摘要
以原子的金属半径、电负性和次内层d电子数为参数,用原子参数—模式识别方法,以及支持向量机等方法求得含碳、氮、硼的三元合金系形成钙钛矿型中间相的判据,并建立了计算钙钛矿型中间相的晶胞参数的半经验回归模型。该模型留一法的预报值正确率均在90%以上,与实验结果符合。
Many carbon, nitrogen or boron-containing ternary alloy phases exhibit perovskite structure are good functional materials, but Goldschmidt's tolerance t cannot be used as the criterion of their formability. In this work, the data of the formation of intermetallic compounds in carbon, nitrogen or boron-containing ternary alloy phases are collected from the Villars's data bank for ternary alloy systems. The atomic metallic radii, electronegativity and the number of d electrons in next outermost shell of the constituent atoms are used as features; the formation and cell constants of these alloy phases are predicted by atomic parameter-pattern recognition method and the support vector machine method. The predictive accuracy of all the semi-empirical models is above 90%, in agreement with the experimental results. This result indicate that not geometrical faetor but more factors, including charge transfer factor and energy band factor should be considered to find an effective regularity of the formation of ternary alloy phases with perovskite structures.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2006年第11期1091-1094,共4页
Computers and Applied Chemistry
基金
国家自然科学基金(20373040)上海市教委项目(214347)
关键词
原子参数
支持向量机
钙钛矿结构
金属间化合物
atomic parameter, support vector machine, perovskite structure, intermetallic compounds