{"id":"3af80b0c-4db5-4af2-aa03-e9f408819282","arxiv_id":"2511.09521","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Wadsley defects and cation disorder both make Li occupy fast-diffusion window sites at lower lithiation, explaining why defect-rich MoNb12O33 outperforms its ordered counterpart.","lead":"Defect-rich MoNb12O33 made at 800 °C stores more lithium and charges faster than a more ordered version made at 900 °C. Computer simulations suggest two kinds of crystal defects both make lithium move through fast pathways earlier, pointing to a defect-engineering strategy for fast-charging battery anodes.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The Wadsley-defect models charge-balance by O vacancies while MNO-800's own EXAFS/STEM show Mo in octahedral/block sites; if real enlarged blocks balance charge via Mo substitution instead, the computed early window-site activation may not transfer.","rationale":"The reader's weakest assumption already identifies the Wadsley-defect charge-compensation mismatch with EXAFS/STEM; I agree and make it the primary concern. This is load-bearing because the concluding sentence attributes MNO-800's improved diffusion to both defect classes, and the Wadsley-defect class is represented only by models whose charge-balancing defect contradicts the experimental evidence. The concern is not fatal: the experimental comparison (MNO-800 vs MNO-900) is extensive and the MLIP-NEB validation against DFT is genuine supporting evidence, so a conditional acceptance remains appropriate rather than rejection. I also considered the MD/Arrhenius-extrapolation and single-disordered-model issues; those affect quantitative confidence but are secondary to the representativeness of the Wadsley-defect models. A targeted recomputation with Mo-in-octahedra block models would settle whether the occupancy shift is robust. Since the reader's verdict is already CONDITIONAL, I recommend UNCHANGED.","tokens_in":22216,"tokens_out":6409,"duration_ms":67902,"concrete_test":"Construct T[3×5] and T[3×4-3×5] supercells with the same metal stoichiometry but Mo placed in octahedral block sites (informed by the STEM-EDS/EXAFS evidence), instead of Mo-only-tetrahedral with O vacancies; relax with DFT, validate the MLIP against DFT-NEB for the new models, and repeat the CE/MC site-occupancy sequence plus 300 K MD D(x) at 25/50/75% Li. If fast Wh/Wv sites no longer populate below x≈6, the Wadsley-defect half of the central claim does not survive.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central mechanistic claim is that Wadsley defects and cation disorder independently lower the lithiation extent at which fast Wh/Wv window sites become occupied. The Wadsley-defect leg of this argument rests on Type-3 models (T[3×5], T[3×4-3×5]) built, as the paper states, by enlarging blocks while keeping Mo only in tetrahedral positions and restoring charge balance with oxygen vacancies. This is not the defect chemistry the same paper measures in MNO-800: Mo K-edge EXAFS cannot be fit with a uniform tetrahedral Mo shell and shows higher average coordination, and STEM-EDS maps Mo throughout [3×8] blocks and along shear planes. If real Wadsley defects incorporate Mo into octahedral block sites rather than generating O vacancies, the local Li-site energetics are different. In the current model the O vacancy itself creates a 528 meV barrier and strongly slows neighboring middle windows, so the early activation attributed to 'block size' may be an artifact of that particular charge-compensation choice. The paper also never models the actual combined defect structure, so the independent 'both defect types' attribution is not uniquely supported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript reports a comparative study of two MoNb12O33 powders from a single sol-gel batch calcined at 800 °C and 900 °C. The defect-rich MNO-800 sample shows larger XRD microstrain, more Wadsley-type block-size variations and mixed T/E interfaces in STEM, higher and less well-defined Mo coordination in EXAFS, higher capacity and rate capability, ~3× larger capacity-weighted diffusivity from ICI, and lower volume expansion on lithiation. To explain the transport enhancement, the authors train a MACE-based MLIP on DFT and use NEB, cluster-expansion/Monte Carlo, and MD to study pristine T[3×4], cation-disordered T[3×4] variants, and enlarged-block T[3×5] and T[3×4-3×5] models with O vacancies. The computations predict that both cation disorder and Wadsley-type block enlargement cause the fast Wh/Wv window sites to become occupied at lower Li content than in the pristine model, matching the earlier onset of high diffusivity in MNO-800. The paper concludes that both defect types contribute independently to the enhanced performance.","tokens_in":22484,"tokens_out":15988,"duration_ms":161550,"significance":"The experimental dataset is rich and internally consistent: same-batch synthesis, multiple structural probes, operando WAXS, and a defensible ICI/SAXS diffusivity protocol. The computational work is mostly independent of the measured transport—no diffusion