About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
Algorithms Development in Materials Science and Engineering
|
| Presentation Title |
A Validated ML-Potential Model to Quantum Algorithms Workflow for Constraint-Aware Defect-Configuration Search in Rare Earth Elements Doped ZrO₂ |
| Author(s) |
Huajing Song |
| On-Site Speaker (Planned) |
Huajing Song |
| Abstract Scope |
Predicting low-energy defect configurations is a combinatorial bottleneck in materials design. We present an end-to-end algorithm that maps fixed-composition defect search in rare earth elements doped ZrO₂, a thermal-barrier-coating system, onto constraint-aware quantum optimization. Energies from the MACE-MPA-0 ML-potential model are fit to a 24-variable quadratic unconstrained binary optimization (QUBO) surrogate reproducing the reference landscape with held-out R²=0.997, validated against exact enumeration of all 448 feasible configurations. The objective is attacked two ways: a constraint-preserving XY-mixer QAOA that stays feasible by construction and places 86% of probability within 1 meV of the optimum, and a fault-tolerant Grover Adaptive Search oracle built layer-by-layer from validated fixed-point reversible arithmetic and resource-estimated at 324 logical qubits and ~4×10⁴ Toffoli gates. This provides a validated, reproducible bridge from machine-learned interatomic potentials to variational and fault-tolerant quantum algorithms for materials optimization. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, Machine Learning, Modeling and Simulation |