The next generation of catalysts
The chemical industry uses catalysts for processes to run efficiently while remaining environmentally friendly, from the transformation of exhaust gases from cars to the production of fertilizers using nitrogen from the atmosphere. Amongst these chemical processes there are some that do not yet have effective catalysts, and these will require solutions in the near future. For example the conversion of carbon dioxide into useful substances to mitigate climate change, and the reaction between oxygen and hydrogen to form water for use in fuel cells. The role of a catalyst is to aid the conversion of chemical substances in a chemical reaction, and an effective catalyst can do this quickly and with small energy loss. It is a great challenge to predict which material will act as a good catalyst for a chemical reaction, and it is exactly this problem that we propose a solution for with a new class of materials, the so-called high-entropy alloys.
High-entropy alloys are a composed of a mixture of five or more metals, having only recently been used as catalysts. We present the first theoretical study of how to systematically benefit from high-entropy alloys to provide the best alloy candidate that can catalyze a desired chemical reaction.
What makes the high-entropy alloys different from other catalysts is that they have a surface with countless local configurations of different atoms giving rise to as many local chemical environments. Imagine a Rubik's Cube: When it is solved, it consists of six faces each with its own color representing the pure metals. Mix the Rubik’s Cube and each face is now composed of many colors. On each face, the six colors can be arranged many different ways. The nine squares represent a local combination of six different metals on the surface of a high-entropy alloy. Some combinations of atoms on the surface will bind the reacting chemical substances weakly, while with others they will bind strongly. For those combinations of atoms where the bond strength is perfect the catalytic activity will be greatest, and these combinations will govern the overall catalytic activity.
By calculating the bond strength of the chemical substances for all configurations of atoms, we can identify the best chemical environments and in what proportion the mixed metals are included at the atomic level. Here, however, we encounter the problem that it would take a lifetime to calculate the bond strengths for all the combinations even with modern quantum mechanical methods. We have solved this problem by calculating the bond strengths of a randomly selected subset of the possible combinations and then used machine learning to calculate the bond strengths for the entire span of combinations in just a few seconds.
When the bond strengths of all local combinations of atoms on the surface are known we are able to tune the ratio of the incorporated metals in order to promote the likelihood that the best bond strengths occur as frequently as possible. This optimal mixing ratio can be calculated and the outcomes are completely new, untested catalysts. The method thus gives us a systematic way of proposing catalysts which only depends on which metals we include. We have used the method to suggest catalysts for the reaction between oxygen and hydrogen forming water but the application is very broad so we are currently working on several other chemical reactions, as well as improving the approximations and assumptions of the method so that we can propose alloys that hopefully exceed the activity of present day catalysts.