Wednesday, December 5, 2012

Process Wednesday: Stopping at the right time

From a recent Org. Process Res. Dev. ASAP [1], a really great point about scaling up and reactions that are sensitive to side reactions. The authors had just developed a successful cyclization with 90% yield and 96% purity to produce 1.5 kilograms of material. And then...:
In order to test the robustness of this process, the reaction was extended for an additional 20 h. While the cyclization was complete within 2 h, the level of 18 increased almost linearly with reaction time, reaching 50% after 22 h at the expense of product 11, indicating that 18 was mainly generated from 11 (Figure 1a) [CJ's note: Left-hand graph above.] Thus, it became critical that the reaction be stopped at the right time point to minimize degradation, a challenge at plant scale. As the de-ethylation is likely acid catalyzed, base additives were considered to slow this process. The addition of 2,6-lutidine (2 equiv) was found to significantly retard the rate of the formation of 18 without a negative impact on the cyclization rate (Figure 1b) [center graph]  
Subsequently, an improved cyclization procedure was developed, using 1 equiv of p-toluenesulfonic acid (p-TSA) buffered with 3 equiv of pyridine as the catalyst and a Dean−Stark trap to remove water generated during the reaction. The cyclization of 10 was complete within 3 h, and 11 precipitated out from the reaction mixture as the p-TSA salt.  
More importantly, once the reaction was complete, no further decomposition of 11 to 18 was detected (Figure 2) [graph to right]; the levels of 11 and 18 remained essentially constant at 94% and 4%, respectively, even after heating the mixture to reflux for an additional 21 h. After cooling to room temperature, 11-TSA was isolated in 97% yield and 95.8% purity (containing 2.6% of 18). Material of this quality can be directly used in the next step. Interestingly, pyridine and p-TSA were found to be the ideal combination. Reactions catalyzed by p-TSA alone or with 2,6-lutidine in place of pyridine were observed to be slower and generated higher levels (4.8−8.8%) of 18.
I can't agree more about the difficulty of stopping a reaction at the right time. I think it's just very difficult to get that sort of thing to take place in a busy chemical manufacturing plant; it's much better to build that robustness in from the start.

1. Shu, L.; Gu, C.; Dong, Y.; Brinkman, R. "Efficient Large-Scale Synthesis of a 2,4,5-Triarylimidazoline MDM2 Antagonist." Org. Process Res. Dev. ASAP. DOI:


  1. Demonstrates the importance of profiling reactions early on in development to avoid a) missing potential hits (eg, fast reaction that isn't sampled soon enough) and b) suprises like this.

    There's a range of automation available to do this for sampling 1-100+ reactions on 0.5-20ml scale which greatly reduces the labour burden (although the analysis and data processing can be fun...). Plus, getting all that kinetic info early on helps reaction understanding and informed process development.

  2. quite recently, my colleagues were running selective acidolytic debenzylations of a substrate in TFA (+ a scavanger): a selective cleavage of benzyl aryl ethers in the presence of benzyl ester. Turns out, this partial deprotection is actually doable if the reaction is monitored and the reaction conditions fine-tuned for maximum selectivity. (I was surprised too).
    But when the batch scale was doubled and the deprotection was run in a reactor (with a solvent swap instead of a rotovap), the selectivity went down (because the vacuum evaporation of TFA from reactor took much longer) and the mix over-reacted. And there was a lot more leftover TFA in the crude product which really complicated the workup

  3. PAT to call the reaction endpoint (perhaps ReactIR)would be useful in situations like 1a/1b. Thorough understanding of variables that impact the kinetic data and a wide design space could also potentially get you away from sampling at all (QbD type process control).


looks like Blogger doesn't work with anonymous comments from Chrome browsers at the moment - works in Microsoft Edge, or from Chrome with a Blogger account - sorry! CJ 3/21/20