Novel fast Li-ion conductors for solid-state electrolytes from first-principles

Loris Ercole; Nicola Marzari; Tushar Singh Thakur

REVIEW 3 major objections 5 minor 2 references

A screen of over 30,000 known lithium crystals predicts that Li7NbO6 — a previously overlooked oxide — conducts lithium ions at about 5 mS/cm at room temperature, which would make it a competitive solid electrolyte.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-03 12:20 UTC pith:BVXDYXR4

load-bearing objection Read this for the workflow and the open dataset, not for the Li7NbO6 headline — the 5 mS/cm claim is a long extrapolation built on a polymorph the paper never tests. the 3 major comments →

arxiv 2601.03151 v1 pith:BVXDYXR4 submitted 2026-01-06 cond-mat.mtrl-sci

Novel fast Li-ion conductors for solid-state electrolytes from first-principles

Tushar Singh Thakur , Loris Ercole , Nicola Marzari This is my paper

classification cond-mat.mtrl-sci

keywords lithium-ion conductorsolid-state electrolytehigh-throughput screeningfirst-principles molecular dynamicspinball modelLi7NbO6Arrhenius extrapolationionic conductivity

verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper claims that a high-throughput computational screen of more than 30,000 experimentally known lithium-containing structures can uncover fast lithium-ion conductors that earlier searches missed. The headline result is Li7NbO6, an oxide the simulations predict to conduct lithium ions at roughly 5 mS/cm at room temperature. That would put it on par with some of the better oxide solid electrolytes and make it an attractive, synthesizable candidate for safer batteries. The claim rests on a computational pipeline that combines a fast approximation to density functional theory (the pinball model) with full first-principles molecular dynamics at several temperatures, and on extrapolating those high-temperature diffusion data down to room temperature.

Core claim

Starting from over 30,000 experimental Li-containing structures, the authors apply structural filters, electronic band-gap screening, and molecular dynamics accelerated by the pinball model to narrow the field to about 1,500 insulators, then to 132 fast diffusers. Of these, 77 are already known conductors, and the remaining candidates are studied with full first-principles molecular dynamics at 1000 K, 750 K, 600 K, and 500 K. Nine materials are identified as fast Li-ion conductors with activation barriers below roughly 0.3 eV. The most prominent is Li7NbO6, whose predicted room-temperature conductivity of about 5 mS/cm comes from fitting an Arrhenius line to diffusion coefficients at those

What carries the argument

The pinball model is the computational engine of the screen: lithium ions are treated as fully ionized point charges ('pinballs') moving in a frozen host-lattice potential, with parameters fitted to DFT forces and refined self-consistently through iterative molecular dynamics runs. It is 200–500 times faster than first-principles MD, which makes scanning ~1,000 insulators feasible. The connection between simulation and measurable conductivity is the Einstein relation (diffusion coefficient from the slope of the mean-square displacement) and the Nernst-Einstein equation, with the Haven ratio set to 1.

Load-bearing premise

The predicted room-temperature conductivities, including the headline 5 mS/cm for Li7NbO6, assume that the diffusion coefficients measured in simulations at 500–1000 K follow a single straight Arrhenius line all the way down to 300 K.

What would settle it

Measure the ionic conductivity of the exact Li7NbO6 polymorph studied here (MPDS identifier S1818764) at room temperature using impedance spectroscopy; if the measured value falls well below 1 mS/cm, the Arrhenius extrapolation on which the headline result is based fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

