Can one scalar field resolve the Hubble, S8, and dark energy tensions together?

One trajectory, three observables moved together

The paper asks whether a single continuous cosmological history can produce the correlated changes that three separate anomalies seem to demand. The proposed field is canonical and minimally coupled, gravity and the matter sectors are untouched, and the potential is an exponential tail with a localized positive feature placed on it. The localized feature supplies a short burst of early energy density; the tail later acts as a slowly evolving dark-energy component.

A calibrated benchmark makes the claim concrete. The scalar reaches a transient peak fraction Ωφ,pk = 0.151 at redshift zpk ≈ 3.50×10³, lowers the drag sound horizon to rd = 141.1 Mpc, and supports H0 = 70.0 km s⁻¹ Mpc⁻¹ along the DESI BAO distance scale. At late times the same trajectory, compressed locally into the CPL form w(a) = w0 + wa(1−a), gives (w0, wa) = (−0.844, −0.176), and solving the growth equation on that same background gives S8 = 0.794.

The important word is same. The early peak, the late expansion and the growth suppression are not three fitted components but three epochs of one numerical solution, which is what makes the construction more restrictive than an early dark-energy sector plus an independently parameterized late dark-energy sector.

The Hubble tension is mostly a sound-horizon problem

The paper's framing of the first target explains why an early modification is needed at all. The cosmic microwave background does not measure H0 directly; it measures the angular acoustic scale θ* = rs(z*)/DA(z*), the ratio of the comoving sound horizon at last scattering to the angular-diameter distance to that surface. If the expansion rate H(z) rises before recombination while the rest of the fit stays under control, the sound-horizon integral shrinks. Because θ* is tightly measured, a smaller rs must be compensated mainly by a smaller distance to last scattering, which pushes the inferred H0 upward.

That logic also sets the difficulty. The modification must be large enough to reduce rd, early enough to affect the acoustic scale, and transient enough not to spoil the later thermal history. The benchmark's Ωφ,pk = 0.151 at z ≈ 3501 is the paper's answer to that three-part requirement, and rd = 141.1 Mpc is the resulting shift, not an independent input.

A hilltop release, then a quintessence tail

The mechanism is easiest to follow through the field's motion. Early on the field sits near a localized maximum of the potential, where the potential's curvature is negative. That negative curvature is a tachyonic episode: it makes the field unstable at the crest and triggers its release once Hubble friction weakens. The stored energy then converts into kinetic motion and the scalar fraction dilutes rapidly, so the standard pre-recombination-to-recombination history is recovered soon after the event.

Later the field rolls along the exponential tail and behaves as thawing quintessence. Because the field is canonical with positive energy density, 1 + wφ = φ̇²/ρφ ≥ 0, so the trajectory is non-phantom and cannot cross w = −1. The preferred quadrant w0 > −1, wa < 0 therefore follows from the kinetic structure rather than being imposed through a CPL fit.

The paper stresses that the crest position and curvature are obtained exactly from the full potential, without a bump-dominated approximation, and that the whole history is one numerical integration of the full scalar equations rather than separate effective proxies for the early and late epochs.

The BAO fit prefers a tail slope near unity

The quantitative confrontation uses the DESI DR2 compressed BAO measurements with their reported uncertainties and within-pair correlations. For the full benchmark, the fit gives χ²_BAO = 9.48 across 13 BAO observables. Profiling the exact late-time scalar background over the tail slope λ returns a best fit λbf = 1.00 with an approximate 68% profile interval 0.51 < λ < 1.26, and χ²_BAO = 8.78 at that minimum, compared with 10.55 for the λ = 0 limit.

Two details matter for reading these numbers correctly. First, the profile is over the exact scalar background, with Ωm0 and c/(H0rd) profiled as well, so λ is constrained by the shape of the late expansion history rather than by a CPL parameterization. Second, the CPL values quoted for the benchmark are a local compression of the numerical wφ(a); distances and growth in the likelihood are integrated from H(z) itself, so a different basis would project the same function differently without changing the dynamics.

