If we've learned anything in the past few months, we've learned that measuring solubility is really tricky
The Open Notebook Science Challenge
has generated 11 answers so far for the solubility of vanillin in methanol:
Rajarshi Guha has provided an extremely handy web query interface
(must use FireFox) to generate these plots. It taps into live data from this GoogleSpreadsheet
and links back to the specific experiments
that generated the data.
Because we have access to the lab notebook pages, these measurements are not equal
. Some of the measurements are based on reports where conditions that later turned out to be important were not reported or controlled. As we learn more about what is important many of these measurements will probably be removed and replaced with more reliable data.
But in the meantime, we're going to use the best possible estimate of the property that we have available. It lets Rajarshi feed his solubility models
and gives us a tight iteration cycle between prediction and experiment. For this purpose, the average value of 3.5 M for the 11 measurements is probably good enough to be part of a training set to allow a rough prediction of solubility. As we get more confident over time we'll improve the model.
Right now, we're not quite ready to do predictions but we should be there soon. The main feedback we're getting now is which compounds we need to focus on to get to that minimum training set (Rajarshi says 50 compounds/solvent and we have about half that number for some solvents). It looks like we'll focus on aromatic aldehydes and aromatic carboxylic acids, mainly because many don't evaporate easily in the SpeedVac (one of the control parameters discussed earlier
Another advantage of aromatics is that we can use UV spectroscopy to determine solubility without using evaporation. Hopefully in the coming weeks this will confirm what Jenny Hale has concluded today in ONSC-EXP011
The results of the calculations give the solubility of vanillin in ethanol as 2.48 M and vanillin in methanol as 4.15 M. This finally gives excellent correlation with exp207, which measured the solubilities as 2.5 and 4.19 M respectively.
It appears that some compounds require significant time and agitation to reach saturation. In this last experiment Jenny carefully recorded what happens over the course of adding vanillin to methanol and periodically vortexing. Inspection of her log
shows several points where someone might have assessed the solution to be saturated when it was just slow to dissolve. It also makes a case for always wearing safety goggles in the lab :)
At this point I am becoming more convinced that the solubility of vanillin in methanol is closer to 4.2 M. If that result is consistently obtained by other students and other methods (such as UV
) using prolonged mixing times then we'll remove from the SolubilitiesSum spreadsheet
the measurements that were obtained from experiments where the mixing time was less or simply not reported.
This evolution of this project also demonstrates the value of the ongoing open peer-review
of an open lab notebook. The judges for the ONS challenge
have provided feedback about future experiments, questioned assertions, pointed out omissions and suggested additional ways of thinking about the experiments. The contributions from the judges shows up in bold in the notebook pages
and can be tracked over time by looking at the wiki page history.
Labels: challenge, open notebook science, solubility