{"id":"a48507a5-11f5-4d51-b9d7-8a7a16c27fae","arxiv_id":"2507.03610","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"In modular inflation models, entropic perturbations frozen during inflation are converted into curvature perturbations after inflation, producing an enhanced power spectrum while preserving the spectral index ns.","lead":"Cosmologists show that in certain two-field inflation models, small fluctuations in a second field, frozen during inflation, can be converted into the density fluctuations we observe in today's cosmic microwave background only after inflation ends, boosting their size. This added boost changes the predicted amount of gravitational waves, so comparing these models with CMB data requires including the first moments after inflation.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed post-inflationary plateau is not established beyond N=3; the paper's own j-function model shows continued growth, and the same long-time test is not reported for the main models.","rationale":"The reader's weakest assumption identifies exactly the load-bearing point: the enhancement plateau must be stable for the derived observables to hold. The manuscript itself supplies the strongest evidence for this concern in Appendix A.1, where the j-function modular model—a model of the same class—is shown to lack a stable plateau for θin=0.3, with growth continuing beyond N=3. The main-text Dedekind and hyperbolic models are only integrated up to N≈3, so the claim that they 'always give rise to a plateau' is an extrapolation. The physical explanation given for saturation is the oscillatory cancellation of the source term and the decay of S; this mechanism is not proven and is contradicted by the j-function case. If the 'plateau' in the main models is actually a transient, the reported enhancement factor E, and the statements ns≈1−2/Nhc and r≈12α/(E Nhc^2), would not be the final predictions. The proposed test would resolve this directly by extending the integration. This does not change the reader's conditional verdict: the paper ought to provide longer-time evidence (ideally with public code) before the plateau and the derived observables are accepted. I agree with the reader's weakest assumption and recommend no change to the verdict.","tokens_in":21801,"tokens_out":7945,"duration_ms":95083,"concrete_test":"Extend the numerical integration of the full perturbation system for the Dedekind (Eq. 2.1) and hyperbolic (Eq. 2.11) models from N=3 to at least N=10–20 for the initial conditions θin=0.3, 0.01 and, for the modular model, the high-enhancement region θin≈0.35. Check whether the curvature power spectrum at N=3 agrees within ~10% with its value at later times and whether the sourced component P_R^s remains flat rather than growing monotonically. As a control, run the same extension for the j-function model (Eq. A.1) at θin=0.3 and verify the reported continued growth. An independent implementation using a public multi-field code (e.g., evolving canonical variables to avoid σ̇→0 coordinate singularities) would settle the issue.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central observational claims (enhanced As with preserved ns and reduced r) depend on the curvature power spectrum reaching an asymptotic plateau within the first few post-inflationary e-folds. The paper stops all numerical evolution at N≈3 (§4.1) and the only extended-time check it reports is negative: for the j-function modular model with θin=0.3, 'both the curvature and isocurvature perturbations continue to grow up to N∼3 and beyond' and 'the plateau ... is not stable if we evolve for a longer amount of e-folds' (Appendix A.1, Fig. 10). For the Dedekind and hyperbolic potentials, the paper asserts that a plateau forms, but no evolution beyond N=3 is shown to confirm it is asymptotic rather than a transient inflection before later growth. The proposed saturation mechanism—the source term η⊥S oscillates and S decays—is heuristic; the j-function model demonstrates that this mechanism can fail in a model of the same class. Since E, ns, and r are read off at N≈3, continued growth would change all three. The absence of public code and the order-of-magnitude reheating treatment (Sec. 4.3) further prevent independent verification of the saturation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies two-field inflationary models on a hyperbolic field-space manifold, namely a modular-invariant model based on the Dedekind eta function and a simpler hyperbolic toy potential. During inflation, the axionic direction θ is frozen and the turning rate is negligible, so the curvature power spectrum follows the single-field