{"id":"30d38999-cd1e-4cdb-b2fc-2a6b021def1d","arxiv_id":"2412.15452","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Nickel doping lowers the computed rate-limiting oxygen evolution barrier on BaTiO3 from about 1.5 eV to 1.2 eV in explicit-water metadynamics simulations.","lead":"This study builds a machine-learned atomistic model of barium titanate with water, then uses metadynamics simulations to trace the oxygen evolution reaction step by step. It finds that replacing one surface titanium with nickel lowers the key free energy barrier from about 1.5 eV to 1.2 eV, explaining why nickel-doped barium titanate is a better catalyst in experiments.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed TS3 barrier is a 2-CV projection of a proton-coupled reaction, yet Section 2.2 admits OOH* is indistinguishable; the quantitative 0.37 eV Ni-doping advantage depends on an unverified assumption that the two elementary steps merge without interfering.","rationale":"The reader's weakest assumption correctly identifies the 2-CV FES and the indistinct OOH* state as the key vulnerability. My analysis agrees and sharpens it: the problem is not merely a labeling ambiguity but the possibility of a projection-induced saddle in a proton-coupled reaction where the proton-transfer coordinate is not biased. The paper provides no committor analysis, no string-method check, and no comparison of the 2-CV barrier against a calculation that explicitly resolves OOH*, so the 0.37 eV barrier reduction is not fully established. That said, the qualitative trend is corroborated by prior DFT and experiment, the MLP is validated on energy/force errors and radial distribution functions, and three independent MetaD runs show a consistent direction. The internal numerical inconsistency between the text (1.57/1.20 eV) and Table C5 (1.48/1.26 eV) further supports the need for the authors to state which numbers are final and to release the dataset and code for independent re-analysis. Therefore, I do not recommend changing the reader's conditional verdict: the paper is plausible and worth publication, but the central quantitative claim requires either an additional validation of the reaction coordinate or a substantially revised, clearly qualified interpretation.","tokens_in":13855,"tokens_out":5848,"duration_ms":57680,"concrete_test":"Using the same MLP, rerun the OER metadynamics for one BTO and one Ni@BTO system with a third collective variable that explicitly tracks proton transfer from the reacting O-H to a water acceptor (e.g., CN(Ow-H) or a coordination-difference CV of the proton), and continue deposition past multiple O2 formation/readsorption events. Compute the committor distribution for configurations taken from the projected TS3 ridge. If OOH* appears as a resolved minimum, if the 3-CV TS3 barrier differs from the 2-CV value by more than roughly 0.1 eV, or if the committor at the projected saddle is not approximately 0.5, the 2-CV TS3 interpretation is a projection artifact and the quantitative barrier difference should be re-derived.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim — Ni doping lowers the rate-determining OER barrier from 1.57 to 1.20 eV — requires that the free energy barrier labeled ΔG‡O→O2 in the two-dimensional FES be the true free energy barrier for oxo-oxo bond formation plus the subsequent proton abstraction. The paper's own Section 2.2 states that the OOH* intermediate cannot be differentiated from O2* on the FES and that the OOH* minimum is indistinct; the subsequent assertion that 'the TS of these steps do not interfere' is an assumption, not a demonstrated property of the model. The two CVs are CN(Os-H) and CN(Os-Oaw). This pair does not include an explicit coordinate for the proton transfer from the nascent OOH* to a solvent water. A 2D projection of a higher-dimensional free energy surface can create a saddle that is not a true transition state, and sequential barriers can appear as a single merged maximum. The FES is also truncated at the first O2 formation event, so the reported barrier depends on when deposition is stopped. If the true mechanism has a resolvable OOH* intermediate or a slow orthogonal water-reorganization mode, the TS3 barrier extracted from this 2-CV FES is not the RDS barrier, and the 0.4 eV difference between BTO and Ni@BTO is not quantitatively established. The qualitative direction is supported by prior DFT and experiment, but the central quantitative claim is not.