{"id":"d7c0e48e-09cd-4a48-99e8-2626afe876e8","arxiv_id":"2608.08546","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A differentiable loss based on the partition function modulus on the Lee-Yang circle refines force fields to match gas-liquid and solid-liquid phase diagrams, with Cu enthalpy and heat capacity improved as out-of-target tests.","lead":"This paper introduces a force field fitting method that uses the magnitude of the partition function, sampled near Lee-Yang zeros, as a target, so only the desired phase diagram is needed. It is worth reading because a system-agnostic phase-diagram optimizer could make interatomic potential fitting faster and easier, including for machine-learned force fields.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The NpT Lee-Yang circle theorem is the least secure link: if realistic enthalpy distributions put zeros off the target circle, the loss minimum need not coincide with the target transition.","rationale":"The paper proposes a genuinely novel, parameter-light loss construction and provides an independent validation in the Cu enthalpy and heat-capacity results, which partially supports the central claim. However, the argument's load-bearing step is the NpT Lee-Yang circle theorem: without it, the evaluation points in Eq. (6) lose their physical meaning and the loss minimum no longer guarantees the target transition temperature. The reader identified exactly this assumption, and I agree. The concern is not that the authors are wrong, but that the geometric premise is imported from overlapping-author work and is not independently verified for metallic solids or for finite-size OPES/MBAR histograms. A direct zero-location check on the refined Cu potential would settle whether the premise holds quantitatively. If the zeros lie on the target circle within the reported accuracy, the central claim survives; if not, the method's success on Cu may be fortuitous or dependent on unstated hyperparameter choices (m, P, angular range, bin count), which the paper itself acknowledges as arbitrary. Because the existing CONDITIONAL verdict already calls for exactly this kind of verification, my read does not change the verdict.","tokens_in":8498,"tokens_out":9917,"duration_ms":122450,"concrete_test":"For the refined Cu EAM potential at p=2.97 GPa, generate a long OPES/MBAR enthalpy histogram, build Zbar(beta_tilde,p) via Eq. (5), and solve for the Lee-Yang zeros. For each transition-associated zero (smallest imaginary part), compute the distance to the ideal circle |T_tilde - T_r/2| relative to T_r/2, or equivalently check whether Re(1/T_tilde)=1/T_r. If the deviation corresponds to more than the reported MAE (~2.4 K), the circle premise is not quantitatively satisfied and the loss minimum may be offset from the target transition; if the deviation is smaller, the premise is supported and the existing CONDITIONAL verdict stands.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that minimizing L(theta) in Eq. (7) drives the predicted Lee-Yang zeros onto the target circle and thereby matches the phase diagram. That inference depends on the NpT Lee-Yang circle theorem, imported from overlapping-author Refs. [18,21], being quantitatively accurate for the systems considered, including Cu EAM. The theorem enters right after Eq. (3) and fixes the evaluation points in Eq. (6): the circle with diameter (0,T_r) is equivalent to the vertical line Re(beta_tilde)=beta_r in the complex inverse-temperature plane. For the polynomial in Eq. (3) with positive coefficients, exact zeros lie on this line only for special coefficient patterns, such as two identical phase peaks. If the two coexisting phases have different enthalpy fluctuation widths, the characteristic-function zeros move off that line, so |Zbar| at the chosen points is not minimized exactly when T_c=T_r. The paper's Cu validation (Fig. 4 inset) is qualitative, with no error bars or a quantitative comparison of zero locations to the ideal circle, so the geometric premise is not independently established for realistic potentials. Thus the least secure step is the assumed identity between the loss minimum and the target transition temperature.