{"id":"2c933316-be83-4919-9e66-ea99504e5307","arxiv_id":"2608.06071","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Integrable inflationary models built from a chosen Hubble function yield power spectra, spectral indices, and tensor-to-scalar ratios, computed beyond slow-roll, that match currently available CMB data.","lead":"This paper constructs 'integrable' inflationary models, potentials built from a chosen Hubble function so the background equations become exactly solvable, and computes primordial perturbation spectra for them without the slow-roll approximation. A generalist reader may care because it offers an analytic laboratory for inflationary predictions and a check of how much the slow-roll approximation misses.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The analytic H(a) inversions (3.7)/(5.17) are only real for a below about e^{-3/4}, yet Section 5 integrates at a_c=e^{10} and e-foldings up to 65; the omitted rescaling constant makes the stated perturbation computation ill-defined.","rationale":"The reader's weakest assumption concerned the physical viability of the negative minimum and the hand-set constant in hyperbolic models. Those are real concerns, but they concern the post-inflationary or auxiliary parts of the construction. The more load-bearing problem is that the analytic inversion used to define the background H(a) during the computed e-foldings is only valid on a bounded interval of a, while the perturbation computation is performed at a_c=e^{10} and e-foldings up to 65. This is not a matter of interpretation or of disagreement with the observational consensus; it is an internal inconsistency in the central calculation. For the Starobinsky-like model, the Lambert-function branch in Eq. (3.5) is real only for a ≤ e^{-3/4}, and Eq. (5.17) inherits this domain. For the T-model, Eq. (3.22) becomes negative for a>e, and Eq. (5.27) is not the correct inverse on that domain. A missing rescaling constant is the natural explanation, but it must appear explicitly in H(a) and be fixed by the e-fold normalization. Without it, the numerical integrations claimed in Section 5 have no well-defined background. If the authors supply the corrected H(a) and the recomputed spectra match Tables 2 and 3, the central claim may stand; otherwise the reported observables belong to a different model. This warrants conditional acceptance pending that check. I therefore disagree with the reader's identification of the weakest assumption, while agreeing that the paper should not be unconditionally accepted as is.","tokens_in":19448,"tokens_out":30698,"duration_ms":270233,"concrete_test":"Evaluate W_{-1}(-a^{4/3}) for α=√(2/3) at a=e^5 and a=e^{10}; if no real value exists, replace Eq. (5.17) by H(a)=H_in[1+1/W_{-1}(-(a/C)^{4/3})] with C fixed by the condition N_e=65 (a_e=e^{65}), recompute the Section 5.2 pipeline, and compare n_s, n_T, and r with Table 2. If the corrected H(a) reproduces the table entries, the omission is cosmetic; if the entries shift, the reported observables are not those of the stated model.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim requires H(a) to be the correct background for the scale-factor range used in the Mukhanov-Sasaki integrations. That fails as written. From Eq. (3.5), e^{-αφ}=-W_{-1}(-a^{2α²}); for α=√(2/3), the real branch of W_{-1} requires a^{4/3} ≤ 1/e, i.e. a ≤ e^{-3/4}≈0.47. However, Section 5.1 defines a=e^N with N counted from φ_start and sets N_e=65 at the end of inflation; Section 5.2 chooses a mode with a_c=e^{10} and integrates from a_in=e^5. At these values the argument -a^{4/3} lies far outside [-1/e,0), so Eq. (5.17) is not real. The same problem affects the T-model: Eq. (3.22) has denominator 1-log²a, which turns negative for a>e, and Eq. (5.27) is only the a<1 branch of the inversion. Introducing an integration constant C (e.g. fixing a(φ_start)=1 or a_e=e^{65}) would make the argument -(a/C)^{4/3}, but no such constant appears in the printed H(a) formulas or in the numerical setup. Therefore the power spectra in Tables 2 and 3 cannot be reproduced from the stated equations. This is an internal inconsistency in the core computation, not a slow-roll or consensus disagreement.