Early Resolution in Cosmology

• Early resolution is a pre-recombination modification that reduces the comoving sound horizon, enabling a higher CMB-inferred H₀ while preserving late-time ΛCDM behavior.
• Axion-like early dark energy, the standard implementation, involves a transient energy injection near recombination with rapid post-trigger dilution to maintain concordance.
• Recent analyses with improved Planck NPIPE data show stringent constraints on the early dark energy fraction, indicating minimal deviation from standard ΛCDM and a residual Hubble tension.

Searching arXiv for papers on early-universe resolutions of the Hubble tension and related “early resolution” usages. Early resolution, in the cosmological literature on the Hubble tension, denotes a pre-recombination modification of the expansion history that reduces the comoving sound horizon $r_s$ and thereby permits a larger CMB-inferred $H_0$ while preserving the observed acoustic angular scale. In this framework, the discrepancy is addressed not by recalibrating late-time distances, but by changing the early-universe ruler itself. Axion-like early dark energy (EDE) is the canonical realization: a transient component contributes a non-negligible fraction of the total energy density before recombination and then rapidly dilutes, leaving late-time $\Lambda$CDM phenomenology approximately intact (Efstathiou et al., 2023).

1. Definition in the context of the Hubble tension

An early resolution changes the expansion rate near recombination or matter-radiation equality, so that the sound horizon is smaller than in $\Lambda$CDM. Since the CMB constrains the acoustic angle,

$\theta_s \equiv \frac{r_s(z_*)}{d_A(z_*)},$

a reduction in $r_s$ can be compensated by a reduction in $d_A(z_*)$, which moves cosmological fits toward a larger present-day Hubble constant. The defining feature is therefore a modification of pre-recombination physics rather than a late-time change in the distance ladder (Efstathiou et al., 2023).

The EDE literature commonly formulates the target phenomenology as a temporary energy injection near the epoch most relevant for CMB calibration. One representative benchmark is $f_{\rm EDE}(z_c)\sim 10\%$ around $z_c\simeq 3500$, followed by dilution as or faster than radiation; in that setting, Planck-era analyses reported that the CMB-inferred $H_0$ could be moved substantially closer to SH0ES than in baseline $H_0$0CDM (Murgia et al., 2020). This establishes the central meaning of early resolution: it is an attempt to solve the tension by altering the early-time ruler, not by revising only low-redshift expansion.

2. Axion-like early dark energy as the standard implementation

The most widely studied construction in the material provided is the axion-like EDE model with potential

$H_0$1

where $H_0$2 is the axion mass, $H_0$3 is the axion decay constant, and $H_0$4 with $H_0$5. The parameterization explored in MCMC analyses uses the critical redshift $H_0$6, the fractional EDE contribution $H_0$7, and the initial field value $H_0$8 (Efstathiou et al., 2023).

The physical picture is that the field is initially frozen by Hubble friction and behaves like dark energy. Near $H_0$9, when the Hubble scale becomes comparable to the field scale, it starts rolling; after that, the extra energy density rapidly dilutes, so the episode is transient. A more general axion-like parameterization uses

$\Lambda$0

with oscillation-averaged equation of state

$\Lambda$1

This form makes explicit why sufficiently large $\Lambda$2 yields rapid post-trigger dilution and therefore preserves late-time cosmology more effectively (Murgia et al., 2020).

This class of models is attractive because it directly implements the early-resolution logic: it raises $\Lambda$3 only in the pre-recombination window that controls $\Lambda$4. A plausible implication is that the success or failure of the framework is tightly tied to the precision of CMB temperature and polarization residuals in exactly that epoch-sensitive range.

3. Observational tests and statistical methodology

The improved constraints highlighted in the recent literature are driven by a revised CMB likelihood based on the Planck NPIPE data release. The central analysis uses a high-$\Lambda$5 temperature and polarization likelihood built from the PR4 NPIPE maps and improved relative to the older 2018 Plik likelihood. It is combined with low-$\Lambda$6 Planck TT and EE, Planck lensing, BAO and $\Lambda$7 from BOSS DR12, low-$\Lambda$8 BAO from 6dFGS and SDSS DR7, Pantheon+ supernovae, and optionally the SH0ES prior on $\Lambda$9 (Efstathiou et al., 2023).

