Skip to main content

Chemical News

Birch Reactions Without the Ammonia

The Birch reduction – there’s an old-school synthetic transformation from you. I thought that when I first did one in 1983, so it must be even more so now, right? You condense liquid ammonia and dissolve a reactive metal in it (sodium or lithium are the usual), giving you a rather unexpected blue solution. That color (weirdly) is due to solvated free electrons, which means that this is one powerful reducing agent. The Birch is famous for being able to knock aromatic rings back down to cycloalkenes, and there are (still) not too many ways to do that.

This new paper (from a group at China Agricultural University) describes an ammonia-free (and indeed, amine-free) variant of the reaction – you use commercial sodium dispersion with a crown ether in isopropanol, and you can do it all at zero degrees C, as opposed to the -30 or so you have (at most) in liquid ammonia. Not having to break out the ammonia tank does sort of lower the barrier to running the reaction, it’s true, although I always regarded that as a sign that one was about to do some serious Birching. And condensing ammonia always seems like a magic trick, as the clear liquid just appears on the dry ice cold finger and drips down into the flask. You get that same something-out-of-nothing effect when you do a distillation, of course, but I think it’s the lack of visibility of the source in the case of the ammonia tank that gives it more of a flourish. And of all the gases that I’ve seen condensed on a cold finger, ammonia is definitely the one that I’d rather work with (as opposed to liquified HCN, for example, which I’ve seen once and have no desire to encounter again).

Another advantage to the Birch is that you typically just let the reaction warm up to get rid of all the ammonia, letting you switch to whatever solvent system you want for the workup. In this case, the reaction is done fairly concentrated (more so than the ammonia ones), and the authors just do an ether/brine extraction without evaporating the isopropanol away. The advantage of that is that the crown ether actually goes into the brine layer, and can be recovered by a second extraction with dichloromethane. This recovery helps, but still is unlikely to make this method of industrial interest (no one wants dichloromethane on that scale). But for bench-scale reactions, this should be pretty convenient. It looks like this reaction does pretty much what the classic Birch does. Of course, sodium in an alcohol solvent is yet another old-school reaction (the Bouveault-Blanc reduction, in the case of esters), and if you don’t add the crown ether, that’s all you’re going to get. But complexing the sodium ions changes the mechanism (outer-sphere versus inner-sphere), and that gives you Birch products.

One thing I wonder is if you can make sodium amide with this system, if you add some catalytic ferric salt at some point in there. That prep starts out just like a Birch, but the iron sends it down a different path. If so, I might have been able to spare myself a painful episode (and I mean that literally) when I first starting doing this sort of chemistry. I’m just glad that there was someone else in the lab to unwind me from the apparatus, is all I can say.

58 comments on “Birch Reactions Without the Ammonia”

  1. A Nonny Mouse says:

    Been done many years ago using Na/silica

    PS My son popped home from university last night and had a grin on his face; seems that he had found “Things I won’t work with”. Also liked your style.

    1. Derek Lowe says:

      The paper does reference these methods, fortunately.

      1. A Nonny Mouse says:

        The good thing about the Na/silica is that it is air stable and can be weighed out accurately. I never like these dispersions of fine metal.

        I have used the Na/silica for a diester coupling and it works very well, but so does Na/xylene if you can get the Na to stay as fine particles (once shook a few hundred grams of Na in 120C xylene by hand….. That was pre-health and safety). The major difference is in the work up in that the Na/Silica can be filtered whereas the Na has to be carefully quenched….

        1. d says:

          My undergrad years were spent making and researching these Na-Silica Gel products (among others). The most interesting to work with was definitely the Cesium variants. Pretty cool to see someone else that has used them in real applications. Funny coincidence!

          1. milkshake says:

            Please can you comment more on the Cs variant? I was reading that the ternary Na-K-Cs liquid alloy stays liquid down to -78C and can generate a stable radical anion from benzene, and only the Cs fraction of the alloy gets transferred into the Cs-benzenide

          2. A Nonny Mouse says:

            A colleague knew the sales person who wanted to get the market going and he offered us a kilo for a very low price, but the shipping to the UK was painfully expensive so we just bought a small amount to try.

