{"id":"5b449e09-2621-4951-ac51-3979c1ad59d9","arxiv_id":"2505.22239","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Finite-size corrections to excess entropy integrals from the RDF are derived, and the running integral plus a corrected RDF is shown to match thermodynamic integration at low densities.","lead":"This paper shows how to compute excess entropy from radial distribution functions with finite-size corrections, and that a simple running integral converges fastest. It also shows that RDF-based excess entropy matches thermodynamic integration at low density only if the RDF is extrapolated to the thermodynamic limit.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equations 2 and 4 omit a factor 1/V in the identity between the double integral and the single-integral weight function, making the central finite-volume derivation internally inconsistent.","rationale":"Read in good faith, the paper's contribution is the finite-volume extrapolation of excess-entropy integrals and an MC check showing that Eq. 1 with g∞(r) reproduces TI within a few percent for ρ<0.1. The MC comparison is supported by six tables and the analytic model is useful. The weakest point is not the statistical claim but the definition of X(L): Eq. 4 conflates the double integral over two particle positions with the single weighted integral that actually defines the finite-volume KB integral. The known result (Krüger et al.) is G_V = V^{-1}∫∫q, so the paper's notation is off by V. This is likely a typographical convention slip rather than a fatal numerical error, because later formulas use the single-integral definition consistently. Nevertheless, since Eqs. 2 and 4 are where the finite-volume expressions are introduced and from which the 1/L Taylor expansion is motivated, the manuscript should be corrected before final acceptance. The empirical '95%' recommendation is not overturned by this finding, so the conditional verdict stands.","tokens_in":17932,"tokens_out":20626,"duration_ms":220893,"concrete_test":"Evaluate both sides of Eq. 4 for q(r)=1 over a sphere of diameter L; the left side is V^2=(πL³/6)^2 and the right side is V=πL³/6. For a nontrivial check, Monte Carlo integrate ∫_V∫_V exp(-|r1-r2|) dr1 dr2 over the sphere and compare with 4π∫_0^L w(r,L) exp(-r) dr. If the ratio is V, the equations need an explicit 1/V, or X(L) must be redefined as G_V; if the ratio is 1, the printed identity is valid. This single test settles the internal consistency of the central derivation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Eq. 4 defines X(L) = ∫_V∫_V q(|r1-r2|) dr1 dr2 = 4π∫_0^L w(r,L) q(r) dr. This equality is false. For q=1, the left side is V^2 = (πL³/6)^2 while the right side is V = πL³/6. The quantity 4π∫_0^L w q dr is the finite-volume Kirkwood-Buff type integral G_V = V^{-1}∫∫q, not the raw double integral. Consequently Eq. 2's first equality differs from its second equality by a factor V, and the Taylor extrapolation in 1/L (Eqs. 6-12) is applied to a quantity whose connection to the double integral is undefined. This is load-bearing because the paper's theoretical result that the running integral X*(L) converges faster for excess entropy rests on these finite-volume definitions. The MC simulations use Eq. 1 directly, so the empirical 95% claim may survive after correcting the definitions, but the derivation as printed is internally inconsistent and needs a stated convention.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper investigates finite-size effects in computing excess entropy Sex from the second-order density expansion, i.e., from the radial distribution function (RDF). It derives finite-volume expressions for Sex and Kirkwood-Buff (KB) integrals using a spherical weight function, proposes Taylor extrapolations in 1/L up to third order, and studies their convergence using an analytic model RDF. It then presents Monte Carlo NVT simulations of the Wang-Ramírez-Dobnikar-Frenkel potential for densities 0.01–0.8 and system sizes N=100 and N=500, comparing Sex from thermodynamic integration with Sex obtained from integrating the RDF with and without the Ganguly–van der Vegt finite-size correction. The central claims are that the running integral X*(L) converges fastest for the excess-entropy integrand, that the corrected RDF g∞(r) is needed for accurate low-density results in small systems, and that Eq. (1) used with g∞(r) captures about 95% of Sex for ρ<0.1, while deviations up to about 20% occur at higher densities due to the low-density approximation inherent in Eq. (1).","tokens_in":18179,"tokens_out":13941,"duration_ms":141884,"significance":"If