Visualization of Large Experimental Space Using Holographic Mapping and Artificial Neural Networks. Benchmark Analysis of Multicomponent Catalysts for the Water Gas Shift Reaction

Abstract

This paper reports the combination of Holographic Mapping (HM) and Artificial Neural Networks (ANNs) in order to provide a benchmark visualization of a multi-dimensional space in two-dimensional forms. In this approach each matrix point in HM represents virtual catalytic data generated by means of ANNs in order to visualize the given multi-dimensional experimental space. A 12-dimensional experimental space related to the composition of catalysts designed for the water gas shift reaction (WGSR) from 12 possible components supported on ZrO2 is visualized. Catalytic data obtained in an earlier combinatorial screening process at 300 °C were used for training of the ANNs. The results show that the visualization of large experimental spaces having more than half a million virtual experimental points can be accomplished. The analysis of synergistic effects between different components revealed that the key components of water gas shift catalysts at 300 °C were Pt, Fe, Eu and V, while Co, Ru, Sb, Ge and Mo had a pronounced negative effect on the activity.