{"id":"7bc86035-a177-41e6-9994-563d930854e6","arxiv_id":"2505.04644","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Simulations show Na+ release from the Preyssler anion is favored through a dissociative pathway, and a single confined water molecule lowers the release barrier by about 2.2 kcal/mol.","lead":"Using machine-learned MACE molecular dynamics and metadynamics, this paper simulates how sodium ions enter and leave the Preyssler anion, a doughnut-shaped polyoxometalate cluster, in water. It finds one trapped water molecule in the cavity weakens the ion's hold and shifts the capture pathway, which matters for designing selective ion-capture materials.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"MACE-MPA-0_medium force-field error on unbenchmarked reaction paths is not quantified; the 2.2 kcal/mol confined-water effect and the 21.3 vs 26.8 kcal/mol mechanism ordering sit on top of an unvalidated surrogate potential.","rationale":"The reader's weakest_assumption identifies exactly the same load-bearing concern: the MACE-MPA-0_medium potential is quantitatively accurate for all three unbenchmarked pathways, and the 2.2 kcal/mol difference is smaller than expected surrogate error. My review agrees with that diagnosis and adds specificity: the single AIMD benchmark covers only the dissociative Na(H2O)@PA path, with no quantitative barrier comparison or error estimate; the associative path and Na@PA path are entirely unvalidated; the CV switching functions are fitted to MACE RDFs rather than AIMD data; and no WT-MetaD convergence analysis (block error, hill-height decay) is provided despite the 'well-converged' claim. These are correctness-risk issues, not disagreements with the general approach or with the qualitative dissociative-mechanism picture. They justify CONDITIONAL rather than ACCEPT or REJECT. I found no internal inconsistency in the methods as described, and the qualitative narrative—confined water shifts Na+ coordination and lowers ejection barrier—is plausible and internally consistent with the CN analysis. The absence of deposited code/data is a secondary concern, not the central one. My concrete test targets the weakest link: an AIMD or alternative-potential recomputation of the Na@PA dissociative barrier, which would directly test whether the 2.2 kcal/mol effect survives surrogate-model error.","tokens_in":18828,"tokens_out":1960,"duration_ms":17334,"concrete_test":"Reproduce the dissociative FES for Na@PA with an alternative validated potential or with short AIMD-based WT-MetaD (even at reduced box size, e.g., 110 H2O, 5 walkers x 15-20 ps) and compare the barrier to the 23.5 kcal/mol MACE value; if the AIMD barrier differs by more than ~2 kcal/mol, the 2.2 kcal/mol water-induced shift and the 21.3 vs 26.8 kcal/mol mechanism ordering are not established. Also compute a per-WTK block error or hill-height-convergence plot for the MACE FESs in Fig. 5 to verify the 21.3/23.5/25.6 kcal/mol values are converged to <1 kcal/mol, not merely 'well-converged' as claimed.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim—that Na+ capture in PA is dissociative with a 21.3 kcal/mol barrier and that one confined water lowers that barrier by ~2.2 kcal/mol (23.5 kcal/mol for Na@PA)—rests entirely on WT-MetaD free energies computed with off-the-shelf MACE-MPA-0_medium (Section II.C, Section III.B). The only AIMD benchmark is a single dissociative pathway for Na(H2O)@PA, run for a total of 75 ps across five walkers (15 ps each, Section II.D), which the text says gives a 'similar' barrier but provides no quantitative comparison, no error estimate, and no published FES in the main text. The 2.2 kcal/mol difference between Na(H2O)@PA and Na@PA is smaller than the expected force-field error of a general-purpose foundation model on a confined, highly charged internal cavity containing W, P, Na, and O, so this difference is not established. Similarly, the 21.3 vs 26.8 kcal/mol preference for dissociative over associative exchange is also within plausible surrogate error. The paper also lacks convergence analyses for the WT-MetaD FES (no block-error or hill-height convergence), and the CVs are CN-based with switching functions fitted to equilibrium MACE MD rather than to AIMD RDFs, so the bias is not calibrated against reference. The internal-consistency argument from CN analysis is supportive but not sufficient: CNs are structural observables, not free energies, and the single AIMD check does not cover the associative path or the Na@PA path. Thus the load-bearing premise—quantitative accuracy of MACE on these reaction paths—is unverified and could shift the mechanism ordering if the error is a few kcal/mol.