the finite-volume definitions are corrected, the paper provides practically useful guidance: for ρ≲0.1, small systems with finite-size-corrected RDFs give excess entropy within a few percent of thermodynamic integration, while at higher densities the second-order density expansion is the dominant source of error. The analytic comparison of convergence orders for qS and qKB is a useful contribution, and the algebra of the Taylor extrapolation in Eqs. (6)–(12) is correct. A strength is that the simulation results are benchmarked against thermodynamic integration, an independent method, and the statistical uncertainties are reported to be below 10^-3. The main weaknesses are the factor-V inconsistency in the central finite-volume definitions, the unclear RDF integration protocol in the simulations, and an incorrect-looking derivative expression in Eq. (17). These issues are fixable but currently prevent the theoretical and numerical parts from being evaluated as a single consistent story.","major_comments":[{"comment":"The relation between the double integral and the single-integral weight function is missing a factor V. With q(r)=1 and a spherical volume of diameter L, the left side of Eq. (4) equals V^2=(πL^3/6)^2, while the right side equals V=πL^3/6. The standard finite-volume Kirkwood-Buff relation is ∫∫ q dr1dr2 = V · 4π∫ w(r,L) q(r) dr, or equivalently X(L)=V^{-1}∫∫ q dr1dr2 = 4π∫ w q dr. As printed, Eq. (2) is also dimensionally inconsistent: -ρ/2∫∫ has the dimension of volume in reduced units, whereas the intended excess entropy per particle is dimensionless. Because Eqs. (6)–(12) and the ordering proof are all formulated in terms of the X(L) defined in Eq. (4), this normalization error is load-bearing: the extrapolated X∞ and the claim that X*(L) converges faster than X∞^1 are statements about the normalized KB-type integral. Please correct the definitions and re-state the results accordingly.","section":"Section 2, Eqs. (2) and (4)"},{"comment":"As typeset, Eq. (17) cannot be the derivative of Eq. (15). It appears as a product of two rational functions, each with denominator (α(1-λ)+r^2)^4, whereas direct differentiation of uWF = λ(1-y)(r_c^2-r^2)^2/y^3 with y=r^2+α(1-λ) yields the single rational function (r_c^2-r^2)^2[y(1-y)+αλ(3-2y)]/y^4. Since Eq. (17) is the quantity averaged in the thermodynamic integration that defines Sex_TI, the printed formula must be corrected or replaced by a reference to the code; without this, the benchmark values in Tables 1–6 cannot be reproduced.","section":"Section 3, Eq. (17)"},{"comment":"The manuscript does not state the upper limit used when evaluating Eq. (1) for the simulated RDFs, nor whether the spherical weight function w(r,L) of Eq. (3) is applied. The theoretical analysis in Section 2 is for a spherical volume of diameter L, but the simulations are carried out in cubic boxes. If Eq. (1) is simply truncated at half the cubic box length without the spherical weighting factor, the convergence properties derived in Section 2 do not directly apply to the numerical data. Please specify the RDF integration cutoff, bin width, and weighting procedure for each system, and state how the box size L is defined; this is needed to assess the '95% for ρ<0.1' claim and to make the simulations reproducible.","section":"Sections 3–4, MC integration protocol"}],"minor_comments":[{"comment":"The word 'identical' overstates the body's results: at ρ=0.1 the absolute percentage errors for g∞(r) are about 3–4% in Tables 1–6. Suggest replacing 'identical' with 'in agreement within a few percent' or a similar quantitative statement.","section":"Abstract and Conclusions"},{"comment":"The statement 'Since 0 < qS(r) < 1' is not generally true: qS = g ln g - g + 1 exceeds 1 for g > e, which can occur near contact in dense fluids. The argument only needs qS(r)>0 to conclude X∞^1(L) < X*(L) < X∞, so the proof is easily repaired, but the stated bound should be corrected.","section":"Section 2, convergence proof"},{"comment":"In the proof after Eq. (12), the inequality '0 < X1(L) < X*(L) < X∞' introduces X1(L), while the object defined in Eq. (10) is X∞^1; please align the notation.","section":"Section 2, notation"},{"comment":"The phrase 'kB is the Boltzmann factor' should read 'kB is the Boltzmann constant'.","section":"Section 1"},{"comment":"The keyword 'Finitie-size effects' contains a typo; it should be 'Finite-size effects'.","section":"Keywords"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the scope of cond-mat.stat-mech. The empirical low-density conclusion is plausible, but the factor-V error in the central definitions and the missing simulation integration protocol need to be fixed before the theoretical and numerical parts can be read as one consistent argument. The incorrect-looking Eq. (17) is also a serious presentation error, though it may be an artifact of typesetting. I do not see a reason to question the novelty or the appropriateness of the self-citations to earlier Kirkwood-Buff finite-volume work."