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a force-field refinement framework based on Lee-Yang theory. In the NpT ensemble, the analytically continued partition function is written as a polynomial in y = exp(-(β̃−β)ΔH), with coefficients taken from MBAR-reweighted OPES enthalpy histograms. The loss function sums |Z̄_θ(β̃_j^r, p_r)|^m over target thermodynamic conditions r and P evaluation points on the target Lee-Yang circle, i.e., the circle in the complex temperature plane with diameter (0, T_r). The authors argue that minimizing this loss drives the predicted Lee-Yang zeros toward the target circle and therefore matches the target phase diagram, without requiring order parameters or response functions. They validate the framework by refining a Lennard-Jones potential against both gas-liquid and solid-liquid reference data, and a Cu EAM potential against simulated and experimental melting curves. The refined force fields substantially reduce the mean absolute error in the target phase boundaries, and the Cu refinement against experiment also improves enthalpy and heat capacity predictions at 1 bar, which are observables outside the loss.","tokens_in":8712,"tokens_out":6262,"duration_ms":72381,"significance":"If the central geometric premise is correct, the method is a genuinely useful contribution: phase-diagram-guided refinement is reduced to minimizing a zeroth-order partition-function modulus, avoiding order-parameter choices, response-function targets, and the numerical ill-conditioning of explicit zero-finding. The construction is coherent and end-to-end differentiable via DMFF/JAX, and the simultaneous refinement of gas-liquid and solid-liquid coexistence with a single set of LJ parameters is a nice demonstration. The independent Cu enthalpy and heat-capacity checks provide evidence that optimizing the phase boundary does not merely overfit the phase boundary. However, the load-bearing assumption that the transition-associated zeros lie exactly on the ideal circle for realistic enthalpy distributions is imported from Refs. [18,21] and not quantitatively verified for the systems studied here, and the reported MAE improvements are given without statistical uncertainties. The empirical results are promising, but the theoretical grounding and the quantitative validation need strengthening before the generality claim is fully supported.","major_comments":[{"comment":"The inference that minimizing L(θ) in Eq. (7) drives the predicted Lee-Yang zeros onto the target circle relies on the NpT Lee-Yang circle theorem being quantitatively accurate for the enthalpy distributions produced by LJ and Cu EAM. For the polynomial in Eq. (3), zeros lie on the unit circle |y|=1 only for special coefficient patterns. For a two-peak enthalpy distribution with unequal weights a and b separated by ΔH, the roots satisfy |y|=(b/a)^{1/ΔH}, which is not 1, and unequal peak widths also move the zeros off the circle. Thus the loss minimum does not automatically coincide with the target transition temperature for a general realistic distribution. Figure 1(b) is presented as a confirmation for Cu, but it is qualitative, and no quantitative comparison of the computed zero locations to the ideal circle is reported. I request either a direct numerical test that computes the actual zeros from the target enthalpy distributions and measures their distance from the ideal circle, or a synthetic test with asymmetric double-peak distributions showing that the loss minimum still tracks T_c for realistic asymmetry.","section":"§2, after Eq. (3) and Eq. (6)"},{"comment":"The central quantitative claims are reported without statistical uncertainties: for LJ gas-liquid, the MAE decreases from 0.0148 to 0.0007 a.u. (95.1% improvement); for LJ solid-liquid, from 0.0246 to 0.0038 a.u. (84.4%); for Cu simulated target, from 36.7 K to 1.4 K (96.1%); for Cu experimental target, from 25.4 K to 2.4 K (90.6%). Since the enthalpy histograms come from stochastic OPES/MBAR sampling, these numbers must be accompanied by error bars, confidence intervals, or at least a statement of the number of independent runs. Without this, the claim that the improvements are significant is not statistically grounded.","section":"§5, quantitative results after Figs. 3 and 4"},{"comment":"The simulated-reference tests in Figs. 3 and 4(a) recover a phase diagram that is itself generated by a well-converged force field. Because L(θ) is built directly from the target transition temperatures, the reduction in MAE after refinement is, to a large extent, a self-consistency check rather than an independent validation of the physical content of the loss. The paper should state this explicitly. The genuinely independent evidence is the Cu experimental-target refinement and the out-of-target enthalpy and heat-capacity predictions in Fig. 5, and the discussion should make clear that those carry the weight of the transferability claim.","section":"§5, simulated-reference benchmarks"},{"comment":"The loss function has several tunable ingredients: the exponent m=2, chosen after preliminary numerical testing; the number of evaluation points P; their angular range; the number of enthalpy bins N; and the target enthalpy interval [Hmin, Hmax]. The statement that the circle geometry introduces no extraneous parameters is therefore only partially accurate. The limitations paragraph acknowledges these choices, but the accompanying assertion that the results are robust to reasonable variations is not supported by any displayed sensitivity analysis. A table reporting the dependence of the