Dallakyan, Maltsev, Chepkasov, Aghamalyan, Hunanyan, Petrosyan, Chobanyan, Sahakyan, Khachatryan, Oganov, and Zakaryan recently published a computational screening study for novel Li₃MX₆ solid-state electrolytes (Journal of Energy Chemistry 112, 2026, 495-504). The paper uses USPEX structure prediction followed by DFT calculations to evaluate 56 candidate compounds across 8 structural prototypes, identifying Li₃InF₆ and Li₃InI₆ as new promising halide electrolytes alongside known references like Li₃YCl₆ and Li₃InCl₆.
I took five representative compounds from their work and ran them through Ouro's MLIP and thermodynamic validation routes to see how a fast machine-learning interatomic potential handles these structures, and whether the Materials Project convex hull agrees with their DFT-based stability assessments.
For each compound, I built an ASE structure from the prototype space group reported in the paper, then executed:
Orb v3 relaxation to check whether the MLIP preserves the crystal symmetry or collapses it (the P1-collapse pattern we have seen in Heuslers, Laves phases, and Cu₂Sb-type compounds)
Materials Project convex hull energy to get an independent thermodynamic stability assessment against the full MP database
Every single Li₃MX₆ structure survived Orb v3 relaxation with its P-31m space group intact. Zero P1 collapse. This is notable because the same MLIP has collapsed structures from multiple families we tested previously: C14 Laves phases, Cu₂Sb-type Mn compounds, and inverse Heusler Li₂YZ compounds all showed varying degrees of symmetry loss or structural collapse under relaxation.
The Li₃MX₆ halide structures are mechanically robust under MLIP relaxation. The energy changes are moderate, ranging from -0.095 eV (Li₃ScF₆) to -0.583 eV (Li₃InI₆), suggesting the starting geometries were already reasonable and the MLIP found modest improvements without breaking symmetry.
Compound | Space Group (in → out) | ΔE (eV) | Relaxation Steps |
|---|---|---|---|
Li₃InCl₆ | P-31m → P-31m | -0.316 | 12 |
Li₃ScF₆ | P-31m → P-31m | -0.095 |
The Materials Project hull tells a consistent story: all five compounds are metastable, sitting within 0.135 eV/atom of the convex hull. None are predicted as thermodynamically stable ground states, which is actually the norm for good solid-state electrolytes. Many high-performance SSEs are kinetically stabilized rather than thermodynamic ground states; the Li₃MX₆ family fits this pattern.
Compound | E above hull (eV/atom) | Formation energy (eV/atom) | MP entries at composition |
|---|---|---|---|
Li₃InCl₆ | 0.062 | -1.461 | 2 |
Li₃InF₆ | 0.067 |
Li₃InCl₆ and Li₃InF₆ sit closest to the hull at 0.062 and 0.067 eV/atom respectively, consistent with the paper's identification of Li₃InF₆ as a promising new candidate. Li₃InI₆, which the paper identified as having the highest predicted conductivity among new compounds (σ = 2.18 mS/cm), has the largest hull distance at 0.135 eV/atom but is still well within the metastable range common for functional electrolytes.
Two things stand out. First, the Li₃MX₆ structural family is mechanically robust under MLIP relaxation in a way that many other structure families are not. Orb v3 preserves P-31m symmetry perfectly here, whereas it collapsed P4/nmm in Cu₂Sb-type compounds, struggled with C14 Laves phases, and even showed issues in some Heusler structures. The halide octahedral framework appears to be a particularly stable arrangement that MLIPs handle well. This is encouraging for anyone wanting to use fast ML potentials for large-scale SSE screening, since you can trust the relaxed structures.
Second, the convex hull results align with the paper's DFT-based assessment. The paper found these compounds to be dynamically stable (no imaginary phonon modes) but did not claim they were global ground states. The MP hull data confirms they are metastable, with hull distances small enough that kinetic stabilization during synthesis is plausible.
All input CIFs, relaxed structures, and route execution details are linked below. The full pipeline ran 10 route executions (5 Orb v3 relaxation + 5 MP hull energy) across five Li₃MX₆ compositions.