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript combines equilibrium MACE-accelerated molecular dynamics with multiple-walker well-tempered metadynamics to study Na+ capture and release in the Preyssler anion {P5W30} in aqueous solution. Two transport mechanisms are compared: an associative ion-exchange path and a dissociative ejection path, and the effect of a single confined water molecule inside the cavity is probed by contrasting Na(H2O)@PA with Na@PA. The principal claims are that Na+ capture proceeds dissociatively with a barrier of 21.3 kcal/mol, that associative exchange is less favorable (26.8 kcal/mol), that removing the confined water raises the dissociative barrier to 23.5 kcal/mol, and that ejecting the confined water itself costs 25.6 kcal/mol. The paper also reports structural analyses (RDFs, coordination numbers, window diameters) that support a picture in which confined water weakens Na+–OPO4 interactions and lowers the barrier to ion release.","tokens_in":19165,"tokens_out":2737,"duration_ms":31396,"significance":"If the quantitative claims are correct, the paper offers a mechanistic explanation for ion capture/release in Preyssler-type POMs and a plausible route toward designing POM-based sorbents with tunable barriers. The work has clear strengths: it benchmarks six MACE models against AIMD in terms of RDFs, presents careful equilibrium CN/RDF analyses distinguishing interior and exterior coordination environments, explicitly tests the role of confined water, and uses multiple-walker WT-MetaD with a pywindow analysis of window contraction/expansion. The CVs are not fitted to the reported free-energy barriers, so there is no direct circularity in the central numbers. However, the central quantitative conclusions rest on a single off-the-shelf machine-learned potential whose error on unbenchmarked reaction paths is not quantified, and the paper provides no statistical uncertainties or convergence diagnostics for the computed free-energy differences.","major_comments":[{"comment":"The well-tempered metadynamics setup is incompletely specified: the bias factor γ in Eq. (2) is never given, nor are the total bias deposition time, the convergence of the deposited hills, or the FES convergence as a function of simulation time. The manuscript describes the runs as \"well-converged\" (Abstract and Section III.B) but provides no hill-height evolution, no block-error analysis, and no comparison of FESs from different walker sets or simulation lengths. Without such diagnostics, the reported barrier heights cannot be distinguished from unconverged estimates.","section":"II.D"},{"comment":"The AIMD validation is not quantitative. The text states that the AIMD benchmark for dissociative ejection in Na(H2O)@PA \"resulted in a similar free energy barrier\" (Section II.D) and later that the 21.3 kcal/mol MACE value \"matches that of the reference AIMD\" (Section III.B.2). However, no AIMD barrier value is reported in the main text, no comparison metric is given, and the AIMD run is only 75 ps total (5 walkers x 15 ps) with Gaussians every 40 steps. This is too short to establish convergence, and the absence of a quantitative number makes the validation claim unfalsifiable as written. At minimum, the AIMD barrier and its uncertainty (or a statement of the uncertainty from the CV time series) must be reported, and at least one additional pathway (e.g., Na@PA or the associative path) should be validated if the barrier ordering is to be supported.","section":"II.D / III.B.2"},{"comment":"The central free-energy barriers (26.8, 21.3, 23.5, 25.6 kcal/mol) are reported as single numbers without statistical uncertainties. The 2.2 kcal/mol difference between Na(H2O)@PA and Na@PA, and the 5.5 kcal/mol difference between associative and dissociative mechanisms, are the load-bearing quantitative claims of the paper. Because the surrogate MACE potential is used for all biased runs and its error on unbenchmarked paths is unknown, these differences could easily be within the combined statistical and systematic error. The CNs in Table II are given with ± one standard deviation, but the ΔF values have no equivalent error bars. The authors should provide error estimates for the barriers, for example from block averaging over walkers or independent replicate WT-MetaD runs, and should temper the mechanistic conclusions accordingly.","section":"III.B / Table II"}],"minor_comments":[{"comment":"The