            We expected the product to take off as in was an intermediate in “the next big thing”, but very little has happened since and I have 100g sitting here waiting for something to happen…

  2. Project Osprey says:

    Derek, I actually used this make sodium amide on an industrial scale at one point. Styrene is plenty cheap, as is ammonia and you get a lot of moles of lithium per Kg. Lithium amide on the other hand is somewhat more expensive but perhaps more importantly seems to age on standing – no matter how carefully you store it. Fresh material gives much better results, which is important if you’re doing a late stage Claisen condensation on something quite expensive.

    1. Derek Lowe says:

      Never worked with lithium amide, but I have encountered aged sodium amide – as it goes orange and (I’m told) becomes somewhat explosive. . .

      1. Project Osprey says:

        Our problem was that it always assayed as being fine even when it looked and smelt wrong (it seemed to give off ammonia, which has low odor threshold – we weren’t intentionally working by ‘sniff-test’). Obvious supplier problems ensued.

        The major barrier on plant is cooling, -50°C is a challenge for most chillers in the lab but at scale its unachievable. The only way is liquid nitrogen, which is more expensive that you might think and also sends most chemical engineers into meltdown.

        1. Cymantrene says:

          We use a Cumulus cryostate from Linde for low temperatures. It cools down methanol to -90 °C with liquid nitrogen. We can keep -80 °C in a 1 m^3 Hastelloy even during reaction (lithiation, Grignard and so on)

  3. Stephen Byrne says:

    I’m quite sceptical of “solvated free electrons”. How exactly is an electron “solvated”? If electrons are in fact “solvated”, how do they give rise to the observed blue colour?

    1. hn says:

      We regularly see blue solvated electrons in synchrotron experiments.

  4. Some idiot says:

    Other things you don’t want to see dripping from your cold finger condenser on a rotovap:

    Some decades ago I made a lot of diazoketones through reaction between an activated acid and diazomethane (yes, I have made serious litres of the stuff…). The product is yellow, and you remove the excess diazomethane by letting it evaporate in the fumehood (adding acetic acid was not a good idea as you also got reduced yields of your diazoketone). After workup I was rotovaping it down on the bench (all our rotovaps were on the bench… those were the days…!) when I noticed that the distillate was bright canary yellow… I yelled to everyone to get out, put my face shield and gloves on, and very, very carefully took the receiver off and put it in the fumehood, and added acetic acid. Then went back to the rotovap, took my flask off, and very carefully washed the whole thing down with acetic acid solution… Never again, thanks, but now at least I know what diazomethane smells like…

    1. Fluorine chemist says:

      I used the cold finger to condense bromotrifluoroethylene, chlorotrifluoroethylene and 1,1,2,3,3,3-hexafluoropropene. The first two were condensed into a stirred mixture of DMF and activated zinc to generate the vinyl zinc reagents; the third was condensed into ether followed by various transformations to yield (E) and (Z) 1-iodo-1,2,3,3,3-pentafluoropropene.

  5. Dick Friary says:


    Were you trying to purify the ammonia by condensing it on a cold-finger? In the mid 1960s my colleagues and I used lithium in liquid ammonia to reduce an alpha, beta-unsaturated ketone to the lithium enolate of the corresponding saturated ketone. We produced liquid ammonia not by distillation but by inverting the ammonia tank and drawing the liquid from the bottom. Our reductions succeeded.

  6. Steve Gonzales says:

    In graduate school, I had a dissolving metal reduction as an early step in a pretty long sequence. The largest one I ran was with ~2L of liquid ammonia in a 5L flask. You had to watch the cooling bath like a hawk lest the ammonia start to boil and blow the flask to pieces, which thankfully I never saw. My heart rate still goes up thinking about it.

