13
$\begingroup$

I'm trying to predict the solubility of long alcohols (e.g. 1-octanol, 1-decanol or 1-dodecanol) in a mixture of water and an organic co-solvent such as DMSO, acetonitrile or ethyl acetate.

For example, decanol in water with 25% DMSO or saturated with ethyl acetate (8.3%). I of course know the solubility of alcohols in water, but not in organic solvents or in mixtures.

I found a server to predict the solubility of any compound in any solvent, but I'm not sure how accurate it is and it only does predictions for pure solvents, not mixtures: http://showme.physics.drexel.edu/onsc/models/multisolvent.php

Any input on this would be greatly appreciated!

$\endgroup$

2 Answers 2

5
$\begingroup$

I think this is very much an open problem. As you may see from the Drexel pages, even solubility prediction in one solvent is still fairly hard, partly because of the lack of experimental data.

The only software that I know capable of making predictions about solubility in mixed solvents is COSMOtherm, based on the COSMO solvation model.

I have never used it, but the capability is mentioned on the website.

$\endgroup$
2
  • $\begingroup$ I had a workshop on COSMO once, iirc the theory is sound, but the predicted values are often far off the experimental results. (But this was several years ago and Klamt was still improving it.) This is probably due to computational limitations. Note that the model used here is COSMO-RS. $\endgroup$ Jun 19, 2015 at 5:59
  • $\begingroup$ @Martin-マーチン I think it's an extremely hard problem and there's obviously not a lot of experimental data to calibrate and test models. My guess is that only in the next 5-10 years will this be possible, and only if there's a lot of open solvation data. $\endgroup$ Jun 19, 2015 at 12:13
2
$\begingroup$

I would be surprised if you found anything predicting solubility in mixtures as there is poor modelling due to non-linear effects. Taking solubilities of two solvents and trying to interpolate does not work. As little as a few hundred ppm to 4% of one solvent in another is enough to confound standard predictive equations without empirical data.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.