Quintessential α-attractors, updated

Abstract: Quintessential $α$-attractor models of single-field inflation and evolving dark energy were constructed about a decade ago. Recently, it was pointed out that some of them might be disfavored due to dark-radiation constraints on gravitational waves and the higher values of $n_s$ favored by ACT. Here we present a class of updated quintessential $α$-attractor models in which a single field describes both inflation and evolving dark energy, yields higher values of $n_s$, and admits reheating scenarios consistent with the dark-radiation bound on $ΔN_{\rm eff}$. Depending on the value of the cosmological constant $Λ$, these models interpolate between $Λ$CDM with $Λ> 0$ (future dS universe), dynamical dark energy with $Λ= 0$ (future Minkowski universe), and dynamical dark energy with $Λ< 0$ (future cosmological collapse). We also study quintessential $α$-attractor models based on an axion-inflaton complex scalar field with hyperbolic geometry, which describe inflation and dark energy of a ``phantom illusion'' type compatible with DESI DR2.

• The paper introduces a continuous parameter Δ that connects de Sitter, Minkowski, and collapsing cosmological futures while interpolating between ΛCDM-like and DESI-motivated dark-energy behavior.
• The paper uses waterfall modulation to raise the spectral index from 0.963 to approximately 0.9735 while preserving the α-attractor relation r ≃ 3(1−n_s)^2 and leaving late-time dark energy unchanged.
• The paper identifies reheating mechanisms that can reduce dark radiation below ΔN_eff ≲ 0.3, but finds that canonical single-field models cannot cross w = −1 and proposes an unverified inflaton–axion extension for that case.

Overview

This paper by Kallosh, Linde, Shmakova, and Yamada revisits the quintessential $\alpha$-attractor program introduced roughly a decade ago [Akrami:2017cir, Dimopoulos:2017zvq], in which a single scalar field drives both inflation and a thawing dark energy stage. The motivation for the update is twofold. First, the combination of Planck/BICEP-Keck data with recent ACT and SPT measurements has pushed the preferred spectral index upward to $n_s = 0.9682 \pm 0.0032$, and joint CMB + DESI DR2 fits yield $n_s = 0.9728 \pm 0.0029$ [Balkenhol:2025wms], placing the simplest plateau models under tension. Second, DESI DR2 favors dynamical dark energy with $w_0 \sim -0.8$, $w_a \sim -0.6$, crossing the phantom divide — behavior that single-field quintessence with a canonical kinetic term cannot reproduce. The authors acknowledge methodological concerns raised about the DESI preference [Ong:2025utx, Afroz:2025iwo] and adopt an explicitly agnostic stance: they construct model families that interpolate between $\Lambda$CDM and the DESI best fit, so as to be testable either way.

Interpolating exponential potentials

The baseline is the exponential ("Exp") class of quintessential $\alpha$-attractors written in hyperbolic half-plane variables $T = e^{-\varphi/\sqrt{2/3}} + i\theta$ with kinetic term $-3\,\partial T\,\partial\bar T/(T+\bar T)^2$. In terms of the canonical field,

$$V_{\rm exp}(\varphi) = M^2 e^{-2g}\left[e^{g\left(\tanh\frac{\varphi}{\sqrt{6\alpha}}+1\right)} - 1\right] + \Lambda .$$

The original constructions fixed two boundary cases: Exp I ($n_s = 0.9682 \pm 0.0032$0, future de Sitter) and Exp II ($n_s = 0.9682 \pm 0.0032$1, future Minkowski). The paper's first contribution is a continuous interpolation parameterized by

$n_s = 0.9682 \pm 0.0032$2

with non-uniformly spaced values of $n_s = 0.9682 \pm 0.0032$3 chosen densely near zero because $n_s = 0.9682 \pm 0.0032$4 runs from its Exp II value to $n_s = 0.9682 \pm 0.0032$5 over the narrow interval $n_s = 0.9682 \pm 0.0032$6. Negative $n_s = 0.9682 \pm 0.0032$7 extends the family to AdS futures with eventual cosmological collapse.

The inflation–dark-energy hierarchy is encoded in $n_s = 0.9682 \pm 0.0032$8–$n_s = 0.9682 \pm 0.0032$9, tuned by bisection so that $n_s = 0.9728 \pm 0.0029$0 with $n_s = 0.9728 \pm 0.0029$1. At $n_s = 0.9728 \pm 0.0029$2, representative values are $n_s = 0.9728 \pm 0.0029$3 for $n_s = 0.9728 \pm 0.0029$4 respectively.

