Pure Axion Inflation: Gauge-Driven Dynamics

• Pure Axion Inflation is a class of models where an axion couples to a pure gauge sector to generate the inflaton potential via nonperturbative dynamics.
• In non-Abelian realizations, the axion couples with pure Yang–Mills dynamics to yield a flattened, plateau-like potential supporting large-field inflation with sub-Planckian decay constants.
• Variants using Abelian gauge sectors and gauge-field production lead to distinctive reheating, oscillon dynamics, and gravitational wave signals constrained by backreaction effects.

Searching arXiv for papers on Pure Axion / Pure Natural Inflation and related developments. Pure Axion Inflation (PAI) denotes a class of inflationary constructions in which the inflaton is an axion or axion-like field and the inflationary potential is generated by a gauge sector in a structurally minimal way, typically through a $\phi F\tilde F$ coupling and nonperturbative gauge dynamics rather than ad hoc scalar self-interactions. In the literature represented here, the label covers two closely related but not identical usages: a non-Abelian large-$N$ “pure natural inflation” framework, where the axion couples to a pure Yang–Mills sector and acquires a flattened single-field potential (Nomura et al., 2017), and a more recent Abelian benchmark, where “pure” means an axion coupled to a pure gauge sector with no charged matter, used to study nonlinear gauge production and sourced gravitational waves (Eckardstein et al., 1 Aug 2025). A further extension identifies the same pure-gauge-generated axion with both the inflaton and a late-time quintessence field (J, 2019).

1. Terminological scope and defining structure

The common core of PAI is the axion–gauge topological interaction

$$\mathcal{L}_{\text{int}} = \frac{1}{32\pi^2} \frac{\phi}{f}\, \epsilon^{\mu\nu\rho\sigma}\, \mathrm{Tr}\,F_{\mu\nu}F_{\rho\sigma},$$

with $\phi$ the axion inflaton, $f$ the microscopic decay constant, and $F_{\mu\nu}$ the field strength of a gauge theory (Nomura et al., 2017). The perturbative action respects the axionic shift symmetry $\phi\to \phi+\text{const.}$, and the potential is generated only by nonperturbative gauge dynamics, so the flatness of the inflaton potential is symmetry-protected rather than imposed by hand (Nomura et al., 2017).

Two recurrent usages appear in the literature.

Usage of PAI	Gauge sector	Representative papers
Pure natural / pure Yang–Mills axion inflation	Pure $SU(N)$ Yang–Mills, large $N$	(Nomura et al., 2017, J, 2019, Gatica et al., 10 Mar 2026)
Pure gauge-sector axion inflation for sourced GW studies	Pure Abelian $U(1)$, no charged matter	(Eckardstein et al., 1 Aug 2025)

In the first usage, “pure” refers to the fact that the axion potential is generated by a pure Yang–Mills theory with no quarks or other charged matter in that sector (Nomura et al., 2017). In the second, “pure” denotes the absence of charged matter coupled to the gauge field, so that Schwinger damping and conductivity effects are absent; that definition is central to the GEF benchmark for sourced gravitational waves (Eckardstein et al., 1 Aug 2025). This suggests that PAI is best understood as an umbrella term for axion inflation models whose dominant structure is an axion coupled to a pure gauge sector, while the detailed realization can be non-Abelian or Abelian depending on the problem being studied.

2. Pure Yang–Mills realizations and large-$N$0 potentials

The canonical non-Abelian realization is pure natural inflation, in which the axion couples to a pure $N$1 Yang–Mills sector and the inflaton potential is extracted from the large-$N$2 $N$3-dependence of pure gluodynamics (Nomura et al., 2017). Witten’s large-$N$4 picture implies a multibranched vacuum structure, with physical $N$5-periodicity restored only after minimizing over branches. On a single branch, however, the potential need not be periodic in the simple cosine sense, which is the crucial departure from standard natural inflation (Nomura et al., 2017).

