Scalar Spectral Index n_s: The Slight Red Tilt of Primordial Fluctuations

What the scalar spectral index actually measures

Inflation is thought to have stretched microscopic quantum fluctuations into cosmic-scale ripples in the density of matter and radiation. The statistics of those ripples are encoded in the primordial curvature power spectrum, written as a power law:

• P_R(k) = A_s (k / k*)^(n_s − 1)

Here k is comoving wavenumber (inverse length scale), A_s is the amplitude at a chosen pivot scale k* = 0.05 Mpc⁻¹, and n_s is the spectral index. The exponent (n_s − 1) is the log-log slope: if you plot power versus scale on logarithmic axes, n_s − 1 is the straight-line tilt.

A value of n_s = 1 is the special Harrison–Zel'dovich case: scale invariance, equal fluctuation power on every scale. The measured value 0.9649 sits just below 1, so power falls gently as you go to smaller scales (larger k). That downward slope on small scales — the long-wavelength end being relatively enhanced — is the red tilt. It is small: n_s − 1 ≈ −0.035, a 3.5% deviation, but it is measured to roughly 8σ significance.

The mechanism: slow-roll inflation predicts the tilt

The red tilt is not an accident of fitting; it is a prediction. During inflation a scalar field (the inflaton) rolls slowly down a potential V(φ). "Slowly" is quantified by two dimensionless slow-roll parameters:

• ε = (M_pl² / 2)(V'/V)² — the steepness of the potential
• η = M_pl² (V''/V) — its curvature

where M_pl is the reduced Planck mass and primes are derivatives with respect to φ. To leading order the spectral index is:

• n_s − 1 = −6ε + 2η

Because inflation must end, the potential cannot be perfectly flat, so ε and η cannot both be zero. Their small nonzero values push n_s away from exactly 1 — and for most well-behaved potentials the combination −6ε + 2η is negative, giving the observed red tilt. Simple models predict n_s − 1 ≈ −2/N, where N ≈ 50–60 is the number of e-folds of expansion before horizon exit. That yields n_s ≈ 0.96–0.967, strikingly close to what is measured. The tilt is thus a direct readout of the inflaton potential's shape.

Key quantities and a worked example

Take the plateau-type model favored by data, Starobinsky (R²) inflation. Its slow-roll parameters at horizon crossing scale as ε ≈ 3/(4N²) and η ≈ −1/N. Plugging into n_s − 1 = −6ε + 2η with N = 55 e-folds:

• ε ≈ 3/(4·55²) ≈ 2.5 × 10⁻⁴ (so −6ε ≈ −0.0015)
• η ≈ −1/55 ≈ −0.0182 (so 2η ≈ −0.0364)
• n_s ≈ 1 − 0.0015 − 0.0364 ≈ 0.962

This lands squarely within the Planck error bar. The same model predicts a small tensor-to-scalar ratio r ≈ 12/N² ≈ 0.004, well below current upper limits. A second-order quantity, the running α_s ≡ d n_s / d ln k, measures how the tilt itself changes with scale; slow-roll predicts |α_s| ~ 1/N² ~ 3 × 10⁻⁴, far too small to detect yet. Planck finds α_s = −0.0045 ± 0.0067, fully consistent with zero — another success, since a large running would have signaled non-standard physics.

How it is observed

n_s is extracted from the angular power spectrum of cosmic microwave background (CMB) anisotropies. The CMB temperature and polarization patterns encode the primordial spectrum, processed by known physics (acoustic oscillations, Silk damping). By fitting the heights and positions of the acoustic peaks — especially the relative amplitudes of the first few peaks near multipole ℓ ≈ 200–800 — cosmologists reconstruct n_s within the six-parameter ΛCDM model.

• WMAP (2003–2013) first hinted at n_s < 1, reaching ~2–3σ evidence.
• Planck (2013–2018) delivered the benchmark 0.9649 ± 0.0042, ruling out n_s = 1 at ~8σ.
• ACT DR6 + Planck (2025) gives a bluer 0.9709 ± 0.0038, rising to 0.9743 with DESI BAO and lensing.

Large-scale structure — galaxy clustering, the Lyman-α forest, weak lensing — probes smaller scales (higher k) and independently supports the mild red tilt, extending the lever arm over which the slope is measured.

How it relates to its close cousins

n_s is one member of a family of primordial parameters; distinguishing it from its relatives clarifies what it does and does not say:

• Amplitude A_s sets how strong the fluctuations are (A_s ≈ 2.1 × 10⁻⁹); n_s sets how that strength tilts with scale. They are independent.
• Tensor spectral index n_t is the analogous slope for primordial gravitational waves; single-field inflation predicts a consistency relation n_t ≈ −r/8, tying it to the tensor-to-scalar ratio r.
• Running α_s is the second derivative — the change in n_s across scales. n_s is the first-order tilt; α_s captures curvature of the log-log spectrum.
• Non-Gaussianity f_NL probes the shape of higher-order correlations, not the two-point slope that n_s describes.

A pure Harrison–Zel'dovich spectrum (n_s = 1, α_s = 0, r = 0) is the scale-invariant baseline against which all these deviations are measured.

Significance and open questions

The red tilt is arguably inflation's most concrete scored prediction. Before the data, inflation generically expected n_s slightly below 1 and a tiny running; the measured 0.9649 confirms both. Ruling out exact scale invariance at ~8σ is a major result: it means the early universe had a preferred, evolving dynamics, not a static symmetric state.

Open questions remain live:

• Which potential? The exact value of n_s (and r) discriminates between models. The 2025 P-ACT shift toward n_s ≈ 0.974 has put mild tension on the long-favored Starobinsky and Higgs-inflation plateau models, reviving interest in others — an active debate.
• Is there running? A confirmed nonzero α_s would point beyond simplest slow-roll; current data say no, but future surveys (CMB-S4, LiteBIRD, 21-cm) will sharpen it by an order of magnitude.
• Is the tilt truly a power law? Features or oscillations in P_R(k) would signal transitions during inflation; searches so far find none, but the constraints are not airtight on all scales.

Scalar spectral index n_s: measurements, model expectations, and how to read the tilt

Case / dataset	n_s value	Meaning
Scale-invariant (Harrison–Zel'dovich)	n_s = 1 (exactly)	Equal power on all scales; ruled out at ~8σ
Planck 2018 (TT,TE,EE+lowE+lensing)	0.9649 ± 0.0042	Red tilt detected; benchmark value
P-ACT (Planck + ACT DR6, 2025)	0.9709 ± 0.0038	Slightly bluer, ~2σ above Planck-only
P-ACT-LB (+ lensing + DESI BAO)	0.9743 ± 0.0034	Higher still; nudges some inflation models
Blue tilt	n_s > 1	More power on small scales; disfavored
Running d n_s/d ln k	−0.0045 ± 0.0067	Scale-dependence of tilt; consistent with 0