Li7NbO6, a previously understudied oxide, is predicted to be a fast room-temperature solid electrolyte, and should be a priority for experimental synthesis and impedance measurement on the exact polymorph screened here.
Li5NaN2, Li4Mo3O8, and the cesium-doped lithium halides (such as Li3CsBr4 and LiCsI2) are predicted to be fast conductors with low activation barriers, giving experimenters a short list of new candidates.
The pinball workflow rediscovers 77 known conductors, suggesting that a similar automated screening can be trusted to flag promising materials in other chemistries without prior chemical intuition.
The 25 materials that diffuse only at 1000 K are not ruled out; the authors note that longer simulations or machine-learned potentials could resolve their lower-temperature behavior.
The full provenance-tracking dataset and screening protocol are released, providing a reusable infrastructure and training data for next-generation machine-learned interatomic potentials for lithium-ion diffusion.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The headline 5 mS/cm is an extrapolation from 500–1000 K simulations; a direct room-temperature measurement on the exact MPDS S1818764 polymorph of Li7NbO6 would either confirm the Arrhenius assumption or expose a slope change that lowers the conductivity to the measured 0.008 mS/cm of the other polymorph.
The paper's own comparison shows the pinball model systematically overestimates diffusion at 1000 K relative to FPMD; if this bias persists at lower temperatures, the absolute conductivities of all nine candidates could be overestimated, even if their relative ranking holds.
Because the screen only starts from experimentally known structures, its new candidates are synthesizable by construction; this avoids the common failure mode of inverse-designed materials that cannot be made, and makes the predicted conductors immediate testing targets.
The authors' public release of first-principles diffusion data could accelerate the training of a 'universal lithium' interatomic potential, which they explicitly identify as a future direction; if that potential becomes accurate enough, it could extend simulations to the nanosecond timescales needed to resolve diffusion at room temperature directly.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Desk Editor's Note

Read this for the workflow and the open dataset, not for the Li7NbO6 headline — the 5 mS/cm claim is a long extrapolation built on a polymorph the paper never tests.

read the letter

Punchline: this is a serious high-throughput screening with a genuinely useful dataset, but the headline Li7NbO6 claim is on thin ice. The room-temperature conductivity is an Arrhenius extrapolation from short FPMD runs, and the paper never tests the polymorph that the only experiment actually made.

What is actually new and good: they screen 30,000+ experimental Li structures, apply sensible filters, run the pinball model with self-consistently fitted coefficients and non-local terms, and land on 982 insulators. Rediscovering 77 known conductors is a solid sanity check. The 55 unexplored candidates get FPMD at multiple temperatures, and the SI has MSD plots for every one of them. The AiiDA provenance and the Materials Cloud deposit are valuable for people training ML potentials, and Fig. 15 (pinball vs FPMD) is an honest calibration: the pinball model overestimates diffusion but rarely misses conductors.

Soft spots, in proportion to how soft they actually are. The abstract's "Li7NbO6 shows ~5 mS/cm at room temperature" is not supported by the evidence presented. That number comes from an Arrhenius fit through three FPMD points (500, 750, 1000 K), where the 500 K point is a 180 ps run. The paper itself admits a change of slope is possible and that 100–200 ps is too short to resolve diffusion for 25 other materials. Table 3 shows no error bars on the extrapolated room-temperature conductivities. More concerning is the polymorph issue: the only experimental report on Li7NbO6 (Feng et al. 2024) measured 0.008 mS/cm at RT on a different space group, and the paper waves this away as "structural differences" without running any calculation on that phase, without an energy-above-hull check, and without showing the MPDS S1818764 polymorph is the synthesizable one. The broader-context claim that all screened materials are "actually synthesisable" is overbroad for exactly this reason.

Novelty of the candidates is also modest. Li7NbO6 was already proposed by He et al., Li5NaN2 and Li4CO4 come from earlier theoretical work, and the Cs-doped halides trace back to Pentin et al. The real contribution is the self-consistent pinball workflow, the systematic FPMD validation, and the 25+18 negative results — that is where the value lies.

For who: this paper is for computational materials scientists doing screening work and for anyone building MLIPs for Li diffusion. It deserves a serious referee, but the authors should be asked to add uncertainty estimates on all tabulated conductivities, run FPMD on the experimentally measured polymorph of Li7NbO6 (or at least assess its stability), and soften the claim to a projection. I would cite it for the workflow and the negative data, not for the headline number.

