CMB Damping Tail: Silk Diffusion and the High-Multipole Fall-Off

What the Damping Tail Is

The CMB temperature power spectrum, C_ℓ, is a sequence of acoustic peaks — snapshots of sound waves frozen into the baryon-photon plasma at recombination (z ≈ 1090, about 380,000 years after the Big Bang). The peaks would in principle continue to arbitrarily high multipole, but instead their amplitude decays under a smooth exponential envelope starting near ℓ ≈ 1000. That envelope is the damping tail.

Physically, the fluid was not a perfect fluid. Photons were tightly but not infinitely coupled to electrons (and, through them, protons) by Thomson scattering. On small scales the coupling leaked: a photon could diffuse across a fluctuation before recombination completed. Hot photons wandered into cold regions and vice versa, and the temperature contrast washed out.

• Large scales (ℓ ≲ 800): fluctuations survive intact — the acoustic peaks.
• Small scales (ℓ ≳ 1000): fluctuations are diffusively erased — the damping tail.
• The transition is set by the photon diffusion length at last scattering.

The Mechanism: A Random Walk Out of the Spots

Before recombination the photon mean free path was λ_C = 1/(n_e σ_T a), where n_e is the free-electron density, σ_T ≈ 6.65×10⁻²⁵ cm² is the Thomson cross-section, and a is the scale factor. A photon takes N ≈ η/λ_C scatterings in conformal time η, and by random-walk statistics its net displacement grows only as the square root of the number of steps.

This gives the governing scaling:

• λ_D ≈ √N · λ_C ≈ √(η · λ_C) — the diffusion length is the geometric mean of the horizon scale η and the tiny mean free path λ_C.
• More carefully, k_D⁻² = ∫ dη / (6 a n_e σ_T) × [R²/(1+R) + 8/9]/(1+R), where R = 3ρ_b/4ρ_γ is the baryon-to-photon momentum ratio; the R terms encode shear viscosity and heat conduction in the imperfect fluid.

Any mode with wavelength shorter than λ_D is suppressed. The net effect on the spectrum is an exponential damping factor of roughly exp[−(k/k_D)²] on the transfer function, which in harmonic space becomes exp[−(ℓ/ℓ_D)²]-like suppression of C_ℓ.

Key Quantities and a Worked Estimate

At recombination the free-electron fraction is dropping fast, so λ_C is lengthening and the diffusion integral piles up right at last scattering. Characteristic numbers:

• Damping wavenumber: k_D ≈ 0.14–0.2 Mpc⁻¹ (comoving) at z ≈ 1090–1200.
• Diffusion length: λ_D ≈ 1/k_D ≈ a few comoving Mpc (often quoted as ~3–7 Mpc; sub-degree on the sky).
• Damping multipole: ℓ_D ≈ k_D · D_A, where D_A ≈ 13.9 Gpc is the comoving distance to last scattering; this puts the onset near ℓ ~ 1300–1500.

Worked estimate. Take λ_C ≈ 2.5 Mpc comoving near recombination and a conformal horizon η* ≈ 280 Mpc. Then λ_D ≈ √(η*·λ_C) ≈ √(280 × 2.5) ≈ √700 ≈ 26 Mpc as a crude upper bound; the proper coefficient (the 1/√6 and the (1+R) weighting, plus averaging over the visibility function) pulls the effective damping scale down to the few-Mpc regime and the observed ℓ_D. The lesson: the ratio r_s/λ_D ≈ 147/λ_D sets how many sharp acoustic peaks you get before diffusion smooths them out.