α-attractor prediction. The main new claim is that after inflation the trajectory turns sharply, the entropic perturbations source the adiabatic ones, and the curvature power spectrum reaches an enhanced plateau within about three e-folds. The paper defines an enhancement factor E, reports that the spectral index remains approximately 1 − 2/Nhc while the tensor-to-scalar ratio is reduced by E, and gives order-of-magnitude reheating constraints. An appendix applies the same analysis to the j-function modular model and finds a plateau only for some initial conditions.","tokens_in":22026,"tokens_out":5378,"duration_ms":68116,"significance":"If the saturation claim is correct, the paper establishes a novel and observationally relevant mechanism: multi-field models with negligible turning during inflation can nevertheless have their curvature power spectrum strongly modified after the end of inflation, with ns preserved but r suppressed by E. The numerical treatment solves the full background and linear perturbation system without slow-roll assumptions, and the super-horizon equations are derived exactly, which are genuine strengths. The paper is also honest about the counterexample in Appendix A.1, where the j-function model does not plateau for θin = 0.3. The significance, however, hinges entirely on the plateau being asymptotic before reheating, and that point is not yet demonstrated for the main models. The paper explicitly stops all evolution at N ≃ 3 and does not report longer-time runs for the Dedekind and hyperbolic potentials.","major_comments":[{"comment":"The central claim that the curvature power spectrum reaches a stable plateau is supported only up to N = 3. All main-text numerical evolutions stop at N ≃ 3, and the only extended-time check reported is negative: for the j-function model with θin = 0.3, the paper states that 'both the curvature and isocurvature perturbations continue to grow up to N ∼ 3 and beyond' and that 'the plateau ... is not stable if we evolve for a longer amount of e-folds'. Since E, ns, and r are read off at N = 3, continued growth after N = 3 would change all three observational predictions. The authors should either evolve the Dedekind and hyperbolic models for enough e-folds to demonstrate an asymptotic plateau, or provide an analytical criterion for saturation and explicitly bound the residual growth after N = 3.","section":"§4.1 and Appendix A.1, Figs. 7 and 10"},{"comment":"The argument that npost_s = nend_s assumes that the enhancement factor E is independent of Nhc. The paper's heuristic estimate of δθ in Eq. (4.6) treats |N| as constant over a Hubble time, which is only a leading-order statement, and the numerical check in Fig. 9 is performed only at N = 3 and shows small deviations from exact N²hc scaling. If E acquires any Nhc dependence or if the spectrum continues to grow after N = 3, the spectral index after enhancement will differ from 1 − 2/Nhc by terms of order 1/N²hc or larger. The authors should quantify these corrections or demonstrate that E is Nhc-independent to the accuracy needed for the claimed ns and r predictions.","section":"§4.2, Eqs. (4.6)–(4.8) and Fig. 9"},{"comment":"The reheating analysis bounds only a perturbative decay rate Γ ≲ H/10 and does not address parametric resonance, which the paper itself notes can occur within O(10) oscillations. Since the claimed plateau forms within the first one to three oscillations, the statement 'we do not expect any reheating-related effects at this early stage' is not a quantitative check. A dedicated estimate of nonperturbative particle production in these models, or at least an explicit statement that the plateau result is conditional on negligible resonant effects during the first few oscillations, is needed before the observational claims can be considered robust.","section":"§4.3, Reheating considerations"}],"minor_comments":[{"comment":"The text reads 'a set ofzweibeins'; this should be 'a set of zweibeins'.","section":"§2.3, after Eq. (2.18)"},{"comment":"The caption refers to the 'analytical formula (first equation in Eq. (2.21))', but Eq. (2.21) defines the turning rate; the analytical large-φ estimates are in Eq. (2.28). The reference should be corrected.","section":"Fig. 4 caption"},{"comment":"The derivation of δθ = (H/2π) sqrt(2/(3α)) Nhc treats |N| as constant over a Hubble time. This should be flagged as a leading-order estimate, since |N| changes by ΔN = 1 and the omitted corrections are of the same order as the deviations visible in Fig. 9.","section":"Eq. (4.6)"},{"comment":"The symbol η is used for conformal time in Eq. (3.23) and also appears in η⊥, ηH, and ηD. Consider using a distinct symbol for conformal time to avoid confusion.","section":"Global notation"},{"comment":"Since the numerical results are a central part of the paper, a short statement on code and data availability would improve reproducibility; no public code is mentioned.