A major methodological issue is prior-volume dependence. Because EDE introduces several new parameters and large regions of its prior space correspond either to poor fits or to effectively $\Lambda$0CDM-like behavior, marginalized posteriors can appear more conservative than best-fit likelihoods. For that reason, the NPIPE analysis explicitly compares Bayesian posteriors with likelihood profiles. In the older Plik setting, the profile likelihood allowed larger $\Lambda$1 and $\Lambda$2 than the posterior, suggesting a possible $\Lambda$3 preference for nonzero EDE in profile space. In the NPIPE setting, by contrast, the difference between profile and posterior is only modest, so the suppression of EDE is not attributable mainly to marginalization artifacts (Efstathiou et al., 2023).

This comparison is central to the meaning of early resolution as a data-driven hypothesis. The issue is not merely whether a particular prior disfavors EDE, but whether the likelihood itself retains a robust preference for a transient pre-recombination component.

4. Quantitative status of the axion-like EDE solution

With the NPIPE likelihood and without SH0ES, the Bayesian posterior yields

$\Lambda$4

over the redshift range $\Lambda$5, together with

$\Lambda$6

at $\Lambda$7 C.L. The data therefore favor a model close to $\Lambda$8CDM and leave a residual $\Lambda$9 tension with the SH0ES Cepheid-based measurement. When SH0ES is included, the fit is pushed toward a nonzero EDE fraction at roughly the $\theta_s \equiv \frac{r_s(z_*)}{d_A(z_*)},$0 level, but the CMB fit worsens by $\theta_s \equiv \frac{r_s(z_*)}{d_A(z_*)},$1, and the combined tension statistic is $\theta_s \equiv \frac{r_s(z_*)}{d_A(z_*)},$2 (Efstathiou et al., 2023).

The likelihood-profile comparison reinforces the same conclusion. For NPIPE, the profile gives

$\theta_s \equiv \frac{r_s(z_*)}{d_A(z_*)},$3

so prior-volume effects are minor. The paper’s conclusion is correspondingly direct: the new CMB likelihood provides no evidence in favor of a significant EDE component (Efstathiou et al., 2023).

This outcome is notable because earlier Planck-based studies had found a milder preference for nonzero EDE. One analysis using Planck 2018 reported $\theta_s \equiv \frac{r_s(z_*)}{d_A(z_*)},$4 at $\theta_s \equiv \frac{r_s(z_*)}{d_A(z_*)},$5 C.L. in a 3-parameter EDE model, but with $\theta_s \equiv \frac{r_s(z_*)}{d_A(z_*)},$6 relative to $\theta_s \equiv \frac{r_s(z_*)}{d_A(z_*)},$7CDM, interpreted as about a $\theta_s \equiv \frac{r_s(z_*)}{d_A(z_*)},$8 preference for nonzero EDE from CMB data alone; adding SH0ES strengthened that preference to about $\theta_s \equiv \frac{r_s(z_*)}{d_A(z_*)},$9, with $r_s$0 km/s/Mpc in the 3-parameter fit and $r_s$1 km/s/Mpc when SH0ES was included in a 1-parameter EDE analysis (Murgia et al., 2020). The NPIPE residual spectra are smaller than Plik’s, are in excellent agreement with $r_s$2CDM, and remove much of the small-scale polarization structure that had previously left more room for EDE (Efstathiou et al., 2023).

The resulting historical trajectory is clear. Early resolution by axion-like EDE remained viable in earlier Planck analyses, but improved high-$r_s$3 temperature and polarization likelihoods substantially weakened that case.

5. Naturalness, coincidence, and neutrino-triggered variants

A longstanding theoretical objection to minimal EDE is the coincidence problem: the scalar must become dynamically relevant near matter-radiation equality even though the underlying scalar-sector parameters are otherwise disconnected from that epoch. Sakstein and Trodden proposed a neutrino-assisted early dark energy scenario in which the EDE scalar couples to neutrinos and receives a transient kick when neutrinos become non-relativistic, $r_s$4. Because an eV-scale neutrino becomes non-relativistic around the relevant epoch, the onset of EDE is tied to neutrino thermal history rather than to a separately tuned scalar mass (Sakstein et al., 2019).

