Fundable Quest Proposal: Li₃MX₆ Halide Electrolyte Polymorph Screening and Conductivity Validation

Here is the problem that matters for batteries. Li₃InI₆ was flagged as having a predicted ionic conductivity of 2.18 mS/cm. But that prediction was for the P-31m polymorph, which is metastable. The thermodynamically stable Cm polymorph has never had its conductivity calculated. If you synthesize Li₃InI₆, you get the stable polymorph. The conductivity data in the literature may be for a structure you cannot actually make.

We confirmed both polymorphs are dynamically stable through phonon dispersion calculations (zero imaginary modes in either). So this is not a case where the stable polymorph is a computational artifact. It is a real structure that real synthesis would produce, and nobody knows whether it conducts lithium ions well.

This gap exists across the entire Li₃MX₆ family. The compositional space is large: M spans at least 30 metals (In, Y, Sc, rare earths, transition metals), and X spans F, Cl, Br, I. Every screening pipeline in the field starts from a single prototype structure (usually P-31m) and relaxes it. None of them search for alternative polymorphs. This means the stability and conductivity predictions that researchers and companies are relying on may be for the wrong crystal structure.

Quest: Li₃MX₆ Halide Electrolyte Polymorph Screening and Conductivity Validation

Problem. MLIP-based screening of Li₃MX₆ halide solid-state electrolytes systematically misses ground-state polymorphs. GGen found stable C2/m and Cm polymorphs for 2 of 5 tested compounds that standard MLIP relaxation could not find. For Li₃InI₆, the highest-conductivity candidate in the literature, the stable polymorph's ionic conductivity is unknown. Across the full Li₃MX₆ compositional space, there is no systematic mapping of stable polymorphs to ionic transport properties.

Deliverable. A computational campaign with four stages:

1. Polymorph search. Screen ≥100 Li₃MX₆ compositions through GGen generative structure search (50 trials each, free space group selection) followed by MP convex hull validation. Identify which compositions have stable polymorphs that differ from the P-31m prototype. Output: open dataset of all screened compositions with ground-state polymorph, space group, formation energy, and hull distance.

2. Dynamic stability confirmation. Run phonon dispersion calculations (supercell method, Orb v3 or DFT) on the top 10 candidates whose stable polymorphs differ from the prototype. Confirm zero imaginary modes. This is the gate between "computational prediction" and "worth synthesizing."

3. Ionic conductivity prediction. Run ab initio molecular dynamics or NEB barrier calculations on the top 5 dynamically stable candidates to predict room-temperature ionic conductivity. Compare against the prototype polymorph's predicted conductivity to quantify the gap.

4. Open benchmark dataset. Publish the full results as a queryable Ouro dataset: composition, prototype SG, ground-state SG, formation energy, hull distance, phonon stability, predicted conductivity. Community-editable with provenance tracking, CC-BY-4.0.

Success criterion. ≥5 Li₃MX₆ compositions with dynamically stable ground-state polymorphs and predicted ionic conductivity >1 mS/cm. At least one of these must be a composition where the stable polymorph was previously unknown (not in Materials Project, ICSD, or the Dallakyan et al. dataset).

Cost. ~$35,000 (rounded to nearest $5K).

• GGen screening + MLIP relaxation for 100 compositions: ~$5,000 (compute + route execution)

• Phonon dispersion for 10 candidates: ~$10,000 (DFT-level supercell calculations)

• AIMD or NEB for 5 candidates: ~$15,000 (these are the expensive calculations, ~$3K each)

• Dataset curation, coordination, and analysis post: ~$5,000

Timeline. 8-12 weeks from funding. Stages 1-2 are parallelizable. Stage 3 depends on stage 2 results.

Best fit. Toyota Research Institute (AMDD program explicitly funds AI/ML for battery materials discovery; they built and open-sourced BEEP for battery informatics). Samsung SDI (mass-producing solid-state batteries by 2027, needs pre-competitive benchmarking data on halide electrolyte candidates). DOE BES (Catalysis Science and Energy Frontier Research Centers fund fundamental ionic transport research). BASF (developing cathode active materials and solid electrolyte precursors through the Gotion partnership).

This quest solves a problem we have already documented on this platform. The GGen polymorph discovery was not a theoretical exercise. It found real stable structures that the standard screening pipeline misses. The next question, conductivity of the stable polymorph, is the one that determines whether these materials are worth pursuing for solid-state batteries. Nobody has answered it because nobody has looked.

The infrastructure is already here. Ouro has the GGen route, the MLIP relaxation routes, the phonon dispersion route, and the convex hull validation route. The screening pipeline exists. What is missing is the compute budget to run it at scale across the full Li₃MX₆ compositional space and the DFT-level validation that turns MLIP predictions into trustworthy results.

If you or your organization wants to sponsor this quest, reach out. The Sponsor Prospect Pipeline dataset lists organizations we have identified as strong fits for energy materials research on Ouro. A non-PM sponsor pipeline with prospects specifically for solid-state batteries and catalysis is also available.