{"id":"629315ee-9774-4610-a4b6-a3cfaeea63f9","arxiv_id":"2604.13659","paper_version":3,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"MACE MLFF simulations of aqueous NaCl and CsI capture experimental water-diffusion anomalies, improve on DeePMD for NaCl via stronger Na+ hydration, and attribute CsI acceleration mainly to I−.","lead":"A MACE machine-learned force field trained on revPBE-D3 data reproduces the experimental ion-specific anomaly of water diffusion (slowed in NaCl, sped up in CsI) and improves on prior DeePMD results. The work supplies a shell-resolved microscopic mechanism that classical force fields have long failed to capture.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"No significant objection identified beyond the reader's already-flagged functional limitation.","rationale":"The manuscript is a careful, well-documented extension of the MLFF program for aqueous electrolytes. All quantitative claims (EoS, RDFs, structure factors, size-extrapolated diffusion and viscosity, shell-resolved diffusivities, PMFs) are backed by SI tables and direct comparison to experiment and to DeePMD-revPBE-D3. The only soft spot that could reverse the strongest claim is the known GGA under-binding of Na+, which the authors themselves cite and which the reader already elevated to the weakest assumption. No independent internal flaw (e.g., unphysical ion pairing after fine-tuning, uncontrolled finite-size effects, or contradictory shell analysis) survives scrutiny. Therefore the CONDITIONAL verdict with high confidence stands; no adjustment is warranted.","tokens_in":24541,"tokens_out":487,"duration_ms":5284,"concrete_test":"Recompute the Na–O first-peak height and the 0.89 m NaCl Dw/D0 with a MACE model fine-tuned on the same configurations but labeled by RPA or DC-SCAN (as in O’Neill et al.); if the relative diffusion rises above ~0.95 or the second-shell contribution vanishes, the claimed improvement over DeePMD is functional-dependent rather than architecture-dependent.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader's weakest assumption (revPBE-D3 under-binding of Na+ relative to RPA/MP2) is correctly identified and already bounds the claim. Within that shared training theory the paper's central claim is internally well-supported: size-extrapolated Dw/D0 (Tables S3–S5, Fig. 6) matches experiment for both salts, the Na–O PMF barrier is higher than DeePMD's (Fig. 17, Tables S6–S7), and the second-shell retardation of Na+ is quantified (Fig. 8d, Table S12). No additional load-bearing inconsistency appears in the shell decomposition, finite-size protocol, or model validation against AIMD. The residual risk remains exactly the one the reader named—whether a higher-level training set would reverse the NaCl ranking—and does not require a further attack.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The manuscript trains a MACE equivariant ML force field on revPBE-D3 energies, forces and stresses for pure water and aqueous NaCl/CsI, then uses classical MD (size-extrapolated Green–Kubo) to compute water self-diffusion and shear viscosity over 0.89–3.56 mol/kg. It reproduces the experimental ion-specific anomaly (Dw/D0 < 1 for NaCl, > 1 for CsI) and reports quantitative improvement over DeePMD trained on the same functional, especially for NaCl. The improvement is attributed to a deeper Na+–O free-energy well and a measurable retarding contribution from Na+’s second hydration shell; for CsI the acceleration is assigned primarily to the diffuse I− shell. Supporting evidence includes RDFs, neutron structure factors, hydrogen-bond counts, shell-decomposed short-time diffusivities, and ion–oxygen PMFs.","tokens_in":24828,"tokens_out":828,"duration_ms":6685,"significance":"If the results hold, the work supplies a concrete, architecture-level demonstration that higher-order equivariant message-passing (MACE) can improve ion–water free-energy landscapes relative to DeePMD at fixed DFT theory, and that this improvement is dynamically consequential for the long-standing anomalous-diffusion problem. The size-extrapolated transport coefficients, block-error analysis, and direct side-by-side comparison with experiment and with DeePMD on the same functional constitute a reproducible benchmark that future MLFFs (including those trained beyond GGA) can be measured against. The shell-decomposition and PMF analysis also give a clear microscopic picture that classical non-polarizable models have systematically missed.","major_comments":[{"comment":"The central claim of quantitative improvement over DeePMD rests on a deeper