Waterfall modulation and the spectral index

To address the elevated $n_s = 0.9728 \pm 0.0029$5, the authors insert a waterfall factor into the potential,

$n_s = 0.9728 \pm 0.0029$6

with $n_s = 0.9728 \pm 0.0029$7, following the waterfall-modulated $n_s = 0.9728 \pm 0.0029$8-attractors of Kallosh et al. [Kallosh:2026kfx]. The modulation modifies the inflationary evolution near the waterfall without touching the dark-energy regime: since $n_s = 0.9728 \pm 0.0029$9, the waterfall contributes only an overall constant $w_0 \sim -0.8$0 to the late-time plateau, absorbable into a redefinition $w_0 \sim -0.8$1 whose shift is negligible compared to the uncertainty in $w_0 \sim -0.8$2 itself.

The key result is that the effective e-fold number $w_0 \sim -0.8$3 induced by the waterfall controls the observables,

$w_0 \sim -0.8$4

so that $w_0 \sim -0.8$5 becomes flexible while the relation $w_0 \sim -0.8$6 is preserved. For $w_0 \sim -0.8$7, $w_0 \sim -0.8$8, $w_0 \sim -0.8$9, varying $w_a \sim -0.6$0 raises $w_a \sim -0.6$1 from $w_a \sim -0.6$2 to $w_a \sim -0.6$3 at $w_a \sim -0.6$4 — covering the ACT+DESI-preferred range. Crucially, unlike earlier mechanisms that raised $w_a \sim -0.6$5 via prolonged kination (which conflicts with reheating constraints), the waterfall achieves this independently of the kination duration.

Dark energy phenomenology versus DESI

Using the autonomous-system formalism of Copeland et al. as implemented in Zhumabek et al., the authors evolve the thawing field from frozen initial conditions $w_a \sim -0.6$6 with $w_a \sim -0.6$7. Two robust findings emerge:

• Thawing relation: in CPL parametrization, the Exp II model satisfies $w_a \sim -0.6$8 robustly against changes in parameters, initial conditions, and small deformations of the potential. The closest achievable point is approximately $w_a \sim -0.6$9, which does not reach the DESI best-fit region.
• No phantom crossing: a canonical single-field model cannot cross $\Lambda$0, so no tuning within this class reproduces the DESI phantom-crossing solution. The authors state this plainly rather than claiming otherwise.

Nevertheless, direct comparison of $\Lambda$1 curves with the DESI confidence band from Jing et al. shows that the interpolating family covers the light-blue $\Lambda$2 region: the best-fit Exp II configuration is $\Lambda$3, $\Lambda$4, reproduced identically with and without the waterfall, confirming that the inflationary modification leaves dark energy untouched. For smaller $\Lambda$5, all $\Lambda$6 curves sit above the best fit. The interpolating models are positioned for upcoming Euclid, Rubin, Roman, and full DESI/DES data if those deviate from DESI toward $\Lambda$7CDM.

Inflaton-axion scenario for phantom-crossing-compatible dark energy

If DESI's phantom crossing survives, the authors sketch an embedding of the axion-dilaton quintessence model of Toomey et al. [Toomey:2025yuy, Chudaykin:2026amr] into hyperbolic inflaton-axion $\Lambda$8-attractors. The essential geometric point is the sign of the axion metric $\Lambda$9: matching the Toomey-Chudaykin kinetic term requires $\alpha$0 in the $\alpha$1 (dark energy) regime, where $\alpha$2. This choice has a dual benefit:

• During inflation at large positive $\alpha$3, the inverse metric suppresses the axion gradient ($\alpha$4), freezing the axion via the "rolling on the ridge" effect [Achucarro:2017ing] — the universality of multi-field $\alpha$5-attractors reduces dynamics to a single field.
• During dark energy at negative $\alpha$6, the same factor grows as $\alpha$7, destabilizing the axion exactly when the two-field quintessence dynamics required by DESI begins.

A shifted potential with $\alpha$8 allows the onset of the dark-energy stage to be placed at any desired $\alpha$9 near the axion destabilization point. The authors are explicit that this remains "a plausible scenario that needs further development" — no full numerical two-field analysis is performed here.

