Why Selection Is Inevitable

The Question

The crystallisation framework introduced a specific functional — F[Φ] = αR + βE − γQ — to model state selection in constrained systems. That functional yields structured attractors, deterministic selection, and testable predictions.

But a natural question follows: is this functional form imposed, or does it arise?

---

Three Minimal Requirements

Consider any system with multiple admissible configurations, constraints on what is allowed, and dynamics that evolve the state over time. The paper identifies three components that a selection process generically requires:

Removing any one of these permits configurations that would not persist under realistic evolution: without R, constraint-violating states pass; without E, energetically unstable states survive; without Q, internally incoherent states are selected.

---

The Functional Emerges

Given these three requirements, the paper shows that selection can be expressed as minimisation of a composite functional:

This is the simplest smooth additive realisation of the required sign structure. The paper does not claim it is unique — nonlinear or higher-order alternatives are possible — but it is the minimal form consistent with the structural requirements.

---

Attractors Are Inevitable

Given a selection functional satisfying these properties, consider gradient flow — the system evolving along the steepest descent of F:

The paper proves two theorems:

• Monotonic Descent. F is monotonically non-increasing along trajectories. This follows directly from the chain rule — dF/dt = −||∇F||² ≤ 0.
• Existence of Attractors. If F is continuous and coercive on a closed, bounded admissible set, then F attains its infimum, trajectories converge to critical points, and local minima with positive-definite Hessian are Lyapunov-stable attractors.

---

Deterministic vs Stochastic Selection

The analysis shows that deterministic gradient flow is sufficient for configuration selection. Stochastic mechanisms (GRW-type random collapse) and threshold-based triggers (Penrose objective reduction) are alternative implementations, but they are not required at the structural level.

The relationship is analogous to Hamiltonian mechanics and statistical mechanics: deterministic dynamics at the microscopic level can produce statistical behaviour at the macroscopic level without requiring randomness as a fundamental ingredient.

---

Connection to the Crystallisation Framework

The crystallisation model introduced in the main paper is one concrete instantiation of this general principle:

• R[Φ] is defined as the squared deviation from an admissible constraint target
• E[Φ] is the expectation value of a coupling operator
• Q[Φ] measures phase alignment across modes
• Dynamics follow gradient descent on the composite F
• Stable configurations correspond to attractors of the flow

The present paper does not derive the crystallisation model. It shows that the model's structure — a balance of constraint satisfaction, cost minimisation, and coherence maximisation — aligns with the minimal requirements for selection in any constrained dynamical system.

---

What This Does Not Claim

• A derivation of the selection functional from specific physical theory — the analysis is at the level of abstract dynamical systems
• That the functional form is unique — it is minimal within the class of smooth additive models
• That the weights α, β, γ are determined — they remain free parameters dependent on the system
• That this mechanism is realised in nature — whether crystallisation produces observable effects remains empirical
• Extension to infinite-dimensional Hilbert spaces — theorems are stated for finite-dimensional configuration spaces

---