The Curvaton Scenario: A Second Field That Seeds the Universe's Structure

What the curvaton is: a spectator field with a memory

A curvaton is a light scalar field σ that is dynamically irrelevant during inflation — its energy density is a tiny fraction of the total, so it does not affect the expansion — yet it carries information about the inflationary era in the form of quantum fluctuations. Because its effective mass m is much smaller than the Hubble rate during inflation (m ≪ H*), the curvaton is essentially frozen, sitting at some field value σ* while inflation stretches its quantum vacuum fluctuations to super-horizon scales. Each Hubble time the field acquires a fluctuation of amplitude δσ ≈ H*/(2π), so it inherits a nearly scale-invariant spectrum, exactly as the inflaton would.

The crucial difference is timing. During inflation these curvaton fluctuations are isocurvature — they perturb the composition of the universe, not its total curvature, because the curvaton contributes negligible energy. Only later, when the curvaton comes to matter energetically, are those fluctuations converted into the adiabatic curvature perturbation ζ that seeds structure. The curvaton is thus a field with a memory: it records inflation and replays it into the density field long afterward.

The mechanism: freeze, oscillate, dominate, decay

The curvaton's life has four acts:

• Freeze (during inflation): with m ≪ H*, σ is stuck at σ* and picks up fluctuations δσ ≈ H*/(2π), giving a relative perturbation δσ/σ*.
• Oscillate (after inflation): once the Hubble rate drops to H ≈ m, the field starts oscillating around the minimum of its potential. A quadratic potential V = ½ m²σ² makes σ behave like pressureless matter, so its energy density redshifts as ρ_σ ∝ a⁻³ — slower than radiation's a⁻⁴.
• Grow in importance: because it dilutes more slowly than the surrounding radiation, the curvaton's fractional energy density climbs. Its density perturbation is δρ_σ/ρ_σ = 2 δσ/σ*, quadratically amplifying the field fluctuation.
• Decay: when the curvaton decays into Standard-Model radiation, it transfers its perturbation to the photon–baryon fluid, producing the adiabatic ζ.

The efficiency of this transfer is set by the dimensionless parameter r_D = 3ρ_σ / (3ρ_σ + 4ρ_radiation) evaluated at decay. If the curvaton fully dominates before decaying (r_D → 1) the transfer is clean; if it is still subdominant (r_D ≪ 1) the conversion is inefficient and leaves strong non-Gaussian and isocurvature fingerprints.

Key quantities and a worked example

The curvature perturbation in the sudden-decay approximation is ζ ≈ (2 r_D / 3) · (δσ/σ*), and its power must match the observed amplitude ζ ≈ 2 × 10⁻⁵ (i.e. the scalar power A_s ≈ 2.1 × 10⁻⁹). The scale dependence — the spectral tilt n_s ≈ 0.965 measured by Planck — comes from the slow variation of H* and the curvaton mass, since in the pure curvaton limit n_s − 1 = 2(H'/H)·(...)|* + 2m²/(3H*²).

The signature prediction is local non-Gaussianity, quantified by

• fNL ≈ (5/4)(1/r_D) − 5/3 − 5 r_D/6 (quadratic potential, sudden decay).

Worked example: suppose the curvaton is still subdominant at decay with r_D = 0.1. Then fNL ≈ 5/(4 × 0.1) − 5/3 − 5(0.1)/6 ≈ 12.5 − 1.67 − 0.08 ≈ +10.7. For r_D = 0.01, fNL ≈ 123 — far above what the CMB allows. In the opposite limit r_D → 1, fNL → 5/4 − 5/3 − 5/6 = −5/4, a small negative value. So the observed near-Gaussianity forces the curvaton to have nearly dominated the energy density before decaying (r_D not too small).