Reheating and the dark-radiation bound

Quintessential inflation lacks perturbative reheating; gravitational production of non-conformal fields during the inflation-to-kination transition is the default mechanism. With $T = e^{-\varphi/\sqrt{2/3}} + i\theta$0, $T = e^{-\varphi/\sqrt{2/3}} + i\theta$1, and $T = e^{-\varphi/\sqrt{2/3}} + i\theta$2 GeV, the minimal setup (four Higgs degrees visible; gravitons plus inflaton-axion fluctuations as dark radiation) gives

$T = e^{-\varphi/\sqrt{2/3}} + i\theta$3

comfortably above BBN temperatures but violating the conservative bound $T = e^{-\varphi/\sqrt{2/3}} + i\theta$4; even gravitons alone give $T = e^{-\varphi/\sqrt{2/3}} + i\theta$5. This is the previously recognized problem with Exp-type quintessential attractors [Jing:2026ymp], compounded there because instant preheating shortens kination and thus removes the prolonged-kination route to higher $T = e^{-\varphi/\sqrt{2/3}} + i\theta$6. The waterfall mechanism resolves the second problem, and the paper surveys four resolutions to the first:

Mechanism	$T = e^{-\varphi/\sqrt{2/3}} + i\theta$7	$T = e^{-\varphi/\sqrt{2/3}} + i\theta$8
Instant preheating via $T = e^{-\varphi/\sqrt{2/3}} + i\theta$9	model-dependent	brought within bound
MSSM visible sector ($-3\,\partial T\,\partial\bar T/(T+\bar T)^2$0, RPV decay)	$-3\,\partial T\,\partial\bar T/(T+\bar T)^2$1–$-3\,\partial T\,\partial\bar T/(T+\bar T)^2$2 GeV	$-3\,\partial T\,\partial\bar T/(T+\bar T)^2$3–$-3\,\partial T\,\partial\bar T/(T+\bar T)^2$4
Higgs curvature coupling $-3\,\partial T\,\partial\bar T/(T+\bar T)^2$5, $-3\,\partial T\,\partial\bar T/(T+\bar T)^2$6	$-3\,\partial T\,\partial\bar T/(T+\bar T)^2$7–$-3\,\partial T\,\partial\bar T/(T+\bar T)^2$8 GeV	$-3\,\partial T\,\partial\bar T/(T+\bar T)^2$9–$$V_{\rm exp}(\varphi) = M^2 e^{-2g}\left[e^{g\left(\tanh\frac{\varphi}{\sqrt{6\alpha}}+1\right)} - 1\right] + \Lambda .$$0
Gravitational production + decay of heavy $$V_{\rm exp}(\varphi) = M^2 e^{-2g}\left[e^{g\left(\tanh\frac{\varphi}{\sqrt{6\alpha}}+1\right)} - 1\right] + \Lambda .$$1	model-dependent	suppressed by entropy injection

The MSSM estimate assumes superparticle masses $$V_{\rm exp}(\varphi) = M^2 e^{-2g}\left[e^{g\left(\tanh\frac{\varphi}{\sqrt{6\alpha}}+1\right)} - 1\right] + \Lambda .$$2 GeV below $$V_{\rm exp}(\varphi) = M^2 e^{-2g}\left[e^{g\left(\tanh\frac{\varphi}{\sqrt{6\alpha}}+1\right)} - 1\right] + \Lambda .$$3, $$V_{\rm exp}(\varphi) = M^2 e^{-2g}\left[e^{g\left(\tanh\frac{\varphi}{\sqrt{6\alpha}}+1\right)} - 1\right] + \Lambda .$$4-parity violation with $$V_{\rm exp}(\varphi) = M^2 e^{-2g}\left[e^{g\left(\tanh\frac{\varphi}{\sqrt{6\alpha}}+1\right)} - 1\right] + \Lambda .$$5 for decay before radiation-kination equality, and dark matter arising outside the LSP sector. The Higgs-curvature result depends on the transition profile and the running of the Higgs quartic coupling. These are order-of-magnitude estimates; detailed analyses are deferred.

Fate of the universe

Extending the interpolation to $$V_{\rm exp}(\varphi) = M^2 e^{-2g}\left[e^{g\left(\tanh\frac{\varphi}{\sqrt{6\alpha}}+1\right)} - 1\right] + \Lambda .$$6 yields futures with negative cosmological constant and eventual collapse, echoing the gauged $$V_{\rm exp}(\varphi) = M^2 e^{-2g}\left[e^{g\left(\tanh\frac{\varphi}{\sqrt{6\alpha}}+1\right)} - 1\right] + \Lambda .$$7 supergravity analysis of Kallosh et al. [Kallosh:2002gf]. Because the autonomous variables require $$V_{\rm exp}(\varphi) = M^2 e^{-2g}\left[e^{g\left(\tanh\frac{\varphi}{\sqrt{6\alpha}}+1\right)} - 1\right] + \Lambda .$$8, collapse is tracked with a second-order system in cosmic time based on the Friedmann-Lemaître form of $$V_{\rm exp}(\varphi) = M^2 e^{-2g}\left[e^{g\left(\tanh\frac{\varphi}{\sqrt{6\alpha}}+1\right)} - 1\right] + \Lambda .$$9, regular through the turnaround where $n_s = 0.9682 \pm 0.0032$00; the Friedmann constraint is imposed only on initial data and monitored thereafter.