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a machine-learned interatomic potential (ANN-based, RPBE+D3 reference) for BaTiO3 and Ni-doped BaTiO3 with explicit water, and uses well-tempered metadynamics to map the free energy surface of the oxygen evolution reaction (OER) through the conventional four-step adsorbate mechanism and O2 desorption. The authors report that the rate-determining step is oxo-oxo bond formation (TS3), with a combined free energy barrier of 1.57 eV for BTO and 1.20 eV for Ni@BTO, and a corresponding theoretical overpotential reduction from 0.34 V to -0.03 V. They conclude that Ni doping lowers both kinetic and thermodynamic barriers, consistent with prior DFT and experimental comparisons. The methodological contribution includes an active-learning dataset of 16,162 configurations and MLP-MetaD simulations at explicit water interfaces.","tokens_in":14266,"tokens_out":2837,"duration_ms":26252,"significance":"If the quantitative barrier reduction is correct, this is a valuable demonstration of MLP-accelerated metadynamics for OER at explicit oxide/water interfaces, going beyond thermodynamic scaling relations to include kinetic barriers and solvent effects. The qualitative conclusion that Ni doping enhances OER activity on BaTiO3 is consistent with earlier DFT and experimental work, and the MLP pipeline with active learning is a reproducible strength. However, the central quantitative claim depends on the validity of a two-collective-variable projection of a proton-coupled reaction, and several internal inconsistencies in the reported barrier values and their statistical treatment currently prevent the quantitative result from being accepted as established.","major_comments":[{"comment":"The central claim that TS3 is the rate-determining step with barriers of 1.57 and 1.20 eV relies on the assertion that the OOH* intermediate is indistinct and that the transition states of oxo-oxo bond formation and proton abstraction 'do not interfere' on the free energy surface. The two collective variables, CN(Os-H) and CN(Os-Oaw), do not include an explicit coordinate for the proton transfer from OOH* to a solvent water molecule. A saddle point in this two-dimensional projection is not necessarily the true transition state of the combined reaction, and sequential barriers can merge into a single apparent maximum. The manuscript states this as an assumption, not as a tested property. A concrete test would be to repeat the MetaD with an additional CV for the O_w-H coordinate, or to perform committor analysis at the putative TS3, before the reported barrier heights can be considered mechanistically quantitative.","section":"Section 2.2 and Section 4.4"},{"comment":"The headline barriers are internally inconsistent. The text reports ΔG‡O→O2 = 1.57 eV for BTO and 1.20 eV for Ni@BTO, but Table C5 lists three individual FES values per system whose means are 1.48 ± 0.07 eV and 1.26 eV, respectively. The reported mean for Ni@BTO is 1.26 eV, not 1.20 eV, and no individual Ni@BTO run in the table gives 1.20 eV (the values are 1.31, 1.20, 1.26; the mean of 1.20, 1.26, 1.31 is not 1.26). The paper must either correct the tabulated statistics or state explicitly which runs and averaging procedure produce the claimed 1.57/1.20 values. As written, the quantitative comparison between theory and experiment is not reproducible from the data presented.","section":"Table C5 and Section 2.2"},{"comment":"The conversion of the activation barrier to a 'theoretical overpotential' by subtracting 1.23 eV yields -0.03 V for Ni@BTO. A negative theoretical overpotential is unphysical for a barrier-based estimate; the overpotential should be derived from the potential-dependent free energy change or from the largest free energy step at U=0, not from subtracting 1.23 eV from an activation free energy at zero applied potential. This step needs to be justified or removed, because the claim of 'close agreement' with the experimental overpotential reduction (0.34 V) currently rests on an undefined quantity.","section":"Section 2.3"},{"comment":"The free energy surfaces are 'calculated only until the O2 formation event occurs', as stated in Section 4.4. Because well-tempered metadynamics estimates free energies from accumulated bias, truncating the simulation at the first O2 formation event means the reported TS3 barrier depends on when deposition is stopped and on the time needed to cross the barrier, rather than on a converged free energy surface. The paper does not provide a convergence check (e.g., hill-to-hill variation or block analysis) for the TS3 region. Given that the central result is a 0.3-0.4 eV barrier difference, the stopping criterion is a load-bearing choice that must be shown not to bias the relative barriers.","section":"Section 4.4 and Figure C3/C4"},{"comment":"The MLP force error (MAE 132-143 meV/Å on the validation set) is large relative to the reported free energy barriers (0.06-0.24 eV for early steps) and to the TS3 difference between BTO and Ni@BTO. The RDF validation in Figure B1 is performed on only 15 ps of AIMD and the caption itself notes the simulations 'may not achieve equilibrium'. The manuscript does not quantify how MLP force and energy errors propagate into the metadynamics free energies. At minimum, the authors should report the MLP ensemble spread for the MetaD barriers (Figure C6 shows an ensemble comparison only for relative energies, not for