{"id":"e9effc39-8bd9-4d7e-acc7-ce54a5aee44f","arxiv_id":"1908.08785","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"In single-field inflation, conservation of the comoving curvature perturbation implies conservation of the unitary and synchronous curvatures, but not the reverse, with observable consequences in ultra-slow-roll and a new braiding model.","lead":"This paper shows that three common ways to define the curvature perturbation in single-field inflation, comoving, unitary, and synchronous, can evolve differently after leaving the horizon, and that a constant comoving curvature forces the other two to freeze as well. It applies this to ultra-slow-roll and a new braiding-ultra-slow-roll model, with consequences for how inflation predictions are mapped to observable fluctuations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Conservation-hierarchy proof in Sec. III discards homogeneous solutions of the first-order relations (37)-(38) and (42)-(43); unless physical initial conditions kill those modes, ζ_c conservation does not by itself imply conservation of ζ_s or ζ_u.","rationale":"The reader's weakest_assumption correctly identifies the ignored integration constants in the conservation proof. My stress-test sharpens this into explicit homogeneous modes: for ζ_s the discarded term -H C makes ζ_s evolve at order ε even when ζ_c is exactly constant, and for ζ_u the discarded mode grows whenever ε/Γ is positive, with ζ_c = 0. These are not mere technicalities: the USR calculation in Section IV.A must retain a boundary contribution to get the nonzero conserved ζ_s, otherwise Eq. (43) with zero initial velocity yields ζ_s = 0. The decisive question is whether the homogeneous modes are physical solutions of the full second-order Mukhanov-Sasaki equation or are eliminated by the vacuum/initial data. The paper does not address this, so the central hierarchy is less secure than the abstract suggests. However, because the paper openly acknowledges the integration-constant caveat and because the concern could be resolved in the authors' favor if the homogeneous modes are unphysical, the appropriate verdict remains conditional rather than reject. I therefore leave the reader's CONDITIONAL verdict unchanged.","tokens_in":21681,"tokens_out":26246,"duration_ms":289054,"concrete_test":"Run a stable Horndeski example with constant αB ≠ 0 and constant ε (for instance αB = 0.5 and ε = 0.01, after checking Qs > 0 and cs^2 > 0 in Eqs. (46)-(47)): numerically solve the full Mukhanov-Sasaki equation (45) with Bunch-Davies initial data, and separately evolve the homogeneous mode of Eq. (31) with ζ_c = 0 and ζ_u ∝ exp(∫ ε/Γ d ln a). If this homogeneous mode is a nonzero solution of the full linearized system, then ζ_c conservation does not imply ζ_u conservation, and the central claim fails as stated. If the mode is excluded only for Bunch-Davies data but allowed for other admissible initial velocities, the paper must state the implication as a statement about particular initial conditions rather than a general property.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that conservation of ζ_c outside the horizon forces conservation of both ζ_s and ζ_u rests on Eqs. (38) and (43), which invert the first-order relations (31) and (32) by integrating ζ'_c. These are particular solutions; the general solutions contain homogeneous modes. For ζ_s the discarded mode is ζ_s = -H C, so for an exactly conserved ζ_c one has ζ_s' = ε(ζ_s - ζ_c). This is order ε, not O(x^2), so ζ_s is not conserved unless ε→0 or C=0. For ζ_u the homogeneous mode satisfies ζ_u ∝ exp(∫ ε/Γ d ln a) with ζ_c = 0, again giving a non-conserved ζ_u alongside a conserved ζ_c. The paper suppresses these modes by the statement in Sections III.A and III.B that 'all perturbations vanish initially, at least in their Hubble time average,' but it does not prove that this is true for Bunch-Davies or any other standard initial state, nor does it check these modes against the full second-order Mukhanov-Sasaki equation (45), which is the only way to decide whether they are physical. The USR example itself shows that boundary terms are not negligible: Eq. (54) retains the full history to obtain the nonzero conserved ζ_s of Eq. (55); if the ignored constants in Eq. (43) are literally set to zero for ζ_c = ζ_u ∝ a^3, one obtains ζ_s = 0 instead. Thus the claimed implication is established only under a non-trivial, unproven initial-condition assumption.