Characterizing the multi-field inflationary potentials for which the curvature perturbation $\mathcal{R}$ approaches a constant as the universe expands

In multi-field inflation, Weinberg's long-wavelength theorem guarantees the existence of an adiabatic solution in which the comoving curvature perturbation $\mathcal{R}$ is constant outside the horizon, but the general solution for generic initial conditions also contains $2N-2$ non-adiabatic modes that can keep sourcing $\mathcal{R}$. Weinberg (2003) noted two classes of potentials — sums of exponentials and rotationally symmetric potentials $V=F(\sum_n\varphi_n^2)$ — in which $\mathcal{R}$ (equivalently the synchronous-gauge quantity $A=-q^2\mathcal{R}$) does approach constants as the scale factor $a(t)$ increases, and posed the open problem of characterizing the general class of multi-field potentials with this property. Later literature has settled the post-inflationary fate of non-adiabatic perturbations, proven that $\mathcal{R}$ is conserved exactly when pressure is a unique function of energy density, and developed trajectory-level criteria (turn rate, effective entropy mass) for reaching the adiabatic limit — but no necessary-and-sufficient characterization of the potentials themselves exists. The problem remains open in its source-faithful form: give a precise classification, or an equivalent dynamical criterion, of the multi-field inflationary potentials for which non-adiabatic contributions decay so that $\mathcal{R}$ and $A$ converge for generic initial data during the inflationary era.

Background

Inflation — an early period of accelerated expansion of the universe — is the leading explanation of the origin of cosmic structure: the observed fluctuations in the cosmic microwave background (CMB) and the large-scale matter distribution are thought to be quantum fluctuations of the fields driving inflation, stretched to cosmological scales. To connect theory with observation one evolves small (linear) perturbations on a spatially flat Friedmann–Robertson–Walker background with scale factor $a(t)$ and Hubble rate $H=\dot a/a$, focusing on modes whose physical wavelength far exceeds the Hubble radius $H^{-1}$ ("super-horizon" modes). This requires fixing a gauge; in the Newtonian (longitudinal) gauge the perturbed metric is characterized by the gravitational potential $\Psi$ (the two scalar metric potentials coincide when anisotropic stress is negligible), and the matter perturbations by the density fluctuation $\delta\rho$ and the velocity potential $\delta u$. The central observable is the gauge-invariant comoving curvature perturbation

$$\mathcal{R} = -\Psi + H\,\delta u,$$

which measures the spatial curvature on hypersurfaces orthogonal to the total matter velocity. Its importance is that in eras with a single dynamical degree of freedom it is conserved for super-horizon modes, so its observed value can be traced back through cosmic history to the quantum fluctuations that seeded it.

A perturbation mode is called adiabatic when the fluctuation of every constituent quantity $S$ is proportional to the background rate of change, $\delta S \propto \dot{\bar S}$ — in particular $\delta\rho_\alpha/\dot{\bar\rho}_\alpha$ is the same for all constituents $\alpha$. Multi-component systems can also carry non-adiabatic (entropy, or isocurvature) perturbations, which change the relative composition rather than the total. Weinberg (2003) proved a general long-wavelength theorem in Newtonian gauge: whatever the constituents of the universe, the perturbation equations always have two independent adiabatic solutions in the zero-wavenumber limit — one with $\mathcal{R}$ equal to a nonzero constant and one with $\mathcal{R}=0$ — alongside possible non-adiabatic solutions. For the case he analyzed in detail, inflation driven by $N$ real scalar fields $\varphi_f$ with canonical kinetic terms (each field entering with the standard quadratic kinetic-energy form) and an arbitrary interacting potential $V(\varphi_1,\dots,\varphi_N)$ (multi-field inflation), the long-wavelength system has $2N+1$ independent solutions: two adiabatic, one removed by a momentum constraint (a constraint equation that eliminates one combination of the field perturbations), and $2N-2$ non-adiabatic. The physical solution for generic initial data generally contains non-adiabatic contributions, and then $\mathcal{R}$ evolves: in the standard adiabatic–entropy decomposition (Sasaki and Tanaka 1998; Gordon, Wands, Bassett and Maartens 2001), perturbations orthogonal to the background field trajectory (entropy modes) source $\mathcal{R}$ whenever the trajectory bends in field space, with the effective mass of the entropy direction controlling the entropy amplitude. Equivalently, Lyth, Malik and Sasaki (2005) proved non-perturbatively that $\mathcal{R}$ is conserved on super-horizon scales if and only if the total pressure is a unique function of the total energy density — a condition multi-field dynamics violates while entropy modes are active. Reviews of this framework include Wands (2008).

