I shouldn’t make bold statements on the subject of future blog posts. I get distracted too easily.
I came across an interesting Nature paper yesterday from Coulhard, Erb and Aggarwal, detailing an improved route to Prostaglandin PGF2a, roughly ten steps shorter than the original Corey work (7 vs 17), tested at gram-scale . Key to shorting the synthesis was cutting through to intermediate 7, which the authors decided to do by condensing succinaldehyde with itself, in tandem aldol reactions.
As would be expected, initial attempts to form the desired product failed, with mostly oligo formation. To determine what was going wrong the authors broke their desired reaction down into individual steps, looking for a point where yields were dropping. In a model system they found that the first aldol went well under proline catalysis, but the second aldol was sluggish at best (5% yield with proline alone). Luckily, one of the three other catalysts they tried worked relatively well (46%, [Bn2NH2][OCOCF3]) .
Murphy’s Law was out in force, and putting the two catalysts together outside of a model system lead to little or no product formation. It seems that the second catalyst poisons the first reaction, and so must be added several hours after mixing of the other compounds (ten hours worked best). A bit of a key point: because the authors had broken their reaction down into model systems they knew that the steps worked individually, and it just became a matter of ensuring that the aldol reactions would work in tandem.
Optimizing solvent, catalyst loading and additives moved the yield up a few percentage points to about 20%, though the yield dropped back down again when they went to large scale (57.5g). A handy trick was to assess yield based on NMR integrations alone, using an internal standard to correct for concentration differences. This definitely sped up data collection, with one set of yields given as an average of five different reactions (Table S10) .
Once the aldol reaction was optimized there was a further five steps to the desired PGF2a, and if you’re interested in that sort of chemistry I’d definitely recommend checking out the paper. For me the real gem was how the authors handled their absurd-on-its-face key reaction, which I think reveals quite a few tools for dealing with complex mixtures.
 One step went up to 285g, requiring them to revise the procedure . The supporting information is huge and is organized semi-chronologically, giving some nice insight into project design.
 This was a pretty good examples of how increasing scale can change the feasibility of a workup. Before a dean stark apparatus saved the day they were extracting their reaction mixtures with 25mL DCM. 70 times per.
 Presumably the intermediate step – hemiacetal formation – goes fairly well.
 The yield-by-NMR tests are done in D6-DMSO, which allows them to add their aliquot directly to the NMR solvent. The reaction solvent could then be removed under high vacuum, without concentrating the reagents. A useful trick.