barriers) and discuss the error budget for the 0.37 eV doping-induced reduction.","section":"Section 4.2, Appendix B, and Figure B1"}],"minor_comments":[{"comment":"The sentence 'There is no any potential impact on its interpretation of oxo-oxo bond formation energy barrier' is grammatically unclear and should be rewritten to state precisely what was tested.","section":"Section 2.2"},{"comment":"The phrase 'averaged over the FESs three MetaD runs' is missing a preposition; it should read 'averaged over the FESs from three MetaD runs'.","section":"Section 2.2 / Figure 5 caption"},{"comment":"The text contains 'as as shown' in the sentence about MACE computational time; this should be corrected.","section":"Appendix B, Table B4"},{"comment":"The repository URL is a placeholder ('https://github.com/atomisticnet/XXXXX'); the actual DOI or repository link should be provided for reproducibility.","section":"Data availability"},{"comment":"The standard deviation row for Ni@BTO reports ±0.00 for several quantities despite visible spread among the individual FES values; this is either a formatting error or a statistical error and should be corrected.","section":"Table C5"}],"recommendation":"major_revision","confidential_remarks":"The qualitative conclusion (Ni lowers the OER barrier on BaTiO3) is plausibly correct and aligns with prior DFT and experiment, but the manuscript's quantitative central claim is not yet supported by the reported data. The text/table inconsistency on the headline barriers, the unphysical negative theoretical overpotential, and the untested two-CV projection are all fixable with additional analysis. I recommend major revision rather than rejection because the methodological framework is useful and the issues appear addressable within the scope of the paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The short version: this paper builds a new machine learning potential for BTO/Ni-BTO/water, runs well-tempered metadynamics with explicit solvent, and concludes that Ni doping lowers the rate-determining OER barrier. The qualitative direction is probably correct—it agrees with three independent runs per system and with prior DFT and experiment. The quantitative claim is shakier.\n\nWhat's genuinely new: the MLP itself, the explicit-water free energy barriers for OER on these surfaces, and the O2 desorption check. The training pipeline is sensible: MACE-mp-0 for configurational sampling, RPBE+D3 relabeling, active learning to cover transition states, and validation against AIMD radial distribution functions. The authors also honestly flag the indistinct OOH* state and the omission of the lattice-oxygen mechanism. That's good practice.\n\nThe soft spots are real and need attention. First, the text reports barriers of 1.57 eV (BTO) and 1.20 eV (Ni@BTO), but Table C5 gives means of 1.48 ± 0.07 and 1.26 eV. The quoted 0.37 eV advantage shrinks to about 0.22 eV using the table, and the Ni@BTO standard deviation is listed as ±0.00, which looks like a rounding artifact. Second, the theoretical overpotential is derived directly from the kinetic barrier, giving a negative value (-0.03 V) for Ni@BTO. That's not a meaningful quantity; overpotential is a thermodynamic concept, and the relation between a barrier and overpotential is not this direct. Third, the central quantitative claim depends on the combined TS3 barrier from a 2-CV projection. Section 2.2 admits OOH* and O2* cannot be distinguished on the FES, and the assertion that the two steps do not interfere is an assumption, not a demonstrated property. If there is a resolvable OOH* intermediate or a slow orthogonal water-reorganization mode, the reported barrier could be an artifact of the CVs and the truncation at O2 formation. I'm not saying the paper is wrong—the qualitative trend is robust across runs and matches earlier work—but the specific barrier difference is not firmly established.\n\nWho should read this: people working on MLP-accelerated electrocatalysis simulations, especially those interested in explicit solvent effects. The MLP and dataset (if actually released) could be useful. The paper deserves a serious referee, but the authors should fix the numerical inconsistencies, remove or thoroughly reframe the overpotential comparison, and either strengthen the CV justification or soften the quantitative interpretation. I would not desk-reject it, but I would ask for major revision before publication.","headline":"A serious MLP+MetaD study with a likely-true qualitative conclusion, but the headline barrier numbers don't match the table, the negative overpotential is an error, and the combined TS3 barrier rests on an unproven assumption about the CV projection.","tokens_in":14726,"tokens_out":2428,"would_cite":false,"duration_ms":23516,"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":"Nickel doping lowers the rate-limiting oxygen evolution barrier on barium titanate from 1.57 eV to 1.20 eV, and identifies oxo-oxo