Monday, February 23, 2015

Worthwhile safety warning on nitrogen in plant-scale Grignard quenches

...Performing a “kill reaction” or a quench of a reactive metal at the bench or at scale is always problematic and requires the skill and close attention of the process chemists and operators. I guess what I’d like to pass on is that nitrogen is not an innocent spectator in the presence of finely divided, activated magnesium. Humid nitrogen can support a combustion reaction to produce nitrided magnesium once preheated to an onset temperature. 
If you mean to kill any reactive residues, it is important to apply the quenching agent in such a manner that the heat generated can be readily absorbed in the quenching medium itself. A good example of a quenching agent is water. Often a reactive must be killed slowly due to gas generation or some particular. Adding a quenching agent to a solution or slurry by slow feed or titration may be your best bet. If you have another vessel available, a feed to a chilled quenching agent will also work.  Dribs and drabs of water on a neat reactive material will lead to hotspots that may be incendive.
Huh, worth considering.  


  1. Lithium and magnesium are two metals that react with nitrogen easily. (The nitrides are dark colored). When you are doing finicky Grignard that is hard to initiate, it is preferable to to pre-activate the dry Mg turnings by dry stirring overnight under Ar plus few small drops of Br2 added to create bromine vapors for the dry stir, before adding ether and substrate. This almost never fails - unless you use nitrogen instead of argon...

  2. The most exciting quench I had to perform was a 300 gal reduction with large excess of Vitride. The quench solution was 10% aqueous NaOH and the reaction medium was toluene, so the beast liked to foam from evolving H2. th'Gaussling has the right idea to slowly add water and my safety solution was to calculate the free reactor volume to limit the feed rate. I figured that the foam is not going to run away if the theoretical volume of evolved H2 would be lower than the free reactor volume.

    The kicker was that water makes some 45 L of gas at 25C per each 18 mL of water, and I had a diaphragm feed pump with about 50 mL stroke volume. I tried to work with two pump clicks and wait for the quench to subside. This worked about a dozen of times and then the air valve got stuck a bit in the open position.

    After picking up some (non-GMP) product from the floor we happily repeated the batch process six more times. Oh, the joys of working for a CRO....

    1. On large scale with LAH and Vitride I would first pre-quench with dropwise EtOAc or acetone addition: the produced Al-alkoxides do not foam or sludge and you can take care of the exotherm while the mix is still reasonably stirrable. Only later you would follow with the standard aqueous NaOH quench. The generated hydrogen volume is much smaller and the mix is less temperamental if you pre-quench in this way.

    2. Hi milkshake,
      Thanks for the idea and I wish I could have used it. This was an unusual process where a tetraester was reduced to a tetraol. We did explore using ethyl acetate and encountered slow phase separation after the quench. With the caustic quench of Vitride there is a limited amount of time to drain the heavy salt layer before it turns solid. Think of a concrete mixer truck stuck in traffic. Next, in the presence of ethanol the follow-up CO2 mop-up of Al wouldn't produce filtrable precipitate while the caustic quench gave fluffy powder of carbonates.

      Also, the actual quench had to be started above 105 C or some form of the product complex with Al would fall out of the solution and turn the reaction mixture into a classic Bingham plastic fluid. The reactor had decent Rushton turbine, but I would need a full anchor to recover from that. On the safety front, at 105 C most of AcOEt resided in the overheads and occasionally burped into the reaction mixture. This produced spectacular visuals even on 5 L scale. I didn't try butyl acetate or anything similar.

    3. this is absolutely terrifying. Maybe MEK quench could work in your case.

    4. Maybe MEK, maybe something heavier. I will keep this in mind.
      The process itself wasn't that terrifying at all! Think about it. The chemistry never failed, we never cemented a reactor, and the only actual failure was mechanical.
      I never properly studied the safety of this process, so it is hard to draw definite conclusions from my experience. I was fresh out of school and pretty much clueless and on my own (OK, I had a helper to move those 600lb drums of 50% caustic).

      Later, I had some diverse experience both in process discovery and in safety engineering. I observed that when a process in development gives consistently high yields and purity of the product there are few concerns to be found about its safety. Conversely, when there purity concerns on scale-up there are safety issues to be found. A well developed process is both stable and reliable.

      What I really needed were tighter mechanical controls and wider margins of safety (free reactor volume, pump shutdowns etc.).

      P.S. I have a confession to make. That moment when the air valve got stuck open and I heard that pump click "50 mL, 50 mL, 50 mL" I was terrified. I was sitting on a 300 gal rocket with near-boiling toluene with a hydrogen booster (think Falcon Heavy), my Tyvek was melting to the reactor dome and one look down the sight glass convinced me that I must have just dropped a few Mentos in that Coke bottle....

    5. "We are going to have a fire" were the last words of the owner of T2 Laboratories, uttered in the control room of a runaway 2500L reactor after the cooling failed

    6. Yup, that CSB video is a good one to watch. That office footage is better than any CGI could do. It is good to remember that the T2 process was inherently unstable. Any loss of cooling must have led to in incident. This was a single factor incident - only one failure was required for a major disaster.

      In the case of a Vitride reduction any single failure was unlikely to lead to a major incident. Even with my pump bump less than 1% of the cold reaction mixture was ejected (more like, oozed out) and Vitride is not pyrophoric.

      I am not sure that would the case with LAH reduction. I have never evaluated one, but from literature LAH is pyrophoric and decomposes with evolution of H2 and pyrophoric Al at about 150 C. The decomposition can be catalyzed.

      Bottom line, my imagination was racing when I saw that foam coming up and later I had a really nasty cleanup to do. In reality, at least one other failure would have to occur at the same time to cause some major consequence.

  3. Still scary as shit. That's just a bit close to failure potential for my liking. Mitigations would be in play for any "what if" problems that could arise.