Spontaneous quantization of the Yang–Mills gradient flow

We formulate a nonsingular loop–space calculus for the Yang–Mills (YM) gradient flow directly in terms of Wilson loop functionals, rather than the underlying gauge fields. All variations act within the manifold of smooth loops via “dot derivatives” that are finite, parametrization–invariant, and free of cusp or backtracking singularities. This yields a closed linear diffusion equation in loop space for Wilson loops. The associated loop operator is universal and trilinear in functional derivatives, and the formulation automatically factors out gauge transformations, exposing the gauge–invariant core of the flow. The construction is valid for any (Abelian or non–Abelian) gauge group.We identify two distinct classes of exact solutions. First, a self–dual (Hodge–dual) matrix–valued minimal surface whose area functional, when exponentiated, solves the fixed–point loop equation exactly, without contact terms or ambiguities; for planar loops the dual area equals 22 times the Euclidean minimal area, providing a geometrically grounded confinement mechanism. We also prove that the ordinary minimal surface in R4 fails to satisfy the fixed–point loop equation, due to a singular nonvanishing contribution from the loop operator.Second, a decaying–flow solution in which the momentum loop executes a periodic random walk on regular star polygons (the “Euler ensemble” known from Navier–Stokes turbulence). Both classes realize spontaneous quantization: the self–dual solution furnishes a stationary quantized state (a fixed manifold of the flow), while the decaying solution describes a quantized trajectory that approaches a pure–gauge vacuum along a universal path.Thus we obtain exact solutions of the Wilson–loop evolution in YM gradient flow. We discuss the emergence of quantum–like Wilson–loop statistics from deterministic classical dynamics, potential implications for confinement in QCD, and the role of these fixed manifolds and trajectories as attractors in the space of YM gradient–flow solutions.