bond formation as the bottleneck.","keywords":["oxygen evolution reaction","barium titanate","nickel doping","machine learning potential","metadynamics","free energy barrier","perovskite oxide","explicit water interface"],"falsifier":"Run the same metadynamics with a third collective variable that distinguishes the OOH* intermediate from the O2* product (for example, the coordination of the reacting water's hydrogen to the surface oxygen), or carry out static transition-state searches for the oxo-oxo bond formation step; if the barrier drops below 1.20 eV or the TS3 assignment changes, the central claim fails.","tokens_in":13679,"feed_emoji":"⚡","tokens_out":8704,"duration_ms":68087,"temperature":0.7,"pith_summary":"This paper tries to establish that replacing a surface titanium atom with nickel lowers the kinetic barrier of the oxygen evolution reaction on barium titanate in contact with liquid water, and identifies which elementary step controls the rate. The authors build a machine-learning interatomic potential trained on density functional theory data, then run well-tempered metadynamics simulations of the full water-covered surface. They find the rate-limiting step is the combined oxo-oxo bond formation and deprotonation that produces an adsorbed O2 species, with free energy barriers of 1.57 eV on pristine BaTiO3 and 1.20 eV on Ni-doped BaTiO3. They also find molecular O2 desorption is easy, about 0.16 eV and 0.13 eV, so it is not rate limiting. If correct, this explains experimentally observed overpotential reduction by Ni doping and provides a mechanistic handle for designing better perovskite OER catalysts.","feed_headline":"Nickel doping cuts barium titanate's OER barrier to 1.20 eV","feed_subtitle":"It matches the measured 0.34 V overpotential drop and identifies oxo-oxo bond formation as the bottleneck.","key_machinery":"The machinery is a machine-learned interatomic potential trained on density functional theory data and used to run well-tempered metadynamics simulations of a BaTiO3(001) slab with 128 explicit water molecules, 592 atoms in total. The reaction is biased along two collective variables: the coordination number between the adsorbed oxygen $O_s$ and hydrogen, $\\mathrm{CN}(O_s{-}H)$, and the coordination number between $O_s$ and surrounding water oxygens, $\\mathrm{CN}(O_s{-}O_{aw})$. These variables separate the $H_2O^*$, $OH^*$, $O^*$, and $O_2^*$ states on the free energy surface, though $OOH^*$ is not resolved as a distinct minimum; step 3 (oxo-oxo bond formation) and step 4 (proton abstraction) are therefore combined into a single TS3 barrier. A separate metadynamics run uses the distance from the O2 center of mass to the surface metal as the collective variable to measure O2 desorption.","core_discovery":"The central claim is that Ni doping lowers the free energy barrier of the oxygen evolution reaction at the BaTiO3(001)/water interface, and that the rate-determining step is the combined oxo-oxo bond formation and proton-abstraction event that produces an adsorbed O2 species. On pristine BaTiO3 the computed barrier is $\\Delta G^{\\ddagger}_{O \\to O_2} = 1.57$ eV; on Ni-doped BaTiO3 it is 1.20 eV. The corresponding theoretical overpotentials are 0.34 V and -0.03 V, and the predicted 0.34 V reduction matches the experimentally measured reduction from 0.80 V to 0.46 V on Ni-doped hollow porous spheres. The paper further claims that O2 desorption is not rate-determining, with barriers near 0.16 eV (BaTiO3) and 0.13 eV (Ni@BaTiO3), and that Ni doping also lowers the endothermicity of O2* formation from 1.37 eV to 0.97 eV.","pith_inferences":["Extension beyond the paper: if the same barrier ordering persists at operating electrode potentials, Ni-doped BaTiO3 could close the activity gap with iridium and ruthenium oxides, since its theoretical rate-limiting barrier of 1.20 eV already sits in the range of good OER catalysts.","Extension beyond the paper: the indistinct OOH* minimum suggests the conventional four-step OER picture may overstate the role of OOH* as a stable intermediate on this surface; descriptors built from OOH* adsorption energies may therefore be the wrong reactivity measure here.","Extension beyond the paper: the data-generation strategy of seeding with a broad pretrained potential and refining with active learning on unstable structures is directly transferable to other doped perovskites, allowing kinetic barriers to be screened without expensive ab initio molecular dynamics."],"forward_implications":["Ni doping reduces the rate-limiting OER free energy barrier from 1.57 eV to 1.20 eV, a drop of 0.37 eV that matches the experimentally observed overpotential reduction of 0.34 V.","O2 desorption is not rate-determining on either surface, with desorption barriers of about 0.16 eV (BaTiO3) and 0.13 eV (Ni@BaTiO3), so product release will not block the active site.","The rate-determining step is the formation of the oxo-oxo bond (step 3), not the later