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a 'fake superpotential' formalism for scalar-field FLRW cosmologies, in which the Hubble function H(φ) determines the potential through V=3H^2−2H_φ^2 and reduces the background dynamics to first-order equations. The authors propose several integrable inflationary models (Starobinsky-like, polynomial, α-attractor/T-model, trigonometric, hyperbolic), invert a(φ) to obtain H(a), and then compute scalar and tensor power spectra by solving the Mukhanov-Sasaki equations in scale-factor time, using numerical and two semi-analytic matching methods. For the Starobinsky-like model they report n_s=0.9653, n_T=−0.0006, r=0.0034, and for the T-model n_s=0.9627, n_T=−0.00254, r=0.0096, comparing these with slow-roll predictions and observational constraints.","tokens_in":19792,"tokens_out":10921,"duration_ms":103435,"significance":"If the computations were valid as printed, the paper would provide a useful non-slow-roll reference class of models whose spectral indices and tensor-to-scalar ratio are genuine predictions, since the only fitted quantity is the overall amplitude H_in, fixed by the CMB normalization. The presentation of three independent solution strategies for the perturbation equations is also a strength. However, the central numerical results are compromised by a domain error in the analytic H(a) inversions used in Sections 5.2 and 5.3, and by an unaddressed negative vacuum energy in several of the proposed models. These issues must be repaired before the reported tables can be taken as valid.","major_comments":[{"comment":"The Lambert-function inversion e^{−αφ}=−W_{−1}(−a^{2α²}) is real only for a^{2α²}≤e^{−1}; with α=√(2/3) this restricts the scale factor to a≤e^{−3/4}≈0.47. However, Section 5.2 defines a=e^N with N counted from φ_start, takes N_e=65 at the end of inflation, and integrates modes with crossing at a_c=e^{10} (see Eq. (5.18) and Figure 6). For these values the argument −a^{4/3} is far outside the real domain of W_{−1}, so Eq. (5.17) is not real on the integration range and the power spectra in Table 2 cannot be reproduced from the stated equations. The integration constant omitted in Eq. (3.4) must be restored (for example, by replacing a with a/C in Eq. (5.17)) and the normalization used in the numerical integration specified.","section":"§3.1, Eq. (3.5); §5.2, Eq. (5.17)"},{"comment":"The T-model inversion has the same domain problem. From Eq. (3.20), cosh(2αφ)=−8α²n log a, so for α=1/4, n=2 the formula (5.27) is real only for log a≤−1, i.e. a≤e^{−1}≈0.37. Section 5.3 nevertheless uses crossing at a_c=e^{10} and integrates from larger values, so the T-model results in Table 3 are not supported by the printed equations. In addition, Eq. (3.22) appears to contain a typo: the correct identity is H/H_in=[(1+8α²n log a)/(8α²n log a−1)]^{n/2} on the inflationary branch, whereas the printed formula uses log² a and has the wrong sign for a<1; this inconsistency between Eq. (3.22) and the special case (5.27) needs to be resolved.","section":"§3.3, Eqs. (3.20)–(3.22); §5.3, Eq. (5.27)"},{"comment":"The Starobinsky-like potential (3.3) is negative at the minimum φ=0, V(0)=−2α²H_in², and the T-model potential (3.19) is likewise negative at φ=0. Since H(0)=0, the background expansion stops at this minimum, and the paper does not discuss the post-inflationary vacuum, reheating, or whether modes are still frozen when the field reaches this region. The definition of the end of inflation by ε_H=1 in Eq. (5.8) does not address these issues. The authors should either modify H(φ) so that V≥0 while preserving integrability, or explicitly justify that the superhorizon perturbation amplitudes are unaffected by the negative-minimum phase and specify a consistent post-inflationary cosmology.","section":"§3.1, Eq. (3.3); §3.3, Eq. (3.19); Figure 2 caption"},{"comment":"The hyperbolic models 'have the problem of infinite inflation', and the proposed remedy is to add a small constant to H and then set it to zero in the analytic inversion because it is 'very little in our work range of values'. This is an uncontrolled approximation: the constant changes the late-time background, the location of ε_H=1, and therefore the number of e-folds and the observable predictions. The authors should quantify the sensitivity of the reported spectral indices to this constant, or exclude the hyperbolic models from the comparison with observations.","section":"§3.5, footnote on page 13"}],"minor_comments":[{"comment":"The text says the final plateau corresponds to 'cosmological constant / “dark energy” (w=1)', but a cosmological constant has w=−1; this appears to be a typo.","section":"§2.4, after Eq. (2.35)"},{"comment":"The