final MAE on m, P, N, and the enthalpy interval would substantially strengthen the universality claim.","section":"Eq. (7) and the limitations paragraph"}],"minor_comments":[{"comment":"The indexing is inconsistent: Eq. (6) defines j = 0, ..., P−1, while Eq. (7) sums j = 1 to P. Please align the two ranges.","section":"Eq. (7)"},{"comment":"The caption of Fig. 1(b) describes a realistic case for Cu but does not say how the zeros were computed or from which simulation. Please provide the computational details in the main text or Supplemental Material.","section":"Fig. 1"},{"comment":"The parameter γ in the well-tempered OPES target p_tg(s) ∝ [P(s)]^{1/γ} is not defined in the main text; please define it in the paragraph introducing Eq. (8).","section":"Sampling details"},{"comment":"The reduced Lennard-Jones units are denoted 'a.u.' in the text without definition. Please define these reduced units in Section 4.","section":"Notation"},{"comment":"The normalization denominator is written as PN−1 k=0 Gp(Hk), which is clear in context, but the notation in the displayed equation is cramped. Please rewrite the summation limits and indices cleanly.","section":"Eq. (4)"},{"comment":"The ambient-pressure melting point in Fig. 5(a) is obtained by extrapolating the experimental optimization targets to 1 bar via the Simon equation. Please state the extrapolation uncertainty, since this point is used to judge the refined force field's accuracy.","section":"Section 5, Cu validation"}],"recommendation":"major_revision","confidential_remarks":"The NpT circle theorem is imported from Refs. [18,21], which share authorship with the present manuscript. An independent derivation or a quantitative verification of the zero-circle correspondence for the LJ and Cu systems would materially increase confidence in the method. A code/data availability statement would also be valuable, as the empirical claims rest on a complex sampling pipeline that is only partially described in the main text."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe new thing here is the loss function: instead of matching densities or solving for Lee-Yang zeros, they minimize the modulus of the NpT partition function evaluated at points on the target Lee-Yang circle. That is a clean, genuinely new optimization target and it avoids order parameters and response properties. The validation is real: for LJ, one parameter set simultaneously improves gas-liquid and solid-liquid coexistence; for Cu, refinement against experimental melting curves also improves enthalpy and heat capacity, which are outside the loss. Those checks are meaningful, and the authors are candid about the arbitrary choices (exponent m, bin count, evaluation-point selection).\n\nWhere I'd push back: the whole target geometry rests on the NpT Lee-Yang circle theorem imported from overlapping-author work (Refs 18 and 21). The paper shows zeros migrating onto the ideal circle after refinement, but it doesn't quantitatively compare the zero locations to the circle or test a case with asymmetric enthalpy peaks. The stress-test worry—zeros drifting off the circle when the two phases have different enthalpy fluctuation widths—is legitimate in principle. If that happens, the loss minimum no longer sits exactly at the target transition temperature. Nothing in the current evidence breaks the central claim, but the universality statement is stronger than the demonstrated support.\n\nMinor: no error bars on the MAE improvements, no code or data shipped, and the loss exponent and evaluation-point density are tuned after preliminary testing. The phase-diagram improvement itself is a fitting outcome, so it can't confirm the method; the Cu enthalpy and heat capacity results are what carry weight.\n\nWho benefits: force field developers and anyone trying to fit potentials against p-T boundaries without inventing order parameters. It deserves a serious referee, not a desk reject. My recommendation: send it out, and ask the authors to (1) quantify zero deviations from the Lee-Yang circle for at least one non-ideal case, (2) report uncertainties on the reported MAEs, and (3) release the workflow. Conditional accept.","headline":"A genuinely new differentiable loss that fits phase diagrams by zeroing the partition function modulus on the Lee-Yang circle; the empirical checks are solid, but the imported circle theorem needs sharper quantitative backing before the universality claim is trusted.","tokens_in":9249,"tokens_out":4301,"would_cite":true,"duration_ms":46185,"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":"Minimizing the partition function modulus on a target Lee-Yang circle moves a force field's predicted phase boundary to the target.","keywords":["Lee-Yang zeros","partition function modulus","force-field refinement","phase diagram","NpT ensemble","melting curve","Lennard-Jones potential","embedded-atom