Tensor-to-Scalar Ratio r: Inflation's Gravitational-Wave Fingerprint

What the Tensor-to-Scalar Ratio Actually Is

Inflation stretched microscopic quantum fluctuations to cosmological scales. Two independent kinds of fluctuation were produced. Scalar perturbations are ripples in the density and curvature of space — they later collapsed under gravity into the cosmic web of galaxies. Tensor perturbations are ripples in the metric itself: primordial gravitational waves, traveling distortions of space that stretch and squeeze in perpendicular directions.

The tensor-to-scalar ratio is defined as the ratio of their power spectra at a chosen reference scale (the pivot, usually k = 0.05 Mpc^-1):

• r = A_t / A_s, where A_s ≈ 2.1 × 10^-9 is the measured scalar amplitude.
• r = 0.05 would mean tensor power is 5% of scalar power.
• r is dimensionless — it is a ratio, so it does not depend on the (huge, uncertain) absolute normalization.

Because A_s is already pinned down precisely by the CMB temperature spectrum, measuring r immediately gives the absolute tensor amplitude A_t, and thus the amplitude of gravitational waves from the Big Bang.

The Mechanism: Why r Encodes Inflation's Energy Scale

In the simplest models, inflation is driven by a single scalar field (the inflaton) slowly rolling down a potential V(φ). The rate of rolling is captured by two slow-roll parameters, ε and η, built from the slope and curvature of V. Quantum fluctuations of the metric during this phase produce a nearly scale-invariant spectrum of gravitational waves whose amplitude is set purely by the expansion rate H — which in turn is set by V.

The consistency relations of single-field slow-roll inflation are:

• r = 16·ε — the tensor-to-scalar ratio measures the first slow-roll parameter directly.
• n_t = −r/8 — the tensor tilt is fixed by r (the consistency relation, a sharp prediction of standard inflation).
• V^(1/4) ≈ 1.9 × 10^16 GeV × (r / 0.10)^(1/4) — the energy scale of inflation.

So r is a rare quantity that translates almost directly into physics near the grand-unification scale. A detection of r ≈ 0.01–0.1 would place inflation at ~10^16 GeV, tantalizingly close to where the strong and electroweak forces are thought to unify — a scale forever beyond particle colliders.

Characteristic Numbers and a Worked Example

Suppose a future experiment measured r = 0.01. What would that tell us?

• Slow-roll: ε = r/16 = 0.01/16 ≈ 6.3 × 10^-4 — inflation was extremely 'slow,' the potential very flat.
• Energy scale: V^(1/4) ≈ 1.9 × 10^16 × (0.01/0.10)^(1/4) ≈ 1.1 × 10^16 GeV. The Hubble rate then was H ≈ 10^13 GeV.
• Field range (Lyth bound): Δφ / M_Pl ≳ (r/0.01)^(1/2). Detectable r generally demands the inflaton move over a super-Planckian distance, Δφ > M_Pl ≈ 2.4 × 10^18 GeV.

By contrast, the measured scalar amplitude A_s ≈ 2.1 × 10^-9 and spectral tilt n_s ≈ 0.965 are known to sub-percent precision. Tensors remain a bound: the current 95%-confidence limit is r < 0.036 (BICEP/Keck 2018), tightening to r < 0.032 when Planck and baryon-acoustic-oscillation data are combined. The measurement precision has reached σ(r) ≈ 0.009 — sharp enough that a signal at r ~ 0.03 would already be showing up.

How r Is Observed: B-Mode Polarization

Gravitational waves cannot be seen directly at these frequencies, but they leave a unique imprint on the polarization of the cosmic microwave background. CMB polarization splits mathematically into two patterns:

• E-modes: a gradient (curl-free) pattern produced by both density and tensor perturbations.
• B-modes: a curl pattern that, on large angular scales, only primordial gravitational waves can produce. Density perturbations cannot make primordial B-modes — this is what makes them a clean tensor signature.

The primordial B-mode signal peaks at angular multipole ℓ ≈ 80 (angular scales of a few degrees), the recombination bump, with a fainter reionization bump at ℓ ≈ 5. The catch: the signal is fantastically weak — for r ~ 0.01 the B-mode amplitude is tens of nanokelvin, against a 2.725 K background. Experiments such as BICEP/Keck at the South Pole, Planck, the Simons Observatory, CMB-S4, and the satellite LiteBIRD stare at the cleanest patches of sky for years to dig it out. Galactic dust and gravitational-lensing B-modes must be carefully subtracted first.

How r Differs From Its Cousins

It is easy to conflate r with related quantities. Distinctions that matter:

• r vs. the scalar spectral index n_s: n_s ≈ 0.965 describes how scalar power tilts with scale and is measured; r describes the tensor amplitude and is only bounded. Together they live on the classic 'n_s–r plane' that discriminates inflation models.
• Primordial vs. lensing B-modes: Gravitational lensing of E-modes by intervening structure also makes B-modes, peaking at ℓ ~ 1000. These are a confusing foreground for the primordial signal and must be 'delensed.'
• Inflationary vs. astrophysical gravitational waves: LIGO/Virgo detect ~100 Hz waves from merging black holes; pulsar timing arrays see ~nanohertz waves from supermassive binaries. Inflationary tensor modes span a vast band but are observed at ~10^-17 Hz through the CMB — a completely different regime.
• r vs. running α: The 'running' dn_s/dln k measures scale-dependence of the tilt, a separate second-order quantity.

Significance, Famous Cases, and Open Questions

r is a make-or-break test of inflation. Many well-motivated models make sharp predictions: R^2 (Starobinsky) inflation and Higgs inflation predict r ≈ 0.003–0.004, sitting comfortably below current limits; simple chaotic m^2φ^2 inflation predicts r ≈ 0.13 and is now ruled out; natural inflation is under heavy pressure.

The most famous episode was BICEP2 in March 2014, which announced a detection of r ≈ 0.2 to worldwide headlines. Within months, joint analysis with Planck showed the signal was Galactic dust polarization, not inflation — a landmark cautionary tale about foregrounds.

• Open question 1: Is r large enough to detect at all? Some models predict r ~ 10^-3 or smaller, near the fundamental floor set by lensing and foregrounds.
• Open question 2: Does the consistency relation n_t = −r/8 hold? A violation would signal exotic (multi-field or non-standard) inflation.
• Open question 3: Can we reach σ(r) ~ 0.001? LiteBIRD and CMB-S4 aim for exactly that in the 2030s, potentially delivering the first detection — or ruling out an entire class of high-scale models.

Scalar vs. tensor perturbations from inflation, and what r ties them to

Property	Scalar perturbations	Tensor perturbations
Physical nature	Density / curvature ripples	Gravitational waves (space strain)
Power amplitude	A_s ≈ 2.1 × 10^-9	A_t = r · A_s (unmeasured)
CMB signature	Temperature + E-mode polarization	B-mode polarization (curl pattern)
Later fate	Seed galaxies, clusters, large-scale structure	Redshift away; leave only CMB imprint
Slow-roll link	n_s − 1 = 2η − 6ε	r = 16ε
Status	Measured to ~0.5%	Only an upper bound: r < 0.036