For $n_s = 0.9682 \pm 0.0032$01, $n_s = 0.9682 \pm 0.0032$02, the outcomes are sharply stratified:

$n_s = 0.9682 \pm 0.0032$03	$n_s = 0.9682 \pm 0.0032$04	Turnaround	Full collapse	Fate
$n_s = 0.9682 \pm 0.0032$05 (Exp I)	$n_s = 0.9682 \pm 0.0032$06	—	—	eternal dS
$n_s = 0.9682 \pm 0.0032$07	$n_s = 0.9682 \pm 0.0032$08	—	—	de Sitter
$n_s = 0.9682 \pm 0.0032$09 (Exp II)	$n_s = 0.9682 \pm 0.0032$10	—	—	Minkowski
$n_s = 0.9682 \pm 0.0032$11	$n_s = 0.9682 \pm 0.0032$12	$n_s = 0.9682 \pm 0.0032$13 Gyr	$n_s = 0.9682 \pm 0.0032$14 Gyr	slow collapse
$n_s = 0.9682 \pm 0.0032$15	$n_s = 0.9682 \pm 0.0032$16	$n_s = 0.9682 \pm 0.0032$17 Gyr	$n_s = 0.9682 \pm 0.0032$18 Gyr	collapse

The diagnostic signature is that collapsing models have $n_s = 0.9682 \pm 0.0032$19 lying above the $n_s = 0.9682 \pm 0.0032$20 curve, mirroring the $n_s = 0.9682 \pm 0.0032$21 pattern. For smaller $n_s = 0.9682 \pm 0.0032$22, even a tiny negative $n_s = 0.9682 \pm 0.0032$23 triggers collapse, and sufficiently negative values would exclude the model by causing collapse before the present epoch. A notable implication: if future observations pin down $n_s = 0.9682 \pm 0.0032$24 (via $n_s = 0.9682 \pm 0.0032$25 and $n_s = 0.9682 \pm 0.0032$26) together with $n_s = 0.9682 \pm 0.0032$27, the position of the observed $n_s = 0.9682 \pm 0.0032$28 relative to the $n_s = 0.9682 \pm 0.0032$29 curve determines whether these models predict eternal expansion or compute a finite lifetime before collapse.

Limitations and open questions

Several caveats are stated in the paper itself. The CMB+DESI value $n_s = 0.9682 \pm 0.0032$30 rests on datasets in significant mutual tension [Ferreira:2025lrd], so the motivation for high $n_s = 0.9682 \pm 0.0032$31 may weaken. The single-field class cannot produce phantom crossing, so its viability against DESI hinges on whether the apparent crossing is an artifact of parametrization or systematics — a question the paper explicitly leaves to Euclid, Rubin, Roman, and full DESI/DES data. The inflaton-axion construction is qualitative: the claim that axion destabilization timing can be matched to the quintessence onset requires numerical verification of the coupled two-field dynamics. The reheating estimates rely on instantaneous-transition approximations, assumed supersymmetric spectra, and model-dependent decay rates. Finally, the collapse-time predictions apply only within this specific potential family and assume the thawing initial conditions adopted from prior work.

Conclusion

The paper updates the quintessential $n_s = 0.9682 \pm 0.0032$32-attractor framework along three axes: a continuous $n_s = 0.9682 \pm 0.0032$33-interpolation between de Sitter, Minkowski, and collapsing futures; a waterfall insertion granting flexible $n_s = 0.9682 \pm 0.0032$34 up to $n_s = 0.9682 \pm 0.0032$35 while preserving $n_s = 0.9682 \pm 0.0032$36 and leaving dark energy invariant; and reheating scenarios satisfying $n_s = 0.9682 \pm 0.0032$37 without requiring prolonged kination. The single-field models span the space between $n_s = 0.9682 \pm 0.0032$38CDM and the DESI best-fit $n_s = 0.9682 \pm 0.0032$39 but cannot cross the phantom divide; a hyperbolic inflaton-axion extension is proposed for that case but remains undeveloped. The framework converts precise future measurements of $n_s = 0.9682 \pm 0.0032$40 and $n_s = 0.9682 \pm 0.0032$41 into a concrete prediction — including, for $n_s = 0.9682 \pm 0.0032$42, the remaining lifetime of the universe before collapse.