method"],"falsifier":"Compute the actual Lee-Yang zeros of the unrefined Cu EAM potential at pressures like 2.97 GPa and compare their locus with the predicted circle of diameter (0,Tpt); if the edge zeros deviate systematically from that circle beyond sampling error, the loss minimum will be biased away from the target melting temperature.","tokens_in":8278,"feed_emoji":"🌡️","tokens_out":7941,"duration_ms":76210,"temperature":0.7,"pith_summary":"The paper proposes a general way to refine interatomic force fields directly against a target pressure-temperature phase diagram, without choosing order parameters or thermodynamic response functions. The idea is to evaluate the complex-temperature partition function at points on the Lee-Yang circle associated with the target transition; when the force field is correct, the modulus of the normalized partition function is small there. Because the enthalpy probability distribution from molecular dynamics supplies the polynomial coefficients, the modulus is differentiable with respect to force-field parameters and can be minimized by automatic differentiation. The authors validate the approach on a Lennard-Jones system for both gas-liquid and solid-liquid coexistence and on a copper embedded-atom potential against simulated and experimental melting curves. Refined force fields reproduce the target phase boundaries, and for copper the refinement also improves enthalpy and heat capacity predictions outside the optimization target.","feed_headline":"Lee-Yang zeros guide force fields to target phase diagrams","feed_subtitle":"A loss built on the Lee-Yang circle fixes gas-liquid and solid-liquid transitions without order parameters.","key_machinery":"The central object is the normalized complex-temperature partition function $\\bar{Z}(\\tilde{\\beta}, p)$, built from the reweighted enthalpy histogram $G_p(H)$ obtained from enhanced-sampling molecular dynamics combined with a multistate reweighting estimator. A Lee-Yang circle is the geometric locus of the transition-associated zeros: a circle in the complex temperature plane with diameter $(0, T_{\\mathrm{pt}})$, so that one target transition temperature fixes one set of evaluation points. Evaluating $|\\bar{Z}|$ at points on the upper arc of that circle yields a differentiable loss whose minimum corresponds to the predicted zeros coinciding with the target circle. The normalization step removes the artifact coming from the total number of microstates so that the loss responds to the zero locations, not to changes in the overall density of states.","core_discovery":"The central claim is that minimizing the normalized partition function modulus $|\\bar{Z}|$ at points on the target Lee-Yang circle drives the predicted Lee-Yang zeros onto that circle, thereby moving the predicted transition temperature to the target value. The paper derives a polynomial representation of the NpT partition function whose coefficients are the discretized enthalpy probability distribution from simulation, then evaluates this polynomial at points $\\tilde{T}_j^r$ on the circle with diameter $(0, T_r)$ for each target pressure $p_r$. The loss aggregates the modulus over all target conditions and evaluation points; no explicit zero-finding is needed. The numerical experiments show that this single loss improves gas-liquid coexistence, solid-liquid melting, and Cu melting simultaneously, and that the refined Cu potential also matches enthalpy and heat capacity data that were not part of the loss.","pith_inferences":["If the circle theorem holds only approximately for real materials, the loss minimum may be systematically biased for transitions with strong non-circulatory zero patterns; a direct test would be to compute the zero set of a known potential and compare its edge locus to the predicted circle.","The same loss could be applied to multi-component or structural transitions where no good order parameter exists, provided the enthalpy histogram has a double-peak structure near coexistence.","Because the loss is built from a reweighted histogram, its gradient noise grows as parameters drift from the sampling reference; an adaptive resampling schedule based on effective sample size would likely make the framework more robust for larger parameter spaces."],"forward_implications":["A single set of refined parameters can improve several phase boundaries at once: the Lennard-Jones refinement simultaneously reduces the gas-liquid and solid-liquid mean absolute errors in transition temperature.","Because the loss uses only the p-T phase diagram as input, the same formulation applies to any discontinuous transition in which a target boundary is known, without system-specific order parameters or response functions.","For the copper EAM potential, refinement against experimental melting data brings the predicted enthalpy and isobaric heat capacity closer to experiment