name 'Muhkanov' should be 'Mukhanov'.","section":"§4, opening paragraph"},{"comment":"The least-squares formula for dlogP/dlogk is displayed in a compressed form; please write the denominator explicitly as Σ_i (log k_i − log k)^2, and clarify the averaging notation.","section":"§5.1, Eq. (5.15)"},{"comment":"In the row for 'Strings, Curvature', the column header 'T able 1' should be 'Table 1', and in the row for vacuum the deceleration parameter q=−1 as given is correct but should be cross-checked with the displayed column order.","section":"Table 1"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern about the domain of the Lambert-function inversion is real and directly affects the two central numerical tables. I still recommend major revision rather than rejection because the formalism is repairable: restoring the integration constant in a(φ) and H(a) and redoing the integrations would put the claims on solid ground. However, the authors should be asked to provide the explicit rescaled background functions and to ensure that the same normalization is used in the perturbation equations and in the reported e-fold numbers."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: useful paper with a reproducibility bug in the printed equations. The reader's report gives a conditional pass and lists secondary issues, but it misses the main one: the H(a) formulas used in Section 5 are not real on the integration domain.\n\nWhat is new: the specific integrable deformations—Lambert-W for the Starobinsky-like and E-models, logarithmic for the T-model—are new closed forms, and the perturbation computation is cross-checked three independent ways (numerical, Hankel matching, parabolic-cylinder matching) that agree. The only fitted quantity is the overall amplitude H_in; n_s and r are genuine predictions. That is a real plus.\n\nThe soft spot: Eqs. (3.5)/(3.7) drop the integration constant in a(phi). With the normalization used in Section 5, a=exp(N), the maximum of a is e^{-3/4} for the Starobinsky-like model, yet the text integrates modes with a_c=e^{10} and counts N_e=65. At those values the argument of W_{-1} is far outside [-1/e,0), so (5.17) is not real. The T-model has the same problem: (5.27) is the a<1 branch, while the integrations use a up to e^{65}. The correct formulas need an explicit rescaling constant C, i.e., -(a/C)^{4/3}. Without it, Tables 2 and 3 cannot be reproduced from the stated equations. This is a load-bearing flaw, but fixable: the numerics must have been run with an implicit C that never appears in print.\n\nSecondary issues, in decreasing order of concern: the Starobinsky-like potential has a negative minimum at phi=0 and the paper does not discuss reheating or the post-inflationary vacuum; the hyperbolic models rely on an unquantified hand-set constant to stop inflation; the Lambda-CDM equation of state (2.35) is inconsistent as written; the pivot-scale matching claim (5.18)-(5.20) lacks a derivation; and the Hankel tensor tilt in Table 2 is positive, which is unphysical for single-field inflation, though the numerical and other semi-analytic results are negative. All of these are addressable.\n\nThe paper is for readers working on analytic inflationary model building and perturbation theory; they will find explicit integrable models and a cross-checked integration method. The central idea holds up, but as printed the headline numbers are not reproducible. Send it to a referee, and expect a major revision to fix the normalization and re-present the results.","headline":"Useful integrable-inflation toolkit, but the printed H(a) formulas omit the rescaling constant and are not real on the integration domain, making the headline spectra irreproducible as written.","tokens_in":20345,"tokens_out":11043,"would_cite":false,"duration_ms":85420,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that for a class of integrable inflationary models built from the Hubble function, scalar and tensor power spectra can be computed beyond the slow-roll approximation, and that the resulting observables are compatible with…","keywords":["inflation","fake superpotential","Mukhanov-Sasaki equations","power spectrum","spectral index","tensor-to-scalar ratio","integrable cosmologies","beyond slow-roll"],"falsifier":"Run an independent numerical integration of the Starobinsky-like model that does not stop at $\\epsilon_H=1$ but lets the field pass through the negative-potential minimum, then compare