Skip to main content

Part 4· Chapter 13

Chapter 13 — Yang–Mills Flow

Prerequisites: Chapters 2--8 (Part I), Chapter 9 (Theorem H).


13.1 Motivation

In Chapters 2--8, the world Wt\mathfrak{W}_t was analysed at a frozen instant. The question now is: how does the gauge connection evolve over time?

The discrete Yang--Mills gradient flow provides the answer. Starting from an arbitrary connection, the flow minimises curvature energy, automatically stabilising the world's essential structure.


13.2 Discrete Yang--Mills Energy

Definition 13.1 (Discrete YM energy)

Let (V,E,T)(V,E,\mathcal{T}) be a graph with triangle set T\mathcal{T} (all 3-cliques). For a GG-valued connection A={gij}(i,j)EA=\{g_{ij}\}_{(i,j)\in E}, the holonomy of triangle =(i,j,k)\triangle=(i,j,k) is:

Ω(A):=gijgjkgkiG.\Omega_\triangle(A):=g_{ij}\cdot g_{jk}\cdot g_{ki}\in G.

The discrete Yang--Mills energy:

E(A):=TdG(Ω(A),e)2.\mathcal{E}(A):=\sum_{\triangle\in\mathcal{T}}d_G(\Omega_\triangle(A),e)^2.

Properties.

  • E(A)=0    \mathcal{E}(A)=0\iff all holonomies are trivial     \iff the connection is flat.
  • E(A)>0    \mathcal{E}(A)>0\iff non-trivial curvature     \iff topological obstruction.
  • E\mathcal{E} is gauge-invariant (Theorem 13.4 below).

13.3 Gradient Flow

Definition 13.2 (Discrete YM gradient flow)

On GEG^E with the product Riemannian structure (i,j),g\sum_{(i,j)}\langle\cdot,\cdot\rangle_{\mathfrak{g}}:

dAsds=gradE(As),\frac{dA_s}{ds}=-\mathrm{grad}\,\mathcal{E}(A_s),

i.e., for each edge (i,j)(i,j):

dgij(s)ds=Egijg.\frac{dg_{ij}(s)}{ds}=-\frac{\partial\mathcal{E}}{\partial g_{ij}}\bigg|_{\mathfrak{g}}.

Theorem 13.1 (Monotone energy decrease)

Along the gradient flow: ddsE(As)=gradE(As)20\frac{d}{ds}\mathcal{E}(A_s)=-\|\mathrm{grad}\,\mathcal{E}(A_s)\|^2\le 0.

Proof. Chain rule: ddsE=gradE,A˙s=gradE2\frac{d}{ds}\mathcal{E}=\langle\mathrm{grad}\,\mathcal{E},\dot A_s\rangle=-\|\mathrm{grad}\,\mathcal{E}\|^2. \square

Since E0\mathcal{E}\ge0, the limit E:=limsE(As)\mathcal{E}^*:=\lim_{s\to\infty}\mathcal{E}(A_s) exists.


13.4 Fixed Points and Stability

Theorem 13.2 (Fixed-point characterisation)

A fixed point AA^* of the flow satisfies gradE(A)=0\mathrm{grad}\,\mathcal{E}(A^*)=0 (critical point). If additionally Hess(E)A\mathrm{Hess}(\mathcal{E})|_{A^*} is positive-definite on the normal space to the gauge orbit, then AA^* is Lyapunov stable.

Theorem 13.3 (Connection to [A][A_\infty])

The gauge class [A][A_\infty] from Construction 7.6 is a stable fixed point of the gradient flow of EF\mathcal{E}_{F^\circ} (flattening energy on the fruit kernel). This follows from Theorem H (Chapter 9).


13.5 Gauge Invariance

Theorem 13.4 (Gauge invariance of YM energy)

For any gauge transformation h:VGh:V\to G: E(Ah)=E(A)\mathcal{E}(A^h)=\mathcal{E}(A).

Proof. Ωh=h(i)Ωh(i)1\Omega_\triangle^h=h(i)\,\Omega_\triangle\,h(i)^{-1}. By bi-invariance of dGd_G: dG(Ωh,e)=dG(Ω,e)d_G(\Omega^h,e)=d_G(\Omega,e). \square

Corollary. The gradient flow is gauge-equivariant: dds(Ah)s=(ddsAs)h\frac{d}{ds}(A^h)_s=(\frac{d}{ds}A_s)^h.


13.6 The U(1)U(1) Case

Theorem 13.5 (U(1)U(1) convergence and uniqueness)

For G=U(1)G=U(1):

  1. The energy E\mathcal{E} is convex on the torus (S1)E(S^1)^{|E|} (each term 2(1cosθ)2(1-\cos\theta) is convex in θ(π,π)\theta\in(-\pi,\pi)).
  2. The global minimum is unique (up to gauge).
  3. The flow converges exponentially to [A][A_\infty] from any initial condition.

Proof. Convexity follows from d2dθ2(1cosθ)=cosθ>0\frac{d^2}{d\theta^2}(1-\cos\theta)=\cos\theta>0 for θ<π/2|\theta|<\pi/2 (near the minimum). The strict convexity of the total energy on the connected quotient space gives uniqueness. Exponential convergence follows from the positive spectral gap of the Hessian (Theorem 7.11). \square


13.7 Non-Abelian Difficulties

SU(2)SU(2) non-convexity

For G=SU(2)G=SU(2), the geodesic distance dSU(2)(g,h)=arccos(tr(gh)/2)d_{SU(2)}(g,h)=\arccos(|\mathrm{tr}(g^\dagger h)|/2) is non-convex on SU(2)ESU(2)^{|E|}:

  • Multiple local minima may exist.
  • Not every critical point is stable.
  • Initial conditions may lead to different limit points.

This is the core reason Conjecture 7.10 (gauge-fixing uniqueness) remains open for non-abelian GG.

Gribov ambiguity

For G=SU(2)G=SU(2), the space of gauge-inequivalent connections modulo gauge may have non-trivial topology (Gribov copies). This complicates:

  • Gauge fixing: no single global gauge choice may work.
  • Counting fixed points.
  • Global properties of the moduli space.

Workaround: compute gauge-invariant quantities (Axes 1--3) directly, bypassing gauge fixing.


13.8 Integration with the Three Axes

After the flow converges to AA_\infty:

  • Axis 1: H(Kα;A)H^*(K_\alpha;A) computed on the stabilised connection.
  • Axis 2: H(Kα,Lα;A)H^*(K_\alpha,L_\alpha;A) reflects the role of doors in the equilibrium state.
  • Axis 3: Barcode of Σ\Sigma reflects the final fractal structure.

The three axes read the world's stable properties---those that survive the dynamical relaxation.