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies three gauge-invariant curvature perturbations in single-field inflation—unitary ζ_u, comoving ζ_c, and synchronous ζ_s—within Horndeski theory and the effective field theory of inflation. It argues that conservation of ζ_c outside the sound horizon implies conservation of ζ_s and ζ_u, whereas the converse holds only under additional conditions. Two models are analyzed: ultra-slow-roll (USR), where ζ_c = ζ_u grows while ζ_s is conserved, and a new braiding-ultra-slow-roll (BUSR) model, where ζ_u and ζ_s are conserved but different while ζ_c grows. The paper then examines consequences for the separate-universe approximation, the δN formalism, and the observable curvature power spectrum, finding in a numerical BUSR example that modes leaving the horizon near a braiding transition acquire a suppressed final amplitude rather than simply inheriting ζ_u at horizon crossing.","tokens_in":22047,"tokens_out":6807,"duration_ms":71293,"significance":"If the core implication is established, the paper provides a useful clarification: the common practice of treating conservation of ζ_c as equivalent to conservation of other curvature variables is too permissive in non-slow-roll or braided single-field models. The USR and BUSR examples are instructive, and the numerical demonstration that the observable power spectrum after inflation can differ from ζ_u at horizon crossing is a nontrivial, falsifiable consequence. The paper also makes a crisp conceptual point that separate universe is tied to synchronous slicing and can fail through the continuity equation even when ζ_s is conserved. The presentation is mostly clear, and the analytic derivations are transparent; the numerical integrations are specified with a full parameter set, making the main quantitative result reproducible in principle. The main weakness is that the general conservation hierarchy in Section III rests on an unproven initial-condition assumption for the homogeneous solutions of the first-order relations.","major_comments":[{"comment":"The derivation that a conserved ζ_c forces conservation of ζ_u and ζ_s solves first-order differential relations by integrating ζ'_c and drops the homogeneous solutions. The paper's statement in Sections III.A and III.B that all perturbations vanish initially, at least in their Hubble time average, is an assumption rather than a derived property of Bunch-Davies or any other standard initial state, and it is not checked against the full Mukhanov-Sasaki equation (45), which is the equation that would decide whether these modes are physical. For a conserved ζ_c, Eq. (32) has the homogeneous solution ζ_s ∝ exp(-∫ ε d ln a), while Eq. (31) admits homogeneous modes ∝ exp(∫ ε/Γ d ln a), which can grow and are not O(x^2). Unless these modes are killed by the physical initial conditions, ζ'_s and ζ'_u will contain O(ε) contributions rather than the claimed O(x^2) corrections. The USR example itself illustrates the role of history/boundary terms: Eq. (54) retains the full solution (52) to obtain the non-zero conserved ζ_s in Eq. (55), whereas simply dropping the homogeneous terms in Eq. (43) with ζ_c = ζ_u ∝ a^3 would give a different result. Thus the central claim that conservation of ζ_c generically implies conservation of ζ_s and ζ_u is established only under a non-trivial, unproven initial-condition assumption.","section":"Section III.A and III.B, Eqs. (37)-(38) and (42)-(43)"}],"minor_comments":[{"comment":"The symbol u is used for the exponential weighting in Eq. (37) and also for the Mukhanov variable in Eq. (45); this notational clash should be fixed, for example by renaming the exponential factor.","section":"Eq. (37)"},{"comment":"The opening sentence says the section studies the relationship between comoving and unitary curvatures, but the section actually treats comoving and synchronous curvatures; the text should be corrected.","section":"Section III.B, first sentence"},{"comment":"There is a duplicated 'the the' near the introduction of the unitary curvature, and 'cosomological' should be 'cosmological' in Section V.B; a careful proofreading pass would remove these errors.","section":"Section I and Section V.B"},{"comment":"The figure's arrow labels are under-explained; the sentence in the text that labels indicate quantities which should be ≲ O(1) would be clearer if each arrow were explicitly defined in the caption.","section":"Fig. 1"},{"comment":"The claim that the δN formalism yields the correct ζ_u in USR is asserted rather than demonstrated; a short explicit computation using Eqs. (67) and (73) would make the claim reproducible and easier to check.","section":"Section V.A"},{"comment":"The numerical example is described for a single parameter set, and the text notes that the x_t^{3/2} scaling is model-dependent; stating the initial field value φ_I and the e-fold interval between the attractor and the transition would strengthen reproducibility.","section":"Section V.B"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid and potentially useful contribution, and I do not see grounds for rejection. The main issue is that the conservation-hierarchy proof needs a rigorous treatment of the boundary/homogeneous modes; this is a fixable gap if the authors can show that standard initial conditions suppress these modes or can reformulate the claim with the necessary conditions stated. The numerical section is illustrative rather than exhaustive, and should be framed accordingly."