Referee Report

3 major / 5 minor

Summary. The paper reports a high-throughput computational screening of more than 30,000 experimental Li-containing crystal structures to identify fast Li-ion conductors for solid-state electrolytes. The pipeline applies structural filters, DFT-PBEsol band-gap screening, and self-consistent pinball-model molecular dynamics, followed by full first-principles molecular dynamics (FPMD) at 1000, 750, 600, and 500 K for the most promising candidates. The authors present nine fast conductors, with Li7NbO6 highlighted as showing a remarkable room-temperature ionic conductivity of approximately 5 mS/cm. They also report that 77 previously known conductors are rediscovered, which they use as a validation of the screening workflow. All data and provenance are made publicly available via AiiDA and Materials Cloud.

Significance. If the room-temperature conductivity prediction for Li7NbO6 is correct, the paper identifies a promising oxide solid electrolyte and demonstrates the value of a high-throughput screening workflow that combines a cheap approximate model (pinball) with FPMD validation. The self-consistent pinball coefficient fitting, the AiiDA provenance tracking, and the public release of first-principles data are concrete strengths that will be useful to the community. However, the headline claims rest on two load-bearing assumptions: Arrhenius linearity over a 500–1000 K to 300 K extrapolation, and the relevance of the specific polymorph used for Li7NbO6 when the only room-temperature experiment reports a much lower conductivity on a different space group.

major comments (3)

[Sec. 3.2.4, Table 3, Fig. 13] The headline claim of ~5 mS/cm at room temperature for Li7NbO6 is obtained by extrapolating an Arrhenius fit to FPMD diffusion coefficients at 500, 750, and 1000 K. Table 3 lists activation energies but gives no error bars for the conductivities or for the extrapolated room-temperature value, despite Sec. 2.3.2 stating that statistical variance and Bayesian error propagation were used. The paper itself warns in Sec. 4 that 'a change of slope is possible' and explains in Sec. 3.2.3 that 100–200 ps runs cannot resolve low-temperature diffusion; the 500 K point for Li7NbO6 comes from a 180 ps run. This is not sufficient support for a quantitative RT conductivity claim. Please provide uncertainty estimates, test Arrhenius linearity with additional lower-temperature data (e.g., longer MLIP-based MD), or explicitly reframe the result as a qualitative ranking rather than a quantitative RT condu
[Sec. 3.2.4, Li7NbO6 paragraph] The only room-temperature experiment for this stoichiometry, Feng et al. (ref. 293), reports 0.008 mS/cm on a different space group. The paper attributes the discrepancy to 'structural differences' but does not test this explanation: no FPMD is run on the experimentally reported phase, and no energy-above-hull or phonon analysis is provided to show that the MPDS S1818764 polymorph used here is the phase that would form under realistic synthesis. The broader-context claim that the materials are experimentally known and 'actually synthesisable' applies to the stoichiometry, not necessarily to the specific polymorph carrying the headline conductivity. This is load-bearing: if the synthesizable product is the Feng phase, the prediction does not apply to the measured material. Please provide a direct comparison of the two phases (FPMD on the experimental phase and/or relative stability analys
[Sec. 3.1, Fig. 15] The validation of the pinball model against FPMD shows all points on or above the identity line, with differences spanning orders of magnitude, indicating systematic overestimation of diffusion by the pinball model. The statement that the pinball model has 'accuracy similar to DFT' (Abstract and Sec. 1) is based on force-component correlations (r² > 0.95), but the diffusion coefficients themselves are not quantitatively accurate. The paper correctly notes that the true number of false negatives is unknown because non-conducting candidates were not followed up (ref. 298), yet the stated upper bound of ~85% predictive rate is used to support the screening methodology. This does not invalidate the final FPMD results, but it means the screening's quantitative ranking and the reported 'predictive rate' are not firmly established. Please either provide a more honest statement of the pinball mo

minor comments (5)

[Supplementary, Figs. S1/S2 and S16/S17] Duplicate captions appear for Figs. S1/S2 (both labeled Li4CO4) and S16/S17 (both labeled Li8SeN2). These should be corrected, and the corresponding figure contents verified.
[Sec. 3.2.1] The formula 'LiH f2(PO4)3' appears to be a typo for LiHf2(PO4)3. Please check the stoichiometry and rendering.
[Fig. S35, Table 2] The caption for Fig. S35 reads 'LiCS(OF)3' while Table 2 lists 'LiCF3SO3' for the same structure. Please make the notation consistent.
[Table 2] The footnote for Li2P2PdO7 states 'at 600 K, since we only performed FPMD simulations at one temperature', but the column header indicates 1000 K. Clarify that this entry is not at 1000 K.
[Fig. 15] The caption refers to a 'bold-grey line' representing the threshold below which MSD convergence cannot be achieved, but the line is not described in the text. State the threshold value or how it is defined.