Input CIFs:
Orb v3 relaxed CIFs:
15 |
Li₃YCl₆ | P-31m → P-31m | -0.361 | 18 |
Li₃InI₆ | P-31m → P-31m | -0.583 | 10 |
Li₃InF₆ | P-31m → P-31m | -0.363 | 17 |
1 |
Li₃YCl₆ | 0.080 | -1.947 | 1 |
Li₃ScF₆ | 0.083 | -3.285 | 2 |
Li₃InI₆ | 0.135 | -0.814 | 1 |
Orb v3 relaxation and MP convex hull validation of 5 Li₃MX₆ halide solid-state electrolytes from Dallakyan et al. (J. Energy Chemistry 2026). All structures preserved P-31m symmetry under MLIP relaxation; all metastable within 0.135 eV/atom of convex hull.
Let's take it a step further and see if we can find more stable polymorphs using ggen.
Ran all five through GGen with 50 trials each, letting it freely choose space groups rather than constraining to P-31m. Two out of five found more stable polymorphs:
Li₃YCl₆: P-31m → C2/m
GGen selected P-3 (SG 147) as the starting space group, relaxed through 401 optimization steps, and the structure settled into C2/m (SG 12). Convex hull comparison:
Polymorph | E above hull (eV/atom) | Formation energy (eV/atom) | Predicted stable? |
|---|---|---|---|
P-31m (original) | 0.080 | -1.947 | No |
C2/m (GGen) | 0.024 | -2.004 | Yes |
3.3x closer to the hull, and the formation energy dropped by 57 meV/atom.
Li₃InI₆: P-31m → Cm
GGen started from P6mm (SG 183), relaxed through 400 steps, and landed in Cm (SG 8). This is the compound Dallakyan et al. flagged as having the highest predicted conductivity (2.18 mS/cm). Convex hull:
Polymorph | E above hull (eV/atom) | Formation energy (eV/atom) | Predicted stable? |
|---|---|---|---|
P-31m (original) | 0.135 | -0.814 | No |
Cm (GGen) | 0.019 | -0.930 |
7.1x closer to the hull. The formation energy dropped by 116 meV/atom. Both polymorphs are now flagged as thermodynamically stable by the MP phase diagram, which the original P-31m structures were not.
The other three (Li₃ScF₆, Li₃InF₆, Li₃InCl₆) all collapsed to P1 under GGen relaxation, so no useful alternative polymorphs there.
GGen CIFs and hull route actions:
The C2/m and Cm space groups are both known structure types for Li₃MX₆ halide electrolytes in the experimental literature, so these aren't artifacts. Worth investigating whether the conductivity predictions from Dallakyan et al. would hold or improve for these lower-energy polymorphs.
Yes |
Assess the thermodynamic stability of a crystal structure by computing its energy above the convex hull of every structure in Ouro's crystal database: Materials Project, Alexandria and GNoME phases, structures generated on Ouro, and structures submitted earlier. Choose the energy surface: Orb v3, where every reference is scored with the model that scores your structure, or PBE, the published DFT energies with your structure scored by Orb v3. MP2020 corrections are applied on both. Upload an already-relaxed CIF so the result refers to that geometry; inputs with max |F| > 0.05 eV/Å are rejected unless is set (which relaxes with Orb v3 first). A structure the database does not have yet is added to it. Also rejects overlapping-atom CIFs unless is set. Returns eabovehull (eV/atom), decomposition products and formation energy; set for an interactive phase diagram too.
Assess the thermodynamic stability of a crystal structure by computing its energy above the convex hull of every structure in Ouro's crystal database: Materials Project, Alexandria and GNoME phases, structures generated on Ouro, and structures submitted earlier. Choose the energy surface: Orb v3, where every reference is scored with the model that scores your structure, or PBE, the published DFT energies with your structure scored by Orb v3. MP2020 corrections are applied on both. Upload an already-relaxed CIF so the result refers to that geometry; inputs with max |F| > 0.05 eV/Å are rejected unless is set (which relaxes with Orb v3 first). A structure the database does not have yet is added to it. Also rejects overlapping-atom CIFs unless is set. Returns eabovehull (eV/atom), decomposition products and formation energy; set for an interactive phase diagram too.