coefficient is fitted to experiment—and the MLIP is checked against DFT-NEB with ~10–15 meV MAE. If the mechanistic conclusion holds, the paper would provide a design principle for defect-enhanced WR anodes and would be of broad interest. However, the computational representation of the defect structures is the weakest link: the Wadsley-defect models are not the defect chemistry measured in MNO-800, and the cation-disorder conclusion rests on one selected motif rather than the SQS representative. These issues are load-bearing because the final attribution of the ~3× diffusivity enhancement to 'both defect types' depends on them.","major_comments":[{"comment":"The Wadsley-defect leg of the central claim is carried by T[3×5] and T[3×4-3×5] models with Mo only in tetrahedral sites and charge balance restored by O vacancies. This conflicts with the paper's own structural evidence for MNO-800: Mo K-edge EXAFS could not be fit to a model and indicates a markedly higher average Mo coordination than ~4.2; STEM-EDS maps Mo throughout [3×8] blocks and along shear planes. The Type-3 formulas also change the Mo:Nb ratio (1:15 and 2:27 vs 1:12). Because the O vacancy itself creates a 528 meV barrier and slows middle windows, the 'block size' effect on early window-site activation is entangled with the assumed charge-compensation mechanism. Please test enlarged-block models with Mo in octahedral/corner sites at the measured composition, or restrict the claim to O-vacancy-bearing block enlargements.","section":"Results – Computational Analysis (Type-3 models, Fig. 5c,d)"},{"comment":"The cation-disorder mechanism is demonstrated with T[3×4-dis-1] (Mo at the block center), but the SQS model T[3×4-dis-5], described as representative of the disordered configuration, does not enhance diffusion relative to pristine T[3×4] and is not used in the site-occupation comparison. The paper should justify why dis-1 is the appropriate representation of disorder in MNO-800 (e.g., by matching the Mo-in-block distribution seen by STEM-EDS) or qualify the conclusion that cation disorder generally lowers the x threshold for activation of fast sites.","section":"Results – Computational Analysis (Type-2 models, T[3×4-dis-5])"},{"comment":"The abstract and conclusion attribute the improvement to 'both defect types,' but no model combines Wadsley block-size variation with cation disorder in a single simulation, and no model represents the actual integrated defect structure of MNO-800. The separate Type-2 and Type-3 results can show that each isolated motif is capable of shifting site occupancies; they do not establish that the effects are independent or additive in the real material. The manuscript itself notes after the STEM-EDS analysis that 'there remains ambiguity as to how each of these defects alters lithium diffusion.' A combined-defect calculation or an explicit discussion of interaction effects is needed before the concluding attribution can be accepted.","section":"Conclusion and abstract"},{"comment":"Table 1 is used to state that D is higher at 25% and 50% Li (order: dis-1 > 3×4-3×5 > 3×5 > pristine). These values are 300 K extrapolations from 2 ns MLIP-MD runs at 500–900 K, but no statistical uncertainty, trajectory convergence, or Arrhenius-fit quality is reported, and each concentration uses only three configurations. Please report error bars or confidence intervals so the ordering and factor differences in Table 1 can be evaluated.","section":"Methods – Molecular dynamics; Table 1"}],"minor_comments":[{"comment":"The text says concentration-dependent MD were 'performed at 300 K' while Methods states D at 300 K was extracted by Arrhenius extrapolation from 500–900 K. Please clarify which values are direct and which are extrapolated.","section":"Methods – Molecular dynamics / Results Table 1 paragraph"},{"comment":"The three parabolic D(x) regions are described by widths and positions but no fitting procedure is given. State whether these are fits or guides and report the fit function/criteria.","section":"Fig. 4e"},{"comment":"Typos and notation: 'lowers' should be 'lower'; 'intensity of the peak reducing with annealing temperature' is ambiguous; non-integer oxygen stoichiometry in MoNb15O40.5 and Mo2Nb27O73.5 should be explicitly defined (per formula unit vs supercell).","section":"General"}],"recommendation":"major_revision","confidential_remarks":"This is a strong experimental study with a promising computational framing, but the current manuscript overstates the computational support for the mechanism. The main revision should focus on making the Type-2 and Type-3 models representative of the measured defect chemistry or, failing that, on softening the 'both defect types' conclusion to a conditional statement. I do not see a need to discard the experimental results; the paper can be publishable after major revision. The citation of the authors' own methodology papers is not problematic, but the novelty claim should rest on the mechanistic separation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing to know: this is a serious, well-executed