Sean Bohun - “Estimating the chemical constituents of modern electrolysis cells”

Problem statement:

• Chemists want to track reactions in electrolysis cells using spectroscopic methods (i.e. by using absorbance data to infer chemical concentrations). However, they don’t know the spectra for the individual species involved in the reaction and they’re not even sure what species
are going to appear.

• Current process is to propose a certain number of species and fit them to the data. It is assumed that the observed spectrum is a linear combination of the species spectra, and requirements of non-negativity (of concentration), closure (mass conservation) and unimodality (of concentration) are enforced in order to determine approximate spectra and approximate concentrations. This process leaves much to be desired
(it’s a bit of a sledgehammer) and is very slow.

• Open question: How do you make any sense of spectroscopic data when you have so little information from other sources? What problems arise at high concentrations?

Suggestions/comments:

• Is unimodality a sensible requirement in the concentration trajectory? Why shouldn’t a species fluctuate?

• It is important to be sure that the reaction has reached thermodynamic equilibrium before performing measurements. Perhaps get a chemical thermodynamics expert to advise on this?

• JRO: The approach seems fine at low concentrations using a series approach. Perhaps Pad´e approximants could be used to extrapolate to higher concentrations.

• Could model-free methods be used?

People interested in follow-up: AAL