金红石型高熵氧化物氢吸附的密度泛函理论计算与图神经网络预测模型OA
Density functional theory calculations and graph neural network prediction model for H2 adsorption on rutile-type high-entropy oxides
传统半导体氢气传感器在低温下难以同时达到快速响应、高灵敏度和高选择性等性能要求,进而无法满足其在氢能安全领域的需求.高熵氧化物(high-entropy oxides,HEOs)为设计高效氢敏材料提供了新途径.氢原子能在室温下进入金红石结构氧化物单晶的晶格使其电阻下降,据此可推测,金红石型 HEOs 在氢气吸附领域有很大的应用潜力.本文针对金红石型 HEOs 的多组分无序特性,采用准随机结构方法构建了标准化结构库,结合密度泛函理论计算定位 Zr-O 桥位为最佳吸附位点,所构建的图神经网络模型实现吸附能预测的决定指数 R2 达 0.889,均方根误差仅为 0.203 eV,平均绝对误差仅为 0.127 eV,为氢敏 HEOs 材料的高通量筛选提供了有力的理论工具.
Traditional semiconductor hydrogen sensors struggle to simultaneously achieve rapid response,high sensitivity,and high selectivity at low temperatures,thus failing to meet the urgent needs in the field of hydrogen energy safety.High-entropy oxides(HEOs)offer a new approach for designing efficient hydrogen-sensitive materials.Studies have reported that hydrogen atoms can enter the lattice of rutile-structured oxide single crystals at room temperature,leading to a decrease in resistance.This suggests that rutile-type HEOs have great potential for hydrogen adsorption applications.Targeting the multi-component disordered characteristics of rutile-type HEOs,this study generated a standardized structure library via the special quasirandom structure method.Combined with density functional theory calculations,the Zr-O bridge site was identified as the optimal adsorption site.A graph neural network model was constructed to predict the adsorption energy,achieving an R2 of 0.889,an root mean square error of 0.203 eV and an mean absolute error of 0.127 eV on the test set.This work provides a powerful theoretical tool for the high-throughput screening of hydrogen-sensitive HEOs materials.
蒋子棱;徐京城
上海理工大学 材料与化学学院,上海 200093上海理工大学 材料与化学学院,上海 200093
化学化工
高熵氧化物密度泛函理论计算图神经网络气体吸附
high-entropy oxidesdensity functional theory calculationsgraph neural networkgas adsorption
《有色金属材料与工程》 2026 (3)
12-18,7
传感器技术全国重点实验室开放课题SKT2502
评论