Derivation 2: Spectral Tilt from Synchronization Cost Gradient

Derivation 2: Spectral Tilt from Synchronization Cost Gradient#

Superseded by Derivation 4. The cost function approach below produces the correct tilt but always gives wrong-sign running (positive instead of negative). A systematic scan (cost_function_scan.py) proved this is a theorem, not a fitting problem. Derivation 4 replaces the cost function with the devil’s staircase of the circle map, which resolves the running sign and leads to the φ² self-similarity result. See INDEX.md for the full chain.

The universe is not scale-invariant. It almost is — and the almost is the whole story.

Claim#

The CMB spectral tilt n_s ≈ 0.965 encodes the curvature of the synchronization cost function across scales. The 3.5% deviation from perfect scale-invariance is the slope of the cost gradient: larger scales had longer to synchronize and therefore lower net cost, producing more power.

Setup#

The primordial power spectrum#

The observed CMB power spectrum is nearly scale-invariant:

P(k) ∝ k^(n_s - 1)

where k is the comoving wavenumber and n_s = 0.9649 ± 0.0042 (Planck 2018). Perfect scale-invariance would be n_s = 1.

Scale as synchronization time#

A mode with wavenumber k corresponds to a physical scale λ = 2π/k. In the synchronization framework, each mode is a synchronization domain. Larger scales (smaller k) have:

  • More participants in the mean field

  • Longer time to synchronize (exited the Hubble horizon earlier)

  • Lower residual synchronization cost

Define the synchronization time for mode k:

τ_sync(k) = τ_exit(k) - τ_0

where τ_exit(k) is the conformal time at which mode k exits the Hubble horizon and τ_0 is the onset of synchronization (beginning of inflation or the equivalent cost-minimization epoch).

For quasi-exponential expansion: τ_exit(k) ∝ -1/k (conformal time is negative and approaching zero). So:

τ_sync(k) ∝ 1/k

Larger scales (smaller k) have longer synchronization time.

Cost function#

The synchronization cost for a mode that has been synchronizing for time τ is:

C(τ) = C_max × K_half / (K_half + τ)

This is the Michaelis-Menten form: saturating, asymptoting to zero at large τ, with half-maximum cost at τ = K_half.

Properties:

  • C(0) = C_max (unsynchronized: maximum cost)

  • C(K_half) = C_max/2 (half-synchronized)

  • C(τ → ∞) → 0 (fully synchronized: zero residual cost)

Power as inverse cost#

The amplitude of a primordial perturbation mode is determined by how much cost remains when the mode freezes out (re-enters the horizon or decouples). Modes with less residual cost have settled more completely — their amplitude reflects the depth of their cost basin:

P(k) ∝ 1/C(τ_sync(k))

This is the key physical statement: power is inverse cost. Modes that synchronized more completely (lower cost) have larger frozen amplitudes because they explored more of their basin.

Derivation#

Substituting τ_sync(k) = τ_0 / k (where τ_0 absorbs the proportionality constant):

C(k) = C_max × K_half / (K_half + τ_0/k)
     = C_max × K_half × k / (K_half × k + τ_0)

So:

P(k) ∝ 1/C(k) = (K_half × k + τ_0) / (C_max × K_half × k)
     = 1/C_max + τ_0/(C_max × K_half × k)

This has two terms:

  • A constant (scale-invariant) term: 1/C_max

  • A 1/k correction (red tilt): τ_0/(C_max × K_half × k)

For the power spectrum index, take the logarithmic derivative:

n_s - 1 = d ln P / d ln k

Let x = K_half × k / τ_0 (dimensionless wavenumber). Then:

P(k) ∝ (x + 1) / x = 1 + 1/x

d ln P / d ln k = d ln(1 + 1/x) / d ln x
                = (-1/x²) / (1 + 1/x)  ×  1
                = -1 / (x(x + 1))
                = -1 / (x² + x)

Evaluating at the pivot scale#

The spectral tilt is measured at a pivot scale k_* where x_* = K_half × k_* / τ_0. The observed value n_s - 1 ≈ -0.035 requires:

1 / (x_*² + x_*) ≈ 0.035

x_*² + x_* ≈ 28.6

x_* ≈ 4.85

This means the pivot scale k_* has synchronized for about 4.85 half-lives of the cost function. This is a reasonable value — the mode is well into the saturating regime but not fully converged.

Physical interpretation of K_half#

K_half is the synchronization timescale — the time for a mode to reach half its maximum synchronization depth. In the KE (Kuramoto-equation) framework — the mean-field theory of coupled oscillators whose order parameter measures collective synchronization:

K_half = π / (K_c × g(0))

where K_c is the Kuramoto critical coupling and g(0) is the frequency distribution width. This connects the spectral tilt to the same synchronization parameters that produce a₀ and H₀.