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThis paper does something worth doing: it separates three curvature variables that are routinely conflated in single-field inflation and shows they do not have to behave the same way outside the horizon. The genuinely new piece is the conservation hierarchy—conservation of comoving curvature implies conservation of unitary and synchronous curvature, while the reverse directions generally fail—and the new BUSR model, a braiding analog of ultra-slow-roll, where ζ_u and ζ_s are each conserved but take different values while ζ_c grows. The USR example is clean: only ζ_s is conserved, and the separate-universe/δN discussion is more careful than what usually appears in the literature. The BUSR transition producing a scale-dependent power suppression for modes leaving the horizon near the braiding transition is a nice, in-principle observable effect.\n\nThe paper is honest about its main soft spot: the general conservation proof in Section III integrates the first-order relations (37)–(38) and (42)–(43) and drops integration constants by assuming all perturbations vanish initially, at least in their Hubble-time average. The stress-test note is right that the discarded homogeneous modes are not obviously O(x^2); with ζ_c exactly conserved, the residual evolution of ζ_s is order ε rather than x^2, and the USR example itself shows that boundary terms matter—equation (54) needs the full history to get the nonzero conserved ζ_s. So the hierarchy is established only under an initial-condition assumption, not from conservation of ζ_c alone. I do not think this sinks the paper: the standard Bunch-Davies state plausibly kills the spurious homogeneous modes, and the examples are solved from the full Mukhanov-Sasaki equation. But a referee should ask for either a proof that standard states satisfy the assumption or a softened statement of the theorem.\n\nThe quantitative BUSR suppression—roughly x^{3/2}, up to 15%—comes from a single tuned parameter set and is explicitly model-dependent. Treat it as an illustration, not a robust prediction. For anyone working on USR, primordial black holes, or EFT inflation, this is a useful reference. I would bring it to a reading group and I would cite it if I were writing on curvature conservation. It deserves serious peer review; the caveat is real but addressable.","headline":"A genuinely useful unpacking of ζ_c, ζ_u, and ζ_s in single-field inflation; the central hierarchy is sound in the examples, but the general proof carries an explicit initial-condition caveat that needs tightening.","tokens_in":22515,"tokens_out":14632,"would_cite":true,"duration_ms":162563,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"In inflation, comoving curvature conservation forces the other two curvatures to freeze.","keywords":["curvature perturbations","single-field inflation","Horndeski theory","ultra-slow-roll inflation","braiding parameter","separate universe","delta N formalism","primordial power spectrum"],"falsifier":"Numerically solve the Mukhanov-Sasaki equation for a single-field Horndeski model with a transient phase in which $\\varepsilon$ decays and $\\alpha_B$ is large, and compare the super-horizon evolution of $\\zeta_u$, $\\zeta_c$, and $\\zeta_s$; if $\\zeta_c$ is conserved while either $\\zeta_u$ or $\\zeta_s$ grows with time, the paper's central hierarchy is refuted. For the braiding-ultra-slow-roll prediction, a direct test is measuring the small-scale scalar power spectrum: the claimed suppression, up to about 15 percent with a slope close to $x_t^{3/2}$, should appear for modes crossing the sound horizon near the braiding transition, and a scale-invariant spectrum across that range would falsify the model.","tokens_in":2103,"feed_emoji":"🌌","tokens_out":2484,"duration_ms":73439,"temperature":0.7,"pith_summary":"This paper asks when super-horizon curvature perturbations are genuinely conserved during single-field inflation. It shows that the three standard gauge-invariant curvatures—unitary, comoving, and