Living versus staticThe single choice that moves 1/α from 143 to 137

Chapter 11 ended with a formula that almost works. Using the BKT critical coupling K_eff = 2/π and the derived variance ratio, the fine structure constant comes out to one of two values — 1/α = 143 or 1/α = 137, depending on how a single integral is evaluated. The inputs are identical: same coupling, same lattice, same topology, same σ² = 1/2. The only difference is whether the coherence transfer factor is averaged along the RG flow (standard lattice field theory) or evaluated at the attractor where the CLR's dynamics have settled (living lattice). This chapter makes that choice visible.

Two ways to sum coherence

The Debye–Waller factor n(l) = exp(−σ²l) decays with RG scale l. It tells you how much of the nearest-neighbour correlation survives one unit of coarse-graining. In the BKT formula for α, the power-law factor is (π/z)ⁿ — and n has to be a specific value, not a function of l. The question is which value.

Static lattice field theory averages the Debye–Waller factor over the full RG trajectory from the lattice cutoff to the crossover scale. For a trajectory of unit length: n_static = ∫₀¹ exp(−σ²l) dl = (1 − exp(−σ²)) / σ². With σ² = 1/2, that comes out to 0.787.
Living lattice evaluates the Debye–Waller factor at the attractor — the endpoint of the flow, where the CLR's dynamics have frozen K at a fixed point (the PLM Lemma). Only the endpoint matters: n_living = exp(−σ²) = 1/√e ≈ 0.607.

Same K. Same σ². Same lattice. Different choice about when in the trajectory the observable is read. Same formula then spits out two different values of the fine structure constant:

Two calculations of n from the same decay curve n(l) = exp(−σ²l), drawn side by side. Left: the full area under the curve from l = 0 to l = 1 — the standard lattice field theory average. Right: a single thin bar at the endpoint l = 1 — the living-lattice evaluation, which is the only scale where the attractor actually lives. Both calculations feed the same BKT formula for α; the two resulting values of 1/α are displayed below each panel. Drag the σ² slider to see that the gap persists across all couplings — at the physical value σ² = 1/2, the gap is exactly 6 integer units of 1/α.

The numbers at the bottom of the figure are the result of feeding each n into the BKT formula and iterating self-consistently. Static integration gives 1/α = 143.1. Living endpoint gives 1/α = 137.03. One of these numbers matches the measured value of the fine structure constant to 29 parts per million. The other does not match anything. The question is why.

Why the endpoint is the right answer

The PLM Lemma (Chapter 5, formally stated in the paper's Lemma 5.2) is the pivot. It says: on a phase-locked mode of the CLR, every alive bond coupling K_ij converges exponentially to a unique fixed point K*_ij. Not asymptotically, not approximately, but exponentially — the distance to the fixed point shrinks by a constant factor every instant. After a finite number of coarse-graining steps, K is frozen at its attractor to any precision you like.

K(t) → K* with rate λ = −η V″(K*) < 0

Click any coloured symbol to see what it means.

In words The bond coupling follows a damped autonomous ODE — gradient descent on a strictly convex potential V(K). The second derivative V″(K*) is positive at the minimum, so the eigenvalue λ of the linearised dynamics is negative. Exponential convergence is automatic. Observables that depend on K therefore have a well-defined asymptotic value — they are evaluated at K*, not integrated over the approach.

This is the clearest distinction between the living lattice and standard lattice field theory. In the standard framework, K is a free parameter: you fix it by hand, compute observables, and integrate over RG scales to compare to experiment. The trajectory average is meaningful because there is no attractor — nothing in the framework tells you to prefer one scale over another.

On the living lattice, K is a dynamical variable: it evolves under the CLR and settles on an attractor determined by the lattice topology (Chapter 11's BKT wall) and the CLR potential. The trajectory is a transient. The attractor is real physics. Observables that depend on K are evaluated where K lives.

Here is that exponential approach, drawn explicitly:

Upper plot: the bond coupling K(t) under the CLR on a single alive bond, released from various initial values, all converging exponentially to the fixed point K*. The dashed horizontal line is K* = K_bulk = 16/π². The decay constant is τ = 1/|λ|, set by the curvature of the CLR potential at its minimum. Lower plot: the Debye–Waller factor exp(−σ²·K(t)) along each trajectory. Regardless of starting point, all curves collapse onto a single asymptotic value exp(−σ²·K*) — the endpoint. Averaging over t is a well-defined operation, but it is not the quantity that observables couple to; observables see the frozen endpoint.

Interpolating between them

The distinction is not an either/or — it is a continuum. If you averaged the Debye–Waller factor over a window of the RG trajectory of width w ending at l = 1, you would get:

n(w) = 1/w · ∫_1−w¹ exp(−σ²l) dl

Click any coloured symbol to see what it means.

In words A trajectory average over the last w units of the RG flow. At w = 1, this is the full static integral (n_static = 0.787). As w shrinks toward 0, the window collapses onto the endpoint and n(w) → exp(−σ²) (n_living = 0.607). Intermediate values of w correspond to partial-trajectory averages with no physical meaning on either the living or static side; they are only shown here to make the geometry of the collapse visible.

As w shrinks, the window collapses onto the endpoint and n falls from 0.787 to 1/√e. The fine structure constant responds continuously. Here is that collapse, with 1/α updating in real time:

Drag the window width slider. Left: the highlighted region of the exp(−σ²l) curve being averaged — a wide band at w = 1, collapsing to a single line at w = 0. Right: the corresponding value of 1/α. The slider continuously walks the coupling from 143 (static) to 137.03 (living). The physical value of 1/α, measured to 12 significant figures, sits at w = 0 — the endpoint, the attractor, where K actually lives.

Why this is the deepest move in the paper

The hingeEvery other calculation in this paper could in principle be done by someone willing to accept a standard renormalization-group framework and a slightly unusual choice of microscopic model. The living-versus-static distinction is the one place where the framework fundamentally departs from standard lattice field theory. It says: the coupling is not a free parameter; it is an order parameter of a dynamical system with an attractor, and observables that depend on the coupling are evaluated at the attractor. This is why the fine structure constant comes out derived, not fit — and it is why every previous attempt at deriving α from a lattice has failed at precisely the same point.

Standard lattice field theory has no PLM Lemma because it has no CLR. There is nothing inside the theory that picks out a preferred scale at which to evaluate observables. The running of K under the RG is all there is, and the sensible thing to do is integrate. You end up with n = 0.787 and 1/α = 143. This is not wrong — given the framework's assumptions, it is the correct answer. It just does not match experiment.

The living lattice has a dynamical principle that governs K: the Coherence Learning Rule. That principle has fixed points, proven via the convexity of V(K) on alive bonds. The existence of those fixed points is not an empirical observation — it is a mathematical consequence of the CLR's convexity and the topological constraint at the BKT wall. When you evaluate observables at the fixed point, 1/α = 137.03. This also is not an accident. It is what the dynamics produces.

Where this leaves us

The formula now reads:

α = R₀(2/π)⁴ × (π/4)^{1/√e + α/(2π)} ≈ 1/137.032

with every piece accounted for: the UV vertex from Chapter 2's von Mises (Chapter 8's z = 4), the power-law base from Chapter 5's variance ratio, the matching exponent 1/√e from this chapter's attractor evaluation, and the Schwinger self-consistency correction α/(2π). That is the whole story, compressed into one line.

The next chapter unfolds this line into its five factors, each clickable, each traced back to its derivation. This is the chapter where the explorable's shape is completed — every ingredient of α has been built from first principles, and you can navigate from any symbol back to the chapter that proved it.