{"id":"28d207d6-55e7-4019-bb7a-f6ada5a9bb4b","arxiv_id":"2501.11474","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"In JT gravity toy models, late-time modular entropy and capacity of entanglement scale inversely with n times the inverse temperature, supporting a thermal reading of the replica parameter.","lead":"This paper studies how the replica parameter n changes the modular entropy and the capacity of entanglement in two toy models of evaporating black holes. The main finding is that at late times both quantities shrink as n grows, suggesting that more replica copies act like a lower temperature and purify the Hawking radiation more effectively.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The island-model results (3.45)–(3.46) and the generalized island formula (1.5) rely on the leading-order conformal welding solution F=G with no estimate of O(κ) corrections; these corrections may shift the claimed 1/(nβ) scaling.","rationale":"The reader's weakest_assumption already identifies the leading-order conformal welding F=G and the single-QES limit as fragile; I agree that this is the most load-bearing technical step in the island-model derivation. My concern sharpens it by pointing out that the matter modular entropy in Eq. (3.39), which is itself derived using the leading-order welding maps, contains a term of order 1/n that is discarded in Eqs. (3.45)–(3.46) along with the stated O(b/(nβ)) terms. Because the capacity of entanglement involves a derivative with respect to n, this discarded matter contribution is not automatically subleading to 2πφ_r/(nβ) unless β is taken strictly to zero. Since the paper uses finite β when comparing with the EoW numerical results, the parametric control of the neglected terms is unclear. This does not prove the result wrong, but it makes the central finite-n island claim conditional on a check that the paper does not provide. I also noted the apparent inconsistency in Eq. (1.4), where C_n is written as −∂_nS_mod but used as −n∂_nS_mod elsewhere; this appears to be a typo, since the calculations consistently use −n∂_nS_mod, so it is not the load-bearing issue. Given that the paper explicitly flags the leading-order nature of the welding solution and lists higher-order corrections as future work, a conditional acceptance remains appropriate pending the proposed verification.","tokens_in":34939,"tokens_out":11882,"duration_ms":116122,"concrete_test":"Solve the conformal welding problem perturbatively in κ: expand F=F_0+κF_1 and G=G_0+κG_1 around Eq. (3.20), impose the welding condition (3.17c) and the boundary EOM (B.6) to first order in κ, and recompute S_island_mod and C_n via Eqs. (3.38)–(3.46) without discarding the matter term. If the O(κ) corrections modify S_mod or C_n by an amount O(1/(nβ)) rather than O(κ/(nβ)), the leading-order result is not under control and the φ_r=2π match with the EoW numerics is not robust. A complementary numerical check is to solve Eq. (B.15) for the boundary mode θ(τ) without setting F=G at κ=0.1, β=3, φ_r=2π and compare the resulting modular entropy with Eq. (3.45).","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central island-model claim—that modular entropy and capacity of entanglement are controlled by nβ (Eqs. 3.45–3.46) and that the island formula generalizes to finite n (Eq. 1.5)—is derived in the high-temperature weak-coupling limit κ≡cβG_N/(24πφ_r)≪1, where the conformal welding problem is solved only at leading order by setting F=G (Eq. 3.20). The paper does not compute the O(κ) corrections to F and G, nor does it show that these corrections are suppressed relative to the leading 1/(nβ) terms. Moreover, in going from Eq. (3.39) to Eqs. (3.45)–(3.46), the matter modular entropy is evaluated at leading order and then partially discarded: the cutoff-dependent term (c/6n)log[β/(πϵϵ_UV^2)] is dropped together with O(b/(nβ)) terms. Since C_n=−n∂_nS_mod, a matter contribution of the form (c/6n)×const contributes to C_n at order 1/n, which is not parametrically smaller than 2πφ_r/(nβ) except in the strict β→0 limit. At the values used to match the EoW numerical results (φ_r=2π, β=3), there is no demonstrated parametric control over either the welding corrections or the discarded matter terms. If the first-order welding correction changes the QES location or the modular entropy by order 1/(nβ), the claimed scaling and the matching with the EoW model fail.