How It Is Observed

The damping tail is a small-angle signal, so it needs high-resolution, low-noise experiments. The history tracks the resolution race:

• Early detections: BOOMERANG, CBI, and ACBAR (2000s) first traced power falling off past the third peak.
• WMAP (2003–2012) measured the first few peaks but ran out of resolution near ℓ ~ 800.
• South Pole Telescope (SPT): the 2011 SPT and 2013 SPT-SZ analyses delivered landmark damping-tail measurements over 650 < ℓ < 3000. ACT (Atacama Cosmology Telescope) reaches ℓ up to ~10,000.
• Planck (2013–2018) mapped TT to ℓ ≈ 2500 from space and stitched in ACT/SPT for the high-ℓ, foreground-dominated regime.
• Polarization: SPTpol and SPT-3G measure the EE damping tail out to ℓ ~ 8000, and the EE tail is cleaner because polarization is only generated at scattering.

Above ℓ ≈ 3000 the primary CMB is buried under foregrounds — the thermal and kinetic Sunyaev–Zel'dovich effects and the cosmic infrared background — which must be modeled and subtracted.

How It Differs From Its Cousins

Three distinct effects all suppress small-scale CMB power, and they are frequently muddled:

• Silk (diffusion) damping happens in the plasma before recombination: photons physically leak out of fluctuations. This is the dominant term and gives the exp[−(k/k_D)²] envelope.
• Finite-thickness / free-streaming smearing happens at recombination: the last-scattering surface has a thickness (Δz ≈ 80), so we average over a shell, blurring small-scale structure along the line of sight. It reinforces the same cutoff.
• Gravitational lensing by intervening large-scale structure smooths the peaks and transfers a little power to very high ℓ, partially filling in the damping tail.

Silk damping is also distinct from baryon acoustic oscillations (BAO) in the matter power spectrum: BAO is the same sound horizon imprinted on galaxies, whereas the damping tail is a photon-diffusion effect with no galaxy analog. And unlike the acoustic peaks (set by geometry and Ω_b), the tail's slope is uniquely sensitive to the expansion rate during recombination.

Why It Matters and What Is Debated

The damping tail is one of cosmology's sharpest tools precisely because its shape depends on physics at recombination, not just the distance to it.

• Relativistic species (N_eff): extra radiation speeds up expansion during recombination, shrinking the sound horizon relative to the damping scale and steepening the tail. The tail is the primary CMB constraint on N_eff, currently N_eff ≈ 2.99 ± 0.17 — beautifully consistent with the Standard-Model prediction of 3.044 for three neutrino species.
• Primordial helium (Y_p): more helium means fewer free electrons per baryon, lengthening λ_C and enhancing damping — letting the CMB alone weigh Big Bang nucleosynthesis.
• Hubble tension: proposals like early dark energy or small-scale clumping at recombination try to shrink r_s to raise H₀, but they generically also distort the damping tail, so precision tail measurements (SPT-3G, ACT DR6, Simons Observatory, CMB-S4) are a key testing ground.

Open questions include whether any residual excess or deficit in the high-ℓ tail signals new light species, non-standard recombination, or unmodeled foregrounds — a live debate as sub-arcminute polarization data sharpen the picture.

Silk damping (diffusion damping) versus the acoustic-peak regime and photon free-streaming — three distinct pieces of CMB physics that are often conflated.

Feature	Silk / diffusion damping	Acoustic peaks	Free-streaming (Landau) smearing
Physical cause	Photon random walk out of hot/cold spots before recombination	Standing sound waves in the baryon-photon fluid	Photons from a finite-thickness last-scattering shell average along the line of sight
Multipole range	ℓ ≳ 1000, strong by ℓ ~ 1500–2000	ℓ ~ 200 (first peak) to ~800 (third peak)	Contributes to the same high-ℓ cutoff
Characteristic scale	λ_D ≈ 3–7 comoving Mpc; k_D ≈ 0.14–0.2 Mpc⁻¹	Sound horizon r_s ≈ 147 Mpc comoving	Shell thickness Δz ≈ 80 (Δη ≈ tens of Mpc)
Effect on C_ℓ	Exponential envelope exp[−(k/k_D)²]	Oscillatory peaks and troughs	Extra smoothing, folded into the damping envelope
Sensitivity	N_eff, primordial helium Y_p, early dark energy	Ω_b, Ω_m, geometry / curvature	Recombination history, reionization