The axion is a hypothetical particle that solves the strong CP problem — why QCD does not violate CP symmetry despite having no apparent reason not to. As a bonus, the axion is a dark matter candidate. But cosmological constraints impose an upper bound on the axion decay constant f_a: if f_a is too large, the axion field carries too much energy density after inflation, overproducing dark matter and overclosing the universe. This upper bound is usually treated as a fundamental constraint on axion models.
Note (2026-11-12; read against the paper's full text): the paper says the early-relaxation scenario, in which the axion relaxes during an epoch of early confinement and so loses any cosmological upper bound on its decay constant, was originally proposed in earlier work, and that its own goal is an explicit implementation inside an SU(5) grand unified theory in inflationary cosmology. It shows that a phase of strong QCD during inflation can be realized by at least two mechanisms (a direct coupling of the inflaton to the gauge fields, or restoration of the SU(5) symmetry during inflation, with strong coupling reached by running of the gauge coupling), which are assumptions about the inflationary epoch, and the essay's statement that early confinement is a natural possibility in SU(5) models is its own. The paper also says the mechanism works for the DFSZ and KSVZ axions and for the two-form gauge axion, and that for DFSZ the phase is a condensate of ordinary quarks, an early-universe analogue of the eta-prime meson.
Dvali, Fitz, and Komisel (arXiv:2603.28620, March 2026) remove the upper bound entirely. The mechanism: during inflation, if an SU(5) grand unified theory confines (rather than remaining in its Coulomb phase), the confinement generates an early axion potential that traps and dilutes the problematic energy density. After inflation ends and SU(5) deconfines, the early potential vanishes and the axion is left with its standard QCD potential — but the dangerous energy density has already been diluted by the inflationary expansion.
The key surprise is what happens to the axion during inflation when the Peccei-Quinn scalar expectation value vanishes — when the field that normally hosts the axion does not exist. The axion survives as the phase of the fermion 't Hooft determinant, a topological quantity that persists even when the underlying scalar field is absent. The degree of freedom that becomes the axion at low energies exists during inflation in a different mathematical guise. It is not born after inflation; it was present throughout, wearing a different identity.
The result: the axion works as dark matter at arbitrarily large decay constants. The cosmological upper bound, thought to be a structural constraint, was a consequence of the assumption that no early confinement occurs. If GUT-scale physics includes a confining phase during inflation — which is a natural possibility in SU(5) models — the constraint simply does not apply.
The structural observation: a cosmological bound that constrains a fundamental parameter is removed by including physics at a different epoch. The bound was real but epoch-dependent: it applied only in cosmologies without early confinement. The parameter space of the theory is larger than the standard cosmological bounds suggest, because the bounds assume a specific thermal history that is not the only possibility.