  7. milkshake says:

    This paper is not entirely practical, compared with the good old Birch (Li/tBuOH/THF/liq. NH3) that is easy to set up and not very moisture sensitive. For crude reductions with Na-liq. NH3 like global benzyl protecting group cleavage, some people do not even bother with condensing liq. NH3 and pour it directly liquid from a baby-sized tank into a wide-mouth Erlenmeyer, then throw some Na until the stuff is blue, no moisture exclusion, no cooling (ammonia high heat of vaporisation takes care of that).

    Now this new protocol requires the use of a somewhat expensive 15-crown-5 (the cheapest source I could find is Oakwood for 120USD/100g) that has MW= 220 and has to be used in large excess (6 to 9 equivalents) in the reaction. Just having to get rid of it, with my 1 g of product in 10g of crown, would give me a pause.

  8. a. nonymaus says:

    It seems like they’re dismissing Benkeser too easily. Calcium/ethylenediamine/BuNH2/THF/t-BuOH gives selectivity on par with what the present article reports. It turns out the key is to add sand as an abrasive:

  9. Mister B. says:

    From Derek “no one wants dichloromethane on that scale”

    Why ?

    Low boiling point, easy to purify again, to remove from reaction mixture and to dry… Recycling solvants is not “green enough” for the green chem’ people ?

    1. CMCguy says:

      Its my understanding that it is chiefly that low volatility that makes CH2Cl2 less tolerable at scale with the main hazard being suffocation without adequate ventilation but also very difficult to prevent all inadvertent releases into the atmosphere (often experienced off gas when open bottle in warm lab). The molecules short term and long-term hazards are know to chemists to be less severe that its siblings, CHCl3 and CCl4, but gets painted with the same brush as far as environmental exposure risks. Not sure if remains this way but there used to be a number of industrial processes “grandfathered” to still allow CH2Cl2 use, as long as could show evidence that alternative solvents unsuccessful.

      In terms of recycling in Pharma almost any solvents, even though selectively practiced, its often found to be easier and cheaper to start fresh, especially if outsourcing to Chindia. Seems to be more talk (with publicity motives) than action to recover and reuse. I trust other chemical industries do implement more consistently.

      1. Wheels17 says:

        Methylene chloride is routinely used on the ton scale for the production of cellulose triacetate films, formerly used for photography, and now used for 4 layers on most LCD displays. Disposal is not an issue, as it is recovered and reused.

        Our process used refrigeration down to -100 degrees F in a recirculating air loop, with a draw to a carbon recovery to bring most of the process slightly below atmospheric pressure. Hundreds of millions of pounds a year were circulated back when photo film was big.

        Try for more background, starting on page 5 of the .pdf.

  10. Calvin says:

    Ahhh the Birch. Never done one, but brings back memories of my undergrad. We were in a teaching lab doing our practicals while a grad student in one of the research labs was doing a Birch. Finding that the valve was sticky he somehow propped the tank (it was a biggie too; no idea how he did it) upside down and “tapped” the valve with a hammer……

    The whole thing sheared off and started to flood the place with ammonia (crappy fumehoods didn’t help). The whole department was evacuated. It was then that I realized that it was a good idea to make sure that all my valuables (like keys) were in my possession at all times (I had them but that was purely accidental) as we were evacuated….. Some people didn’t get back in until the next day so had to stay with friends.

  11. Dominic Ryan says:

    “…as opposed to liquified HCN, for example, which I’ve seen once and have no desire to encounter again…”
    A long time ago in grad school we used to run reactions in liquid H2S. Perhaps that does not sound as bad as HCN until you look up the toxicity, about 1000x HCN and the same mechanism. The difference of course is that your nose can detect H2S at 100000x lower concentration.

    1. Derek Lowe says:

      Absolutely no urge to experience that one, either.

    2. Design Monkey says:

      > liquid H2S. Perhaps that does not sound as bad as HCN until you look up the toxicity, about 1000x HCN

      Ewwww. That is just freaking misleading lies. Toxicity of H2S is roughly on the same level as HCN. And nowhere near of your invented “1000x” more.