title contains a typo: \"T ransport\" should read \"Transport\".","section":"Title"},{"comment":"The sentence \"the cutoff distance r c and the switching functions were fitted based on the first solvation shell peaks from the RDF analyses of the final 200 ps of the equilibrium MACE MD simulations\" (Section II.D) is unclear because it implies the switching functions themselves were fitted, but only the cutoff distance appears to be adjusted. Clarify how the functional form of Eq. (3) was chosen.","section":"II.D"},{"comment":"The use of superscripts and subscripts in the TS columns (e.g., \"TS 1−→2 / 0−→6\") is not defined in the table caption or the text. Please define the notation explicitly, since it is central to reading the CN changes at the transition state.","section":"Table II"},{"comment":"The sentence \"The higher calculated barrier of 4.3 kcal/mol for water ejection likely arises from its larger kinetic diameter (≈2.65) compared to bare Na+ (≈2.02 Å)\" gives a plausible physical rationale but is presented without a quantitative test. If the kinetic diameter argument is intended as an explanation, it would benefit from a direct comparison of the window size to the two diameters.","section":"III.B.3"},{"comment":"The text refers to \"three distinct minima\" in MS2 with CN values of approximately 2, 3, and 5, and then states that a 2.1 kcal/mol barrier separates them. It is not clear whether this 2.1 kcal/mol is the barrier between the deepest and shallowest of those minima or between adjacent minima; please state the value precisely and identify the minima on the FES (Fig. 5c).","section":"III.B.4"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses a timely and interesting problem, and the structural analysis is careful. My recommendation of major revision is driven by the mismatch between the precision of the central quantitative claims (e.g., a 2.2 kcal/mol water effect) and the lack of reported uncertainties or a quantitative AIMD benchmark. The authors should be encouraged to add error bars, convergence diagnostics, the missing bias factor, and at least one additional AIMD check; without these, the paper would not meet the standard expected for a central mechanistic claim. I see no ethical or scope concerns."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the headline: this is the first free-energy calculation of Na+ transport through the Preyssler cavity, and the core claim—confined water lowers the ejection barrier by shielding Na+ from the PO4 oxygens—is plausible and internally consistent. But the quantitative barriers come with no error bars, and the 2.2 kcal/mol water effect sits right on top of the expected uncertainty of an off-the-shelf MACE potential.\n\nWhat's genuinely new: no prior simulation in the cited literature reconstructs the FES for Na+ capture/release in this POM with and without confined water. The associative vs dissociative comparison and the window-diameter analysis are also new. The paper does real benchmark work: six MACE models are compared against AIMD RDFs, and the chosen model reproduces the main structural features. The equilibrium CN/RDF analysis is careful, and the interpretation that water weakens Na-OPO4 coordination is backed by both equilibrium and biased runs. The two-step heating/capture design rule is stated with an appropriate caveat.\n\nThe soft spots are quantitative. The three barriers (21.3, 23.5, 26.8 kcal/mol) are reported without uncertainties, and the text calls the WT-MetaD 'well-converged' without showing hill-height or block-error convergence. The AIMD check covers only the dissociative path in Na(H2O)@PA, runs just 75 ps total, and the paper only says the barrier is 'similar'—no numbers. That's a thin benchmark for a foundation model in a highly charged, confined W/P/Na/O environment. The CV switching functions are fit to equilibrium MACE RDFs rather than AIMD data, so the bias and the model share the same reference. No input files or trajectory data are deposited.\n\nAre these fatal? Not for the qualitative mechanism. The 5.5 kcal/mol gap between dissociative and associative is large enough that it might survive a few kcal/mol of model error. The 2.2 kcal/mol water effect is the fragile number; I wouldn't trust it without a dedicated AIMD or an error-informed MLIP check. The paper's own caution about generalizing is appropriate.