Na+–O PMF (Fig. 17, Tables S6–S7) and second-shell retardation (Fig. 8d, Table S12). The manuscript itself cites O’Neill et al. that revPBE-D3 underestimates the Na–O first-peak height relative to RPA/MP2. Because the training data and the DeePMD reference share this functional, the residual error could still reverse the NaCl ranking once higher-level data are used. A short, explicit discussion of this risk (or a limited higher-level single-point check on representative Na–O configurations) is needed to bound the claim.","section":null},{"comment":"AIMD validation trajectories are only ~10 ps (Appendix B). While structural RDFs and short-time VACFs are compared (Figs. 10–13), the transport coefficients that form the paper’s main result require nanosecond sampling. The authors should state more clearly that the AIMD comparison validates only local structure and short-time dynamics, not the long-time Dw/D0 values themselves.","section":null}],"minor_comments":[{"comment":"Fig. 3 ion–ion RDFs at low concentration are noisy; a brief note that the noise is statistical (few ions) would help readers.","section":null},{"comment":"Hydration-shell cut-offs are defined via the inflection of n(r) (Fig. S3). The numerical values should also appear in the main text or a table for reproducibility.","section":null},{"comment":"The multi-head fine-tuning protocol (30 epochs, 10 k foundation configurations) is described only in the Appendix; a one-sentence summary in §II would improve accessibility.","section":null},{"comment":"Typographical inconsistencies appear in a few places (e.g., “Nos´ e-Hoover”, “˚A”); a light copy-edit pass is warranted.","section":null}],"recommendation":"minor_revision","confidential_remarks":"The work is a solid, well-executed incremental advance over Avula et al. (same functional, better architecture). Fit for a specialized physical-chemistry journal is clear; novelty relative to the DeePMD predecessor is real but modest. No citation or ethical concerns."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"This is a careful, well-executed extension of the MLFF program for the classic ion-specific water-diffusion anomaly. What is actually new is not the anomaly itself (Avula et al. already got the qualitative signs with DeePMD on the same functional) but a quantitative tightening of the NaCl Dw/D0 curve, a measurably deeper Na–O PMF barrier than DeePMD’s, a non-negligible second-shell retardation for Na+, and a pure-shell versus overlap decomposition that cleanly attributes CsI acceleration mainly to I−.\n\nThey do the hard parts properly. Size-extrapolated Green–Kubo diffusivities and viscosities (Tables S3–S5, Figs. 6–7), block-error analysis, neutron structure-factor comparisons, and direct side-by-side PMFs against DeePMD all line up. The fine-tuning step that killed the spurious Na–Na and Cs–Cs pairing is documented, and the shell cut-offs are defined from the inflection of n(r) rather than by hand-waving. Within the shared revPBE-D3 training theory the central claim holds.\n\nThe soft spot is exactly the one the reader named and the stress-test confirmed: revPBE-D3 is already known (O’Neill et al.) to under-structure the Na–O first peak relative to RPA/MP2. The paper cites that work but never checks whether a higher-level training set would reverse the NaCl ranking. That is a real residual risk, not a fatal one; it simply bounds how far the mechanism can be trusted beyond GGA. Minor additional notes: AIMD validation trajectories are short (~10 ps), model weights are not released, and the highest-concentration shell populations start to show cooperative effects that the low-c analysis does not fully capture. None of these overturn the reported trends.\n\nThis is for people who already care about MLFFs for aqueous electrolytes or about the microscopic origin of Hofmeister dynamics. It is not a paradigm shift, but it is honest, reproducible work that a serious referee should see. I would send it out.","headline":"Solid MACE-revPBE-D3 work that improves the NaCl Dw/D0 curve over DeePMD and supplies a clean pure-shell/overlap decomposition; the functional limitation is already flagged and does not break the internal claim.","tokens_in":25471,"tokens_out":541,"would_cite":true,"duration_ms":6112,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"A many-body machine-learned force field trained on revPBE-D3 data reproduces the experimental ion-specific anomaly of water diffusion and shows that Na+ retards while I− accelerates.","keywords":["anomalous water diffusion","aqueous electrolytes","machine-learned force