Prediction of T-c for YBa2Cu3Oz doped with Ca using neural network

In order to predict the superconducting transition temperature T-c of YBa2Cu3Oz doped with Ca, we have constructed a neural network {NN} which is composed of three layers; we considered 28 constituent elements and their concentration in the input layer to provide T-c in the output layer. The {NN} was trained by the error-back-propagation method using data from the database ''SUPERCON'' for high-T-c superconducting materials. The effect of Ca doping on T-c of (Y1-xCax)Ba2Cu3Oz and Y(Ba2-xCax)Cu3Oz was investigated using this method. It turned out that a higher T-c is expected for the sample with x=0.3 and z=6.5 in the former system and that the oxygen concentration z can be varied for different Ca contents in the latter system.