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies how the modular entropy and the capacity of entanglement depend on the replica parameter n in two JT-gravity settings: the End-of-the-World (EoW) model and the eternal two-sided black hole coupled to a thermal bath (the island model). For the EoW model it derives n-dependent Page-curve-like evolution in the microcanonical and canonical ensembles, with late-time canonical results S_mod^late = S0 + 4π^2/(nβ) and C_n^late = 4π^2/(nβ). For the island model it constructs a finite-n replica geometry, computes the modular entropy and capacity in a high-temperature, weak-coupling limit, obtains late-time expressions of the form 2πφ_r/(nβ) plus subleading terms, and proposes a finite-n generalization of the island formula, S_R(n) = min ext[Σ(S0+φ_n(∂I))+S_mod(R∪I)]. The n→1 limits are stated to reduce to known results, and the late-time analytic expressions in the canonical EoW model are checked against numerical plots.","tokens_in":35363,"tokens_out":10076,"duration_ms":101887,"significance":"If the main claims hold, the paper would extend the replica-wormhole and island formalism beyond the n→1 limit and provide a suggestive statistical-mechanics interpretation of the replica parameter n as an inverse temperature, potentially giving a practical finite-n island formula. The paper has clear strengths: the EoW microcanonical and canonical calculations are explicit, the n→1 limits reproduce known results, and the analytic late-time approximation in the canonical EoW model is consistent with the numerical curves. However, the central island-model results—the 1/(nβ) scaling and the proposed finite-n island formula—are derived under a leading-order conformal-welding approximation and rely on discarding n-dependent matter terms whose suppression is not demonstrated. As a result, the main island-model claims should be regarded as provisional until these approximations are controlled.","major_comments":[{"comment":"The matter modular entropy contribution retained in (3.39) contains a term c/(6n) log[β/(πϵ ϵ_UV^2) e^{-2πb/β}] (or with the sign of the b-exponent possibly reversed; see minor comment). In passing to (3.45) and (3.46) this term is discarded as O(b/(nβ)). The n-dependence of this log term is not O(b/(nβ)): the constant part log[β/(πϵ ϵ_UV^2)] is independent of b and yields a contribution to C_n of order c/(6n), which is not parametrically smaller than the claimed leading term 2πφ_r/(nβ) except in the strict β→0 limit. At the parameters used to match the EoW results (φ_r=2π, β=3), there is no demonstrated suppression. The claimed equality between the island-model results and the EoW canonical results, and the universal 1/(nβ) scaling, are therefore not established.","section":"§3.4, Eqs. (3.45)–(3.46)"},{"comment":"The finite-n island calculation is performed at leading order in the high-temperature, weak-coupling limit by setting the conformal welding functions F=G (Eq. 3.20) and by using the resulting QES location a→∞ (Eq. 3.44). No estimate is given for the O(κ) corrections to F, G, the QES position, or the modular entropy. Since κ≡cβG_N/(24πφ_r)≪1 is the only small parameter, and since the central results (3.45), (3.46), and the proposed island formula (1.5) are obtained at this order, the 1/(nβ) scaling is not robust against the first corrections. The paper explicitly acknowledges this limitation in Section 4, but the issue is load-bearing and should be addressed with an explicit error estimate or by reformulating the claims as leading-order results.","section":"§3.3–3.4, Eqs. (3.20) and (3.44)"},{"comment":"The capacity of entanglement is defined as C_n = -n∂_n S_mod under the assumption that the gravitational saddle used to define the modular entropy remains smooth under infinitesimal changes of n. This is an additional assumption beyond the quantum-extremal-surface prescription: for non-integer n the replica geometry is a Z_n orbifold with conical singularities, and it is not established that the dominant saddle varies smoothly or that the analytic continuation in n is well-behaved. Since C_n is a central output of the paper, this assumption should be justified, or at least its failure modes should be discussed.","section":"§3.2, Eq. (3.15)"}],"minor_comments":[{"comment":"The sign of the b-dependent exponent in the log term appears inconsistent with the large-a limit of Eq. (3.39): taking cosh(2π(a+b)/β) - 1 ≈ (1/2)e^{2π(a+b)/β} and sinh(2πa/β) ≈ (1/2)e^{2πa/β} gives a factor e^{+2πb/β} inside the logarithm, not e^{-2πb/β}. Please check whether the sign is a typo and ensure the final expressions are consistent.","section":"§3.4, Eq. (3.45)"},{"comment":"The microcanonical expression for t∈[0,1] is extended to t≥1 simply by replacing t with 1/t. A short justification would be helpful, for example by noting that the delta-function contribution λ'=0 vanishes for n>0 in the integrals (2.23)–(2.24), so the replacement is consistent with the density (2.21b); as written, the step is stated without proof.","section":"§2.2, Eq. (2.26)"},{"comment":"The early-time