{"id":"3f3ed2e0-cab3-41f3-9ba8-cc660dc3689c","arxiv_id":"1908.03285","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Solvent HOMO renormalization in electrolytes correlates with Gutmann Donor and Acceptor numbers, allowing a fitted two-parameter model to replace expensive solvation calculations.","lead":"This paper shows that a solvent's oxidation stability loss in a battery electrolyte can be estimated from two simple chemical quantities instead of expensive calculations. A generalist might read it because cheap screening could speed up design of higher-voltage lithium-ion and lithium-oxygen batteries.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Transfer of the descriptor model to real electrolytes rests on an untested additivity assumption: single-pair renormalizations are summed over solvation-shell species, with no validation for multi-species, cooperative, or solvent-solvent effects.","rationale":"Reader identified essentially the same weakest assumption; I agree. The central claim as stated is a two-part claim: (1) descriptors correlate with single-pair renormalization, and (2) these can be summed to predict real electrolyte stability. The evidence supports (1) in-sample, but (2) is the practically important part and is asserted without test. The paper's own text explicitly invokes the shell-composition invariance and the summation procedure, so this is not an invented concern. The proposed cluster DFT test would settle whether the additive model transfers. If it passes, the conditional can be lifted; if it fails, the descriptor model becomes only a qualitative trend. I do not see other concerns that are more load-bearing: the fit/evaluation on the same dataset is a limitation but standard for descriptor papers; co-oxidation exclusion is acknowledged and bounded; the IP benchmark (MAE 0.12 eV over 34 solvents) gives independent support to the method. Therefore the verdict should remain CONDITIONAL, unchanged.","tokens_in":9174,"tokens_out":5409,"duration_ms":56260,"concrete_test":"Perform PBE/GPAW DFT on explicit multi-species clusters for a representative electrolyte, e.g., EC with LiPF6: (i) one Li+, one PF6-, and three EC molecules arranged per MD-derived coordination; (ii) one Li+, one PF6-, and one EC. Compute the cluster IP and the renormalization relative to isolated EC. Compare against the additive prediction using Eq. (3): C + alpha*sqrt(AN_EC*DN_PF6) + N_EC * alpha*sqrt(AN_EC*DN_EC), with N_EC the number of additional EC molecules. If the cluster DFT renormalization differs from the additive sum by more than ~0.2 eV (the model MAE), the additivity assumption in the paragraph following Eq. (3) is falsified and the electrolyte-transfer claim needs revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Equation (3) is trained on 32 single-solvent/single-Li-anion complexes (Table 2), yet the paper's practical claim—estimating renormalized oxidative stability in a real electrolyte—requires the step in the paragraph after Eq. (3): 'the net renormalization due to the entire solvation shell ... by summing the individual specie contribution.' This additivity assumption is load-bearing because a real solvation shell contains multiple anions, multiple solvent molecules, ion pairs, and cooperative polarization; the fit says nothing about these. The authors assert that 'the number of species in the solvation shell of the salt anions is not significantly different' to justify using the fit, but no coordination-number analysis, MD, or cluster DFT is provided for the studied systems. Moreover, the same model is applied outside its training domain to solvent-solvent pairs (EC/EC, giving 1.14 V), where the donor species is a neutral solvent rather than an anion; the alpha calibrated on anion-solvent interactions need not transfer. If the additive single-pair picture fails, the headline R^2 = 0.86 does not rescue the screening prediction for realistic electrolytes. This is a correctness risk, not a circularity: the internal fit is plausible, but the generalization step is unsupported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a descriptor-based model for the solvation-induced renormalization of solvent HOMO levels in lithium-ion battery electrolytes. The authors first validate PBE-level ionization potentials (IPs) of isolated solvent molecules against 34 experimental values, reporting an MAE of 0.12 eV. They then compute IPs for solvent–salt complexes and propose two models for the renormalization: a linear model in the Gutmann acceptor number (AN) of the solvent and donor number (DN) of the anion, and a geometric-mean model proportional to sqrt(AN*DN). Both models are reported to achieve R^2=0.86 and MAE=0.16 eV on 32 