      1. Dominic Ryan says:

        Thank you for poking me here. I was recirculating some data I recalled from quite a long time ago having to do with the Kd of CN-, CO, H2S (perhaps HS-) to heme. The order of binding placed H2S as much tighter than CN. That was not too surprising to me given the binding to Fe. Unfortunately I cannot put my hand on references for that.
        Having said that, I recognize that I have not been keeping up with that literature and the role of H2S in both toxicity and endogenous signalling is much more complex than assumed years ago.
        I welcome correction or divergent opinion, that is science. I object to being tarred with ‘misleading lies’. To spit out such a characterization actually says more about you than me, especially when behind a pseudonym, and is counter productive. Rather, please cite leading literature that might help illuminate the situation. I just came across this reference for example:
        I don’t know how reputable this journal is but a quick scan looks interesting.
        I would love to learn about an authoritative paper that points to CN and H2S having the same toxicity. Could you cite that please?

        1. milkshake says:

          H2S being as toxic as HCN is a bunch of crock and ridiculous baloney invented by trial lawyers suing by compensation – the H2S is at least an order of magnitude weaker poison than HCN. And while H2S can kill a lone worker quickly – by loss of consciousness in a poorly ventilated sewer for example – is not an insidious mitochondrial poison whose effect can build up gradually, like with HCN.

          In the old days, we had a Kipp apparatus on the bench in the primitive tiny highschool lab (without a fume hood) and we precipitated metal sulfides by bubbling a fast stream of H2S gas through it: It stank like hell so we opened the window, but nobody passed out and nobody was hurt. I cannot recommend doing this H2S gas work of course, it was not safe, but imagine what would happen if you tried to release a quick stream of HCN gas in unventilated highschool lab

          1. Dominic Ryan says:

            Please cite literature around the relative toxicities, that sort of experience does not provide real evidence. I genuinely would like to be updated.

          2. Chester says:

            HCN has a 10 min AEGL-3 of 27 ppm, while that for H2S is 76 ppm; 1 hr values are 15 and 50 ppm respectively. Quite a bit less toxic but not quite an order of magnitude difference.

          3. milkshake says:

            AEGL-3 is the concentration when someone can suddenly collapse and die without help. H2S poisoning is reversible with fresh air and resuscitation whereas HCN poisoning needs a rapid administration of nitrite+thiosulfate antidote, and the recovery is quite problematic

            By the way, the safe way to test whether you can smell cyanide (some people can’t) is to take off a cap from bottle of solid, and carefully sniff at the cap liner (not the bottle content!). HCN traces produced by hydrolysis have unpleasant pyridine-like bitter burnt acrid smell

    3. A Nonny Mouse says:

      Once was doing an azide displacement reaction and wondered what the volatile liquid in the condenser was; forgot about the acid in the molecule and this was hydrazoic acid distilling. Pretty big scale, but escaped without damage. Used the ester after that.

  12. Retro sinner says:

    The point about DCM volatility is correct, it can be recovered but you can lose a lot with standard water chilled systems. The environmental impact is also a concern and the volatility doesn’t help with that.

    What should worry chemists more than it does is the reactivity. 3y amines, pyridines, thiolates… all react with DCM and the first displacement generally accelerates the second. The presence of a spurious NMR singlet at about 5.5 ppm integrating to 2 hydrogens in the ‘correct product’ is a giveaway, it might not be residual solvent. Sometimes with 3y amines, it stops at the first substitution but you wouldn’t know because no one looks for it – until the presence of a potential mutagen is uncovered by an inquisitive analyst.

    It can be a slow reaction but I saw a colleague experience untested DCM washing of a filter cake by a contractor that gave over 20 per cent conversion – ouch! Even as recently as 2010, the reaction of pyridines with DCM could get you a paper in J. Org. Chem (

    My view is that every process chemist learns this eventually, the secret is in doing it as early as possible in your career on as unimportant a reaction as possible. Mine was in a failed avenue of my PhD so not the worst timing.