expansion of S_n^(early) uses the relation ∫dλ D(λ)(λ-1/k)^n = n Z2/Z1^2 in (2.47). The validity of this relation for the canonical-ensemble density is not fully explained, and the order of the expansion in (2.46) would benefit from a more explicit derivation.","section":"§2.3, Eqs. (2.46)–(2.47)"},{"comment":"The manuscript contains numerous typos and minor language issues, including \"calcluation\" in §2.1, \"satae\" in Eq. (2.12), \"dose\" in §3.1, and \"recovers\" versus \"recover\" in §4. A careful proofread is needed before publication.","section":"General"},{"comment":"The statement that n replica AdS2 disks at inverse temperature β glue into a single disk at inverse temperature nβ is an interpretation, not a derived consequence of the calculations. The paper should clearly distinguish this heuristic picture from the explicit results, especially when it is used to motivate the finite-n island formula (1.5).","section":"§4, Figure 11"}],"recommendation":"major_revision","confidential_remarks":"The paper's central island-model claims are close to those in Ref. [79] (Hollowood, Kumar, Piper), which the authors cite for the strict proof of the modular generalized entropy. The present derivation in Appendix B is complementary, but the overlap and the precise division of credit should be clarified. Given the acknowledged neglect of O(κ) conformal-welding corrections and the problematic treatment of the n-dependent matter term, the main claims of the island section need substantial revision before the paper can be accepted."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a useful, honest, incremental paper. The n-dependent EoW curves and the explicit late-time island expressions are new relative to the n=1 literature, and the reduction to known results works. But the island-model core rests on a leading-order conformal welding solution and a dropped matter term that is the same order in 1/(nβ) as the claimed leading term; I would want that fixed or argued before trusting the scaling.\n\nWhat it does well: the microcanonical and canonical EoW calculations are carried through carefully, with n-dependent Page curves and capacity curves that are not in the earlier n=1 papers. The paper is self-aware, lists higher-order corrections and multi-QES configurations as future work, and correctly credits [79] for the generalized modular entropy and the finite-n island construction. The n=1 limits reproduce known results, and the φ_r=2π match to the EoW numerics is a nice cross-check. I did not rederive every hypergeometric identity, but the checks I did are consistent.\n\nWhere I would push: (i) Eq. (3.20) sets F=G at leading order in κ, with no estimate of the first correction to the modular entropy or the QES location. (ii) The log term in Eq. (3.45) has a 1/n coefficient, and its cutoff-dependent part contributes to C_n at order 1/n, the same parametric order as 2πφ_r/(nβ) unless β is taken strictly to zero. Calling it O(b/(nβ)) does not settle the issue, since b/β can be order one. (iii) The nβ coupling in the island answers partly comes from the normalization φ̃_r=φ_r/n, so the thermal interpretation of n is more built in than derived. (iv) The CoE requires the saddle to remain smooth under n-variation, which is assumed.\n\nNone of these make the paper incoherent. The EoW half stands on its own, and the island half is a plausible extension that needs a sharper error budget. This paper is for people working on replica wormholes, modular entropy, and capacity of entanglement; it is not a broad-audience piece. I would send it to a serious referee and ask them to focus on the conformal welding corrections and the dropped matter term. Conditional accept, not reject.","headline":"A solid, incremental extension of the replica wormhole program whose island-model 1/(nβ) scaling needs an error estimate before it is taken at face value.","tokens_in":35806,"tokens_out":3018,"would_cite":true,"duration_ms":36357,"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":"This paper establishes that at late times the modular entropy and the capacity of entanglement in both the End-of-the-World model and the island model of JT gravity are controlled by the product $n\\beta$, falling as $1/(n\\beta)$, and uses…","keywords":["replica wormholes","modular entropy","capacity of entanglement","JT gravity","island formula","Rényi entropy","Hawking radiation","Page curve"],"falsifier":"Compute the modular entropy including the next-order corrections to the conformal welding maps $F$ and $G$ in Eq. (3.20), or include a second quantum extremal surface in the extremization of Eq. (1.5). If the prefactor of $1/(n\\beta)$ in $S_{\\rm mod}^{\\rm island}$ or $C_n^{\\rm island}$ acquires an $n$-dependent