Binary Mixture Data Feasibilty Study Meeting Notes

24/03/2020

- Discussion of the main ‘question’ for the study and which sub questions we want to answer

• can we replace pure data with mixture data? I.e with mixture data there is more available, and it allows you to more readily vary the different interactions i.e different types of mixtures, being able to titrate the relative amounts of interactions by varying the concertaions of mixtures etc.

OM: Good overall question, but broad. Important to tease out specific effects.

Maybe look at do a 1 for 1 swap of Hvap with Hmix, and rho(pure) with rho(x)?

MRS: Would be good at looking at how sensitive the opt is to starting parameters. Start from perturbed parameters.

SB: Can do this at the end, using only the prop types which we decide should be proposed going forward.

MRS: Benchmark against properties from a single mixture over a wide range of composition

MRS: Look at homogeneity in fluids - solvation shells, etc.

SB: Automated test for aggregation - good project for Peter / undergrad?

- Updated plots with bootstrapped uncertainties which now also include std dev in estimated values

Open

- How to generalise the results of the alcohol-ester study?

??? Extend the study to mixtures of polar-polar, less polar-polar, apolar-apolar h-bond acceptor - h-bond donor)

We would drop studies involving Vexcess(x) as fitting against that proved difficult, and didn’t seem to offer any performance benefits over rho(x)

• MRS: Should write this up in the study plan, and will give feedback.