Saturday, June 18, 2016

Weekend mediumreads: the chemistry of recording

Also in this week's C&EN, I have been remiss in not mentioning a really great article by Matt Davenport about the chemistry of recording, starting with Thomas Edison (a boyhood hero of mine) and wax cylinders, going forward to modern vinyl records:
Using these documents, Monroe is tracking how Aylsworth and his colleagues developed waxes and gaining a better understanding of the decisions behind the materials’ chemical design. For instance, in an early experiment, Aylsworth made a soap using sodium hydroxide and industrial stearic acid. At the time, industrial-grade stearic acid was a roughly 1:1 mixture of stearic acid and palmitic acid, two fatty acids that differ by two carbon atoms. 
That early soap was “almost perfection,” Aylsworth remarked in his notebook. But after a few days, the surface showed signs of crystallization and records made with it started sounding scratchy. So Aylsworth added aluminum to the mix and found the right combination of “the good, the bad, and the necessary” features of all the ingredients, Monroe explains. 
The mix of stearic acid and palmitic is soft, but too much of it makes for a weak wax. Adding sodium stearate adds some toughness, but it’s also responsible for the crystallization problem. The aluminum stearate prevents the sodium stearate from crystallizing while also adding some extra toughness. 
It would be interesting to take a look at the notes of early chemical formulators like Aylsworth to see how much chemical intuition/understanding they had, and how much was sheer trial-and-error. 


  1. Formulation chemistry! Something that is not taught nearly enough in school, probably because it is so ambiguous. I'd bet that a LOT was trial and error early on, but as they built a library of compounds that could work, their sophistication grew rapidly.

  2. FYI (

    "In Excellence R Us: University Research and the Fetishisation of Excellence (commentable version here) the authors "...examine how excellence rhetoric combines with narratives of scarcity and competition and show that hypercompetition that arises leads to a performance of 'excellence' that is completely at odds with the qualities of good research." Inverse interviews one of the authors. Times Higher Education interviews another."

  3. I agree wholeheartedly with ClutchChemist. My education was designed to train me for either a professor position or some academic-like industry job, say the old Central R&D at DuPont, or organic synthesis at Merck, or something like that. Those kind of jobs just plain don't exist anymore, but manufacturing companies will always need formulators.

    There's still a lot of trial and error and intuition to this day. Say a paint, shampoo, skin cream, electrical insulating resin, or whatever has 10 ingredients in it. They can interact with each other 10! ways (10x9x8x7 etc). Running 10! experiments isn't feasible, so I rely heavily on intuition and experience.

    1. KT: I have a similar background, trained in organic synthesis / process chemistry, for a job in pharma, but I learned very early on about statistically designed experiments (DoE). I started my career at DuPont and all new Ph.D. chemists were required to take an in-house course in DoE within their first two years. I used it once in my five years at DuPont but did not do so once I moved to pharma for most of my career. Currently, I am working far from pharma in the polymer area. While my work still involves small molecule synthesis, much of what other people do here is formulation. DoE should be a big part of what they do but as you suggest, they were never taught it in school. I have been pushing for it and am starting to see some people take a real interest in it. In the meantime, they still rely on “intuition and experience”.

      To give you an idea of the time / resource savings when using DoE, let’s start with the 10! possible interactions which is 3,628,800 possible interactions. A full two-level factorial design for ten variables requires 1024 experiments not including replicates or center points. A ¼ fractional factorial design is down to 256 experiments not including replicates or center points. A Box-Behnken design requires about 170 experiments including replicates and center points. Finally, a Plackett-Burman design requires 12 experiments not including replicates. As the number of experiments decreases, the design becomes more a screening design to determine which factors (ingredients) are the most important as opposed to a response surface exploration but once you know the ingredients that really need to be optimized, I bet using intuition and experience for the remaining ones would suffice.

      Everyone doing formulation work should be required to learn to use DoE.