Data and code supporting the paper "Competing adsorption of H and CO on Pd-alloy surfaces: Mechanistic insight into the mitigating effect of Cu on CO poisoning"

This record contains machine-learned interatomic potential (MLIP) and cluster expansion (CE) models for the AuCuPd:CO,H system along with reference data from density functional theory (DFT) calculations. Technical details are provided in Pernilla Ekborg-Tanner and Paul ErhartCompeting adsorption of H and CO on Pd-alloy surfaces: Mechanistic insight into the mitigating effect of Cu on CO poisoningdoi: https://doi.org/10.48550/arXiv.2603.00776 MLIP models Models based on the neuroevolution potential (NEP) architecture are provided in the nep-*.txt files. The "full" model has been trained against all available data whereas the "split" models have been trained against a random selection of 90% structures from the available structures. Models based on the MACE architecture are provided in the MACE-*.model files. CE models Models are provided for (111), (110), and (100) surfaces in the *.ce files along with reference structures in the *.xyz files Reference data Reference data from DFT calculations is provided in the reference-structures.db file, which is a sqlite database generated using the ase package.本数据集包含针对AuCuPd:CO,H体系的机器学习原子间势（machine-learned interatomic potential, MLIP）与簇展开（cluster expansion, CE）模型，以及来自密度泛函理论（density functional theory, DFT）计算的参考数据。相关技术细节参见： Pernilla Ekborg-Tanner与Paul Erhart的论文《Pd合金表面H与CO的竞争性吸附：Cu缓解CO中毒的机制解析》，DOI：https://doi.org/10.48550/arXiv.2603.00776 ## 机器学习原子间势模型 基于神经进化势（neuroevolution potential, NEP）架构的模型存储于nep-*.txt文件中。其中“完整”模型使用全部可用数据训练得到，而“拆分”模型则从所有可用结构中随机选取90%的结构进行训练。 基于MACE架构的模型存储于MACE-*.model文件中。 ## 簇展开模型 针对(111)、(110)及(100)晶面的簇展开模型存储于*.ce文件中，配套的参考结构存储于*.xyz文件内。 ## 参考数据 密度泛函理论计算的参考数据存储于reference-structures.db文件，该文件为使用ase工具包生成的SQLite数据库。