coefficient, or if the extremum shifts away from $a\\to\\infty$, the claimed universal scaling and the finite-$n$ island formula would be contradicted.","tokens_in":34708,"feed_emoji":"🕳️","tokens_out":15941,"duration_ms":135173,"temperature":0.7,"pith_summary":"This paper asks what happens to entanglement measures of black hole radiation when the number of copies $n$ in the replica trick is kept finite instead of being sent to $1$ at the end of the calculation. Working in two tractable models of black hole evaporation in Jackiw–Teitelboim (JT) gravity—the End-of-the-World brane model and the eternal two-sided black hole coupled to a bath—it computes the modular entropy $S_{\\rm mod}$ (a replica-parameter generalization of entropy) and the capacity of entanglement $C_n$ (a measure of how widely the entanglement spectrum is spread) as functions of $n$. The central result is that at late times both quantities are set by the product $n\\beta$: they decrease as $1/(n\\beta)$, so increasing the number of replica copies acts like lowering the temperature. Because the ordinary entanglement entropy is only the $n\\to1$ limit, keeping $n$ finite exposes how each additional replica copy reshapes the radiation state, information the von Neumann entropy cannot see. On this basis the paper reads $n$ as an inverse temperature in the replica ensemble and proposes a finite-$n$ island formula that reduces to the standard one at $n=1$.","feed_headline":"Adding extra copies shrinks late-time black hole entanglement","feed_subtitle":"Two models agree: modular entropy and capacity of entanglement fall as one over replica count times temperature.","key_machinery":"The load-bearing objects are the replica trick at general $n$, the modular entropy $S_{\\rm mod}=n^2\\partial_n[(n-1)S_n/n]$, and the capacity of entanglement $C_n=-n\\partial_n S_{\\rm mod}$. In the gravitational path integral, $\\mathrm{Tr}(\\rho_R^n)$ is evaluated by summing over replica geometries, and the competition between the fully disconnected saddle and the fully connected replica wormhole saddle controls the Page-like behavior. For the island model, the finite-$n$ generalized entropy is $S_{\\rm gen}(n)=\\sum_{\\partial I}(S_0+\\phi_n(\\partial I))+S_{\\rm mod}^{\\rm CFT}(R\\cup I)$, extremized over the island boundary to locate the quantum extremal surface, with the conical-singularity dilaton $\\phi_n=2\\pi\\phi_r/(n\\beta)\\tanh(2\\pi\\sigma/(n\\beta))$ replacing the area term. The technical bottleneck is the conformal welding problem—finding the holomorphic maps $F$ and $G$ that glue the gravitational disk to the bath—solved at leading order by setting $F=G$ in the high-temperature limit $\\kappa=c\\beta G_N/(24\\pi\\phi_r)\\ll1$. The interpretive identity is that $n$ disks at inverse temperature $\\beta$ glue into a single disk at inverse temperature $n\\beta$.","core_discovery":"The paper establishes a late-time statement shared by two models: the modular entropy and the capacity of entanglement are controlled by $n$ times the inverse temperature $\\beta$. In the canonical End-of-the-World model it obtains $S_{\\rm mod}^{\\rm late}=S_0+4\\pi^2/(n\\beta)$ and $C_n^{\\rm late}=4\\pi^2/(n\\beta)$ (Eqs. (2.54)–(2.55)), while in the single-island configuration of the eternal JT black hole it obtains $S_{\\rm mod}^{\\rm island}=S_0+2\\pi\\phi_r/(n\\beta)+O(b/(n\\beta))$ and $C_n^{\\rm island}=2\\pi\\phi_r/(n\\beta)+O(b/(n\\beta))$ (Eqs. (3.45)–(3.46)). The capacity of entanglement therefore does not decay to zero at late times in these settings; it saturates at a positive value that decreases as $n$ or $\\beta$ grows. The paper interprets the coupling as geometric: $n$ replicated AdS$_2$ disks, each at inverse temperature $\\beta$, glue along their boundaries into one disk at inverse temperature $n\\beta$, so $n$ acts as an inverse temperature in the replica ensemble. Using that interpretation, it generalizes the island formula to finite $n$ as $S_R(n)=\\min\\,\\mathrm{ext}\\big[\\sum_{\\partial I}(S_0+\\phi_n(\\partial I))+S_{\\rm mod}(R\\cup I)\\big]$ (Eq. (1.5)), which returns to the standard island formula when $n\\to1$.","pith_inferences":["Beyond the paper: if the $n\\beta$ identification is exact, the finite-$n$ island formula could be recast as a statement about a single black hole at temperature $nT$; a direct test would be to compute the modular entropy at fixed $n\\beta$ while varying $n$ and check whether all data collapse onto one curve.","Beyond the paper: because the capacity of entanglement behaves like a heat capacity in the replica ensemble, it is a sharper diagnostic of the Page transition than the von Neumann entropy; its microcanonical peak rises with $n$, a signature that could be sought in random-matrix or