complexes. The paper applies these models to estimate renormalization in EC/LiPF6 electrolytes and for superoxide in Li-O2 batteries, and suggests the descriptors as a basis for high-throughput screening of stable solvents.","tokens_in":9495,"tokens_out":6780,"duration_ms":64160,"significance":"The isolated-molecule IP validation is a solid benchmark, and the descriptor idea is attractive because, if it generalizes, it would allow a cheap estimate of solvation-induced HOMO shifts without expensive quantum chemistry. The authors also correctly note the need to eventually compute DN and AN from first principles. However, the central screening claim currently rests on in-sample fit quality and an untested additivity assumption; the evidence provided does not yet establish predictive accuracy for new solvents or realistic multi-species electrolytes.","major_comments":[{"comment":"The coefficients of both models are trained on the same 32 DFT complexes that are used to report R^2=0.86 and MAE=0.16 eV. Because no cross-validation, held-out test set, or uncertainty quantification is provided, these metrics demonstrate in-sample agreement only, not predictive accuracy. The screening claim in the abstract requires the model to predict renormalization for solvents and salts not used in the fit; I request leave-one-out cross-validation (or an external test set) and error bars on the model coefficients and predictions.","section":"Section following Eq. (3), Fig. 4"},{"comment":"The manuscript states that the goodness of the fit 'proves that for all the binary salt and solvent combinations, the number of species in the solvation shell of the salt anions is not significantly different' and then proposes to sum individual renormalizations to obtain the net solvation-shell effect. The fit is performed on single anion–solvent (or Li-ion-pair–solvent) complexes and cannot by itself establish constancy of the solvation-shell composition in a real electrolyte. No coordination-number analysis, molecular dynamics, or multi-species cluster DFT calculations are presented. This additivity assumption is load-bearing for the EC/LiPF6 and screening estimates. In particular, the EC/EC estimate (1.14 V) applies the model to a neutral solvent donor, whereas the alpha coefficient was calibrated on anion donors; that transfer requires separate validation.","section":"Paragraph after Eq. (3) and paragraph beginning 'Using the models'"},{"comment":"The abstract states that the method uses 'fast GGA-level DFT calculations compared to previously used expensive, experimental data dependent methods,' but the descriptor model requires experimental Gutmann AN and DN for each solvent and salt of interest. The concluding paragraph acknowledges that first-principles computation of DN and AN is needed for unexplored molecules. As a result, the current method is not yet a purely computational screening tool for new species; the claims should be rephrased to make the role of experimental descriptor data explicit.","section":"Abstract and concluding paragraph"},{"comment":"The model is trained exclusively on complexes classified as solvent oxidation; 8 of the 41 complexes in Table 2 involve salt oxidation or co-oxidation and are excluded from the fit. The Bader charge thresholds (solvent charge greater than 0.6 e, anion charge less than -0.4 e) are applied without sensitivity analysis. In a screening application one must know whether the solvent or the salt is the limiting oxidized species, but the model only provides a renormalization when solvent oxidation is assumed. Please justify the thresholds, test their sensitivity, and clarify how the model should be used when the anion oxidizes first.","section":"Table 2 and Bader analysis paragraph"}],"minor_comments":[{"comment":"The caption refers to 33 solvents, while the text and SI Table 1 list 34 solvents; this inconsistency should be corrected.","section":"Figure 2 caption"},{"comment":"Table 1 lists six anions but omits SCN-, which is included in the complex calculations of Table 2. Please add the DFT and experimental IPs for SCN- and its Li+ ion pair.","section":"Table 1"},{"comment":"The text says that seven solvents were simulated, but Table 2 contains no water complexes and lists only six solvents. Please clarify whether water was included and, if so, why its complexes are absent from Table 2.","section":"Text and Table 2"},{"comment":"The sign convention for the renormalization is not stated explicitly. It should be defined as the difference between the isolated IP and the complex IP, with a comment on how a positive value corresponds to reduced oxidative stability.","section":"Equation (1) and renormalization definition"},{"comment":"The