Predicting the photosynthetic ammonia on nanoporous cobalt zirconate via graph convolutional neural networks

•Nanoporous Co2ZrO5 system is exploited by us to improve the efficiency of ammonia photosynthesis.•Deep-learning algorithms are used to expand a way to reveal the structure–property relationship and optimize catalytic activity on Co2ZrO5 catalysts.•Graph Convolution Neuron networks (GCNN) are used as the most efficient network with a high accuracy on statistical significance.•Combining the interdisciplinary means, the best ammonia yield is evaluated depending on the dataset collected by us aided by 3D contour model.