| OBSERVATION'S |
PROCESS CHEMISTRY is usually equated with scale-up, but characterizing process chemistry simply as the scale-up of a synthetic route does a grave disservice to the organic chemists who have chosen to focus their creative efforts in this field. "Process research and development is not about making large amounts of compound," The synthesis of large quantities of a compound is really just a "by-product" of the real job, which is the acquisition of "process knowledge," Process chemistry has evolved from what author calls a "cottage industry"--in which each project was approached individually as a "once off"--to a "business process." In this business process approach, the goal is to identify aspects of process development that are common to all molecules. Author doesn't have an exact number for those attributes, but he guesses that it's somewhere around 70% of the whole. That part of the development can then be attacked in the same manner for each molecule, allowing the chemists to focus their creativity where it's needed: on the 30% that is unique to the compound being developed. For example, much process research is now done with automated parallel experimentation, analogous to combinatorial chemistry in drug discovery. A type of mathematical modeling called design of experiments is used to set up the experiments. Multiple variables are changed simultaneously to find the optimum process, as opposed to the more classical approach of changing one variable at a time (OVAT) and finding the best value for one parameter before moving on to the next. Two things are missed by optimizing variables sequentially rather than simultaneously. First, you are never totally certain that you have found the global, as opposed to a local, maximum. The reason for this is that the maximum is determined usually not only by the sum of the individual factors, but also by second- and higher order interactions of variables, which cannot be determined in the OVAT approach. Second, you don't know what happens if you subsequently vary a factor, information critical in determining how tightly that factor must be controlled in the final process. "In one of the simpler experimental designs, you pick a low, medium, and high value for each variable," "Then through a randomization process, you set up multiple parallel experiments in which combinations of variables are assessed for their impact on the desired output, typically reaction yield. You then get the computer to do a regression analysis of the impact of the variables on the experimental outcomes. You get a lot more information for the same number of experiments. You not only are sure of having the absolute maximum, but you also know how critical the variables are. Are you sitting on top of the Eiffel Tower, where variable control in the manufacturing setting will need to be very tight, or on the Sydney Opera House, where there are multiple optima and close control of some variables is more important than others?" |
| UPDATE | 05.02 |
| AUTHOR | This data is not available for free |
| LITERATURE REF. | This data is not available for free |
|
Want more information ? Interested in the hidden information ? Click here and do your request. |