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a serious numerical study with an honest statement of its main limitation in Appendix A.1. The missing piece is a demonstration that the plateau is asymptotic rather than a transient for the Dedekind and hyperbolic models. This is fixable within the paper's scope, so I recommend major revision rather than rejection. The heavy reliance on the authors' earlier work is appropriate given the direct continuation of Ref. [13]."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The short version: this is a credible, clearly written claim that in modular two-field models where the inflaton follows a geodesic during inflation, the turning that happens after inflation ends converts frozen entropic perturbations into an enhanced curvature power spectrum, with ns preserved and r suppressed by the enhancement factor E. The physics is plausible and the numerics look careful. But the plateau, which is the load-bearing part, is only demonstrated to N≈3, and the paper's own j-function appendix (A.1, Fig. 10) shows a case in the same model class where the plateau is not stable. That makes the observational conclusions conditional, not settled.\n\nWhat is actually new: the general sourcing of adiabatic from isocurvature modes is known multifield lore, and the citations there are fair. The new content is the demonstration that this sourcing becomes efficient after inflation in SL(2,Z)/modular potentials where the turning during inflation is negligible. The super-horizon equations (3.12)-(3.13) are derived without slow-roll assumptions and remain valid for ϵ>1, which is the right tool. The j-function appendix deserves credit: the authors show a failure case and say plainly that the plateau is not stable there.\n\nThe soft spots, in proportion. First, plateau stability: the numerical evolution stops at N≈3 in §4.1, and no long-time check is reported for the Dedekind or hyperbolic models. The j-function case shows the same mechanism can keep growing. The saturation argument, that the source η⊥S oscillates and S decays, is plausible but heuristic, and their own appendix shows it can fail within the same class. I would want a run to at least N≈10–20, past several oscillations, before trusting E, ns, and r as stated. Second, the enhancement E is purely numerical; an analytical handle on the saturation would help considerably. Third, no code or data is released, so independent verification of the coupled-mode numerics is not possible. Fourth, the reheating treatment is order-of-magnitude, which the authors admit; it establishes that a viable reheating window exists, which is enough for the main claim but not a sharp prediction.\n\nThe self-citation pattern is not a problem here: this is a direct follow-up to Ref. [13], and the post-inflationary computation is done independently rather than assumed.\n\nWho this is for: people working on modular inflation, alpha-attractors, and multifield dynamics. If the effect survives longer evolution, it changes how future CMB experiments interpret r limits in these models.\n\nRecommendation: send it to peer review. The central question is well-posed and a referee can usefully push on the exact soft spots above. Conditional acceptance is the right frame, not rejection.","headline":"Credible, clearly written demonstration of post-inflationary adiabatic enhancement in modular models, but the load-bearing plateau is only checked to N=3 and the paper's own j-function appendix shows the same mechanism failing to saturate.","tokens_in":22547,"tokens_out":6168,"would_cite":true,"duration_ms":62931,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["98.80.Cq"],"model":"deepseek-v4-flash","headline":"The paper claims that in multi-field modular inflation with negligible turning during inflation, entropic perturbations can be efficiently converted into curvature perturbations after inflation ends, producing a new enhanced plateau in…","keywords":["modular cosmology","SL(2,Z) invariance","alpha-attractors","multi-field inflation","adiabatic-entropic decomposition","curvature power spectrum","isocurvature perturbations","post-inflationary evolution"],"falsifier":"Evolve the same perturbation system for ten or more e-folds after inflation, including a small dissipative term to mimic reheating, and check whether the curvature power spectrum remains flat; in particular, for the $j$-function model at $\\theta_{in}=0.3$ the paper's own Figure 10 shows the power spectrum still growing at $N=3$, so a longer run that continues to grow would falsify the plateau claim.","tokens_in":2084,"feed_emoji":"🌌","tokens_out":2893,"duration_ms":109487,"temperature":0.7,"pith_summary":"Cosmological models with two scalar fields usually convert entropic (isocurvature) perturbations into curvature perturbations while the field-space trajectory turns during inflation. This paper claims that the conversion can instead happen after inflation, in models where the trajectory is geodesic during inflation and the entropic direction is nearly massless, because the turning rate becomes large once the fields begin oscillating toward their minima. In the modular-invariant and hyperbolic potentials studied numerically, the curvature power spectrum grows into a new, enhanced plateau within a few oscillations after inflation. If correct, the CMB amplitude $A_s$ is enhanced by a model- and initial-condition-dependent factor, the spectral index stays at the $\\alpha$-attractor value $n_s \\approx 1 - 2/N_{hc}$, and the tensor-to-scalar ratio is correspondingly reduced.","feed_headline":"After inflation, curvature spectrum jumps by orders of magnitude","feed_subtitle":"Frozen entropic modes convert into curvature after inflation, keeping the spectral tilt and lowering r.","key_machinery":"The load-bearing object is the adiabatic-entropic decomposition of the two-field system, with unit vectors along and orthogonal to the background trajectory, together with the turning rate $\\eta_\\perp$, which measures how much the field-space trajectory bends per Hubble time. On super-horizon scales the curvature perturbation obeys $\\dot{\\mathcal R} = 2H\\eta_\\perp \\mathcal S$ up to a term suppressed by $a^3$, so any significant turning converts isocurvature $\\mathcal S$ into curvature $\\mathcal R$. The models are built so that $\\eta_\\perp$ is exponentially small during inflation but jumps to large values after inflation, making the post-inflationary conversion efficient. The potentials are a modular-invariant one based on the Dedekind eta function, with double-exponential flatness in the axion direction, and a simpler hyperbolic-tangent toy model that reproduces the same qualitative behavior; the perturbations are evolved numerically from Bunch-Davies initial conditions through the first three e-folds after inflation.","core_discovery":"The central claim is that multi-field inflationary models with negligible turning in field space during inflation can nevertheless generate a large, effective sourcing of adiabatic from entropic perturbations immediately after inflation ends. The paper demonstrates this with two scalar fields in a hyperbolic field-space metric: during inflation the axion-like field $\\theta$ is frozen because its potential is extremely flat, so its perturbations are isocurvature modes that stay decoupled and retain an amplitude proportional to $N_{hc}^2$; once $\\varphi$ oscillates around its minimum and $\\theta$ rolls to its own minimum, the turning rate $\\eta_\\perp$ becomes of order 10 to 100, converting isocurvature into curvature power through the super-horizon relation $\\dot{\\mathcal R} = 2H\\eta_\\perp \\mathcal S$. The resulting curvature power spectrum reaches a plateau roughly one to three e-folds after inflation, with an enhancement $\\mathcal{E} = P_{\\mathcal R}^s/P_{\\mathcal R}^0$ that depends strongly on the initial value of $\\theta$. The authors conclude that $A_s$ is set by the enhanced value, that the spectral index remains $n_s \\approx 1 - 2/N_{hc}$ because both vacuum and sourced pieces scale as $N_{hc}^2$, and that $r$ is suppressed by $\\mathcal{E}$ relative to the single-field $\\alpha$-attractor prediction.","pith_inferences":["Beyond the paper, if post-inflationary conversion of this kind is generic, then the standard single-field attractor predictions for $A_s$ and $r$ should be re-derived for any model with a light, frozen entropic direction, not only modular or hyperbolic potentials.","Beyond the paper, the plateau-stability assumption is the most exposed link: running the perturbation evolution past $N=3$ with reheating friction would either confirm that the plateau is a genuine attractor or reveal that the continued growth seen for the $j$-function model at some initial angles is the generic behavior, which would alter the predicted $n_s$ and $r$.","Beyond the paper, the strong sensitivity to $\\theta_{in}$ suggests that the initial axion angle, often treated as an irrelevant detail, becomes a physical parameter controlling both the amplitude and the level of