statement that the predicted EC renormalization values (0.65 V and 1.14 V) are 'very close' to quantum-chemistry values is qualitative. Please provide the actual values from refs 12 and 22 and include the model uncertainty in the comparison.","section":"EC/LiPF6 comparison paragraph"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses a timely and important problem, and the IP benchmark is solid. The central screening claim, however, is currently supported only by an in-sample fit and an untested additivity assumption. The revision should focus on adding a proper validation of the descriptor model (cross-validation or an external test set) and on either providing direct evidence for the additivity assumption or substantially tempering the screening claims. The manuscript is likely within scope for the journal if these issues are resolved."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"One thing to know: this paper delivers a genuinely cheap descriptor model for how much solvation lowers a solvent HOMO, and the isolated-molecule IP validation is honest and solid. I'd send it to a referee; I would not treat the screening numbers as settled.\n\nWhat is actually new: Eq. (3), renormalization ~ C + alpha*sqrt(AN*DN), fitted to GGA-DFT delta-SCF calculations of one-solvent/one-Li-anion complexes. That specific two-parameter correlation is not present in the cited references. The benchmark on 34 solvents (MAE 0.12 eV) is a real strength; it independently validates the IP method. The AN/DN space coverage is also well chosen.\n\nSoft spots, in order:\n- The central model is fit and evaluated on the same ~32 points. R^2 = 0.86 is in-sample. Co-oxidation cases are excluded, and the model says nothing about those cases.\n- The stress-test concern is the important one. The paper's end use is the net renormalization in a real electrolyte, obtained by summing single-pair contributions. The fit gives no evidence for multi-anion, ion-pair, solvent-solvent, or cooperative effects. The claim that the number of species in the solvation shell is not significantly different is asserted, not shown. The EC/EC estimate uses a neutral solvent as donor, outside the anion-donor calibration. This is a generalization gap, not circularity, because AN/DN and the DFT renormalization are independent. But it is load-bearing for the screening claim.\n- The 1.14 V EC/EC number relies on that untested additivity and domain transfer. It matches QC values in the literature, which is reassuring for carbonates, but it is not a validation of the general screening method.\n- No code or input structures are provided. For a descriptor paper, that hurts reproducibility.\n\nWhat I would tell a referee: accept conditional on (i) separating training from evaluation somehow, (ii) testing the additivity premise with a two-solvent or two-anion cluster, and (iii) releasing inputs. None of these is fatal; the descriptor is plausible and useful. The paper is for computational electrolyte screeners who need a cheap pre-filter, not for people needing quantitative oxidation potentials. The citation pattern looks fine; the self-citations to refs 41/42 are relevant and not padding.\n\nRecommendation: worth peer review, with the additivity validation requested before publication.","headline":"Useful cheap descriptor for solvent HOMO renormalization, with a solid IP benchmark, but the transfer to real electrolytes leans on an untested additivity step.","tokens_in":9999,"tokens_out":2160,"would_cite":true,"duration_ms":22860,"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":"This paper argues that the oxidative stability of battery solvents under solvation is controlled by the donor number of the anion and the acceptor number of the solvent, expressible as…","keywords":["oxidative stability","HOMO renormalization","donor number","acceptor number","electrolyte solvation","lithium-ion battery","high-voltage electrolyte","lithium-oxygen battery"],"falsifier":"Measure the oxidation onset potential of ethylene carbonate in a LiPF6 electrolyte and compare the observed HOMO downshift with the model's predicted total of about 1.79 eV (0.65 eV from $\\mathrm{PF}_6^-$ plus 1.14 eV from a neighboring EC molecule). If the measured shift deviates by more than about 0.2 eV, the additive single-species picture is wrong.","tokens_in":8980,"feed_emoji":"🔋","tokens_out":8186,"duration_ms":77982,"temperature":0.7,"pith_summary":"The paper asks what makes a solvent in a lithium-ion battery resist oxidation at high-voltage cathodes, and argues that what matters is the solvent's highest occupied molecular orbital (HOMO) level inside the electrolyte, not in vacuum. It shows that the solvent's HOMO is