Weinberg also introduced an equivalent bookkeeping quantity in synchronous gauge (a coordinate choice in which the perturbed metric has no time–time component),

$$A \equiv a^2 H\psi - 4\pi G a^2\delta\rho^{(S)} - q^2 H\,\delta u^{(S)},$$

with $q$ the comoving wavenumber. It satisfies $\dot A = -q^2 H\dot{\delta u}^{(S)}$, so $A$ is conserved at $q=0$ for every mode, and is related to $\mathcal{R}$ by $A=-q^2\mathcal{R}$; hence for any finite $q$ the fractional time-variation of $\mathcal{R}$ equals that of $A$. Whether the general solution has $A$, and hence $\mathcal{R}$, approaching constants as $a(t)$ increases is therefore exactly the question of whether the non-adiabatic contributions decay relative to the adiabatic one.

Two classes of potentials are known to have this property during multi-field inflation. The first is the family of separable exponential potentials $V=\sum_n g_n\exp(-\lambda_n\varphi_n)$ studied by Malik and Wands (1999) in "assisted inflation": the background flows to a power-law scaling solution (an attractor along which the ratios of the fields freeze, even when each field alone is too steep to drive inflation), and the entropy perturbations about this attractor decay. The second, noted by Weinberg without citation, consists of rotationally symmetric potentials $V=F(\sum_n\varphi_n^2)$ with arbitrary $F$, where conservation of field-space angular momentum renders the motion asymptotically radial. Conversely, for generic potentials with light entropy directions or persistently turning trajectories, $\mathcal{R}$ keeps evolving on super-horizon scales — and with an ultra-light entropy mode (an entropy fluctuation whose effective mass is far below the Hubble rate $H$) coupled to the adiabatic direction it can even grow linearly in time (Achúcarro, Atal, Germani and Palma 2017). Weinberg closed his analysis by asking for the general characterization: "It would be interesting to characterize the general class of potentials for multi-field inflation for which $A$ and $\mathcal{R}$ approach constants as $a(t)$ increases."

Problem Statement

For multi-field inflation — $N$ real scalar fields $\varphi_1,\dots,\varphi_N$ with canonical kinetic terms, minimally coupled to gravity, interacting through an arbitrary potential $V(\varphi_1,\dots,\varphi_N)$ — characterize the general class of potentials for which the quantity $A$, and hence the comoving curvature perturbation $\mathcal{R}$ (related by $A=-q^2\mathcal{R}$), approaches constants as the scale factor $a(t)$ increases, for the general solution of the coupled background and super-horizon perturbation equations, i.e. for generic initial conditions in which both the adiabatic mode and the $2N-2$ non-adiabatic modes are present.

The characterization has four components: (1) define the class — or an equivalent dynamical criterion, e.g. in terms of invariants of $V$, of the attractor structure of the background trajectories, or of trajectory-level data such as the turn rate (the angular speed of the background trajectory in field space) and the effective entropy-mass spectrum relative to $H$ — precisely; (2) prove that every potential in the class has the non-adiabatic contributions to $\mathcal{R}$ decay relative to the adiabatic contribution, so that $\mathcal{R}$ (equivalently $A$, after appropriate normalization) converges, for generic initial data as $a(t)$ increases during the inflationary era; (3) establish the complementary side — that potentials outside the class fail the property, exhibiting the mechanism (persistent light or ultra-light entropy modes, sustained turning of the trajectory) that prevents convergence, or prove that no potential-level classification exists because the property is irreducibly trajectory- or initial-data-dependent; and (4) verify that the known member classes (separable exponential sums $V=\sum_n g_n\exp(-\lambda_n\varphi_n)$ and rotationally symmetric potentials $V=F(\sum_n\varphi_n^2)$) and known non-members (persistent-turn and ultra-light-entropy models) fall on the correct side of the characterization.

When a potential-level class, or an equivalent dynamical criterion, is defined, these four components form the characterization; if instead the alternative in (3) is established — that no potential-level classification exists, the property being irreducibly trajectory- or initial-data-dependent — that result replaces (1) and (2), and (4) is then checked against the equivalent dynamical criterion that the impossibility argument supplies rather than against a potential-level class. Since the trajectory-level criteria already available in the literature (for example the effective entropy-mass condition of Ref. 10) are conditions obtained model by model rather than the necessary-and-sufficient characterization for generic initial data, restating them is not by itself an answer to (1) or (2).