A widely used effective potential is

$N$6

with $N$7 the inflationary scale, $N$8 the effective field-range parameter, and $N$9 a shape parameter encoding the nonperturbative Yang–Mills dynamics (Nomura et al., 2017). In the holographic large-$$\mathcal{L}_{\text{int}} = \frac{1}{32\pi^2} \frac{\phi}{f}\, \epsilon^{\mu\nu\rho\sigma}\, \mathrm{Tr}\,F_{\mu\nu}F_{\rho\sigma},$$0 benchmark one obtains $$\mathcal{L}_{\text{int}} = \frac{1}{32\pi^2} \frac{\phi}{f}\, \epsilon^{\mu\nu\rho\sigma}\, \mathrm{Tr}\,F_{\mu\nu}F_{\rho\sigma},$$1 and

$$\mathcal{L}_{\text{int}} = \frac{1}{32\pi^2} \frac{\phi}{f}\, \epsilon^{\mu\nu\rho\sigma}\, \mathrm{Tr}\,F_{\mu\nu}F_{\rho\sigma},$$2

so that $$\mathcal{L}_{\text{int}} = \frac{1}{32\pi^2} \frac{\phi}{f}\, \epsilon^{\mu\nu\rho\sigma}\, \mathrm{Tr}\,F_{\mu\nu}F_{\rho\sigma},$$3 for $$\mathcal{L}_{\text{int}} = \frac{1}{32\pi^2} \frac{\phi}{f}\, \epsilon^{\mu\nu\rho\sigma}\, \mathrm{Tr}\,F_{\mu\nu}F_{\rho\sigma},$$4 (Nomura et al., 2017). The microscopic decay constant can remain sub-Planckian even when the effective field range is $$\mathcal{L}_{\text{int}} = \frac{1}{32\pi^2} \frac{\phi}{f}\, \epsilon^{\mu\nu\rho\sigma}\, \mathrm{Tr}\,F_{\mu\nu}F_{\rho\sigma},$$5, which is one of the main structural advantages of the model.

The large-field tail is plateau-like rather than cosine-like. For $$\mathcal{L}_{\text{int}} = \frac{1}{32\pi^2} \frac{\phi}{f}\, \epsilon^{\mu\nu\rho\sigma}\, \mathrm{Tr}\,F_{\mu\nu}F_{\rho\sigma},$$6,

$$\mathcal{L}_{\text{int}} = \frac{1}{32\pi^2} \frac{\phi}{f}\, \epsilon^{\mu\nu\rho\sigma}\, \mathrm{Tr}\,F_{\mu\nu}F_{\rho\sigma},$$7

while near the minimum the potential is approximately quadratic (Nomura et al., 2017). This flattened behavior is also present in the quintessence extension “Pure Natural Quintessential Inflation,” where the large-$$\mathcal{L}_{\text{int}} = \frac{1}{32\pi^2} \frac{\phi}{f}\, \epsilon^{\mu\nu\rho\sigma}\, \mathrm{Tr}\,F_{\mu\nu}F_{\rho\sigma},$$8 pure Yang–Mills construction is represented by the single-branch form

$$\mathcal{L}_{\text{int}} = \frac{1}{32\pi^2} \frac{\phi}{f}\, \epsilon^{\mu\nu\rho\sigma}\, \mathrm{Tr}\,F_{\mu\nu}F_{\rho\sigma},$$9

with $\phi$0 and $\phi$1 (J, 2019). In that model the quadratic term remains $\phi$2-independent, as required by Witten’s conjecture, while the large-field regime again approaches a plateau (J, 2019).

The defining contrast with standard natural inflation is therefore not merely quantitative. Standard natural inflation assumes

$\phi$3

whereas PAI in its pure Yang–Mills form replaces the cosine by a large-$\phi$4-motivated flattened branch potential (Nomura et al., 2017).

3. Inflationary dynamics and current CMB status

Inflationary predictions are obtained with standard single-field slow-roll formulae,

$\phi$5

evaluated on the Yang–Mills-generated plateau (Nomura et al., 2017). Because higher-order terms remain relevant when $\phi$6, the tensor amplitude is reduced relative to quadratic or cosine models while $\phi$7 stays in the Planck-preferred range (Nomura et al., 2017).