The λ = 0 comparison is the informative contrast: it shows what the data prefer when the exponential tail is switched off, and the benchmark's χ²_BAO sits below that limit.

Growth is suppressed by about three percent

The S8 target is handled separately from distances because it concerns how efficiently matter perturbations grow, not how far light travels. With rest-frame sound speed cs² = 1, scalar fluctuations are smooth on the sub-horizon scales used by redshift-space distortions and weak lensing, so the matter growth factor follows the standard growth equation on the same background used for the sound horizon and BAO calculations.

Solving that equation from z = 50 with the same high-redshift normalization as the matched reference, σ8,ref = 0.811, gives Dφ(1)/DΛCDM(1) = 0.9696, σ8(0) = 0.786 and S8 = 0.794. The paper describes this as a 3.0% suppression of the present growth factor relative to the matched ΛCDM case, and notes that the DES Year-6 cosmic-shear value S8 = 0.798 (+0.014/−0.015) for its preferred analysis lies essentially on the benchmark prediction.

The matched normalization is the condition that makes this comparison meaningful: the suppression is measured against a ΛCDM case sharing the same high-redshift amplitude, not against an arbitrary reference.

Stability is established, agreement is not yet

The paper formulates the canonical scalar perturbation equations and shows that the quadratic kinetic term has the standard positive sign with cs² = 1, so the model has neither a ghost nor a gradient instability. The negative curvature at the early crest is a finite tachyonic-curvature episode that ends once the field leaves the localized feature, not a pathology of the kinetic sector.

This is a statement about theoretical consistency, and it should not be read as a data-fit result. The author is explicit that the present analysis separates what the evaluated observables can test from what requires a Boltzmann-code likelihood. The DESI-compressed BAO likelihood constrains the late-time slope and distance scale, the sound-horizon integral constrains the early feature through rd, DES Y6 and RSD test the growth suppression, and the local H0 measurement tests whether the reduced rd moves the inverse-distance-ladder calibration in the required direction.

What is missing is decisive: the early scalar component changes pre-recombination gravitational potentials and therefore cannot be reduced to rd alone. Cosmic microwave background temperature, polarization and lensing spectra must be fitted with the perturbation equations implemented in a Boltzmann solver, jointly with BAO, Pantheon+, RSD, weak lensing and a local-H0 likelihood.

A calibrated mechanism, not a completed model selection

The paper's own claim is deliberately narrower than its title's question. What the calculation establishes is that one exact canonical trajectory can generate a localized early fraction and reduce rd, fit the DESI DR2 BAO distance ratios quantitatively with an order-unity, observationally profiled λ, and suppress late growth into the DES-Y6 S8 range while remaining perturbatively well posed.

The author frames this as the nontrivial prerequisite that a full joint likelihood must test rather than assume. Because the mechanism already yields a single trajectory with the required early and late signatures, the next calculation is a direct falsifiability test of the model rather than a search for whether the desired phenomenology can be engineered at all.

The distinction the paper draws between a reduction in a tension and a statistically established resolution is the right lens for the result. The benchmark's numbers are internally consistent and calibrated against compressed data, but the CMB-spectrum test that would decide the question has not been performed.

The restrictive part is the shared trajectory

The most decision-relevant feature is not any single number but the coupling between them. In this construction the exponential slope is measured by the late expansion history, the localized feature controls the early fraction and rd, and the same background fixes the growth history. A model builder cannot tune the early dark-energy fraction to fix rd and separately tune a late dark-energy sector to fix w(a) and S8, because there is only one field and one potential.

That is also where the model is most exposed. The benchmark's agreement with DES Y6 and with the DESI DR2 BAO scale is conditional on the matched high-redshift normalization and on the compressed likelihood, and the early feature's effect on pre-recombination potentials is exactly what a full CMB-spectrum fit would probe. The paper's own priority ordering follows from this: the decisive next step is implementing the background and perturbation equations in a Boltzmann code and running the joint CMB, lensing, BAO, supernova, RSD, weak-lensing and local-H0 analysis, which would test the single trajectory as a whole rather than one observable at a time.

Sources used in this answer