field","MACE","revPBE-D3","hydration shells","ion-specific effects","potentials of mean force"],"falsifier":"Recompute the same concentration series of relative water diffusion coefficients with a MACE (or equivalent) force field trained on RPA or MP2 data for the same ions; if the NaCl slowdown or CsI speedup disappears or reverses relative to experiment, the central claim fails.","tokens_in":25444,"feed_emoji":"💧","tokens_out":715,"duration_ms":6403,"temperature":0.7,"pith_summary":"Water molecules move more slowly when NaCl is dissolved and more quickly when CsI is dissolved—an ion-specific anomaly that classical force fields have long failed to capture. This paper trains an equivariant many-body machine-learned force field (MACE) on density-functional theory data (revPBE-D3) and uses it for nanosecond-scale classical molecular dynamics of both salts over a wide concentration range. The simulations recover the experimental trends and improve quantitatively on earlier machine-learned models trained on the same electronic-structure method, especially for the NaCl slowdown. The improvement is traced to a deeper free-energy well for Na+–water and a measurable retarding effect that extends into Na+’s second hydration shell; the CsI speedup is shown to be driven mainly by the diffuse, weakly bound shell of I−. Shell-resolved diffusivities and ion–oxygen potentials of mean force supply a coherent microscopic picture of how structure-making and structure-breaking ions reshape water mobility.","feed_headline":"ML force field captures why NaCl slows water and CsI speeds it","feed_subtitle":"Stronger Na+ binding plus I−’s loose shell explain the experimental ion-specific anomaly","key_machinery":"Shell-decomposition of short-time water diffusivities (pure first shells, overlapping shells, and second shells) together with the corresponding ion–oxygen potentials of mean force; these quantities map local free-energy barriers onto the measured mobility changes.","core_discovery":"Classical molecular dynamics driven by a MACE force field trained on revPBE-D3 energies, forces and stresses quantitatively reproduces the experimentally observed concentration dependence of the relative water diffusion coefficient—suppression in NaCl, enhancement in CsI—and improves on DeePMD results obtained with the same functional, the gain arising from a stronger Na+–water interaction in the first shell plus a non-negligible retarding contribution of the second hydration shell of Na+, while the CsI acceleration is primarily driven by the diffuse hydration shell of I−.","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["MACE force field maps why NaCl retards water but CsI accelerates it","Stronger Na+ shells slow water in NaCl; I−'s loose shell speeds CsI","MACE captures ion-specific water diffusion anomaly: NaCl slow, CsI fast","How Na+ second shell retards water while I− drives acceleration in CsI","MACE MLFF reproduces anomalous water diffusion in NaCl versus CsI"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"The claim rests on the premise that the revPBE-D3 functional, known to underestimate the height of the Na–O first peak relative to higher-level electronic-structure methods, is still accurate enough that residual errors do not reverse the sign or ranking of the ion-specific diffusion anomalies.","fun_headline_variants_meta":{"raw":{"variants":["MACE force field maps why NaCl retards water but CsI accelerates it","Stronger Na+ shells slow water in NaCl; I−'s loose shell speeds CsI","MACE captures ion-specific water diffusion anomaly: NaCl slow, CsI fast","How Na+ second shell retards water while I− drives acceleration in CsI","MACE MLFF reproduces anomalous water diffusion in NaCl versus CsI"]},"model":"grok-4.5","effort":"low","cost_usd":0.007254,"raw_usage":{"total_tokens":1869,"prompt_tokens":894,"num_sources_used":0,"completion_tokens":111,"cost_in_usd_ticks":72540000,"prompt_tokens_details":{"text_tokens":894,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":864,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":894,"tokens_out":111,"duration_ms":7457,"temperature":1.0,"reasoning_tokens":864,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-12T20:39:19.128337+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Recompute the same concentration series of relative water diffusion coefficients with a MACE (or equivalent) force field trained on RPA or MP2 data for the same ions; if the NaCl slowdown or CsI speedup disappears or reverses relative to experiment, the central claim fails.","supporting_citations":[],"review_version":2}