pushed upward (destabilized) by solvation, and that this shift is described by two empirical acidity/basicity parameters: the acceptor number of the solvent and the donor number of the anion or other solvating species. A geometric-mean formula, $\\mathrm{RN}=C+\\alpha\\sqrt{\\mathrm{AN}\\times\\mathrm{DN}}$, fits 32 DFT calculations with $R^2=0.86$ and a mean absolute error of 0.16 eV, after only cheap semilocal DFT is used for the bare ionization potential. If the model holds, screening candidate solvents for high-voltage electrolytes reduces to a look-up of donor/acceptor numbers plus one fast calculation, with no expensive quantum-chemistry treatment of the full solvation shell.","feed_headline":"Two simple numbers predict solvent oxidation in batteries","feed_subtitle":"A square-root formula using donor and acceptor numbers fits 32 DFT results within 0.16 eV.","key_machinery":"The central mechanism is the empirical acid–base interaction between a solvent and the species around it, quantified by the donor number (DN, a Lewis-basicity scale of how strongly a species donates electron density) and acceptor number (AN, a Lewis-acidity scale of how strongly it accepts it). The relative shift of the solvent HOMO is modeled as a function of the product $\\sqrt{\\mathrm{AN}\\times\\mathrm{DN}}$, on the electrostatic argument that the interaction energy scales with the product of the two charge-related parameters. The machinery consists of computing the bare solvent ionization potential by semilocal DFT, computing the ionization potential of a solvent–anion or solvent–ion-pair complex, and taking the difference; a partial-charge partitioning analysis decides whether the solvent or the anion is the oxidized species. The model's fitted coefficients then allow renormalization contributions to be summed over solvation-shell components, so no further DFT is needed for the full mixture.","core_discovery":"The central claim is that the oxidative stability of a solvent in a lithium-ion battery is set not by the isolated molecule's HOMO but by the HOMO renormalized by the solvation shell, and that this renormalization is controlled by two empirical acid–base parameters: the donor number of the anion and the acceptor number of the solvent. For a given solvent, anions with higher donor numbers lower the solvent's ionization potential more; for a given salt, solvents with higher acceptor numbers show larger reductions. Both a linear model and a geometric-mean model, $\\mathrm{RN}=C+\\alpha\\sqrt{\\mathrm{AN}\\times\\mathrm{DN}}$, fit the 32 DFT data points with $R^2=0.86$ and MAE=0.16 eV, and the authors use the fitted model to estimate that $\\mathrm{PF}_6^-$ shifts EC's HOMO down by 0.65 V and a neighboring EC molecule shifts it by 1.14 V, consistent with quantum-chemistry calculations. The same model applied to $\\mathrm{O}_2^-$ predicts a 1.25–1.75 V destabilization of solvents in Li-O2 batteries.","pith_inferences":["Beyond the paper: because donor and acceptor numbers are known for many organic molecules, the same descriptors could link oxidative stability to existing electrolyte property databases, producing a library-scale map of renormalized HOMO levels for untested solvents before any new DFT is run.","Beyond the paper: the additive single-species assumption implies a testable prediction that HOMO renormalization grows linearly with salt concentration (more anions in the shell) until ion pairing or aggregation changes the shell composition; a concentration-dependent oxidation-onset measurement would discriminate the additive model from cooperative models.","Beyond the paper: the same geometric-mean logic may transfer to cathode-surface-induced HOMO renormalization by replacing donor/acceptor numbers with a surface charge or surface-site Lewis acidity parameter, a direction the authors flag for future work."],"forward_implications":["Low-AN solvents (ethers, esters, carbonates) and low-DN salts (LiPF6, LiTFSI) should be preferred for high-voltage stability, while high-DN components that improve salt solubility tend to reduce oxidative stability.","For Li-O2 batteries, superoxide ($\\mathrm{O}_2^-$), with a donor number above 40 kcal/mol, is predicted to reduce solvent oxidative stability by 1.25–1.75 V, so solvents must be stable to at least about 4.5 V vs Li/Li+ or superoxide must be managed by redox mediators.","The model provides a route to high-throughput screening: compute the bare ionization potential with fast DFT, look up donor/acceptor numbers, and estimate the renormalized HOMO for arbitrary solvent, salt, and additive combinations without expensive coupled-cluster or GW calculations.","The Li/Li+ reference potential itself shifts by 0–0.75 