study that actually tries to separate two confounding defect types, and it mostly succeeds. The experimental comparison of two samples from one sol-gel batch is clean, and the characterization depth (XRD, XAS, STEM, ICI, SAXS, operando WAXS) is well above the field's norm. The central mechanistic idea — that both Wadsley defects and cation disorder shift Li occupation into fast Wh/Wv window sites at lower lithiation — is new relative to prior reports, which treated defect-enhanced diffusion more qualitatively. The MLIP-MD work is also properly grounded: the potential is trained on DFT, not on measured transport, and the NEB/MD results are cross-checked against DFT-NEB. That is real evidence, not curve-fitting. I think the reader's assessment is about right: conditional, moderately confident.\n\nSoft spots, in order of importance. First, the Wadsley-defect models charge-balance enlarged blocks with oxygen vacancies while keeping Mo tetrahedral, but the paper's own EXAFS shows MNO-800 Mo is not uniformly tetrahedral and STEM-EDS places Mo inside blocks. The authors acknowledge this tension; they even note the real structure requires both oxygen vacancies and octahedral Mo. But that means the type-3 models are idealized end-members, and the 'block-size effect' they isolate is partly an oxygen-vacancy effect. The stress-test note is fair on this point, though it slightly overstates the damage: the cation-disorder leg of the argument (type-2 models) is less affected, and both defect types independently shift occupation toward fast sites, so the qualitative conclusion survives even if the quantitative decomposition is model-dependent.\n\nSecond, the 300 K MD diffusivities are Arrhenius extrapolations from 500–900 K with no error bars. For trend comparisons this is acceptable, but the three-fold ratios in Table 1 should not be read literally. Minor.\n\nThird, the cation-disorder narrative leans heavily on one of five models (T[3×4-dis-1]) without showing the others in the main text. The SI apparently has the details, but the main-text presentation makes the 'disorder activates fast sites' claim look more universal than the model set supports. Also minor.\n\nThe self-citations are appropriate — they are prior method papers by the same group, and citing them here is legitimate, not a red flag.\n\nBottom line: the paper deserves peer review. The mechanistic decomposition is plausible, the experiments are thorough, and the computational work is reproducible in principle even without shipped code. A good referee would push on the Wadsley-defect charge-compensation assumption and ask for the other disorder models to be shown, but these are revision-level concerns, not fatal flaws. I'd bring it to reading group and would likely cite it if I work on WR anodes. Verdict: accept for review, expect conditional acceptance after revision.","headline":"Strong experimental/computational pairing that plausibly explains defect-enhanced Li diffusion in MoNb12O33, with the main caveat that the Wadsley-defect models rely on a charge-compensation assumption the paper's own EXAFS/STEM do not support for the real MNO-800 sample.","tokens_in":23036,"tokens_out":751,"would_cite":true,"duration_ms":11170,"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":"This paper establishes that the faster lithium diffusion in defect-rich MoNb12O33 comes from a change in which lithium sites fill first, not from faster hopping in any single channel. In the ordered block structure, lithium first occupies s","keywords":["Wadsley–Roth niobates","MoNb12O33","lithium-ion diffusion","cation disorder","Wadsley defects","cluster expansion","machine-learned interatomic potential","anode materials"],"falsifier":"Measure the occupancy of fast window sites at low lithiation (e.g., x≈3–5) in MNO-800 using neutron diffraction or 7Li NMR; if fast Wh/Wv sites are not occupied early, the proposed mechanism is wrong. Alternatively, compute the defect-rich model with Mo placed in octahedral sites and no oxygen vacancies; if the fast sites no longer activate at low x, the conclusion is tied to the assumed charge-balance mode.","tokens_in":22058,"feed_emoji":"🔋","tokens_out":4545,"duration_ms":41902,"temperature":0.7,"pith_summary":"This paper establishes that the faster lithium diffusion in defect-rich MoNb12O33 comes from a change in which lithium sites fill first, not from faster hopping in any single channel. In the ordered block structure, lithium first occupies slow pocket sites and only reaches fast window sites late in lithiation. In computational models containing either cation disorder or enlarged Wadsley-defect blocks, the fast window sites become occupied at much lower lithium content, matching the earlier onset of rapid diffusion seen experimentally. The result matters because it turns defect engineering into a transferable design strategy: defects that reorder site occupancy can speed up intercalation materials without changing their intrinsic per-site mobility.","feed_headline":"Defects triple lithium diffusivity in MoNb12O33, modeling shows why","feed_subtitle":"Both Wadsley