An algebraic map separates inflation’s motion from its growing perturbations

Classifying histories, not just models

In single-field inflation, a scalar field called the inflaton evolves through a potential while the universe expands. Ordinary slow roll describes a familiar regime in which the potential’s slope largely determines the field’s velocity. But the same model can pass through phases that depart substantially from this balance.

In their theoretical preprint, Jinn-Ouk Gong and Wei-Chen Lin develop a classification for a canonical single scalar field in general relativity. Their framework combines two questions: where the physical velocity sits among the roots of an exact algebraic equation, and whether a particular difference between curvature perturbations grows or decays on very large scales.

The distinction matters because these two classifications do not generally change at the same point.

A cubic equation with one physical trajectory

The authors rewrite the background equations as a depressed cubic: a cubic equation without a squared term. Its variable is a dimensionless measure of the inflaton’s velocity.

For monotonic downhill motion on a nonzero potential slope, the cubic has two positive roots, counted with multiplicity, after choosing a sign convention. The physical velocity can occupy either the lower or higher position in this algebraic ordering. At a boundary where the roots merge, that ordering changes.

This is not a sudden jump in the field’s motion. Nor do the roots represent independent physical solutions with the same background data. The background constraint selects the physical root, and its trajectory remains regular through the merger.

The analysis also identifies a separate limit on how quickly the field can slow down while moving downhill. The exactly flat-potential case lies on this boundary. If the flat region persists, it supports ultra-slow-roll evolution, in which the velocity decreases as the inverse cube of the scale factor.

That downhill restriction is not a universal bound on all motion. A field temporarily climbing its potential can slow down faster, because the potential force and expansion-related friction both oppose its motion.

What actually grows?

The perturbation criterion concerns two curvature quantities: the comoving curvature perturbation, R, and the curvature perturbation on uniform-density surfaces, ζ. They describe curvature on two physically specified choices of time slicing.

At linear order, their difference is proportional to the change in R per e-fold of expansion: ζ − R = −(1/3)dR/dN. Here, N counts the logarithmic growth of the scale factor.

In the strict infinite-wavelength limit, the momentum conjugate to R is conserved. This makes the evolution of the curvature difference depend on the background coefficient a³Hε, where a is the scale factor, H the expansion rate and ε measures its fractional change. If this coefficient decreases, a nonzero curvature difference grows; if it increases, the difference decays.

There is an essential qualification: a difference that starts with zero conserved momentum remains zero. And growth of the difference does not, by itself, establish growth of the complete amplitude of R or enhancement of its final power spectrum. Those depend additionally on the initial perturbation state and the preceding evolution.

A chart with distinct boundaries

The authors express their classification using ε and a second quantity, χ, defined as the square root of the ratio between the potential slow-roll parameter and ε. This compares the potential’s slope with the field’s actual motion.

The curvature-difference growth boundary becomes the horizontal line χ = 1/2. The algebraic branch-swap boundary instead follows χ = 2ε/(3 − ε). The flat-potential boundary is χ = 0.

For ε below 3/5, a region exists where the physical velocity occupies the lower algebraic branch while the curvature difference grows. Above 3/5, the ordering reverses, allowing the higher branch to coexist with a decaying difference. Neither branch is therefore a general synonym for slow roll or ultra-slow roll.

The chart gives an instantaneous classification. Determining a trajectory through it also requires the potential’s curvature, which the authors use to derive a criterion for the direction of branch crossing.

Illustrations and limits

Toy potentials demonstrate the separation. In a smooth transition to a shallower slope, one example develops a growing curvature difference without crossing the branch-swap boundary; a stronger flattening crosses it. A quadratic correction to a constant potential also supplies two approximate solutions on the lower branch with opposite curvature-difference behaviour.

These are theoretical illustrations, not observational tests. Moreover, momentum conservation is exact only at zero wavenumber. Finite-wavelength corrections can matter even on large scales, particularly near the growth boundary or when the curvature difference is small. The preprint therefore offers a map of dynamical regimes, not a stand-alone prediction of the final primordial spectrum.