    1. Derek Lowe says:

      Exactly – it’s more trouble than it’s worth, in reactivity, waste disposal, etc. No process chemist I know will touch the stuff unless their back is absolutely to the wall.

      1. David Edwards says:

        I gather you have other reasons to be averse to dichloromethane … “Chiracel column” springs to mind … 😁

    2. Mister B. says:

      I never thought about that ! Thank you very much for this great answer, the included link. With a keen interest in process chemistry, I keep a close eye on anything I can learn on this very topic !

      Thanks a lot !

    3. milkshake says:

      the funny thing, even carboxylates react with DCM, all it takes is a little base and a quaternary ammonium at reflux, and with unhindered aliphatic acids you get CH2(OCOR)2
      Worse yet, the students are not told nowadays that azides do not mix well with chlorinated solvents, several people blew up their rotovaps in Hruby lab at U of Arizona because Evans group published a method using tetramethylguanidinium azide in DCM, and the substrates in Hruby lab were more hindered and needed an extended reflux, and diazidomethane is volatile enough to be missed on TLC analysis, and it waits patiently in the reaction mix until you concentrate it on rotovap…

  13. Stephen says:

    I found that some birch reactions could be done in ethylamine at O C. More convenient in the lab than condensing ammonia.

    1. milkshake says:

      the trouble with anhydrous EtNH2 is that it is not as cheap. It invites curiosity from DEA. And because of the higher reaction temperature (at least in case of Li) you get an over-reduction to tetrahydro compounds because the produced EtNHLi is basic enough to cause isomerization of the produced 1,4-cyclohexadienes to conjugated 1,3-diene that is reduced to cyclohexene

  14. Anonymous says:

    I did my first Birch under the tutelage of a fantastic post-doc who called it The Ty-D-Bol Reaction; since then, so do I. 🙂 It worked like a charm.

    There is a difference of opinion on when to invert your NH3 tank to draw off liquid vs gas. Iron salts or iron (rust) particles may be entrained when drawing off liquid directly, so you might get an unexpected result from iron catalysis. If you can get away with liquid, go for it. I’ve been in labs with cylinder holders for the ~10″ diameter x 30″ tall (I forget my tank sizes) tanks. It supported the tank upright OR upside down, with plenty of room for a valve, so you can easily draw off liquid from a stable upside down tank. (NH3 is currently shipped in aluminum tanks, not steel tanks?)

    As I recall from Birch’s autobiography, “To See the Obvious” (1995), I think he said that Djerassi named it the Birch Reduction and helped to spread the word to others and that it definitely helped his career. Birch thought of it as the Wooster / Godfrey Reduction because he based his work on their precedents. I also think that Birch mentioned that some people couldn’t reproduce some early results and accused him of fraud or incompetence or other bad behaviors. It was subsequently determined that you need a trace of Na in your Li in order to get the reduction to go and that’s why you can buy Li + 0.5% Na from SIAL, et al. Do you need a trace of Li when you do a Na Birch? I don’t remember.

    The trick to make fine sodium particles by melting a chunk in boiling xylenes, stoppering the flask and giving some good shakes is described in Fieser, Organic Experiments as something to do in the undergrad lab. Better them than me.

    I looked up Ty-D-Bol on wikipedia and picked up this tidbit. In 1992, Ty-D-Bol sponsored a survey about Spring Cleaning that asked “1,006 American adults “if they had the power to throw out what exists and start all over again”, what would they choose? 49 percent picked the U.S. Congress, 23 percent the IRS tax code.” Those days, it might be said, do come again and again.

    1. milkshake says:

      A presence of trace amount of Na in Li metal is essential if you want to make BuLi from BuCl in hexane.

      I once worked on Birch reduction of dehydroabietane skeleton with Li/tBuOH/THF in refluxing liquid ammonia – it worked but it was painfully slow (36 hours to completion, even with huge excess of Li that formed liquid shiny Li bronze solution at the high concentration at the top layer, separating from the less Li rich deep blue layer bellow. I thought the trialkyl substituted electron rich benzene in the substrate was to blame for the slow rate but now I realize maybe I used too pure lithium metal

    2. A Nonny Mouse says:

      As I mentioned above, I shook a few hundred grams of sodium to make the sodium sand for an Org Syn reaction which I didn’t read correctly (cooled before the reaction had started then added too much to start it).