The question concerns the multi-field inflationary era itself, in Weinberg's idealized setting in which that era persists as $a(t)$ increases; the fate of non-adiabatic perturbations after inflation ends was settled separately by Weinberg (2004) under the assumption of reheating into local thermal equilibrium, and a restatement of that post-inflationary result is not by itself an answer. Extensions to non-canonical kinetic terms or curved field-space metrics (which add a universal geometrical contribution to the entropy mass) may be included as generalizations, but the flat, canonical case above is the source-faithful core.

References

1. Steven Weinberg, "Adiabatic modes in cosmology", Phys. Rev. D 67, 123504 (2003). DOI: 10.1103/PhysRevD.67.123504; arXiv:astro-ph/0302326; https://arxiv.org/abs/astro-ph/0302326
2. Karim A. Malik and David Wands, "Dynamics of Assisted Inflation", Phys. Rev. D 59, 123501 (1999). DOI: 10.1103/PhysRevD.59.123501; arXiv:astro-ph/9812204; https://arxiv.org/abs/astro-ph/9812204
3. Misao Sasaki and Takahiro Tanaka, "Super-Horizon Scale Dynamics of Multi-Scalar Inflation", Prog. Theor. Phys. 99, 763 (1998). DOI: 10.1143/PTP.99.763; arXiv:gr-qc/9801017; https://arxiv.org/abs/gr-qc/9801017
4. Christopher Gordon, David Wands, Bruce A. Bassett, and Roy Maartens, "Adiabatic and entropy perturbations from inflation", Phys. Rev. D 63, 023506 (2001). DOI: 10.1103/PhysRevD.63.023506; arXiv:astro-ph/0009131; https://arxiv.org/abs/astro-ph/0009131
5. David H. Lyth, Karim A. Malik, and Misao Sasaki, "A general proof of the conservation of the curvature perturbation", JCAP 05 (2005) 004. DOI: 10.1088/1475-7516/2005/05/004; arXiv:astro-ph/0411220; https://arxiv.org/abs/astro-ph/0411220
6. Steven Weinberg, "Must Cosmological Perturbations Remain Non-Adiabatic After Multi-Field Inflation?", Phys. Rev. D 70, 083522 (2004). DOI: 10.1103/PhysRevD.70.083522; arXiv:astro-ph/0405397; https://arxiv.org/abs/astro-ph/0405397
7. Ki-Young Choi, Jinn-Ouk Gong, and Donghui Jeong, "Evolution of the curvature perturbation during and after multi-field inflation", JCAP 02 (2009) 032. DOI: 10.1088/1475-7516/2009/02/032; arXiv:0810.2299; https://arxiv.org/abs/0810.2299
8. David Wands, "Multiple field inflation", Lect. Notes Phys. 738, 275 (2008). DOI: 10.1007/978-3-540-74353-8_8; arXiv:astro-ph/0702187; https://arxiv.org/abs/astro-ph/0702187
9. Joel Meyers and Navin Sivanandam, "Non-Gaussianities in Multifield Inflation: Superhorizon Evolution, Adiabaticity, and the Fate of $f_{NL}$", Phys. Rev. D 83, 103517 (2011). DOI: 10.1103/PhysRevD.83.103517; arXiv:1011.4934; https://arxiv.org/abs/1011.4934
10. Sébastien Renaux-Petel and Krzysztof Turzyński, "On reaching the adiabatic limit in multi-field inflation", JCAP 06 (2015) 010. DOI: 10.1088/1475-7516/2015/06/010; arXiv:1405.6195; https://arxiv.org/abs/1405.6195
11. Ana Achúcarro, Vicente Atal, Cristiano Germani, and Gonzalo A. Palma, "Cumulative effects in inflation with ultra-light entropy modes", JCAP 02 (2017) 013. DOI: 10.1088/1475-7516/2017/02/013; arXiv:1607.08609; https://arxiv.org/abs/1607.08609
12. Diego Cruces, Cristiano Germani, and Adrian Palomares, "An update on adiabatic modes in cosmology and $\delta N$ formalism", JCAP 06 (2023) 002. DOI: 10.1088/1475-7516/2023/06/002; arXiv:2212.05112; https://arxiv.org/abs/2212.05112