Wednesday, December 11, 2013

Process Wednesday: a LiHMDS mystery

Peng et al. OPRD
In the midst of 3 very interesting papers [1] on the process development for the compound filibuvir, a fascinating comment from some Pfizer process chemists on Dieckmann cyclization (see figure to right) yield variation due to the source of lithium hexamethyldisilazide:
We also observed a source-dependence for the LHMDS, similar to that observed with the oxazolidinone substrate. (CJ's note: the authors had an oxazolidinone in place of the ethyl ester in structure 23 in a previous route.) With LHMDS prepared from n-BuLi and HN(TMS)2 the yield was consistently 88−90%. However, in the case of one supplier using a lithium metal process (Li metal, 2-methyl-1,3-butadiene, HN(TMS)2), a consistently lower yield of 81% was observed.
Similar observations with LDA prepared from n-BuLi vs Li metal/styrene have also been reported. In the course of investigating the cause of this variation, we identified two additional suppliers who used the lithium metal process, but with those sources the higher yield was consistently obtained (88−90%). We also found that if the “low olefin content” grade of LHMDS was used from the initial supplier, the yield improved to 85−87%. 
We studied several potential culprits, including the presence of residual 2-methyl-2-butene and small quantities of other metal contaminants (e.g., sodium amide bases, LiCl, and Li-alkoxides), but none of these accounted for the observed discrepancy. In all cases where the yield decreased, an increased level of elimination impurities was observed (i.e., the total mass balance was consistent). While these results continue to intrigue us, our successful identification of at least three viable commercial suppliers (Optima, BASF, and Chemetall “Low Olefin Content” LHMDS) for the reagent and the relatively modest yield variations have attenuated our concern with this unexplained LHMDS source variation.
Below, the proposed elimination pathways and accompanying text that were in the oxazolidinone route (which was covered in the first paper in the series):

Singer et al. OPRD 
The two elimination pathways are shown in Scheme 6. The first involves elimination of acetic acid from the cyclization substrate 24, to form a mixture of acrylamides 28. An E1CB mechanism via the lithium enolate is shown, but this could also proceed through an E2 elimination in the presence of a weaker base such as the lithiated oxazolidinone generated during the cyclization. The second pathway is a ring-opening elimination of the β-keto lactone product to generate a β-keto acid, which undergoes decarboxylation to generate a mixture of enone isomers 29. The top elimination pathway predominates in “normal addition” mode, i.e. addition of LHMDS to the substrate. We estimate the pKa values of the relevant protons at ∼25 for C3′ and ∼23 for C2. The use of a strong, hindered base such as LHMDS and the low reaction temperature (−20 °C) favor kinetic deprotonation at the more sterically accessible position (C3′). Nonetheless, direct enolization at C2 remains a possibility, and intramolecular proton transfer from the desired C3′ enolate to C2 is also feasible.
I don't have any good answers for their questions either. Why would the method of preparation of LiHMDS matter? What is in there that is promoting elimination? Why is potassium and lithium tert-butoxide much worse? Readers, any ideas?

1. (a) Singer, R.A.; Ragan, J.A.; Bowles, P.; Chisowa, E.; Conway, B.G.; Cordi, E.M.; Leeman, K.R.; Letendre, L.J.; Sieser, J.E.; Sluggett, G.W.; Stanchina, C.L.; Strohmeyer, H.; Blunt, J.; Taylor, S.; Byrne, C.; Lynch, D.; Mullane, S.; O’Sullivan, M.M.; Whelan, M. "Synthesis of Filibuvir. Part I. Diastereoselective Preparation of a β‑Hydroxy Alkynyl Oxazolidinone and Conversion to a 6,6-Disubstituted 2H‑Pyranone." Org. Process Res. Dev. ASAP (b) Peng, Z.; Ragan, J.A.; Colon-Cruz, R.; Conway, B.G.; Cordi, E.M.; Leeman, K.; Letendre, L.J.; Ping, L.-J.; Sieser, J.E.; Singer, R.A.; Sluggett, G.W.; Strohmeyer, H.; Vanderplas, B.C.; Blunt, J.; Mawby, N.; Meldrum, K.; Taylor, S. "Synthesis of Filibuvir. Part II. Second-Generation Synthesis of a 6,6-Disubstituted 2H‑Pyranone via Dieckmann Cyclization of a β‑Acetoxy Ester." Org. Process Res. Dev. ASAP (c) Ide, N.D.; Ragan, J.A.; Bellavance, G.; Brenek, S.J.; Cordi, E.M.; Jensen, G.O.; Jones, K.N.; LaFrance, D.; Leeman, K.R.;  Letendre, L.K.; Place, D.; Stanchina, C.L.; Sluggett, G.W.; Strohmeyer, H.; Blunt, J.; Meldrum, K.; Taylor, S.; Byrne, C.; Lynch, D.; Mullane, S.; O’Sullivan, M.M.; Whelan, M. "Synthesis of Filibuvir. Part III. Development of a Process for the Reductive Coupling of an Aldehyde and a β‑Keto-lactone" Org. Process Res. Dev. ASAP


  1. I would have figured LiCl, but that was obvious to them too. Maybe the alkene/Li method generates radicals via radical anion intermediates - if there are low-level metal impurities, organometallic radicals might promote side reactions (but there might not be enough around to do that). Why they would promote elmination I don't know - maybe radical H abstraction, electron transfer to regenerate the anion, and elimination from the internal anion would cause elimination.