synchronous—obey a one-way hierarchy: conservation of the comoving curvature forces both unitary and synchronous curvatures to be conserved, but conserving either of the other two does not force the comoving curvature to freeze. In ultra-slow-roll inflation only the synchronous curvature is conserved, which clarifies why the separate-universe picture is only mildly violated even though the comoving curvature grows. In a new braiding-ultra-slow-roll model, unitary and synchronous curvatures freeze to different values while the comoving curvature grows, and the power spectrum observable after inflation is set not by the unitary curvature at horizon exit alone but by how the braiding phase ends. Choosing the 'wrong' curvature therefore changes inflationary predictions for the primordial spectrum.","feed_headline":"Comoving curvature conservation rules the other two in inflation","feed_subtitle":"Ultra-slow-roll and braided models show why the choice of gauge changes the predicted primordial spectrum.","key_machinery":"The moving parts are the three gauge-invariant curvatures—unitary $\\zeta_u$ (spatial curvature on slices where the inflaton is unperturbed), comoving $\\zeta_c$ (on slices comoving with the total fluid), and synchronous $\\zeta_s$ (seen by free-falling observers initially at rest with the expansion)—together with the two exact Horndeski relations $\\zeta_c = \\zeta_u - (\\Gamma/\\varepsilon)\\zeta_u'$ and $\\zeta_c = \\zeta_s + \\zeta_s'/\\varepsilon$, where $\\varepsilon$ is the slow-roll parameter and $\\Gamma = \\alpha_B/(2-\\alpha_B)$ encodes the braiding coupling. These identities convert a conservation statement for one variable into a differential relation for another. The conservation hierarchy follows from integrating these relations in the long-wavelength limit, so that frozen $\\zeta_c$ leaves no room for $\\zeta_u$ or $\\zeta_s$ to drift, while frozen $\\zeta_u$ or $\\zeta_s$ leaves an unfixed integration that allows $\\zeta_c$ to grow. The paper also uses the Mukhanov-Sasaki equation and the $\\delta N$ formalism to connect these gauge differences to observables.","core_discovery":"The paper's central claim is that among the three curvature variables $\\zeta_u$, $\\zeta_c$, and $\\zeta_s$, conservation outside the sound horizon is not a property of the perturbations themselves but of the chosen gauge. Within Horndeski single-field theories, if $\\zeta_c$ is constant then $\\zeta_s$ and $\\zeta_u$ are constant as well, because conservation of $\\zeta_c$ integrates to fix the other two; the reverse implication fails, and can fail dramatically. In ultra-slow-roll inflation, $\\zeta_u = \\zeta_c$ grows as $a^3$ while $\\zeta_s$ freezes; in the braiding-ultra-slow-roll model the authors construct, $\\zeta_u$ and $\\zeta_s$ both freeze but take different values, $\\zeta_c$ grows, and the comoving curvature that seeds structure after inflation differs from the unitary curvature at horizon crossing, with a scale-dependent suppression of power near the sound horizon at the braiding transition.","pith_inferences":["An implication the authors leave implicit is that in any non-slow-roll phase with a time-dependent sound speed, the common practice of equating the horizon-exit value of $\\zeta_u$ with the post-inflation observable should be re-checked mode by mode, since exit and freeze-in can occur at different epochs.","The scale-dependent suppression of the braiding-ultra-slow-roll power spectrum is a testable signature: a feature in the small-scale scalar spectrum, with slope close to $x_t^{3/2}$ for the example parameters, could be searched for in primordial black hole abundance or spectral distortion constraints.","The one-way hierarchy suggests a practical diagnostic for model builders: if the comoving curvature is frozen, gauge subtleties in translating variables are harmless, whereas if it is not frozen, each observable must be computed in the gauge in which it is defined."],"forward_implications":["If $\\zeta_c$ is conserved in a single-field Horndeski model, then both $\\zeta_u$ and $\\zeta_s$ are conserved and approach the same value, making the usual translation from inflationary variables to post-inflation initial conditions safe.","In ultra-slow-roll inflation, using $\\zeta_c = \\zeta_u$ as the super-horizon variable gives a growing curvature, and the separate-universe picture fails in the continuity equation even though $\\zeta_s$ is conserved.","Despite that mild separate-universe violation, the $\\delta N$ formalism still gives the correct $\\zeta_u$ power spectrum in ultra-slow-roll, but it breaks down for the uniform-density curvature $\\zeta_\\rho$.","In braiding-ultra-slow-roll