In that framework the scalar couples conformally to neutrinos, the neutrino stress-energy trace becomes nonzero as the species transitions away from the relativistic regime, and the field is displaced from the minimum of an effective potential. The proposal therefore aims to make the timing of the early-resolution mechanism natural. A later analysis defended this $r_s$5EDE construction against negative claims by arguing that the critique relied on incorrect equations, an unduly restrictive notion of naturalness, and an incomplete parameter scan. That paper further emphasized that the natural initial condition is generally $r_s$6, not $r_s$7, and that $r_s$8EDE remains a natural and cosmologically interesting potential resolution that merits fuller observational study (González et al., 2023).

These variants do not overturn the empirical status of axion-like EDE under NPIPE. They instead reframe early resolution as a broader class of models in which the pre-recombination energy injection is triggered by known microphysics, particularly neutrino thermodynamics, rather than by an apparently ad hoc scalar clock.

6. Relation to structure growth and later extensions

A persistent complication is the relation between early resolution of the $r_s$9 tension and the late-time $d_A(z_*)$0 tension. The EDE literature emphasizes that EDE tends to help $d_A(z_*)$1 by shrinking $d_A(z_*)$2, but CMB fits often compensate by increasing $d_A(z_*)$3 and sometimes $d_A(z_*)$4, which pushes $d_A(z_*)$5 upward. One detailed reassessment concluded that current weak-lensing and large-scale-structure data did not rule out EDE, even though EDE predicts somewhat higher $d_A(z_*)$6 than $d_A(z_*)$7CDM; it also cautioned against treating combinations of statistically inconsistent datasets as though they produced unambiguous exclusions (Murgia et al., 2020).

A recent extension combines EDE with an interacting dark energy–dark matter sector, with EDE supplying the early-time increase in $d_A(z_*)$8 and the DE–DM interaction suppressing structure growth to reduce $d_A(z_*)$9. For the combined dataset Planck 2018 + DESI + DES + Pantheon+ + SH0ES, that mixed model gives $f_{\rm EDE}(z_c)\sim 10\%$0, $f_{\rm EDE}(z_c)\sim 10\%$1, $f_{\rm EDE}(z_c)\sim 10\%$2, and $f_{\rm EDE}(z_c)\sim 10\%$3 Mpc, with $f_{\rm EDE}(z_c)\sim 10\%$4 and $f_{\rm EDE}(z_c)\sim 10\%$5. The raw fit improves by $f_{\rm EDE}(z_c)\sim 10\%$6, but the information-criterion advantage is negligible, $f_{\rm EDE}(z_c)\sim 10\%$7, and the model does not fully solve either tension (Yashiki, 29 May 2025).

The explanation given is physically transparent: both EDE and the interacting dark-sector component independently favor a higher present-day matter density, and that upward shift in $f_{\rm EDE}(z_c)\sim 10\%$8 reduces $f_{\rm EDE}(z_c)\sim 10\%$9, limiting how much EDE can shrink $z_c\simeq 3500$0 without degrading the CMB fit. This suggests that the challenge for early resolution is no longer only how to raise $z_c\simeq 3500$1, but how to do so without inducing compensating parameter shifts that spoil concordance elsewhere.

7. Present assessment

The current state of the subject is therefore mixed but sharply defined. As a mechanism, early resolution remains conceptually clear: a temporary pre-recombination energy component increases $z_c\simeq 3500$2, reduces the sound horizon, and raises the CMB-inferred $z_c\simeq 3500$3. As an empirical proposal, however, the improved Planck NPIPE likelihood finds no evidence for a significant axion-like EDE component and instead favors a cosmology close to $z_c\simeq 3500$4CDM, with a residual $z_c\simeq 3500$5 discrepancy with SH0ES (Efstathiou et al., 2023).

At the same time, the broader program has not disappeared. Earlier Planck analyses, weak-lensing reassessments, neutrino-triggered variants, and mixed early-plus-late extensions show that the idea of an early-universe resolution continues to organize both model building and statistical tests of the Hubble tension (Murgia et al., 2020). The most conservative synthesis is that early resolution remains a well-defined and theoretically motivated strategy, but in its standard axion-like EDE form it is not strongly favored by the improved CMB data currently available (Efstathiou et al., 2023).