proton abstraction, which is barrierless on the computed free energy surface.","Ni doping lowers the endothermicity of O2* formation from 1.37 eV to 0.97 eV, making the overall OER both kinetically and thermodynamically more favorable.","The machine-learning potential trained for this five-element water/slab system provides a basis for future simulations of the lattice-oxygen-mediated mechanism, which the authors explicitly leave for later work."],"supporting_citations":[{"why":"Provides the initial BTO and Ni@BTO slab structures and the static DFT overpotential comparison that this work extends with explicit kinetics.","marker":"[13]"},{"why":"Experimental study of Ni-doped BTO hollow porous spheres; supplies the measured overpotential reduction from 0.80 V to 0.46 V used as the comparison target.","marker":"[19]"},{"why":"Pretrained general-purpose machine learning potential used to generate the initial configurational dataset for the bespoke potential.","marker":"[24]"},{"why":"Defines the exchange-correlation functional used for the DFT reference energies and forces that label the training data.","marker":"[29]"},{"why":"Supplies the dispersion correction combined with the functional in the DFT reference calculations.","marker":"[30]"},{"why":"Establishes the well-tempered metadynamics method used to compute the free energy surfaces and barriers.","marker":"[48]"},{"why":"Motivates checking whether O2 desorption could be the rate-determining step on oxide surfaces, which the paper's desorption simulations directly address.","marker":"[36]"},{"why":"Reports O2 desorption barriers on rutile oxide surfaces, providing the quantitative framing for the desorption results on BTO and Ni@BTO.","marker":"[37]"}],"fun_headline_variants":["Ni doping drops BaTiO3 OER barrier to 1.20 eV","Ni doping cuts BaTiO3 OER barrier by 0.37 eV","Machine learning reveals Ni doping OER mechanism on BaTiO3","Ni doping matches measured OER overpotential drop on BaTiO3"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the two reaction coordinates chosen for the metadynamics—how many hydrogens are bound to the surface oxygen and how many water oxygens are bound to it—fully capture the path of the rate-limiting step, so that the computed 1.20 eV and 1.57 eV barriers are real and not an artifact of how the simulation was biased.","fun_headline_variants_meta":{"raw":{"variants":["Ni doping drops BaTiO3 OER barrier to 1.20 eV","Ni doping cuts BaTiO3 OER barrier by 0.37 eV","Machine learning reveals Ni doping OER mechanism on BaTiO3","Ni doping matches measured OER overpotential drop on BaTiO3"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.002194,"raw_usage":{"total_tokens":8517,"prompt_tokens":988,"completion_tokens":7529,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":604,"completion_tokens_details":{"reasoning_tokens":7451}},"tokens_in":604,"tokens_out":7529,"duration_ms":47346,"temperature":1.0,"reasoning_tokens":7451,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T11:25:12.668557+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same metadynamics with a third collective variable that distinguishes the OOH* intermediate from the O2* product (for example, the coordination of the reacting water's hydrogen to the surface oxygen), or carry out static transition-state searches for the oxo-oxo bond formation step; if the barrier drops below 1.20 eV or the TS3 assignment changes, the central claim fails.","supporting_citations":[{"cited_title":"Physical Chemistry Chemical Physics 18(42), 29561–29570 (2016)","cited_arxiv_id":null,"evidence_quote":"Provides the initial BTO and Ni@BTO slab structures and the static DFT overpotential comparison that this work extends with explicit kinetics."},{"cited_title":"Applied Surface Science 618, 156599 (2023) 34","cited_arxiv_id":null,"evidence_quote":"Experimental study of Ni-doped BTO hollow porous spheres; supplies the measured overpotential reduction from 0.80 V to 0.46 V used as the comparison target."},{"cited_title":"The Journal of chemical physics 132(15) (2010)","cited_arxiv_id":null,"evidence_quote":"Supplies the dispersion correction combined with the functional in the DFT reference calculations."},{"cited_title":"Physical review letters 100(2), 020603 (2008) 39","cited_arxiv_id":null,"evidence_quote":"Establishes the well-tempered metadynamics method used to compute the free energy surfaces and barriers."},{"cited_title":"Energy & Environmental Science 15(6), 2519–2528 (2022)","cited_arxiv_id":null,"evidence_quote":"Motivates checking whether O2 desorption could be the rate-determining step on oxide surfaces, which the paper's desorption simulations directly address."},{"cited_title":"The Journal of Physical Chemistry C 121(21), 11455–11463 (2017)","cited_arxiv_id":null,"evidence_quote":"Reports O2 desorption barriers on rutile oxide surfaces, providing the quantitative framing for the desorption results on BTO and Ni@BTO."}],"review_version":1}