By Kilian Kobl - Project Manager at Ypso-Facto
Progress in catalysis has enabled major breakthroughs in domains as various as agriculture, health and daily life, to only cite a few. Although new approaches of studying and developing catalytic processes emerge from time to time, people generally use one of three ways:
- Fully experimental – either “trial and error” or testing all possible combinations. It becomes rapidly apparent that the latter approach is bound to failure. For example, take a process with 3 reaction steps, each with 5 key operating parameters which in turn can take 3 values. If 1 experiment lasts 1 week, one would need ten thousands of years to test all possibilities!
- Statistical models. Mathematical relations (e.g., linear, quadratic) are used to correlate operating parameters (e.g., pH, temperature) with responses (e.g., purity, yield). This can allow reducing the number of experiments and be applied to any process. However, it suffers from a lack of predictive capability, i.e. there is no way to predict what happens outside tested conditions.
- Mechanistic models. The physico-chemical phenomena at stake (e.g., reaction kinetics, diffusion, …) are identified and described by a set of equations solved numerically. This approach requires some understanding of the underlying mechanisms and thorough analytical work but is extremely powerful. Once the model parameters have been determined within the experimental range, the model can be used in software to explore by simulation new conditions, which have not been experimentally studied, including different scales.