人工神经网络用于金属离子-乙二胺四乙酸络合物稳定性研究

Study on the Stabilities of Ethylenediamine Tetraacetic Acid Complex of Metal Ions by Using Artificial Neural Network

采用函数连接型网络,以金属离子的电荷、离子半径、价电子结构、电负性及适配价轨道数因子为描述参数,成功地关联了近60种已知的金属-EDTA络合物稳定常数logK值,并在此基础上,预报了包括所有锕系元素在内的20种金属-EDTA络合物的logK值.对学习样本的正确选择、高价金属离子与EDTA络合物logK值的报道差异及成因、Co^(3+)离子的“额外”稳定性进行了分析讨论.

A functional-link net(FLN) was employed to correlate a set of atomic structural descriptors, such as electric charges ( Z), ionic radii ( r), electron numbers in the valence layer (f, d, s) , principal quantum numbers in the outer-shell ( n ) , electronegativity (xp ) , and the factor of adaptive valence orbit number(w) of the metal ions with known stable constants (logK) of ethylenediamine tetraacetic acia complexes of 60 metal ions. Stable constants (log/0 data of 22 metal ions, including the actinide elements, lake of experimental data are predicted with the trained FLN mode. Additionally, extensive analyses and discussions were made on the correct selection of training samples, the differences and their causes of logK values reported of EDTA complexes of high valance metal ions, and the 'extra' stability of Co3+.