复向量丛上的度量与结构群：规范理论的数学视角探究

Great question—let's break this down systematically, since connecting Yang-Mills theory to complex vector bundles and algebraic geometry requires grounding both the Lie group/functional analysis side and the geometric vector bundle perspective.

First, let's anchor the core Yang-Mills setup before shifting to complex geometry:

• For a principal $G$-bundle $P \to M$ over a smooth manifold $M$, a connection $A$ on $P$ induces a curvature 2-form $F_A \in \Omega^2(M, \text{ad}(P))$, where $\text{ad}(P)$ is the adjoint bundle associated to the Lie algebra $\mathfrak{g}$ of $G$.
• The Yang-Mills functional is defined as:

  YM(A) = \int_M |F_A|^2 \, \text{dvol}
  

  Compactness of $G$ is critical here: compact Lie groups admit a bi-invariant Haar measure, which lets us define a $G$-invariant inner product on $\mathfrak{g}$. This inner product extends to a pointwise inner product on $\text{ad}(P)$, making the norm $|F_A|^2$ well-defined and gauge-invariant (invariant under transformations $g: M \to G$ that act on connections via $A \mapsto gAg^{-1} + g , dg^{-1}$). Without compactness, we lose this invariant inner product, and the functional fails to be gauge-invariant.

When moving to algebraic geometry, we focus on complex vector bundles $E \to X$ over a complex manifold (or algebraic variety) $X$. The structure group of $E$ is initially $\text{GL}(n, \mathbb{C})$, but we care about reducing this group to smaller subgroups—this is where metric-related questions come into play:

• A Hermitian metric $h$ on $E$ corresponds exactly to a reduction of the structure group from $\text{GL}(n, \mathbb{C})$ to $\text{U}(n)$ (the unitary group), since $\text{U}(n)$ is the subgroup of $\text{GL}(n, \mathbb{C})$ that preserves the Hermitian inner product.
• A structure group is reducible if it can be further reduced to a product subgroup like $\text{U}(k) \times \text{U}(n-k)$, which corresponds to $E$ decomposing as a direct sum $E = E_1 \oplus E_2$ (where $E_1$ has rank $k$, $E_2$ has rank $n-k$). More generally, reducible groups correspond to $E$ splitting into a direct sum of subbundles.

Here’s where the algebraic geometry connection becomes concrete, focusing on metrics and their relationship to bundle stability and curvature:

• Chern Connections and Curvature: For a Hermitian vector bundle $(E, h)$, there’s a unique connection $\nabla_h$ (the Chern connection) that is compatible with both the complex structure of $E$ and the metric $h$. Its curvature $F_{\nabla_h}$ is a Hermitian anti-symmetric 2-form ($F_{\nabla_h}^* = -F_{\nabla_h}$). If $E$ splits as $E_1 \oplus E_2$ with metric $h = h_1 \oplus h_2$, the Chern connection decomposes as $\nabla_{h_1} \oplus \nabla_{h_2}$, and the curvature becomes block-diagonal with no cross terms between $E_1$ and $E_2$.
• Hermitian-Einstein Metrics and Stability: A key bridge to algebraic geometry is the Donaldson-Uhlenbeck-Yau (DUY) Theorem, which states that a holomorphic vector bundle over a compact Kähler manifold admits a Hermitian-Einstein metric (a metric where the curvature satisfies $\Lambda F_{\nabla_h} = \lambda \cdot \text{Id}_E$ for some constant $\lambda$) if and only if the bundle is polystable—meaning it’s a direct sum of stable bundles with the same slope (slope = $\text{deg}(E)/\text{rank}(E)$, defined via Chern classes). For reducible bundles (direct sums), this means each summand must be stable and have equal slope to admit such a metric.
• Metric Deformations for Reducible Bundles: If $E$ has a reducible structure group, the space of Hermitian metrics preserving the direct sum decomposition is isomorphic to the product of Hermitian metric spaces on each summand. However, if you allow metrics that don’t preserve the decomposition, you can deform the bundle’s structure—this ties into questions of bundle moduli spaces, a core topic in algebraic geometry.

To tie these threads together, focus on:

• Chern Classes: Topological invariants of complex vector bundles, defined via curvature integrals:

  c_k(E) = \frac{1}{(2\pi i)^k} \int_X \text{tr}(\wedge^k F_{\nabla_h})
  

  For reducible bundles, $c(E_1 \oplus E_2) = c(E_1) \cdot c(E_2)$ (where $c(E)$ is the total Chern class), linking bundle decompositions to topological invariants.
• Invariant Theory: Compact Lie group representation theory tells us how structure group reductions correspond to bundle decompositions—for example, $\text{U}(n)$ representations decompose into irreducible components exactly when the vector bundle splits into stable summands.
• Elliptic PDEs: Yang-Mills and Hermitian-Einstein equations are elliptic partial differential equations; their solvability (and the existence of metrics) is deeply tied to the algebraic geometry of the base manifold and the bundle’s stability properties.

内容的提问来源于stack exchange，提问作者Benighted