inflation, modes that leave the sound horizon less than about one e-fold before braiding vanishes receive a suppression in the final power spectrum, up to about 15 percent for the parameters shown.","When braiding vanishes before the end of inflation, $\\zeta_u(t_{\\rm end}) = \\zeta_c(t_{\\rm end})$, so the post-inflation comoving curvature is still determined by unitary curvature, but only after evolving it through the braiding transition rather than freezing it at horizon exit.","Conservation of $\\zeta_c$ is the most restrictive condition: it implies conservation of both $\\zeta_u$ and $\\zeta_s$, whereas the reverse does not hold."],"supporting_citations":[{"why":"Establishes the Effective Field Theory of inflation in unitary gauge, the framework for defining $\\zeta_u$.","marker":"[3]"},{"why":"Defines the Horndeski action, the general single-field class for which the conservation relations are derived.","marker":"[17, 18]"},{"why":"Introduces the EFT parameters $\\alpha_B$, $\\alpha_K$, $\\alpha_M$, $\\alpha_T$ used to express the braiding coupling.","marker":"[19]"},{"why":"Provides the conserved-unitary-curvature condition $\\zeta_u' / \\zeta_u = c_s^2 O(x^2)$ used as the long-wavelength starting point.","marker":"[20]"},{"why":"Defines ultra-slow-roll inflation, the model where $\\zeta_u = \\zeta_c$ grows while $\\zeta_s$ is conserved.","marker":"[11]"},{"why":"Gives the standard USR result that $\\zeta_u$ grows and the associated $\\delta N$ squeezed bispectrum relation.","marker":"[21]"},{"why":"Links the synchronous curvature $\\zeta_s$ to the separate-universe condition seen by free-falling observers.","marker":"[4]"},{"why":"Supplies the separate-universe validity conditions used to show that USR mildly violates the continuity equation.","marker":"[12]"},{"why":"Gives the standard $\\delta N$ prescription that predicts the correct $\\zeta_u$ but the incorrect $\\zeta_\\rho$ in USR.","marker":"[13]"},{"why":"Motivates the braiding transition by showing that persistent braiding can spoil reheating.","marker":"[14, 15]"}],"fun_headline_variants":["Gauge choice decides which curvature freezes in inflation","Comoving conservation is the strictest: it freezes the other two","USR inflation: only synchronous curvature survives the horizon","Braiding USR: two curvatures freeze, but with different values","Which curvature seeds the universe? Inflation says it depends"],"cache_read_input_tokens":24704,"weakest_assumption_plain":"The proof that conservation of $\\zeta_c$ propagates to $\\zeta_u$ and $\\zeta_s$ assumes that all perturbation fields vanish initially, at least in their Hubble-time average, so the integration constants dropped in equations (37) and (42)–(43) are zero; if a model starts with non-vanishing long-wavelength perturbations, boundary terms could break the claimed hierarchy.","fun_headline_variants_meta":{"raw":{"variants":["Gauge choice decides which curvature freezes in inflation","Comoving conservation is the strictest: it freezes the other two","USR inflation: only synchronous curvature survives the horizon","Braiding USR: two curvatures freeze, but with different values","Which curvature seeds the universe? Inflation says it depends"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000286,"raw_usage":{"total_tokens":1726,"prompt_tokens":1030,"completion_tokens":696,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":646,"completion_tokens_details":{"reasoning_tokens":612}},"tokens_in":646,"tokens_out":696,"duration_ms":7404,"temperature":1.0,"reasoning_tokens":612,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:39:15.967592+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Numerically solve the Mukhanov-Sasaki equation for a single-field Horndeski model with a transient phase in which $\\varepsilon$ decays and $\\alpha_B$ is large, and compare the super-horizon evolution of $\\zeta_u$, $\\zeta_c$, and $\\zeta_s$; if $\\zeta_c$ is conserved while either $\\zeta_u$ or $\\zeta_s$ grows with time, the paper's central hierarchy is refuted. For the braiding-ultra-slow-roll prediction, a direct test is measuring the small-scale scalar power spectrum: the claimed suppression, up to about 15 percent with a slope close to $x_t^{3/2}$, should appear for modes crossing the sound horizon near the braiding transition, and a scale-invariant spectrum across that range would falsify the model.","supporting_citations":[{"cited_title":"Separate Universes beyond General Relativity","cited_arxiv_id":"1612.02454","evidence_quote":"Gives the standard $\\delta N$ prescription that predicts the correct $\\zeta_u$ but the incorrect $\\zeta_\\rho$ in USR."}],"review_version":1}