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the main thing to know: this paper's practical recommendation—use the Ganguly–van der Vegt corrected RDF and a truncated running integral for $S_{ex}$ at low densities—is backed by solid Monte Carlo data. The theoretical claim that the running integral converges faster for the entropy integrand $q_S$ than for KB integrals is new, and the argument is correct once you read $X(L)$ as the volume-normalized finite-volume integral. That's also where the main flaw sits: Eq. 4 defines $X(L)$ as the raw double integral and then equates it to the single integral with the weight function, which is off by a factor $V$. For $q=1$, the left side is $V^2$ and the right is $V$. The rest of the derivation only makes sense if $X(L)$ is the normalized version $(1/V)\\int\\int q$. So it's a notation error, easy to fix, but it is in the central equation and should be corrected. The stress-test note is right that the printed equality is false; it's wrong to say the whole derivation collapses, because the Taylor expansion and the ordering proof $X_\\infty^1 < X^* < X_\\infty$ are unchanged by making the normalization explicit. Still, it must be fixed.\n\nWhat's genuinely good: the algebraic derivation of the 1/L corrections, the analytic model comparisons for KB vs entropy integrals, and the systematic Monte Carlo study—six tables, two potential cutoffs, two system sizes, thermodynamic integration as an independent benchmark. The APE data consistently show the corrected RDF gives under ~4% error at $\\rho \\le 0.1$, and that N=100 is enough if you use $g_\\infty(r)$. That is practically useful for people doing excess-entropy scaling.\n\nSoft spots besides Eq. 4: the abstract says 'identical' while the body reports 0.5–4% differences; soften that. The MC section never states the integration upper limit for Eq. 1 or whether the spherical weight function was applied. For cubic boxes, the theory assumes a spherical volume with diameter $L$, so the connection between the analytic $L$ and the box cutoff is unclear. They should report the cutoff and, ideally, show how the result changes with the weight function. Also, the running integral $X^*$ is used in Eq. 1 directly, but the theoretical extrapolations $X_\\infty^1 \\dots X_\\infty^3$ are not applied to the simulation data—the paper should say that explicitly, or include a comparison with $X_\\infty^1$ to show which recommendation actually holds. The self-citation cluster is heavy, but the cited results are established and not constructed for this paper.\n\nBottom line: a competent, useful methods paper. It deserves peer review and should be published after the normalization issue and the simulation-detail gaps are addressed. A careful referee should request those revisions, but the empirical conclusions are likely to survive. I'd bring it to a reading group focused on simulation methodology.","headline":"A useful finite-size correction study for RDF-based excess entropy; the central equation has a missing 1/V factor, but the empirical recommendations survive.","tokens_in":18682,"tokens_out":6465,"would_cite":true,"duration_ms":64114,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["82B30","82B80"],"pacs":[],"model":"deepseek-v4-flash","headline":"Corrected radial distribution function reproduces excess entropy within 5% below density 0.1.","keywords":["excess entropy","radial distribution function","finite-size effects","Kirkwood-Buff integrals","thermodynamic integration","Wang-Ramirez-Dobnikar-Frenkel potential","running integral"],"falsifier":"Run the same NVT protocol at $\\rho=0.01$, $N=100$, $r_c=2$ but compute $X(L)$ with the spherical weight $w(r,L)$ of Eq. 3 and compare it with a simple truncation of Eq. 1 at the cubic box edge; the paper does not state which of these was used, and a difference beyond statistical error would show that the claimed convergence ordering does not describe the actual simulation geometry. A second check is to take the analytic model with $\\chi=20$ and a cubic-box running integral without spherical weighting; if $X^*(L)$ no longer lies between $X^1_\\infty$ and $X_\\infty$, the faster-convergence proof fails for that