\n\nWho gets value: computational chemists working on POM separations, and anyone testing foundation-model MLIPs on confined reactive environments. It deserves peer review: the question is timely, the workflow is sound, and the weaknesses are fixable. I'd ask for error bars, a convergence analysis, at least one more AIMD benchmark (ideally the Na@PA dissociative path), and a data/code deposit. Conditional accept is the right call, not desk reject.","headline":"First free-energy map of Na+ transport through the Preyssler cavity, with a plausible water-shielding mechanism but barriers that need error bars before the 2.2 kcal/mol effect is trusted.","tokens_in":19773,"tokens_out":2727,"would_cite":true,"duration_ms":28620,"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":"Ion release from a Preyssler anion cage proceeds by direct ejection, not exchange","keywords":["Preyssler anion","polyoxometalate","ion transport","sodium ion","free energy barrier","metadynamics","machine learning potential","confined water"],"falsifier":"Run the associative pathway and the water-free dissociative pathway at the same ab initio level used for the benchmark, or measure Na+ release rates from Na(H2O)@PA versus Na@PA in temperature-controlled experiments; if the associative barrier is not higher than the dissociative one, or if removing water does not raise the barrier, the central mechanism claim is wrong.","tokens_in":18631,"feed_emoji":"🧪","tokens_out":5681,"duration_ms":58444,"temperature":0.7,"pith_summary":"The paper asks how sodium ions get in and out of the Preyssler anion, a doughnut-shaped polyoxometalate cluster whose internal cavity makes it a candidate sorbent for recovering metals from saline water. Using equilibrium and biased molecular dynamics driven by a machine-learned potential benchmarked against ab initio simulations, it computes free-energy surfaces for two competing pathways. It claims that releasing an encapsulated Na+ by simply ejecting it into solution costs 21.3 kcal/mol, whereas exchanging it for an incoming Na+ costs 26.8 kcal/mol, so the dissociative route is favored. It also claims that a single pre-encapsulated water molecule lowers the dissociative barrier by about 2.2 kcal/mol by shielding the ion from the cavity's phosphate oxygens, while ejecting that water itself costs 25.6 kcal/mol. If correct, this gives a concrete design rule: regenerate the sorbent by heating in deionized water, and expect the ion to leave before the confined water does.","feed_headline":"Na+ leaves the Preyssler cage by ejection, not exchange","feed_subtitle":"Simulations place the release barrier at 21.3 kcal/mol and show one confined water lowers it by 2.2.","key_machinery":"The argument is carried by the Preyssler anion itself—a {P5W30} cluster with a ~5 Å internal cavity and two windows—together with coordination-number collective variables that track how many phosphate oxygens and water oxygens surround the encapsulated Na+. Multiple-walker well-tempered metadynamics, using an off-the-shelf machine-learned interatomic potential benchmarked against ab initio MD for structure and for one barrier, maps the free-energy surfaces; window diameters measured along the trajectories connect cage flexibility to the mechanism.","core_discovery":"The central claim is that ion capture in the Preyssler anion operates through a dissociative mechanism: the encapsulated Na+ passes through a window and is solvated by bulk water without needing a second ion to enter. The computed free energy barrier for this process is 21.3 kcal/mol in Na(H2O)@PA, versus 26.8 kcal/mol for the associative ion-exchange pathway, and the difference shows up in coordination-number changes at the transition state. A pre-encapsulated water molecule acts as a modulator: it hydrogen-bonds to the phosphate oxygens, reduces the Na+ coordination number from roughly 6.5 to 4.9, and lowers the dissociative barrier by 2.2 kcal/mol relative to the water-free Na@PA complex. The same water is harder to eject than the ion (25.6 kcal/mol), and the two windows of the cage contract and dilate by about 0.05–0.1 Å during transport, coupling molecular breathing to ion passage.","pith_inferences":["The same water-shielding logic likely applies to other polyoxometalate clusters and frameworks with internal cavities, meaning the hydration state inside the cavity should be treated as a tunable variable in sorbent design rather than a fixed crystallographic detail.","Because the 2.2 kcal/mol difference between Na(H2O)@PA and Na@PA is smaller than typical errors of surrogate machine-learned potentials, the quantitative ordering between these two systems remains to be confirmed by direct ab initio free-energy calculations or temperature-dependent