in both solid and liquid phases, even though those observables are not in the loss.","Avoiding explicit polynomial root-finding lets the optimization proceed smoothly under automatic differentiation and automatically weights the evaluation points where physical zeros concentrate.","The framework scales naturally to more complex potentials, with extension to machine-learning force fields identified in the paper as the next target."],"supporting_citations":[{"why":"Foundational Lee-Yang theory: phase transitions appear as partition-function zeros approaching the real axis, motivating the use of zero locations as a refinement signal.","marker":"[14–16]"},{"why":"Supplies the discretization of the NpT partition function into a polynomial whose coefficients are the enthalpy probability distribution, and the practical scheme for computing Lee-Yang zeros from molecular dynamics simulations.","marker":"[17, 18]"},{"why":"States the Lee-Yang circle theorem for double-peak enthalpy distributions: transition-associated zeros lie on a circle with diameter (0,Tpt), which determines the evaluation points in the loss.","marker":"[21]"},{"why":"Enhanced-sampling method used to cross the solid-liquid free-energy barrier and recover unbiased enthalpy distributions at the target conditions.","marker":"[22]"},{"why":"Multistate reweighting estimator that combines multiple sampling ensembles to obtain the enthalpy distribution at any target temperature and pressure.","marker":"[23]"},{"why":"Automatic-differentiation simulation framework through which gradients flow from the partition-function-modulus loss to the force-field parameters.","marker":"[8]"},{"why":"Experimental copper melting data used as the target for the practical refinement and for extrapolating the ambient-pressure melting point.","marker":"[27]"},{"why":"Experimental isobaric heat capacities used to validate that the refined copper force field improves observables outside the optimization target.","marker":"[28]"}],"fun_headline_variants":["Partition function modulus guides force fields to phase diagrams","Lee-Yang loss refines force fields without order parameters","Universal force-field tuning via Lee-Yang zeros","Phase-diagram loss from Lee-Yang circle fixes transitions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The method assumes that every discontinuous transition with a double-peaked enthalpy histogram has its transition-related partition-function zeros on a circle whose diameter runs from zero to the transition temperature; if real potentials deviate from that circle, the loss minimum will no longer sit at the target transition.","fun_headline_variants_meta":{"raw":{"variants":["Partition function modulus guides force fields to phase diagrams","Lee-Yang loss refines force fields without order parameters","Universal force-field tuning via Lee-Yang zeros","Phase-diagram loss from Lee-Yang circle fixes transitions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000256,"raw_usage":{"total_tokens":1545,"prompt_tokens":883,"completion_tokens":662,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":499,"completion_tokens_details":{"reasoning_tokens":607}},"tokens_in":499,"tokens_out":662,"duration_ms":6685,"temperature":1.0,"reasoning_tokens":607,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:32:28.385205+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the actual Lee-Yang zeros of the unrefined Cu EAM potential at pressures like 2.97 GPa and compare their locus with the predicted circle of diameter (0,Tpt); if the edge zeros deviate systematically from that circle beyond sampling error, the loss minimum will be biased away from the target melting temperature.","supporting_citations":[{"cited_title":"Tracking metastable phases by complex Lee-Yang zeros","cited_arxiv_id":"2606.08004","evidence_quote":"States the Lee-Yang circle theorem for double-peak enthalpy distributions: transition-associated zeros lie on a circle with diameter (0,Tpt), which determines the evaluation points in the loss."},{"cited_title":"Invernizzi and M","cited_arxiv_id":null,"evidence_quote":"Enhanced-sampling method used to cross the solid-liquid free-energy barrier and recover unbiased enthalpy distributions at the target conditions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Multistate reweighting estimator that combines multiple sampling ensembles to obtain the enthalpy distribution at any target temperature and pressure."},{"cited_title":"Wanget al., J","cited_arxiv_id":null,"evidence_quote":"Automatic-differentiation simulation framework through which gradients flow from the partition-function-modulus loss to the force-field parameters."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Experimental isobaric heat capacities used to validate that the refined copper force field improves observables outside the optimization target."}],"review_version":1}