V depending on the effective donor number of the electrolyte mixture, giving an additional degree of freedom for widening the stability window.","The fitted contour map of renormalization as a function of acceptor number and donor number allows quick estimates for common electrolyte components, such as the EC/LiPF6 case, without any new electronic-structure calculation."],"supporting_citations":[{"why":"Establishes the donor number concept used as the anion basicity descriptor in the model.","marker":"[30]"},{"why":"Supplies the acceptor number values for the solvents in the training set.","marker":"[31]"},{"why":"Provides the standard identification of the HOMO level with the negative of the ionization potential.","marker":"[33]"},{"why":"Supports the use of DFT total-energy differences for accurate ionization potentials without more expensive GW or MP2 calculations.","marker":"[36]"},{"why":"Defines the exchange-correlation functional used for all DFT calculations in the paper.","marker":"[37]"},{"why":"Provides the experimental donor numbers of the anions used in fitting the descriptor model.","marker":"[40]"},{"why":"Offers a way to estimate solvation-shell composition so renormalization contributions can be summed for real electrolyte mixtures.","marker":"[42]"},{"why":"Provides the quantum-chemistry reference values that the EC/LiPF6 renormalization estimates are compared with.","marker":"[22]"}],"fun_headline_variants":["Donor and acceptor numbers predict solvent oxidative stability","Two acid-base numbers forecast battery solvent stability","HOMO renormalized by solvation: a two-number predictor","Square-root rule links donor and acceptor numbers to oxidation","Simple descriptors renormalize solvent HOMO for battery design"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that a real solvation shell renormalizes the solvent HOMO as the sum of independent single-species contributions, and that the number of solvating species does not vary significantly across the systems studied.","fun_headline_variants_meta":{"raw":{"variants":["Donor and acceptor numbers predict solvent oxidative stability","Two acid-base numbers forecast battery solvent stability","HOMO renormalized by solvation: a two-number predictor","Square-root rule links donor and acceptor numbers to oxidation","Simple descriptors renormalize solvent HOMO for battery design"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001467,"raw_usage":{"total_tokens":5889,"prompt_tokens":926,"completion_tokens":4963,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":542,"completion_tokens_details":{"reasoning_tokens":4886}},"tokens_in":542,"tokens_out":4963,"duration_ms":39261,"temperature":1.0,"reasoning_tokens":4886,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:18:46.061621+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the oxidation onset potential of ethylene carbonate in a LiPF6 electrolyte and compare the observed HOMO downshift with the model's predicted total of about 1.79 eV (0.65 eV from $\\mathrm{PF}_6^-$ plus 1.14 eV from a neighboring EC molecule). If the measured shift deviates by more than about 0.2 eV, the additive single-species picture is wrong.","supporting_citations":[{"cited_title":"Solvent effects on the reactivities of organometallic compounds","cited_arxiv_id":null,"evidence_quote":"Establishes the donor number concept used as the anion basicity descriptor in the model."},{"cited_title":"The acceptor number -— A quantitative empirical parameter for the electrophilic properties of solvents","cited_arxiv_id":null,"evidence_quote":"Supplies the acceptor number values for the solvents in the training set."},{"cited_title":"\\ \"U about the assignment of wave functions and eigenvalues ​​to the single electrons of an atom","cited_arxiv_id":null,"evidence_quote":"Provides the standard identification of the HOMO level with the negative of the ionization potential."},{"cited_title":"W.; Thygesen, K","cited_arxiv_id":null,"evidence_quote":"Supports the use of DFT total-energy differences for accurate ionization potentials without more expensive GW or MP2 calculations."},{"cited_title":"F.; Taha, A","cited_arxiv_id":null,"evidence_quote":"Provides the experimental donor numbers of the anions used in fitting the descriptor model."},{"cited_title":"M.; Pande, V.; Khetan, A.; Viswanathan, V.; McCloskey, B","cited_arxiv_id":null,"evidence_quote":"Offers a way to estimate solvation-shell composition so renormalization contributions can be summed for real electrolyte mixtures."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the quantum-chemistry reference values that the EC/LiPF6 renormalization estimates are compared with."}],"review_version":1}