Org Prep Daily

June 10, 2015

The power of blunder – based optimization

Filed under: industry life, procedures — milkshake @ 8:46 pm

I have been trying to optimize a difficult reaction; I thought a presence of zinc chloride might help so I gave this a try and there was an improvement: The results were getting better, week after week.

Some time later – by now with improved product purity and filtrability – I begun to wonder if the zinc chloride effect was real, or maybe something else was going on, so I finally got around to run a control. And sure enough, the reaction worked even better without zinc chloride. So, after many tries with quantities of reagents and additives, I arrived at optimized procedure which looked almost exactly like the one that I started with, except few minor details – the little changes that were incidentally co-introduced because of the ZnCl2 addition – few small changes that make a difference… I would have never tried these changes without it. And I would have given up if I had run the control experiments earlier and found out it does nothing.

It is delightful to read methodology papers, the observations and explanations arranged neatly, flowing like a good detective story, with a chain of clear logical reasoning based on the experimental evidence. But I suspect it is mostly fictional (There is no good place in a process paper to explain that after very slow reagent addition because of a clogged valve that no-one cared to inspect before the pilot run, the impurity profile improved and the troublesome sideproduct from the second step no longer buggers up the recrystallization). I worry that reading published accounts of process research can give the management a very unrealistic impression what a normal project should look like.


  1. Hey Milkshake,
    Nice to see you back.

    Do you have any tricks or know of any literature on quickly getting crystalline derivatives of small molecules. A have a mostly aliphatic substance, with a lactone and a free primary hydroxy group. I have heard that bromo-benzoates or nitro-benzoates might often help, but they did not in my case. I have also tried an acetyl mandelate, to no avail. Do you know of other esters with a large propensity to form crystals?

    Comment by Young Padawan — June 12, 2015 @ 3:55 am

  2. Then I would try 4-phenylbenzoic acid ester (p-biphenyl-substituted compounds are notoriously crystalline). Also a plain old trityl group (Trityl chloride, NEt3, cat. DMAP) is worth trying.

    Comment by milkshake — June 12, 2015 @ 8:16 am

  3. Honestly I think you’re on to something here. I think there maybe *is* a place for it in blogging, if someone would be courageous enough to get the ball rolling (which maybe you just did…?). It’s like the chemistry version of mommy blogging: telling about all your mistakes and the difficulties of day to day work in the industry, but showing how you produce good work despite and sometimes because of the mistakes.

    Comment by Charlie Kilian — June 12, 2015 @ 11:06 am

    • There was Blog Syn, which was a lesson learned in how much time is required for a project of that nature.

      Comment by croon — June 14, 2015 @ 7:19 pm

      • I was trying to say something else, about process optimization: When you are trying to change only one parameter, you have to worry what else changes too.

        It can be surprisingly hard to run the series of ten or twelve experiments over a period of few weeks in exactly the same way. The execution of the first experiment is probably not going to be quite the same as with the last one – an experiment worked up late at night is maybe precipitating and filtering little longer than experiment done in the morning, and there is less sunlight, etc. You don’t even realize that you are making changes. When results keep improving and “make sense”, you easily convince yourself that the improvement comes from the variable that you are trying to optimize (rather than how many coffees you had already and how much you had to rush the precipitation to finish in time for a meeting). There is always a bias in research – if the results are going the way you like, you don’t look too close. But if it is not, you can find an excuse why it did not work out…

        One more example of this problem: I have been trying to optimize a heterogenous reaction, with the starting material and product both insoluble in the reaction media: a transformation from slurry to slurry. The reaction worked fine but it was taking over 2 days to complete, and I did not want to heat it because the material decomposes readily, I needed to shorten the reaction time and at the same time I worried about the reproducibility because I had to limit carryover of the staring material (which contained a problematic heavy metal) into the final product.
        So I though I should run 24 hour experiments and vary the concentration of the reagent, to see which concentration range works best for one day run…

        I was going to test concentration of the reagent from 0.5M all the way to 1.6M, and since I could not manage to set up 12 experiments in parallel (I needed to use mechanical stirrers, and we don’t have that many stirrers, and the workup was labor intensive), I set only two experiments on each day.

        I set two experiments a day in parallel but out of sequence: For example, 0.5M + 1.6M concentration, the next day 0.8M + 1.2M etc. Finally, when everything was finished, isolated and analyzed, the results in table: A complete mess. More than half of all experiments showed incomplete conversion – but the analysis results for the starting material carryover into the product did not make sense, it did not correlate with the concentration of the used reagent at all – the only correlation that I could find was with the result from the other experiment run on the same day, and also according to the batch of the used starting material… So whatever role the reagent concentration played, it was completely overshadowed by some other factors which I did not appreciate and therefore did not control.

        This botched series of experiments was useful because it pointed to a major complication: There was a problem with wetting the solid particles of the poorly soluble greasy starting material with a concentrated solution of the reagent (which was aqueous, and has a pretty high surface tension). That’s why using material with a different particle size gave a different result… I was able to notice this only because the data points – the individual experiments – were run out of sequence. If I had run this in sequence (for example 1.6M + 1.5M, next day 1.4 + 1.3M etc, in regular intervals), the second part of the series with lower concentrations would be worse than the first half, and the table with results would show a clear but false trend and I would convince myself that the reaction can work only at higher concentrations.

        [Nowadays I am adding a little bit of surfactant to help with the wetting – the problem solved.]

        Comment by milkshake — June 14, 2015 @ 8:32 pm

        • Changing one parameter at the time would be a pleasant and welcome situation, at my shop. Most of the times a well constructed DOE is the only way – giving solutions that no rational reaction design could ever achieve.

          Comment by processchemist — June 15, 2015 @ 5:05 am

RSS feed for comments on this post. TrackBack URI

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Blog at

%d bloggers like this: