求翻译成可以用代数几何检验的稳定性条件。
  • String theory: heterotic compactifications on Calabi–Yau manifolds need gauge bundles solving the Hermitian–Yang–Mills equations. DUY turns that physical requirement into stability, checkable by algebraic geometry.
  • 一种范式:"稳定 ⇔ 典范度量"这一思想后来推广到希格斯丛(Hitchin、Simpson),也启发了丘成桐–田刚–唐纳森猜想:法诺流形上凯勒–爱因斯坦度量的存在性等价于 K-稳定性,这一猜想已于 2015 年得到证明(陈秀雄–唐纳森–孙崧;田刚)。
  • A paradigm: "stability ⇔ canonical metric" later extended to Higgs bundles (Hitchin, Simpson) and inspired the Yau–Tian–Donaldson conjecture — Kähler–Einstein metrics on Fano manifolds exist iff K-stability holds — proved in 2015 (Chen–Donaldson–Sun; Tian).
References

原文与延伸阅读Original papers & further reading

  • 1986K. Uhlenbeck, S.-T. Yau, On the existence of Hermitian–Yang–Mills connections in stable vector bundles, Comm. Pure Appl. Math. 39(S1), S257–S293. DOI ↗
  • 1985S. K. Donaldson, Anti self-dual Yang–Mills connections over complex algebraic surfaces and stable vector bundles, Proc. London Math. Soc. 50, 1–26.
  • 1987S. K. Donaldson, Infinite determinants, stable bundles and curvature, Duke Math. J. 54, 231–247.
  • 1965M. S. Narasimhan, C. S. Seshadri, Stable and unitary vector bundles on a compact Riemann surface, Ann. of Math. 82, 540–567.