Circularity Check

0 steps flagged

No significant circularity: the headline Li7NbO6 conductivity is produced by independent FPMD, not by the fitted pinball coefficients or by the authors' prior results.

full rationale

The central claim—Li7NbO6 conductivity of ~5 mS/cm at room temperature—is derived from full first-principles molecular dynamics, not from the pinball model or from literature values. Diffusion coefficients are obtained from MSD slopes via Eq. (1), converted to conductivities via the Nernst-Einstein relation Eq. (2), and extrapolated to 300 K by a linear Arrhenius fit (Sec. 2.3.2, Table 3). The pinball model and its self-consistently fitted α/β coefficients are used only for pre-screening; the 9 final candidates are re-run with FPMD, so the headline conductivity does not reduce to any fitted parameter. The main self-citation (ref. 120 for the pinball model's accuracy) is a method benchmark external to this paper and is not load-bearing for the final FPMD conductivity values. The paper also discloses the extrapolation risk ('a change of slope is possible', Sec. 4) and notes that 100–200 ps runs may be too short to resolve low-temperature diffusion for other materials; these are scientific caveats, not circular steps. The discrepancy with the experimental Li7NbO6 value of 0.008 mS/cm is attributed to different space groups; this is an untested external-falsifiability concern, not a self-referential reduction of the predicted quantity to the input.

Axiom & Free-Parameter Ledger

3 free parameters · 7 axioms · 0 invented entities

The central claims rest on standard statistical-mechanical relations (Einstein, Nernst-Einstein, Arrhenius) plus several domain assumptions about frozen hosts, Haven ratio = 1, and very short simulation times. No new physical entities are postulated. The pinball coefficients are fitted per material to DFT forces, and the Arrhenius parameters are fitted to FPMD data; neither is a universal first-principles constant.

free parameters (3)

Pinball coefficients α1, α2, β1, β2 = not listed in paper; converged to r² > 0.95 against DFT forces
Eq. (3) introduces four phenomenological coefficients per material, fitted by linear regression to DFT forces on rattled and MD-generated configurations (§2.3.1). The entire pinball diffusion ranking depends on these fits.
Activation energy Ea (and pre-exponential D0) from Arrhenius fit = 0.15–0.28 eV for the 9 fast conductors (Table 3)
Ea and D0 are fitted from FPMD D(T) at 500–1000 K via linear regression (§2.3.2). The room-temperature conductivities are then extrapolated from these two fitted parameters.
Conductivity threshold 1 mS/cm at 1000 K = 1 mS/cm
Chosen by hand as the pinball-stage cutoff to select candidates for FPMD (§3.1). This threshold determines which structures are ever validated and is not derived from a target performance.

axioms (7)