defects and cation disorder shift lithium into high-speed window sites at low charge.","key_machinery":"The analysis rests on classifying lithium sites in Wadsley–Roth blocks into slow five-coordinate pocket sites and fast four-coordinate window sites (horizontal Wh and vertical Wv), then computing two things with a machine-learned interatomic potential trained to density functional theory: (1) per-site hopping barriers and diffusion coefficients, and (2) the ground-state sequence of lithium site occupancy at increasing lithiation via cluster expansion and Monte Carlo. Comparing pristine T[3×4], cation-disordered T[3×4], enlarged T[3×5], and mixed T[3×4–3×5] models isolates the two defect effects.","core_discovery":"In the ordered T[3×4] block structure, lithium first fills the slow five-coordinate pocket sites and only populates the fast four-coordinate window sites (horizontal Wh and vertical Wv) late in lithiation. In models with either cation disorder (Mo displaced from tetrahedral sites into the block) or enlarged Wadsley-defect blocks (with charge balance restored by oxygen vacancies), the fast Wh and Wv sites become occupied at substantially lower lithium concentrations. Since per-site hopping barriers change only modestly, the main effect of both defect types is to reorder the sequence of site occupancy rather than to speed up any individual channel.","pith_inferences":["A testable extension is to partially prelithiate the ordered MNO-900 to x≈3–5 before high-rate cycling: if the site-occupancy sequence is the controlling factor","this could emulate the defect-rich early activation and recover some of MNO-800's rate performance.","The computational separation of defect types rests on charge balancing enlarged blocks with oxygen vacancies","but the paper's own EXAFS shows Mo in MNO-800 is not uniformly tetrahedral","if the real defect-rich structure instead places Mo in octahedral sites","the predicted early-activation effect would need to be re-evaluated for that charge-balance mode.","The same site-occupancy-reordering mechanism may operate in other intercalation hosts with multiple lithium site types of differing mobility","where defect engineering could be screened by computing only relative site energies rather than full diffusivities."],"forward_implications":["MNO-800's higher capacity at 0.1C (307 vs ~293 mAh/g) and its 200 mAh/g at 10C follow from faster solid-state diffusion, which ICI overpotential analysis identifies as the dominant rate-limiting step.","The ~3x capacity-weighted diffusivity of MNO-800 is attributed to early activation of fast window sites; the measured D(x) curves converge at high lithiation, matching model predictions at ~75% Li.","Wadsley defects and cation disorder are predicted to be independently beneficial, so a material containing both should retain the advantage of each.","The mechanism provides a design principle: defects that lower the energy of fast window sites relative to slow pocket sites can accelerate rate capability across Wadsley–Roth anode materials."],"fun_headline_variants":["Disorder and defects triple MoNb12O33 Li diffusion by reordering sites","Both defect types flip fast-site filling early in MoNb12O33","MoNb12O33 defects triple Li diffusion by unlocking fast sites sooner","Defect-driven site reordering triples Li speed in MoNb12O33","Cation disorder and Wadsley defects both reroute Li to fast paths"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The computational models of Wadsley defects keep Mo only in tetrahedral sites and restore charge balance with oxygen vacancies, but the experiments show Mo in the defect-rich sample is not uniformly tetrahedral; if the real defect-rich structure balances charge differently, the predicted early activation of fast lithium sites may not apply.","fun_headline_variants_meta":{"raw":{"variants":["Disorder and defects triple MoNb12O33 Li diffusion by reordering sites","Both defect types flip fast-site filling early in MoNb12O33","MoNb12O33 defects triple Li diffusion by unlocking fast sites sooner","Defect-driven site reordering triples Li speed in MoNb12O33","Cation disorder and Wadsley defects both reroute Li to fast paths"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00074,"raw_usage":{"total_tokens":3169,"prompt_tokens":799,"completion_tokens":2370,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":543,"completion_tokens_details":{"reasoning_tokens":2269}},"tokens_in":543,"tokens_out":2370,"duration_ms":15512,"temperature":1.0,"reasoning_tokens":2269,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T22:35:34.650340+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the occupancy of fast window sites at low lithiation (e.g., x≈3–5) in MNO-800 using neutron diffraction or 7Li NMR; if fast Wh/Wv sites are not occupied early, the proposed mechanism is wrong. Alternatively, compute the defect-rich model with Mo placed in octahedral sites and no oxygen vacancies; if the fast sites no longer activate at low x, the conclusion is tied to the assumed charge-balance mode.","supporting_citations":[],"review_version":1}