Predictions and tensions#

  1. Running of the spectral index: The Michaelis-Menten form predicts:

    dn_s/d ln k = +(2x + 1) / (x² + x)²

    At x_* ≈ 4.85: dn_s/d ln k ≈ +0.013 (positive running)

    Planck 2018 measures: dn_s/d ln k = -0.0045 ± 0.0067

    The sign is positive: n_s approaches 1.0 (scale-invariance) at smaller scales (larger k), because those modes had less time to synchronize and therefore sit closer to the unsynchronized (flat) regime. The Planck central value is negative but consistent with zero at ~1σ.

    This is a genuine tension. The MM cost function predicts that short-wavelength modes should be more scale-invariant than long-wavelength modes. The data weakly prefers the opposite.

    The Hill generalization (cooperative synchronization, n > 1) makes this worse — sharper transitions increase the running magnitude. This rules out simple cooperative extensions.

    What could resolve it: A cost function with a non-monotonic derivative — one that has an inflection point near the pivot scale. Physically: synchronization cost that increases before decreasing, as when modes must first overcome a barrier before settling. This is the cost equivalent of activation energy in chemistry, and it would produce negative running over the range where the cost function is concave-up.

  2. Tensor-to-scalar ratio: In this framework, tensor modes (gravitational waves) are synchronization cost fluctuations of the mean field itself, while scalar modes are cost fluctuations of the participants. The ratio r encodes the coupling strength between these two cost channels. Prediction: r is set by the ratio of mean-field cost to participant cost at the pivot scale.

  3. Scale-dependent tilt: n_s < 1 at all scales (red tilt), with the deviation from scale-invariance increasing at larger scales (smaller k). At k << k_*, the cost function dominates and the spectrum steepens dramatically — consistent with the low-ℓ CMB anomalies (quadrupole suppression). These modes are outside the synchronization horizon: they couldn’t afford to synchronize before last scattering.

Connection to lattice results#

The Stribeck lattice shows differential attenuation: high-frequency modes (ω_d) dissipate while low-frequency modes (ω₀) propagate. This is exactly the spectral tilt mechanism — the cost gradient across frequencies, made visible in a chain of oscillators.

In the lattice:

  • ω_d is the “unsynchronized” mode (high cost, slip regime)

  • ω₀ is the “synchronized” mode (low cost, stick regime)

  • The ratio P(ω₀)/P(ω_d) increases with chain length

This mirrors the CMB: larger scales (more synchronization stages) have more power (lower cost, deeper basin occupation).

Status#

The Michaelis-Menten cost function recovers n_s ≈ 0.965 with one free parameter (the dimensionless pivot x_* ≈ 4.85).

The running prediction (+0.013) is a discriminator: the sign and magnitude are testable. The MM form predicts positive running. If future data confirms negative running, the cost function must have non-monotonic derivative structure (activation barrier). This would be a strong constraint on the cost functional’s form.

See spectral_tilt_numerical.py for the numerical companion.

Open: Derive x_* from the KE parameters rather than fitting it. The connection K_half = π/(K_c × g(0)) provides the route but needs the frequency distribution g(ω) of primordial oscillators.

Open: The positive running is a tension point. What physical mechanism would introduce an activation barrier into the synchronization cost? Candidate: the self-consistency condition itself — modes must first constitute a local mean field before they can synchronize against the global one, and constituting the local field costs energy.

Open: Does the self-consistency condition on the cost functional select the Michaelis-Menten form, or is it one of several consistent choices?

Superseded by Derivation 04#

A systematic scan (cost_function_scan.py — which tested Michaelis-Menten, Hill, stretched-exponential, and power-law cost functions, finding that all monotone forms produce the wrong sign of running) showed that every monotonically decreasing cost function C(τ) with P ∝ 1/C produces positive running. This is structural: ln(1/C(τ₀/k)) is convex in ln k for any monotone C.

The issue is ontological: “cost” carries optimization structure that obscures the mechanics. Derivation 04 (04_spectral_tilt_reframed.md) replaces the cost function with the direct mechanical observable:

P(k) ∝ g(ω) / (Kr)²

where g(ω) is the frequency distribution shaped by mode-locking (Arnold tongues, devil’s staircase). The staircase has both convex and concave regions — the running sign depends on the pivot’s position on the staircase, not on the monotonicity of an envelope.

Numerical exploration (mode_locking_spectrum.py — which computed the power spectrum from the devil’s staircase of the circle map across coupling strengths and identified where tilt and running have the observed signs simultaneously) confirms that regions with simultaneous red tilt and negative running exist just past rational plateaus (e.g., near ω = 2/3).