Closure Dynamics: Gradient Flow on the Constraint Landscape

Start Here

Paper XII Closure Dynamics (PDF) Paper XI Correspondence Theorems Paper VII Closure Dynamics & Constraint Operators Repository GitHub — vfd-crystallisation

Epistemic status: minimal dynamics — gradient flow and variational extension. The evolution parameter t is NOT physical time. No relativistic invariance, no QFT, no scattering amplitudes.

From Structure to Dynamics

Papers V–XI were structural: they specified which states are admissible, what properties they carry, how they correspond to observed particles. But no paper addressed evolution — how a state moves through configuration space over time.

This paper introduces the simplest dynamical extension compatible with the existing framework: gradient flow of the closure functional F. States descend F, constraint satisfaction increases, and the flow terminates at closure-stable fixed points.

The Gradient Flow

The dynamical law is projected gradient descent on the closure functional, constrained to remain within the admissible domain:

dψ/dt = −Π_{S_adm}(∇F(ψ))

Projected gradient flow. The state descends F while remaining within the admissible domain S_adm.

The projection Π ensures the trajectory never leaves the admissible set. The gradient ∇F provides the direction of steepest constraint violation reduction. Together they define a first-order dynamical system on S_adm.

Monotonic Descent

The central result of the gradient flow formulation: F cannot increase along any trajectory.

dF/dt = −‖Π(∇F)‖² ≤ 0

The closure functional decreases monotonically. Equality holds only at fixed points where the projected gradient vanishes.

Monotonic Descent

The closure functional decreases monotonically along every trajectory. Closure-stable states (F = 0) are fixed points. The flow is irreversible relaxation toward constraint satisfaction.

Fixed Points and Stability

The fixed points of the gradient flow are exactly the elements of S_cl — the closed state set. At these points the projected gradient vanishes and the state is stationary.

Stability is determined by the constrained Hessian H(ψ) of F evaluated at each fixed point. If H is positive semi-definite, the fixed point is a stable attractor: nearby perturbations flow back. If H is indefinite, the fixed point is an unstable saddle: perturbations along negative-curvature directions carry the state away.

Attractor Basins

Each stable fixed point has a basin of attraction: the set of all initial configurations whose gradient-flow trajectories converge to that fixed point. Distinct basins correspond to distinct particle classes.

The particle type is determined by which attractor the initial configuration flows toward. Classification is not imposed by hand — it emerges from the topology of the flow. Two initial states in the same basin are the same particle; two in different basins are different particles.

Confinement in Dynamics

Disconnected configurations — states with support on non-adjacent shells — have F > 0 due to the gap penalty in the closure functional. Under gradient flow they are not fixed points: the projected gradient is nonzero and the state evolves.

These configurations flow toward connected composites where the gap penalty vanishes. Isolation is dynamically unstable. Binding is the gradient’s destination.

Confinement as Gradient

Confinement is not a force — it is the gradient of the constraint functional. Disconnected configurations sit on a slope; the flow carries them toward connected composites where the functional reaches zero.

Two Modes of Motion

Motion decomposes into two geometrically distinct modes relative to the constraint manifold M_cl:

Transverse motion — toward or away from M_cl. Associated with the closure residual and therefore with mass. Gradient flow governs this mode: it drives states toward the manifold.
Tangential motion — along M_cl. Associated with internal geometry, geodesic-like motion on the constraint surface. The gradient flow is silent here: F is constant along M_cl.

Gradient flow handles the transverse mode. A Lagrangian formulation is needed for the tangential mode.

The Lagrangian Extension

The gradient flow is first-order and dissipative. A second dynamical formulation extends the framework to conservative, time-reversible evolution:

L(ψ, ψ̇) = ½‖ψ̇‖² − F(ψ)

Minimal Lagrangian: kinetic energy minus the closure functional as potential.

The Euler–Lagrange equation gives second-order dynamics:

ψ̈ = −∇F(ψ)

Second-order, conservative, time-reversible. States oscillate around closure-stable configurations rather than dissipating toward them.

Both formulations share the same fixed points but differ in their approach dynamics:

Property	Gradient Flow	Lagrangian
Order	First-order	Second-order
Character	Dissipative	Conservative
Approach to fixed points	Monotonic relaxation	Oscillation
Time reversal	Irreversible	Reversible
Fixed points	S_cl	S_cl

Stated Limitations

The evolution parameter t is not physical time — it parameterises constraint relaxation, not spacetime dynamics
No relativistic invariance — the flow is defined on an abstract configuration space without Lorentz structure
No quantum dynamics — no path integral, no superposition of trajectories, no measurement theory
No scattering or decay amplitudes — interactions between distinct trajectories are not modelled
No coupling constants — the flow has no free parameters beyond the closure functional itself
The Lagrangian is minimal — no gauge fields, no fermion kinetic terms, no symmetry-breaking potential
Basin structure is not fully characterised — the number and topology of attractor basins remain open

The framework breathes: states evolve, attractors stabilise, disconnected configurations dissolve.

Paper open-access. Part of the crystallisation programme at github.com/vfd-org/vfd-crystallisation.