      Only fire I have ever had (in ether!) and was lucky to escape without damage as I walked away to get a fire extinguisher as it went up spraying sodium sand and ether which ignited. Gone out before I arrived with the extinguisher.

      Anyway, I now make it with a fast overhead in xylene at 120C. For some reason it clumps together on cooling if you use toluene

      1. milkshake says:

        clumping in toluene: It is probably just a surface tension and viscosity effect. Surface tension is what keeps the solidifying molten sodium particles nice and round, prevents them from sticking together. Surface tension and viscosity typically starts dropping rather fast as you get closer to the boiling point. Melting point of sodium is fairly close to toluene boiling point

    3. Greg says:

      An easy solution to this used to be buying the ammonia cylinder with an Eductor tube – I did that for multiple 12L reactions back in my kilolab days.

  15. Jose says:

    With the photoassisted Cristol-Firth-Hunsdiecker reaction, I had the twice-in-a-lifetime spookiest-thing-ever non-pleasure of watching dark red Br2 vapour refluxing in CCl4. Over a slurry of Hg salts. I still have nightmares….

    1. 10 Fingers says:

      I used to run this one all the time in grad school (double carboxylate-to-bromide or carboxylate-to-iodide reactions were a mainstay on the way to various tortured small organics). It can be made to work with Iodides as well, though iodosobenzene diacetate based transformations can be better. Good gentle fun, these.

  16. TIWWW fan says:

    “liquified HCN, for example, which I’ve seen once and have no desire to encounter again”

    Hopefully you can write about that in more depth soon?

    1. milkshake says:

      a colleague generated toluene solutions in situ from TMSCN and iPrOH. When cooled to -78C, HCN precipitates from toluene as beautiful needles

      1. Derek Lowe says:

        Well, live and learn! I had never heard of that one, nor of (re)crystallizing HCN from solution in general.

  17. Andy says:

    Whilst diazomethane is a hazard, it’s as important to be aware of the potential for generating diazidomethane, through exposure of DCM to azide during workup. As explosive (and volatile) as you’d imagine.

    1. Barry says:

      Bouveault Blanc reduction (sodium metal, liquid ammonia, ethanol) is pretty damned green chemistry. The Ammonia is biodegradable (also known as “fertilizer) and the side product is NaCl, Na2SO4, or NaOAc depending on how you quench it.
      And if you’re doing it on scale as Kowalski did for the Tagamet alcohol, you can distill the ammonia back and forth between two reactors for years.
      This is not the transformation in need of a “Green” alternative.

  18. loupgarous says:!divAbstract is the abstrtact for the actual paper, “Laboratory-scale photoredox catalysis using hydrated electrons sustainably generated with a single green laser” Robert Naumann, Christoph Kerziga and Martin Goez*

  19. Barry says:

    Stan Hall gave us a useful bit of technology whereby in one pot, one lithiates Ar-Br to Ar-Li (with high Na Lithium wire), adds that to an aldehyde and–by just condensing in NH3–achieves the reductive cleavage of the benzylic C-O bond. Overall, Ar-Br –>Ar-R with no transition metals.

  20. Daniel says:

    Glad you still have the humor and humility to mention that time you got all wound up.

    If you’re correct about that Iron route, this may well become “Bench Birch” while the ammonia tank is wheeled out for scaled up and industrial applications…

  21. Francis says:

    Used Na-SG (I) for the first time last Friday to do a desulfonation of a secondary amine. Must have been pretty humid here because even though I did everything under a funnel of dry nitrogen my new 5g bottle sparked and proceeded to quench itself. I got what I needed before that happened and the reaction worked, but man that was a quick way to waste the rest of the bottle.

Comments are closed.