geometry.","tokens_in":17788,"feed_emoji":"⚛️","tokens_out":11637,"duration_ms":104068,"temperature":0.7,"pith_summary":"The paper asks how badly the standard excess-entropy formula, $S_{\\rm ex}/(k_B N) = -2\\pi\\rho \\int_0^\\infty [g(r)\\ln g(r)-g(r)+1]\\,r^2\\,dr$, is affected by computing the radial distribution function (RDF) in a finite simulation box, and how to correct for it. It derives finite-volume versions of both the excess-entropy integral and the Kirkwood-Buff integral using a spherical weight function, and extrapolates to the thermodynamic limit with a $1/L$ Taylor expansion. With an analytic model RDF it shows that the excess-entropy integrand is essentially positive and small, so the running integral $X^*(L)$ converges faster than the corrected Kirkwood-Buff integrals. In Monte Carlo simulations of the Wang-Ram\\'irez-Dobnikar-Frenkel (WF) fluid, truncating the formula at the box size and using the finite-size-corrected RDF reproduces thermodynamic-integration values within 5% for densities $\\rho < 0.1$, even with 100 particles. This matters because it makes excess entropy a cheap, simulation-friendly quantity for dilute fluids, although at higher densities the low-density approximation itself causes errors of roughly 20% or more.","feed_headline":"Corrected RDF yields 95% of excess entropy at low densities","feed_subtitle":"Truncated integral plus corrected RDF matches thermodynamic integration for dilute systems, even with 100 particles.","key_machinery":"The load-bearing object is the geometric weight function for a spherical volume, $w(r,L)=4\\pi r^2(1 - \\frac{3r}{2L} + \\frac{r^3}{2L^3})$, which reduces the double integral over two particle positions to a one-dimensional integral of $q(r)$. A Taylor expansion of the finite-volume integral in $1/L$ (Eq. 6, kept to third order) produces corrected estimates $X^1_\\infty$, $X^2_\\infty$, $X^3_\\infty$; for excess entropy the running integral $X^*(L)=4\\pi\\int_0^L q_S(r)r^2\\,dr$ is shown to be the fastest-converging estimator because $q_S(r)\\approx (g(r)-1)^2$ is positive and small. For simulation RDFs, the finite-size correction of Eq. 18 rebuilds $g(r)$ so that it approaches 1 at the cut-off, and this corrected $g_\\infty(r)$ is what makes the truncated integral accurate at small $N$.","core_discovery":"The paper's central claim is that finite-size effects in the excess-entropy integral are manageable and differ in kind from those in Kirkwood-Buff integrals. For the excess-entropy integrand $q_S(r)=g(r)\\ln g(r)-g(r)+1$, the running integral $X^*(L)=4\\pi\\int_0^L q_S(r)r^2\\,dr$ satisfies $0<X^1_\\infty(L)<X^*(L)<X_\\infty$ and therefore converges to the thermodynamic limit faster than the first-order corrected estimate. Consequently, Eq. 1 can be truncated at the simulation box edge provided the RDF is first corrected to the thermodynamic limit. The Monte Carlo results support this: for $\\rho<0.1$ and for both $r_c=2$ (Lennard-Jones-like) and $r_c=1.2$ (colloid-like) Wang-Ram\\'irez-Dobnikar-Frenkel fluids, the corrected-RDF route agrees with thermodynamic integration to within a few percent at $N=100$ and $N=500$, while uncorrected RDFs can be off by more than 20% at the lowest densities in small systems. Above $\\rho=0.1$ both RDF-based values agree with each other but deviate from thermodynamic integration by 20\\textendash 25%, which the paper attributes to the second-order density expansion itself.","pith_inferences":["The paper's ordering argument uses $q_S(r)>0$; a strongly associating fluid where $q_S(r)$ is negative over a wide range would not automatically inherit the running-integral advantage, so the protocol should be re-tested before transferring it to such systems.","The derivation assumes a spherical integration volume, while the simulations use cubic boxes; whether the same convergence ordering holds for a cubic-box cutoff without spherical weighting is a testable extension the paper does not report.","The 95% low-density result suggests a practical screening workflow: estimate $S_{\\rm ex}$ from one 100-particle run and feed it into entropy-scaling transport correlations, reserving thermodynamic integration for dense states."],"forward_implications":["For $\\rho<0.1$, excess entropy can be computed from one short NVT simulation with as few as 100 particles by integrating the corrected RDF to the box edge, with no thermodynamic integration needed.","The finite-size-corrected RDF should be used in all excess-entropy calculations; at very low density the uncorrected RDF in a 100-particle box can give errors above 20% for colloid-like interactions.","For