小波神经网络用于原子半径的研究

Prediction of the Atomic Radii Using an Wavelet Neural Network

采用原子的第一电离能与原子的核外电子分布作为输入参数，使用已知的半径值作为训练样本，对小波神经网络进行了训练并成功地预报了８６种元素的原子半径，较完整地补充了共价半径、金属半径与范德华半径标度方法中所缺的相应值．结果表明，对于原子及分子物理的研究，小波神经网络是一种很有潜力的工具．

Wavelet Neural Network(WNN) is successfully applied to the prediction of the atomic radii in this paper, with the atomic first ionization potentials and the atomic electronic configurations selected as the input parameters. An excellent correlationship between the predicted values and the measured values is obtained, and the missing values in valence radii, metallic radii and Van der Waals radii scales are also predicted by the method reasonably. The results indicate that the WNN promises to be a high performance method for studies in atomic and molecular physics.