domain assumption Li atoms are completely ionized and the host lattice (non-Li atoms and frozen charge density) remains fixed at equilibrium positions
The pinball Hamiltonian, Eq. (3), is built on these two assumptions (§2.3.1). If host dynamics or charge reorganization matter for diffusion, the screening ranking can be biased.
domain assumption Haven ratio H = 1 and full ionization Z = 1 in the Nernst-Einstein equation
Eq. (2) converts tracer diffusion to ionic conductivity assuming dilute non-correlated motion. The authors note H is often < 1 and say this means they do not overestimate conductivities, but it is still an unverified simplification.
domain assumption DFT-PBEsol band gap > 1 eV identifies an electronic insulator suitable for solid electrolytes
Used in the electronic filter (§2.2). PBEsol underestimates gaps, and the 1 eV cutoff is a rule of thumb; materials with gaps just above 1 eV may still have electronic leakage issues.
domain assumption Experimental geometries from databases are adequate for diffusion MD without full relaxation
FPMD is run on experimental geometries; only 25% of structures were relaxed for band-gap checks (§2.2). Errors in experimental coordinates or pressure-dependent phases directly affect simulated diffusion.
domain assumption Li-ion diffusion follows a linear Arrhenius law between 500 K and 1000 K, and this law can be extrapolated to 300 K
Activation barriers are extracted from four temperatures and then extrapolated to room temperature (§2.3.2, Table 3). The authors explicitly warn in Sec. 4 that 'a change of slope is possible' for the Arrhenius plot.
domain assumption 100–180 ps FPMD runs are long enough to estimate diffusivity at 500 K
The paper itself shows in Sec. 3.2.3 that 100–200 ps is insufficient to resolve diffusion for 25 materials at lower temperatures, so the same timescale limitation applies to the 9 'fast' conductors.
domain assumption Database structures represent experimentally known, synthesisable phases
The screening deliberately uses only experimental structures (§2.1). But polymorphism matters: Li7NbO6's predicted 5 mS/cm corresponds to one space group, while Feng et al. measured 0.008 mS/cm on another phase of the same composition (Sec. 3.2.4).

pith-pipeline@v1.3.0-alltime-deepseek · 31675 in / 10392 out tokens · 96701 ms · 2026-08-03T12:20:42.066022+00:00 · methodology

0 comments

read the original abstract

We present a high-throughput computational screening for fast lithium-ion conductors to identify promising materials for application in all solid-state electrolytes. Starting from more than 30,000 Li-containing experimental structures sourced from Crystallography Open Database, Inorganic Crystal Structure Database and Materials Platform for Data Science, we perform highly automated calculations to identify electronic insulators. On these ~1000 structures, we use molecular dynamics simulations to estimate Li-ion diffusivities using the pinball model, which describes the potential energy landscape of diffusing lithium with accuracy similar to density functional theory while being 200-500 times faster. Then we study the ~60 most promising and previously unknown fast conductors with full first-principles molecular dynamics simulations at several temperatures to estimate their activation barriers. The results are discussed in detail for the 9 fastest conductors, including $Li_7NbO_6$ which shows a remarkable ionic conductivity of ~5 mS/cm at room temperature. We further present the entire screening protocol, including the workflows where the accuracy of the pinball model is improved self-consistently, necessary to automatically running the required calculations and analysing their results.

discussion (0)

Reference graph

Works this paper leans on

2 extracted references · 1 linked inside Pith

[13]

1–43 | 11 Table 3 The most promising structures that were found to be conducting with FPMD at lower temperatures

These discrepancies arise from structural differences, as the two structures possess different space groups and lattice param- **at 600 K, since we only performed FPMD simulations at one temperature due to high computational costs for this structure. 1–43 | 11 Table 3 The most promising structures that were found to be conducting with FPMD at lower temper...

1980
[1999]

universal-Li

Our FPMD simulations indicate that both materials possess low activation barriers, with the yttrium-doped version perform- ing slightly better. Based on the activation energies, we estimate the ionic conductivities at room temperature to be 0.2 mS/cm. We strongly recommend further experimental studies to validate these findings and confirm the potential a...

Pith/arXiv arXiv 1935

[1] [13]

1–43 | 11 Table 3 The most promising structures that were found to be conducting with FPMD at lower temperatures

These discrepancies arise from structural differences, as the two structures possess different space groups and lattice param- **at 600 K, since we only performed FPMD simulations at one temperature due to high computational costs for this structure. 1–43 | 11 Table 3 The most promising structures that were found to be conducting with FPMD at lower temper...

1980

[2] [1999]

universal-Li

Our FPMD simulations indicate that both materials possess low activation barriers, with the yttrium-doped version perform- ing slightly better. Based on the activation energies, we estimate the ionic conductivities at room temperature to be 0.2 mS/cm. We strongly recommend further experimental studies to validate these findings and confirm the potential a...

Pith/arXiv arXiv 1935