kinetic measurements.","A testable extension is to vary the encapsulated ion or the window size; the model predicts that ions with a larger kinetic diameter than Na+ should make the associative pathway comparatively more favorable as window dilation becomes rate-limiting."],"forward_implications":["PA-based sorbents should be regenerated by heating in deionized water, since direct ejection is 5.5 kcal/mol cheaper than ion exchange.","Na+ will be released before the confined water leaves, because water ejection costs 25.6 kcal/mol versus 21.3 for the ion; the cavity retains its water during regeneration.","The confined water is not a passive occupant: it lowers the ejection barrier by about 2.2 kcal/mol by cutting Na+ coordination to the phosphate oxygens from 6.5 to 4.9.","The two windows of the cage open and close measurably during transport, so window flexibility is part of the capture and release mechanism.","Ion capture at room temperature is a rare event, consistent with the stability of the encapsulated ion observed over 1 ns of equilibrium simulation."],"supporting_citations":[{"why":"Supplies the experimental crystal structure of PA with encapsulated Na+ and water, giving the initial Na(H2O)@PA configuration and the ~5 Å cavity picture.","marker":"[25]"},{"why":"Provides the machine-learned interatomic potential architecture and foundation model used for all equilibrium and biased MD simulations.","marker":"[54,55]"},{"why":"Establishes the multiple-walker metadynamics approach that lets five walkers jointly reconstruct the free-energy surfaces.","marker":"[90]"},{"why":"Defines well-tempered metadynamics, the enhanced-sampling method whose convergence properties underlie the reported barriers.","marker":"[91,92]"},{"why":"Supplies the software interface for defining collective variables and depositing bias in both the MACE and ab initio simulations.","marker":"[93,94]"},{"why":"Furnishes the pore-analysis tool used to measure PA window diameter changes during ion and water transport.","marker":"[96]"}],"fun_headline_variants":["Ion ejects from Preyssler cage more easily than water swap","Single water reduces Na+ escape barrier in Preyssler by 2.2 kcal","Preyssler cage prefers ejecting Na+ to ion exchange","Caged water lowers dissociative barrier for Na+ release","Ejection, not exchange, is the Na+ exit from Preyssler"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the off-the-shelf machine-learned potential remains quantitatively accurate for all three computed free-energy pathways, especially the associative route and the 2.2 kcal/mol water-shielding difference, since only the structure and the single dissociative barrier were checked against ab initio reference data.","fun_headline_variants_meta":{"raw":{"variants":["Ion ejects from Preyssler cage more easily than water swap","Single water reduces Na+ escape barrier in Preyssler by 2.2 kcal","Preyssler cage prefers ejecting Na+ to ion exchange","Caged water lowers dissociative barrier for Na+ release","Ejection, not exchange, is the Na+ exit from Preyssler"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001018,"raw_usage":{"total_tokens":4348,"prompt_tokens":1045,"completion_tokens":3303,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":661,"completion_tokens_details":{"reasoning_tokens":3208}},"tokens_in":661,"tokens_out":3303,"duration_ms":22002,"temperature":1.0,"reasoning_tokens":3208,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:39:02.549635+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the associative pathway and the water-free dissociative pathway at the same ab initio level used for the benchmark, or measure Na+ release rates from Na(H2O)@PA versus Na@PA in temperature-controlled experiments; if the associative barrier is not higher than the dissociative one, or if removing water does not raise the barrier, the central mechanism claim is wrong.","supporting_citations":[{"cited_title":"\\ Kim , author M","cited_arxiv_id":null,"evidence_quote":"Supplies the experimental crystal structure of PA with encapsulated Na+ and water, giving the initial Na(H2O)@PA configuration and the ~5 Å cavity picture."},{"cited_title":"Raiteri , author A","cited_arxiv_id":null,"evidence_quote":"Establishes the multiple-walker metadynamics approach that lets five walkers jointly reconstruct the free-energy surfaces."},{"cited_title":"Miklitz \\ and\\ author K","cited_arxiv_id":null,"evidence_quote":"Furnishes the pore-analysis tool used to measure PA window diameter changes during ion and water transport."}],"review_version":1}