excess entropy, the raw running integral $X^*(L)$ is the right estimator, so existing codes that report running integrals are already using the faster-converging quantity.","At $\\rho>0.1$, discrepancies of 20\\textendash 25% against thermodynamic integration are not a finite-size artifact but the low-density approximation in Eq. 1; higher-order density expansions or thermodynamic integration are needed there."],"supporting_citations":[{"why":"derives the second-order density expansion of the entropy that defines the excess-entropy formula studied here.","marker":"[1]"},{"why":"provides the spherical weight function for finite-volume integrals, which is the starting point of the truncation analysis.","marker":"[21]"},{"why":"introduces the finite-size RDF correction whose use is the paper's main practical recommendation.","marker":"[22]"},{"why":"established the 1/L Taylor expansion of finite-volume Kirkwood-Buff integrals that the paper extends to third order.","marker":"[24]"},{"why":"supplies the analytic model RDF used to compare convergence of the estimators.","marker":"[27]"},{"why":"introduces the WF pair potential used in the Monte Carlo validation.","marker":"[32]"},{"why":"provides the higher-order density-expansion approximation used to interpret high-density deviations.","marker":"[19]"}],"fun_headline_variants":["Corrected RDF matches thermodynamic integration at low density","Extrapolated RDF hits thermodynamic-limit entropy at low density","Excess entropy from RDF converges faster than Kirkwood-Buff","RDF extrapolation solves finite-size entropy error at low density"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The central assumption is that the finite-volume entropy integral has a convergent $1/L$ Taylor expansion, which is accurate only when correlations in $q(r)$ decay over a finite length; for long-ranged correlations the system sizes tested here would not capture 95% of the excess entropy.","fun_headline_variants_meta":{"raw":{"variants":["Corrected RDF matches thermodynamic integration at low density","Extrapolated RDF hits thermodynamic-limit entropy at low density","Excess entropy from RDF converges faster than Kirkwood-Buff","RDF extrapolation solves finite-size entropy error at low density"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000955,"raw_usage":{"total_tokens":4104,"prompt_tokens":1011,"completion_tokens":3093,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":627,"completion_tokens_details":{"reasoning_tokens":3022}},"tokens_in":627,"tokens_out":3093,"duration_ms":24031,"temperature":1.0,"reasoning_tokens":3022,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T13:12:21.481157+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same NVT protocol at $\\rho=0.01$, $N=100$, $r_c=2$ but compute $X(L)$ with the spherical weight $w(r,L)$ of Eq. 3 and compare it with a simple truncation of Eq. 1 at the cubic box edge; the paper does not state which of these was used, and a difference beyond statistical error would show that the claimed convergence ordering does not describe the actual simulation geometry. A second check is to take the analytic model with $\\chi=20$ and a cubic-box running integral without spherical weighting; if $X^*(L)$ no longer lies between $X^1_\\infty$ and $X_\\infty$, the faster-convergence proof fails for that geometry.","supporting_citations":[{"cited_title":"Laird and A","cited_arxiv_id":null,"evidence_quote":"derives the second-order density expansion of the entropy that defines the excess-entropy formula studied here."},{"cited_title":"Kr¨ uger and T.J.H","cited_arxiv_id":null,"evidence_quote":"provides the spherical weight function for finite-volume integrals, which is the starting point of the truncation analysis."},{"cited_title":"Ganguly and N.F","cited_arxiv_id":null,"evidence_quote":"introduces the finite-size RDF correction whose use is the paper's main practical recommendation."},{"cited_title":"Kr¨ uger, S.K","cited_arxiv_id":null,"evidence_quote":"established the 1/L Taylor expansion of finite-volume Kirkwood-Buff integrals that the paper extends to third order."},{"cited_title":"Verlet, Computer experiments on classical fluids","cited_arxiv_id":null,"evidence_quote":"supplies the analytic model RDF used to compare convergence of the estimators."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"introduces the WF pair potential used in the Monte Carlo validation."},{"cited_title":"Huang and M","cited_arxiv_id":null,"evidence_quote":"provides the higher-order density-expansion approximation used to interpret high-density deviations."}],"review_version":1}