The ChemArtificer

Artemisinin – photochemical flow synthesis 5 months on

The semi-synthesis of the anti-malarial drug artemisinin from dihydroartemisinic acid, announced in an article by Levesque and Seeberger (Angew. Chem. Int. Ed. 2012, 51, 1706), was followed by extensive coverage in the the scientific and non-scientific press, as well as through social media. The synthetic sequence was a nice piece of lab work utilising flow chemistry to enable reactions that are normally a bit capricious in batch mode i.e. photochemistry and oxidation with molecular oxygen. The flow operation incorporated four reactions: photochemical oxidation, Hock cleavage, oxidation and cyclisation. The nice touch was to use singlet oxygen – activated by light in a reactor set-up based on that reported originally by K. Booker-Milburn – and then later in the sequence the remaining oxygen (in the triplet state) oxidised the intermediate, leading to the all important incorporation of the endoperoxide functionality. This permitted all reactions to be performed sequentially in a few minutes to give artemisinin (after chromatography) in 39% yield.

The unfortunate irony here seems to be that despite using such an enabling technique as flow chemistry, the weak link appears to be the standard (in the context of this synthesis) TFA mediated Hock cleavage. In addition, the batch reactions gave an overall yield of 37.5%, so the flow yield has been hit hard by lack of regioselectivity. The flow process can only be as good as the chemistry going into it.

However, taking nothing away from the endeavour, this is a nice exhibition of lab scale flow.

The quote in follow up articles was that since the original lab development, the process had been improved so only 150 (down from one thousand or so) of these reactors would be needed to produce the annual requirement of artemisinin.

And here comes the point.

To number up this many reactors in this type of process would not normally be attempted by practioners in fine chemical or pharmaceutical manufacturing, irrespective of the techniques used, batch or flow. As a starter, how would you exercise cGMP on 150 separate reactors for a single process? These could all be classed as separate production streams. Even 10 would be a challenge. Could running 150 reaction streams into one pot be classed as blending? In batch it would so why would should flow be classed differently?

Using 150 reactors needs 150 bulbs, 150 pumps, 150 oxygen lines, the list goes on. If GMP was being observed, each reactor stream would have to be monitored and analysed separately, lending huge weight to the PAT paradigm, but causing friction burn to the wallet hand 150 times over.

So back to the bulbs. What’s the lifetime of a medium pressure mercury bulb? Let’s be generous and say 1200h (but I concede I may not have found ones with more longevity), which equates to 50 days continuous operation. All those bulbs need changing. How many electrical engineers does it take to change 150 light bulbs? Punchlines to the blog please.

Or rather, what are the logistics of organising the preventative maintenance of 150 bulbs all operating simultaneously? And do the lamps all need to be supplied with step-up transformers?

It’s easy to pick holes, but at the end of the day, this is a lab process and not as it stands, the basis for a commercial one. No-one suggested numbering up Kugelrohrs to carry out production scale distillation.

I re-iterate. Great piece of lab work, but the barriers to commercialisation are numerous.

So where am I going with this?

Flow photochemistry can be carried out efficiently at production scale without numbering up lots of reactors. The heat exchange issues and bulb issues can be dealt with simply that could be subjected to GMP without it being a nightmare – no FEP tubing required and less than 150 reactors, or even 10 for that matter. I promise. The ChemArtificer has some ideas how to do it. Interested? Contact us at Or feel free to comment.