How it is tested: the CMB bispectrum

The curvaton leaves two observable calling cards, both measured with the cosmic microwave background. First is the bispectrum: because ζ = ζ_Gaussian + (3/5)fNL·ζ_Gaussian², the perturbations acquire a three-point correlation peaked in squeezed triangles (one long wavelength modulating two short ones). Planck's 2018 analysis of temperature and E-mode polarization gives fNL^local = −0.9 ± 5.1 (68% CL). This rules out curvaton models with r_D ≲ 0.1 unless the potential is tuned, and it is fully consistent with r_D near 1.

Second is isocurvature. If the curvaton does not dominate at decay, some of its perturbation survives as a residual entropy (isocurvature) mode in cold dark matter or baryons, correlated or anti-correlated with the adiabatic mode. Planck constrains the CDM isocurvature fraction to a few percent, further squeezing the parameter space. Future experiments — CMB-S4, LiteBIRD, and large-scale-structure surveys probing scale-dependent halo bias — aim to reach σ(fNL) ≈ 1, deep into curvaton territory. A confirmed detection of fNL ≈ few would be a smoking gun for a curvaton-like multi-field origin.

Curvaton versus its cosmological cousins

The curvaton belongs to a family of mechanisms that generate ζ after inflation from a field other than the inflaton. It is worth distinguishing them:

• Single-field inflation: the inflaton alone seeds ζ; predicts negligible local non-Gaussianity (Maldacena's consistency relation, fNL ≈ (5/12)(n_s − 1) ≈ −0.02). Any robust detection of fNL ≳ 1 would falsify all single-field models.
• Modulated reheating (Dvali–Gruzinov–Zaldarriaga): a second field modulates the inflaton's decay rate Γ, varying the reheating temperature from place to place. It also produces local fNL but through the decay rate rather than the field's own energy.
• Inhomogeneous end of inflation: ζ arises because inflation ends at different times in different regions.

What unifies them is the δN formalism: ζ equals the perturbation in the number of e-folds of expansion, ζ = δN, so any field that changes how much a region expands can source curvature. The curvaton is the archetype because its physics — freeze, oscillate, decay — is so clean and its fNL prediction so sharp.

Why it matters, and what is still open

The curvaton scenario matters because it decouples the amplitude of perturbations from the energy scale of inflation. In single-field models, the measured ζ ≈ 2 × 10⁻⁵ pins the inflationary Hubble rate to H* ≈ 10¹³–10¹⁴ GeV for observable gravitational waves. With a curvaton, H* can be far lower, which rescues many particle-physics inflation models that would otherwise overproduce or underproduce perturbations, and it weakens the link between a small tensor-to-scalar ratio r and the inflaton potential.

Open questions remain. Identity: no confirmed particle-physics candidate for the curvaton exists, though axions, moduli, flat directions of the MSSM, and the Higgs itself have all been proposed. Naturalness: keeping m ≪ H* and avoiding thermal corrections requires care. Fine-tuning of r_D: Planck's tight fNL bound pushes r_D toward 1, which some regard as tuned. And a famous tension is that curvaton models generically predict detectable non-Gaussianity and isocurvature — so far unseen. The scenario is alive and well, but it is on the clock: the next generation of CMB and galaxy surveys will either detect its fingerprints or drive it into an increasingly narrow corner.

Curvaton scenario versus standard single-field inflation for generating primordial perturbations

Property	Single-field inflation	Curvaton scenario
Source of ζ	Inflaton fluctuations δφ	Curvaton fluctuations δσ, converted after inflation
Local fNL	≈ 0, slow-roll suppressed (|fNL| ≲ 1)	Can be large; fNL ≈ 5/(4 r_D) for small r_D
Isocurvature modes	None generically	Can be significant, correlated/anti-correlated with adiabatic
Tensor-to-scalar r	Tied to spectrum via Lyth bound	Decoupled; r can be very small
Inflation energy scale	Fixed by amplitude of ζ	Can be much lower (H* not set by ζ alone)
Field dominant during inflation?	Yes (drives expansion)	No (energetically negligible spectator)