    I haven't read the paper, but perhaps the shorter version is "I don't know".

  2. There is plausible explanation for both the isoprene detrimental role in LiN(TMS)2 preparation and poor results with LiOtBu. Tert-alkoxides for Dieckmann condensations and other similar carbonyl enolizations are typical reversible deprotonation systems that lead to thermodynamically equilibrated mixture of enolates.In this case the undesired enolate undergoes facile beta elimination so most of the material is channelled into the eliminates sideproduct. Whereas LiN(TMS)2 is both more basic and hindered and should provide kinetic control enolate on the less hindered alpha (acetate) that equilibrates much slower than it cyclizes to the desired product.

    Li + isoprene and R2NLi + isoprene are inherently dirty systems so the quality of the obtained LiN(TMS)2 was inherently poor. Isoprene provides with Li metal anion radical that is quite prone to dimerization to a dilithium species (that can further set polymerization of isoprene, especially if the used Li contains 1-2% of Na) . Other nasties in the mix include dimethylallyl lithium, which is apparently quite stable and can be silylated on carbon. Finally, R2NLi bases react readily with 2 equivs of butadiene to provide 1:2 telomeric adducts (neryl-substituted tertiary amines), especially upon heating. This process is useful for feedstock material in the Noyori asymmetric synthesis of (R)-citronellal, a precursor of synthetic menthol

    1. milkshake, you rock! (No surprise here, though.)

  3. The Collum group works with problems like this regularly, and from personal experience I can say that the source and purity of lithium amides makes a monumental difference in the behavior of a system. In LDA kinetics, trace impurities (< 1 mol% LiCl) can change the deaggregation equilibrium in such a way to either increase or decrease the rate of lithiation depending on which aggregate is favored for the reaction. While most of our work has focused on the LDA dynamics, it is not unlikely that there is some impurity in the system that promotes the aggregate or mixed aggregate needed for the elimination or the one needed for the cyclization, resulting in different impurity profiles depending on base source.

    How we get around this problem of impurities from commercial base is to synthesize our own base via lithium metal/isoprene and then recrystalize it. This method, if done with extraordinary care under rigorously air-free conditions, will give you lithium base with no detectable impurities. A write up is posted on our website (linked in name). It's best suited for bench top ~10 g scales, so it's not a remedy for Pfizer's problems, but for anyone in academia who runs across problems like this and truly needs an answer (and is willing to put in the time and specialized glassware) getting impurity free base is do-able.

    1. Ack - I meant to tweet this to your boss!

      Aggregation! Of course, why didn't I think of that?!?

      The authors did make reference to a crystallization, but I am not sure it's the same as what you're suggesting. From the first paper in the series:

      "Lower-yielding LHMDS batches could be remediated by crystallization of the LHMDS at low temperature and reformulation in fresh solvent."

  4. BuLi aggregates were a big deal in the kinetics of the cationic polymerization process of the polymer plant I used to work at. The position of the Li changed how many molecules formed a stable aggregate in solution.

  5. Are there good and reliable sources of commercial LDA these days? Many years ago I know one couldn't purchase stuff that was of sufficient quality for even most bench work much less for scale-up which is why people would make their own (and still fought variability).

    I worked on a project that went to kilo scales and found could not use 1N BuLi in hexanes as retarded and induced precipitates during the reaction. Although found MeLi or tBuLi worked the best in lab those could only be obtained in Et2O which incurred highly prohibitive approval nightmare that seemed almost redundant based on the nature of the other reagents involved hence we ultimately had to use 10M BuLi in pilot plant that brought big concerns for viscosity in transfers and the more extreme reactivity of any residuals to deal with.

    1. Since LDA is unstable in THF on long storage (few months) but it is so easy to make in situ from BuLi and iPr2NH which store well, (instant non-exothermic formation at 0C in THF) I think there is no reason for trying try to fix batch quality problem by finding a reliable LDA supplier.

    2. Non-exothermic? I never made LDA at 0C (always <-70) but my thermocouple said that it was very exothermic. It should be, given the pKa difference.

  6. I should mention that you can buy mid-concentration BuLi solutions in like 2.5M in hexane or 2M in cyclohexane. (the common 1.6M BuLi is a bit too diluted for large scale procedures). Or you could start with 10-11M BuLi in a small vessel and dilute it down with toluene just before use


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