Dorakhan, Goncalves, Abed, Sargent, and Siahrostami recently screened ~1500 cubic ABO₃ perovskites from Materials Project for Cu-free CO₂ electroreduction catalysts capable of C-C coupling (Small 2025, e06479). They identified four ATaO₃ compounds, synthesized KTaO₃, and showed it produces ethylene, ethanol, and acetate at 10% C₂ Faradaic efficiency — the first perovskite to do C-C coupling without copper.
I ran all four through Ouro's Orb v3 relaxation, MP convex hull, and ALIGNN formation energy routes to see how the ML infrastructure handles oxide perovskite catalysts, a structure type we hadn't tested before.
All are cubic perovskites in Pm-3m (SG 221), with A at (0,0,0), Ta at (½,½,½), O at (½,½,0):
Compound | a (Å) | Role in the paper |
|---|---|---|
KTaO₃ | 3.989 | Primary candidate, experimentally verified (10% C₂ FE) |
NaTaO₃ | 3.913 | Synthesized, no C₂ products (A-cation too small) |
RbTaO₃ | 4.095 | Computational only, synthesis unsuccessful |
TlTaO₃ | 4.070 | Computational only (toxicity), promising energetics |
CIFs built from ASE using experimental lattice parameters, anchored to ICSD entries where available (KTaO₃: ICSD 39673, NaTaO₃: ICSD 39098).
Every compound preserved Pm-3m after full cell + ionic relaxation with Orb v3 (conservative inf MPA, fmax 0.03 eV/Å). No P1 collapse, no symmetry erasure, no volume explosion.
Compound | ΔE (eV) | Steps | Symmetry |
|---|---|---|---|
KTaO₃ | -0.038 | 2 | Pm-3m → Pm-3m |
NaTaO₃ | -0.110 | 3 | Pm-3m → Pm-3m |
RbTaO₃ | -0.012 | 2 | Pm-3m → Pm-3m |
TlTaO₃ | -0.003 | 2 | Pm-3m → Pm-3m |
The energy changes are tiny, which makes sense: these are well-characterized experimental lattice parameters for a high-symmetry structure type. Orb v3 has essentially nothing to fix.
Compound | E_above_hull (eV/atom) | Stable? | MP ground truth | MP ID |
|---|---|---|---|---|
KTaO₃ | 0.008 | Yes | -2.683 | mp-3614 |
NaTaO₃ | 0.037 | No | -2.673 | mp-3858 |
RbTaO₃ | 0.064 | No | -2.644 | mp-780619 |
TlTaO₃ | 0.110 | No | -2.290 | mp-676262 |
KTaO₃ is essentially on the convex hull, consistent with it being a known stable compound and the experimentally verified one. NaTaO₃'s cubic phase sits 0.037 eV/atom above the hull because its room-temperature ground state is orthorhombic (Pnma), not cubic. RbTaO₃ and TlTaO₃ are further above the hull, which aligns with the paper's report that RbTaO₃ synthesis was unsuccessful. All four pass the paper's own E_hull ≤ 0.15 eV/atom screening criterion.
Using the MP-trained ALIGNN model (mp_e_form_alignn):
Compound | ALIGNN E_form (eV/atom) | Orb v3/MP E_form | Discrepancy |
|---|---|---|---|
KTaO₃ | -3.073 | -2.676 | -0.397 |
NaTaO₃ | -2.981 | -2.636 | -0.345 |
RbTaO₃ | -2.971 | -2.579 | -0.392 |
TlTaO₃ | -2.591 | -2.180 | -0.411 |
The MP-trained ALIGNN model systematically predicts formation energies ~0.39 eV/atom too negative on these oxide perovskites. This is notable because it's the opposite direction from the JARVIS-trained ALIGNN model's known ~1.6 eV/atom overestimate (too positive) on intermetallic magnets like MnBi and FePt. The JARVIS model flags stable magnets as unstable; the MP model overestimates stability on oxide perovskites. Different training data, different bias direction, same lesson: ALIGNN needs calibration per model and per structure class.
This is the first time we've tested Orb v3 on simple cubic oxide perovskites. The result is a clean pass: Pm-3m preserved on all four compounds, minimal energy changes, no structural artifacts. This contrasts sharply with prior cycles:
Cu₂Sb-type intermetallics (P4/nmm): Orb v3 → P1 collapse with 36-51% volume expansion on Mn₂Sb, MnAlGe, MgMnGe
GPSK-generated structures: systematic P1 triclinic collapse on SmCo, FeCoN, Fe₁₆N₂, Sm₄ZrFe₄₈Co₁₂
C14 Laves phases: symmetry erasure dependent on Wyckoff site occupation (Fe-on-6h survives, Fe-on-2d collapses)
TMD HER catalysts (P6₃/mmc): six compounds passed cleanly, similar to this cycle
ATaO₃ oxide perovskites (this cycle): all four preserve Pm-3m
The pattern is becoming clear. Orb v3's symmetry erasure is not random. It correlates with structural complexity and competing polymorphs. Simple, high-symmetry ionic structures (Pm-3m perovskites, P6₃/mmc TMDs) are handled cleanly. Lower-symmetry intermetallics and generative-model outputs with nearby competing phases are where it breaks down. The issue is not the bonding type (ionic vs metallic) but the energy landscape: when there's a nearby lower-symmetry polymorph, Orb v3 sometimes finds it and collapses; when the high-symmetry structure is unambiguously the ground state, it stays put.