the slope of the frozen scalar power spectrum with the reported $n_s=0.9653$; a shift larger than the few-units-in-the-last-digit oscillations quoted in the paper would falsify the claimed beyond-slow-roll result.","tokens_in":19200,"feed_emoji":"🌌","tokens_out":16254,"duration_ms":140994,"temperature":0.7,"pith_summary":"This paper claims that a broad class of inflationary models becomes exactly integrable when the scalar potential is repackaged through the Hubble function $H(\\phi)$, which acts as a fake superpotential. In those models the background equations reduce to a first-order system, and in scale-factor time the Hubble parameter $H(a)$ is known in closed form, so the Mukhanov-Sasaki equations for scalar and tensor perturbations can be integrated numerically and semi-analytically without assuming slow roll. For the Starobinsky-like model the computation gives $n_s=0.9653$, $n_T=-0.0006$ and $r=0.0034$, and for the T-model $n_s=0.9627$, $n_T=-0.00254$ and $r=0.0096$, both inside the observational bounds the paper quotes. If the results are correct, these are viable models and the method supplies exact reference spectra against which slow-roll approximations can be measured.","feed_headline":"Inflation spectra computed beyond slow roll match CMB data","feed_subtitle":"The Starobinsky-like benchmark gives n_s=0.9653 and r=0.0034 by direct integration, not slow-roll shortcuts.","key_machinery":"The load-bearing object is the Hubble function $H(\\phi)$, promoted to a fake superpotential by the relation $V(\\phi)=3H(\\phi)^2-2H_\\phi(\\phi)^2$, which turns the Friedmann equations into the first-order system $\\dot a/a=H$ and $H_\\phi=-\\dot\\phi$. The second piece is scale-factor time $a$: with $J(a)=1/(a^2H(a))$, scalar and tensor perturbations both obey $u_k''(a)+\\bigl(k^2J(a)^2-Z''(a)/Z(a)\\bigr)u_k(a)=0$, where $Z_S(a)=a^2\\sqrt{-2aH'(a)}$ and $Z_T(a)=a^2\\sqrt{H(a)}$. This form has only Fuchsian singularities at $a=0$, so Bunch-Davies initial data can be evolved numerically and also matched to local Hankel or parabolic-cylinder solutions around the horizon-crossing and freezing epochs. For each integrable model the closed-form $H(a)$ makes the entire pipeline explicit.","core_discovery":"The central claim is that giving the Hubble function $H(\\phi)$ rather than the potential $V(\\phi)$ makes a large family of scalar-field cosmologies solvable at the background level. With $V(\\phi)=3H(\\phi)^2-2H_\\phi(\\phi)^2$, the Friedmann equations become $\\dot a/a=H$ and $H_\\phi=-\\dot\\phi$, and for the model families studied the resulting $a(\\phi)$ can be inverted in closed form to give $H(a)$. The paper then rewrites the Mukhanov-Sasaki equations in scale-factor time $a$, where both scalar and tensor perturbations obey $u_k''(a)+\\bigl(k^2/(a^4H(a)^2)-Z''(a)/Z(a)\\bigr)u_k(a)=0$, with $Z$ built from $H(a)$ and its first derivative. Numerically and through piecewise matching with Hankel and parabolic-cylinder functions, it computes the frozen power spectra from Bunch-Davies initial conditions and extracts the spectral indices beyond slow roll. For the Starobinsky-like model it fixes $H_{\\mathrm{in}}=7.817\\times10^{-7}$ from the observed scalar amplitude and finds $n_s=0.9653$, $n_T=-0.0006$, $r=0.0034$; for the T-model it finds $n_s=0.9627$, $n_T=-0.00254$, $r=0.0096$. The paper concludes that these models are viable and that the slow-roll formulas reproduce the exact predictions closely.","pith_inferences":["An extension the paper leaves open is to continue the perturbation integration through the negative-potential minimum of the Starobinsky-like model; that would test whether the reported $n_s$ and $r$ survive the reheating phase.","Because the pipeline only needs $H(a)$ in closed form, it can be applied to models with intermediate plateaus to compute enhanced small-scale power, the route the paper mentions for primordial black holes.","The exact spectra offer a clean calibration of slow-roll truncations: comparing exact and slow-roll values of $n_s$ across the model families would show where the approximation breaks down."],"forward_implications":["The Starobinsky-like model is observationally viable: fixing $H_{\\mathrm{in}}=7.817\\times10^{-7}$ from the scalar amplitude makes the predicted $n_s$, $n_T$ and $r$ fall inside the current bounds quoted in the paper.","The T-model is also viable but predicts a tensor-to-scalar ratio about three times larger than the Starobinsky-like model ($r=0.0096$ versus $r=0.0034$), so future B-mode measurements can separate the two.","The exact beyond-slow-roll indices differ from the slow-roll values by small but quantifiable amounts, giving model-specific estimates of the error in the slow-roll approximation.","The piecewise analytic solutions reproduce the frozen numerical spectra, so a model can be checked with matched local solutions instead of a full numerical integration once analytic $H(a)$ is known.","The same first-order scalar-field description covers inflation, kination, radiation, matter and dark-energy eras through $\\rho(a)=3H(a)^2$, providing one integrable language for the whole cosmological history."],"supporting_citations":[{"why":"Defines the original Starobinsky model whose plateau the integrable Starobinsky-like model is constructed to resemble.","marker":"[1]"},{"why":"Provides the CMB observational constraints on the scalar spectral index and tensor-to-scalar ratio used in the comparison tables.","marker":"[11]"},{"why":"Supplies the scale-factor time formalism and the semi-analytic perturbation techniques that this paper adapts.","marker":"[16]"},{"why":"Gives the standard scalar-field Starobinsky potential that motivates the choice $H(\\phi)=H_{\\mathrm{in}}(1-e^{\\alpha\\phi})$.","marker":"[17]"},{"why":"Is the natural-inflation model generalized by the paper's trigonometric class of integrable potentials.","marker":"[25]"},{"why":"Derives the gauge-invariant scalar perturbation equation used as one of the Mukhanov-Sasaki equations.","marker":"[26]"},{"why":"Provides the quantum treatment of cosmological perturbations behind the tensor equation and Bunch-Davies initial conditions.","marker":"[27]"},{"why":"Introduces the fake superpotential in the domain-wall context that motivates the first-order formulation and its name.","marker":"[30]"}],"fun_headline_variants":["Beyond slow roll: exact inflation spectra from Hubble function","Hubble function unlocks exact inflation perturbations","Integrable inflation: spectra beyond slow roll match data","Exact Mukhanov-Sasaki solutions beyond slow roll","Fake superpotential yields exact inflation observables"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that whatever happens at the end of inflation and during reheating does not change the perturbation amplitudes the model predicts, even though the Starobinsky-like potential is negative at its minimum and the hyperbolic models require a tiny stabilizing constant to be set to zero.","fun_headline_variants_meta":{"raw":{"variants":["Beyond slow roll: exact inflation spectra from Hubble function","Hubble function unlocks exact inflation perturbations","Integrable inflation: spectra beyond slow roll match data","Exact Mukhanov-Sasaki solutions beyond slow roll","Fake superpotential yields exact inflation observables"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00071,"raw_usage":{"total_tokens":3263,"prompt_tokens":1075,"completion_tokens":2188,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":691,"completion_tokens_details":{"reasoning_tokens":2116}},"tokens_in":691,"tokens_out":2188,"duration_ms":28076,"temperature":1.0,"reasoning_tokens":2116,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T15:29:55.631947+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run an independent numerical integration of the Starobinsky-like model that does not stop at $\\epsilon_H=1$ but lets the field pass through the negative-potential minimum, then compare the slope of the frozen scalar power spectrum with the reported $n_s=0.9653$; a shift larger than the few-units-in-the-last-digit oscillations quoted in the paper would falsify the claimed beyond-slow-roll result.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the standard scalar-field Starobinsky potential that motivates the choice $H(\\phi)=H_{\\mathrm{in}}(1-e^{\\alpha\\phi})$."},{"cited_title":"Freese, J","cited_arxiv_id":null,"evidence_quote":"Is the natural-inflation model generalized by the paper's trigonometric class of integrable potentials."},{"cited_title":"Sasaki,Large Scale Quantum Fluctuations in the Inflationary Universe,Prog","cited_arxiv_id":null,"evidence_quote":"Derives the gauge-invariant scalar perturbation equation used as one of the Mukhanov-Sasaki equations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the quantum treatment of cosmological perturbations behind the tensor equation and Bunch-Davies initial conditions."}],"review_version":1}