For the holographic $\phi$8 benchmark, pure natural inflation is consistent with current data at the $\phi$9 C.L. provided

$f$0

for $f$1, and

$f$2

for $f$3; the corresponding $f$4 C.L. limits are

$f$5

(Nomura et al., 2017). The model therefore realizes viable large-field inflation with sub-Planckian microscopic $f$6 and $f$7 (Nomura et al., 2017).

The quintessential extension exhibits a similar observational pattern. For $f$8, it gives $f$9 across the scanned parameter space and allows

$F_{\mu\nu}$0

depending on $F_{\mu\nu}$1, with explicit examples reaching $F_{\mu\nu}$2 for $F_{\mu\nu}$3 while keeping the fundamental decay constant sub-Planckian for $F_{\mu\nu}$4 (J, 2019).

The most recent CMB reassessment is the ACT-focused letter “Pure Natural Inflation Passes the ACT,” which finds that under both the instantaneous reheating hypothesis and standard assumptions for reheating, a non-trivial fraction of parameter space is ruled in (Gatica et al., 10 Mar 2026). That analysis also studies a phenomenological extension in which $F_{\mu\nu}$5 is allowed to be fractional or mildly negative. Under instantaneous reheating, $F_{\mu\nu}$6 retain parameter points in the ACT $F_{\mu\nu}$7 region, while a particularly favorable point occurs at $F_{\mu\nu}$8, where the minimum-$F_{\mu\nu}$9 point gives

$\phi\to \phi+\text{const.}$0

and lies very close to the center of the ACT-preferred region (Gatica et al., 10 Mar 2026).

4. Reheating, oscillons, and non-cold post-inflationary dynamics

In the original pure natural inflation analysis, reheating can proceed through axionic couplings to gauge bosons with perturbative decay width

$\phi\to \phi+\text{const.}$1

and a representative reheating temperature

$\phi\to \phi+\text{const.}$2

for typical parameters, high enough for thermal leptogenesis (Nomura et al., 2017). This is the standard perturbative picture when the post-inflationary condensate remains approximately homogeneous.

A different regime appears when the effective decay constant is sufficiently small. For the $\phi\to \phi+\text{const.}$3 pure natural inflation potential, the post-inflationary condensate fragments into oscillons when

$\phi\to \phi+\text{const.}$4

because the potential is shallower than quadratic around the minimum and supports narrow parametric resonance (Hong et al., 2017). Lattice simulations using a modified LatticeEasy code show localized, quasi-spherical lumps forming at

$\phi\to \phi+\text{const.}$5

for $\phi\to \phi+\text{const.}$6, with the early-time profile matching the I-ball solution

$\phi\to \phi+\text{const.}$7

in $\phi\to \phi+\text{const.}$8 dimensions (Hong et al., 2017). In this oscillon-dominated regime, reheating is no longer spatially homogeneous, and the onset of homogeneous radiation is bounded by

$\phi\to \phi+\text{const.}$9

or equivalently

$SU(N)$0

with the illustrative estimate

$SU(N)$1

for $SU(N)$2 values of the remaining factors (Hong et al., 2017).

A separate caveat concerns the assumption that axion inflation is cold. When generic axion–gauge couplings are treated dynamically, warm inflation can emerge over a large region of parameter space. For a generic pure $SU(N)$3 gauge sector without light fermions, cold inflation along the “cold-fit” line is only possible if

$SU(N)$4

whereas below that threshold the system typically enters weak warm, strong warm, or thermally broken regimes (DeRocco et al., 2021). This does not invalidate pure natural inflation, but it implies that cold single-field analyses are not automatically self-consistent once generic gauge couplings are included (DeRocco et al., 2021).

5. Gauge-field production, gravitational waves, and the Abelian PAI benchmark

A second, more recent meaning of PAI arises in the study of gauge-field amplification during axion inflation. In this benchmark, the matter Lagrangian is

$SU(N)$5

with a linear coupling

$SU(N)$6

and, in the benchmark scan, a quadratic inflaton potential

$SU(N)$7

(Eckardstein et al., 1 Aug 2025). “Pure” here means pure Abelian gauge sector: no charged matter, no extra scalars, and hence no Schwinger damping (Eckardstein et al., 1 Aug 2025).