tensor-network models of evaporation.","Beyond the paper: applying the finite-$n$ island formula (1.5) to other dilaton-gravity, cosmological, or de Sitter island setups would predict the same $1/(n\\beta)$-type saturation; observing a different $n$ dependence in those settings would delimit how much of the effect is special to JT gravity."],"forward_implications":["In the canonical EoW model, the late-time modular entropy and capacity of entanglement both decrease as either the replica number $n$ or the inverse temperature $\\beta$ increases, with the explicit forms $S_{\\rm mod}^{\\rm late}=S_0+4\\pi^2/(n\\beta)$ and $C_n^{\\rm late}=4\\pi^2/(n\\beta)$.","In the microcanonical EoW ensemble, the modular entropy follows an $n$-dependent Page curve and saturates to $S_0$ at late times, while the capacity of entanglement peaks at the Page time and then decays to zero, with the peak value growing with $n$.","For the single-island JT configuration, the modular entropy and capacity of entanglement saturate at $S_0+2\\pi\\phi_r/(n\\beta)$ and $2\\pi\\phi_r/(n\\beta)$, respectively, matching the canonical EoW late-time behavior when $\\phi_r=2\\pi$.","The proposed finite-$n$ island formula (1.5) reduces to the standard island formula as $n\\to1$ and gives a concrete handle on how the island phase purifies black hole radiation at finite replica number.","Across both models, configurations with more connected replica wormholes—or with an island—produce lower modular entropy and capacity, which the paper reads as evidence that more $n$ copies mean more effective purification of thermal radiation."],"supporting_citations":[{"why":"Supplies the End-of-the-World model, the planar resolvent equations, and the disconnected/replica-wormhole saddle competition that this paper extends to general $n$.","marker":"[13]"},{"why":"Supplies the replica wormhole island construction and the conformal welding problem whose leading-order solution is used in the finite-$n$ calculation.","marker":"[14]"},{"why":"Supplies the two-sided eternal JT black hole coupled to a bath and the coordinate frames on which the island computation is built.","marker":"[16]"},{"why":"Supplies the capacity-of-entanglement calculation in a related JT setting that the $n\\to1$ limit of this paper recovers.","marker":"[77]"},{"why":"Supplies the generalized modular entropy formula for JT gravity that underlies the finite-$n$ island generalization.","marker":"[79]"},{"why":"Supplies the definition of modular entropy and its relation to Rényi entropy used throughout the paper.","marker":"[125]"},{"why":"Supplies the holographic identification of modular entropy with the area of a cosmic brane, motivating the dilaton-area term in the modular generalized entropy.","marker":"[132]"}],"fun_headline_variants":["Replicas shrink late-time entanglement in JT gravity","Entanglement capacity decays as 1/(n beta) in two models","Modular entropy and capacity fall with extra copies","Island formula extended to finite replica number","Late-time entanglement tamed by n-fold replicas"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The late-time results assume a hot, weakly coupled black hole in which the gluing of the replica copies is handled only to leading order and a single island boundary sits exactly at the horizon; if higher-order gluing corrections or additional island boundaries matter, the $1/(n\\beta)$ scaling and the proposed finite-$n$ island formula could fail.","fun_headline_variants_meta":{"raw":{"variants":["Replicas shrink late-time entanglement in JT gravity","Entanglement capacity decays as 1/(n beta) in two models","Modular entropy and capacity fall with extra copies","Island formula extended to finite replica number","Late-time entanglement tamed by n-fold replicas"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000637,"raw_usage":{"total_tokens":3021,"prompt_tokens":1115,"completion_tokens":1906,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":731,"completion_tokens_details":{"reasoning_tokens":1829}},"tokens_in":731,"tokens_out":1906,"duration_ms":14122,"temperature":1.0,"reasoning_tokens":1829,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T18:14:42.309602+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the modular entropy including the next-order corrections to the conformal welding maps $F$ and $G$ in Eq. (3.20), or include a second quantum extremal surface in the extremization of Eq. (1.5). If the prefactor of $1/(n\\beta)$ in $S_{\\rm mod}^{\\rm island}$ or $C_n^{\\rm island}$ acquires an $n$-dependent coefficient, or if the extremum shifts away from $a\\to\\infty$, the claimed universal scaling and the finite-$n$ island formula would be contradicted.","supporting_citations":[],"review_version":1}