non-Gaussianity; future CMB measurements of $r$ and $f_{\\rm NL}$ could in principle constrain it."],"forward_implications":["If the plateau is real, the scalar amplitude must be matched to the COBE normalization after enhancement, so the underlying potential scale $V_0$ is smaller by a factor $\\mathcal{E}$ than in single-field models with the same $A_s$.","The scalar spectral index remains $n_s \\approx 1 - 2/N_{hc}$, identical to single-field $\\alpha$-attractors, because both the vacuum and sourced curvature amplitudes inherit the same $N_{hc}^2$ scaling.","The tensor-to-scalar ratio is predicted to be $r \\approx 12\\alpha/(\\mathcal{E} N_{hc}^2)$, suppressed by the enhancement factor; a measurement of $r$ below the standard $\\alpha$-attractor prediction, with unchanged $n_s$, would be the observational signature.","The enhancement is strongly initial-condition dependent; for the modular potential, the sharp dependence at $\\theta_{in} \\lesssim 0.4$ suggests potentially large non-Gaussianity, which the authors leave to future work.","Reheating can be neglected during the first oscillations for weakly coupled fields, but the analysis constrains perturbative reheating temperatures to roughly $10^{12}$-$10^{13}$ GeV for the runs at $N=3$ to be valid, and non-perturbative resonance, if present, would occur after the plateau forms."],"supporting_citations":[{"why":"Supplies the analytic expectation of geodesic motion and double-exponential flatness in the axion direction that this paper tests numerically and extends beyond the end of inflation.","marker":"[13]"},{"why":"Introduces the SL(2,Z) cosmological attractor setup and the class of modular potentials used here.","marker":"[11]"},{"why":"Provides the axion-stabilization mechanism and potentials used in Appendix A.2 to contrast stabilized trajectories against the enhanced-sourcing scenario.","marker":"[16]"},{"why":"Defines the adiabatic-entropic decomposition of multi-field perturbations that underlies the entire analysis.","marker":"[40]"},{"why":"Supplies the relation between the comoving curvature perturbation and the adiabatic field perturbation, and the associated power-spectrum formalism.","marker":"[41]"},{"why":"Provides the super-horizon evolution and sourcing equations for curvature and isocurvature perturbations that the paper uses beyond inflation.","marker":"[43]"}],"fun_headline_variants":["Curvature boost after inflation: frozen isocurvature converts","Post-inflationary e-folds amplify curvature power","Modular inflation: entropic modes feed curvature after end","Curvature spectrum plateau from post-inflationary conversion","Frozen isocurvature modes become curvature after inflation"],"cache_read_input_tokens":24704,"weakest_assumption_plain":"The argument stands on the assumption that the conversion from entropic to curvature perturbations stops growing and stabilizes within the first few oscillations after inflation; the numerical runs stop at about three e-folds after inflation, and the paper itself notes a case in the $j$-function model where the power spectrum is still rising at that point, so if the plateau is not stable the predicted enhancement and observables change.","fun_headline_variants_meta":{"raw":{"variants":["Curvature boost after inflation: frozen isocurvature converts","Post-inflationary e-folds amplify curvature power","Modular inflation: entropic modes feed curvature after end","Curvature spectrum plateau from post-inflationary conversion","Frozen isocurvature modes become curvature after inflation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00071,"raw_usage":{"total_tokens":3188,"prompt_tokens":930,"completion_tokens":2258,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":546,"completion_tokens_details":{"reasoning_tokens":2176}},"tokens_in":546,"tokens_out":2258,"duration_ms":20037,"temperature":1.0,"reasoning_tokens":2176,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:05:38.040789+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evolve the same perturbation system for ten or more e-folds after inflation, including a small dissipative term to mimic reheating, and check whether the curvature power spectrum remains flat; in particular, for the $j$-function model at $\\theta_{in}=0.3$ the paper's own Figure 10 shows the power spectrum still growing at $N=3$, so a longer run that continues to grow would falsify the plateau claim.","supporting_citations":[{"cited_title":"Bartolo, S","cited_arxiv_id":null,"evidence_quote":"Supplies the relation between the comoving curvature perturbation and the adiabatic field perturbation, and the associated power-spectrum formalism."}],"review_version":1}