Thursday, August 29, 2013

Reflecting on Yield, and the Need for a Qualitative Assessment of Reactions

Here's some thoughts:

1. I categorize all reaction yields qualitatively. 
What does this mean? Generally speaking, I don't really believe in yields other than 0%, 25%, 50%, 75%, or quantitative. I also sometimes break this down as 1.) Doesn't work, 2.) Works a little, 3.) Goes halfsies, 4.) Pretty alright, 5.) The bomb diggity. Now, I might conform to the standards of society and report a 62% yield for a publication or presentation, but for the purposes of a practicing-academic-chemist-grad-student, my breakdown is good enough. I readily acknowledge that this view of looking at yields is generally flawed, but it gets the job done on a day-to-day basis.

2. I don't record yields on reactions less than 10 mg.
This isn't entirely true. If I've rigorously run a reaction, and then repeated it, I'll believe a yield on 10 mg scale. But generally speaking, there are too many ways for the yield to go wrong: trace solvent, water, silica, or dust? add 10-20%. Did any manipulation of the material? subtract 10-20%. If I do record a yield on this scale, its most likely going to be as defined in section 1, rounded to the nearest 25%. 

3. TLCs of reactions in SI
Why doesn't anyone do this? We record our TLCs in our notebooks and report Rf in the SI. But in my opinion, the TLC is the most important qualitative and rapid way to assess a reactions "goodness." If your reaction has an 75% yield, and an Rf = 0.40, but an inseparable side product at Rf = 0.37, then I hate your reaction (but I'll still use it if I need it). This is useful qualitative data! Far more useful than melting point. Believe it or not, you can optimize reactions on tiny scale by TLC alone. Its not ideal, but it definitely works when material is scarce.

4. Yield ranges on multiple scales
Reported at 95%! Awesome! Except, maybe that was the best yield you got out of 15 experiments, the reaction usually sits around 75% (and once you got a 30%). Maybe you know the reason for variabliliy (great! put it in the discussion), maybe you don't (thats ok too). Do good science: report yield ranges on multiple scales. If its not possible because its the front line of your 72-step linear synthesis, I forgive you. If its a methodology on simple starting materials, no excuses. 

5. Yield is not important.
I'm use hyperbole for effect. But honestly, my yield doesn't ever really matter, in total synthesis there are two quantities of material: enough and not enough. There is a need in the literature to deemphasize the importance of yield. Its been said before that emphasizing yield is bad. If it were me, I would do so in favor of honest reporting of pros and cons of reactions, with a discussion of more qualitative aspects: ease of running and purifying, cost of catalysts, etc.

Those are my thoughts. Have a discussion. Or not. Now if you'll excuse me, I've got some scale-up to get back to. 


  1. Dane:

    Some interesting thought-fodder here. Allow me to expand on some of your points.

    1. I, too, am a "qualitative yield" guy. I have even fewer benchmarks: traces, half, or full conversion. I only calculate out for publications.

    3. Ever since the Blog Syn days, I now take pictures of all my reactions and TLC plates, and tape them into the notebook next to the prep. Other chemists have responded favorably, as they did when you guys did this for your ingenol paper.

    4. I have never reported a "single experiment" yield. They're all averages of 2-3 reactions, which is as it should be. Someone (Greg Fu?) actually reports averages with (n= x) for the number of times it was run.

    5. The only place yield DOES matter is Process Development. There, purity, quality, and number (%) really have an impact on the bottom line.

    1. We have scanners in lab to photocopy our TLC plates, I'd say most people do it, and keep them in their notebooks. With as cheap ($ and space) as digital information is, people should be including photos and TLCs in their SI.

      I agree completely about process chem., I had a sentence to that effect in my original post, but it must have gotten lost while editing.

  2. Definitely important points - consistency of reactions, especially on scale up or down are fundamental to the application of methodology in synthesis. If you want to develop useful chemistry that people will/want to employ, it doesn't have to be brilliant i.e. spot to spot with quantitative yield, it just has to "do what it says on the tin". The responsibility lies with the authors to clarify such issues, making it as easy as possible for other to use the chemistry. After all, isn't that a major reason for developing methods? For further naive and overly optimistic ramblings on this issue you can see here

  3. Karl,
    "A straightforward solution would be the inclusion of this additional information in a ‘hints and tips’ section for example, within the supporting information of a report."

    I like this concept a lot. It would be interesting to see how it works in practice. Org. Syn. procedure footnotes have this type of stuff already, and everyone loves Org. Syn.

    1. There have been a few cases when I have wanted to include more detail about an impportant synthetic procedure than I was able to in the experimental section of a paper. In such cases I have published reactions at ChemSpider SyntheticPages . CSSP is a really nice forum for synthetic chemists who want to share their work simply for the sake of spreading knowledge and not for boosting their ego or reputation. As a result I think procedures and results are reported more honestly than in peer reviewed journals.

  4. Regarding number 2,
    I would like to add that on screening reactions with a few mgs of material, it's actually very helpful to use internal standard to quantify (or semi-quantify) your yields, either with NMR or GC

    1. I would also add that when you use an internal standard, if possible, use one that is a solid to make sure your amount added is known as accurately as possible on small scale. Also, if you can, try to integrate protons of the same type (sp3 vs sp2, etc) so that relaxation differences are mitigated. If you can't do that, make sure to amend your NMR delay so that the integration is more accurate.