The key instability parameter is

$SU(N)$8

which governs tachyonic amplification of one helicity of the gauge field and the sourced tensor spectrum (Eckardstein et al., 1 Aug 2025). The system is evolved in the gradient expansion formalism (GEF), which captures nonlinear gauge backreaction while neglecting axion gradients (Eckardstein et al., 1 Aug 2025). The central result is that GW signals within the reach of future GW interferometers can only be realized in parameter regions that also lead to strong backreaction and that are in conflict with the upper limit on $SU(N)$9, i.e. the allowed energy density of dark radiation (Eckardstein et al., 1 Aug 2025).

The companion fermionic study sharpens the contrast. In the Abelian benchmark without charged matter, PAI predicts a strongly blue-tilted GW spectrum, and in the GEF benchmark all parameter regions leading to observable GWs violate the upper limit on $N$0 (Eckardstein et al., 29 Sep 2025). Once charged fermions are included, Schwinger pair creation damps the gauge-field growth, attenuates the GW signal, and can place the resulting background within the sensitivity reach of LISA and ET without violating the $N$1 bound (Eckardstein et al., 29 Sep 2025). This establishes that the pure-gauge limit is a useful benchmark, but also that its most optimistic GW region is highly constrained.

6. Extensions, related constructions, and recurrent misconceptions

One recurrent misconception is that all axion-based inflation models qualify as PAI. Several prominent PQ-based models do not. In “Unifying inflation and dark matter with the Peccei–Quinn field,” the inflaton is the radial PQ field $N$2, while the QCD axion is the angular mode and a spectator during inflation; the paper explicitly distinguishes this from pure angular axion inflation (Fairbairn et al., 2014). The same structural point holds in the non-supersymmetric $N$3 model, where the inflaton is the radial mode of a PQ singlet and the axion supplies dark matter and isocurvature rather than driving inflation (Boucenna et al., 2017). More recently, “Inflation models with Peccei-Quinn symmetry and axion kinetic misalignment” realizes pure PQ inflation through the radial PQ mode near a kinetic pole, while the angular axion acquires kinetic misalignment; this again comes close to PAI in sectoral minimality, but not in the strict sense of an angular axion inflaton (Lee et al., 2024).

A second misconception is that “pure axion inflation” is synonymous with simple cosine natural inflation. A broad review of axion inflation already emphasized that axion models include natural inflation, monodromy, aligned multi-axion models, gauge-friction scenarios, chiral tensor production, oscillatory correlators, and PBH production, all tied together by the shift symmetry rather than by a single potential template (Pajer et al., 2013). Within that broader landscape, pure natural inflation is one specific UV-motivated branch, distinguished by its pure Yang–Mills origin and plateau rather than by periodic cosine dynamics (Nomura et al., 2017).

A third issue is whether pure axion inflation remains complete once gravitational corrections are taken seriously. “Natural-Scalaron Inflation” argues that the quantum effects generating a PNGB potential also generically generate an $N$4 term, so that once the scalaron is included, observational agreement improves substantially relative to pure-natural inflation alone (Salvio, 2021). This does not negate PAI, but it does place it within a broader EFT context in which additional gravitational degrees of freedom may be radiatively unavoidable.

Finally, the quintessential extension shows that the same pure-gauge-generated axion can, in principle, interpolate between primordial inflation and late-time dark energy. In that construction, the inflaton later behaves as a thawing quintessence field with

$N$5

and present-day potential energy

$N$6

while keeping the inflationary potential form unchanged (J, 2019). This is not the generic fate of PAI models, but it demonstrates that the pure-gauge axion framework extends beyond the inflationary epoch.

Taken together, these results place PAI in a precise technical niche: it is an effectively single-field axion inflation framework in which the potential is generated by a pure gauge sector, most cleanly realized in pure Yang–Mills large-$N$7 models and, in a distinct benchmark sense, in pure Abelian gauge-sector studies. Its central phenomenological features are plateau inflation with sub-Planckian microscopic decay constant, nontrivial reheating and fragmentation dynamics at small $N$8, and, in gauge-sourced variants, a sharp interplay between observable GW production and backreaction constraints (Nomura et al., 2017).