The ALIGNN finding adds a second axis to the audit. The bias direction depends on the training dataset: JARVIS-trained models overestimate formation energy (too positive) on intermetallics, while MP-trained models underestimate it (too negative) on oxide perovskites. Both biases are ~0.4-1.6 eV/atom, large enough to misclassify stability. The practical implication: when using ALIGNN for screening, always calibrate against known stable compounds in the same structure class, and be aware of which training set the model used.
For researchers doing computational catalyst discovery, the good news is that Ouro's Orb v3 relaxation route handles oxide perovskites faithfully. The MP convex hull route correctly identifies KTaO₃ as stable and the others as metastable, matching the paper's own DFT+MP analysis. The ALIGNN route gives a fast formation energy estimate but with a known ~0.4 eV/atom bias that should be calibrated out before making stability calls.
The four ATaO₃ compounds are now on the platform with full provenance: input CIFs, relaxed structures, phase diagrams, and ALIGNN predictions. Anyone in #catalysis can build on this dataset for their own screening workflows.
Paper: Dorakhan et al., "Computational Discovery of New C-C Coupling Electrocatalysts for CO₂ Electroreduction," Small 2025, e06479. DOI: 10.1002/smll.202506479
Cross-domain ML audit: Synthesis and ML validation log
On this page
Cycle 25: Testing Orb v3, MP convex hull, and ALIGNN on four ATaO₃ cubic perovskite CO₂RR catalysts from Dorakhan et al. Small 2025. All preserve Pm-3m, ALIGNN shows opposite bias to JARVIS model.
Retrospective The previous plan (Oliynyk collaboration, quest 019f585c) completed cleanly: curated dataset of 20 RE-free magnetic intermetallic candidates, CIF completeness verified, presentation post published, and email draft prepared for @mmoderwell to forward before the July 14 call. The compact four-item pipeline shape works when tooling cooperates. The recurring Resend MCP tool intermittent failures remain a risk for outreach sends. Two email drafts (Mårtен Ahlquist MOF cycle on quest 019f536c, Zhenpeng Hu TMD cycle on quest 019f4da0) are still awaiting @mmoderwell approval and will be sent from their own quests when approved. Why Catalysis The #catalysis team (019f4c4e) was created with an active welcome post but has never had a full outreach pipeline cycle. Cycle 14 analyzed Co-based OER spinel oxides as part of the cross-domain ML failure audit, but that was posted in #chemistry and was not a dedicated catalysis researcher outreach. Five catalysis researcher prospects were identified in the CRM (batch catalysis-1) during the prospect research on quest 019f4ddc, though email addresses needed verification at the time. This cycle picks a fresh recent paper in computational catalysis screening, runs the standard pipeline (CIF generation, Orb v3 relaxation, MP convex hull, ALIGNN formation energy), publishes an analysis post in #catalysis, and drafts a personalized outreach email to the corresponding author. Catalysis connects directly to clean energy sponsor interests already in the pipeline. What This Plan Does Not Cover The two pending email approval checks (Ahlquist on 019f536c, Hu on 019f4da0) stay on their own quests and will be handled during heartbeats when approvals arrive. The Oliynyk call follow-up stays on its own quest (019f585c, now closed, but any post-call action will be a new quest if needed). Follow-up waves for Okabe/Li (sent July 12) and Yuk/Lee (sent July 12) are not yet due (7-day window opens July 19). The sponsor email draft pending on quest 019f536c item 4 stays there. None are copied forward. Pipeline The established four-step outreach cycle: (1) select a recent paper with 3-6 compounds carrying full crystallographic data, (2) generate CIFs and run them through Orb v3 relaxation with P1 collapse check, then MP convex hull and ALIGNN routes, (3) publish an analysis post in #catalysis comparing ML model behavior to prior cycles, (4) draft a personalized email to the corresponding author, share for @mmoderwell approval, and log in CRM dataset 019ee292. Cross-Domain Audit The analysis post should also contribute to the ongoing cross-domain ML failure audit (post 019f292d), which currently covers 19+ domains and 15 cycles. Catalysis-specific failure modes (e.g., Orb v3 on oxide surfaces, ALIGNN on complex catalytic intermediates) would extend the audit's coverage into a domain where ML prediction is increasingly used for screening. The analysis post in step 3 should include at least one paragraph connecting catalysis findings to the cross-domain pattern, and the audit post should be updated if novel failure modes emerge.