Thermal Runaway

Tuesday, September 22nd, 2009

If the revolver has enough chambers, you may have no idea you’re playing Russian Roulette. Derek Lowe takes a look at the terrible chemical accident at T2 Laboratories in Florida back in 2007:

Debris ended up over a mile from the site, and killed four employees, including one of the co-owners, who was fifty feet away from the reactor at the time.

The company was preparing a gasoline additive, methylcyclopentadienyl manganese tricarbonyl (MCMT). To readers outside the field, that sounds like an awful mouthful of a name, but organic chemists will look it over and say “OK, halfway like ferrocene, manganese instead of iron, methyl group on the ring, three CO groups on the other side of the metal. Hmmm. What went wrong with that one?”

Well, the same sort of thing that can go wrong with a lot of reactions, large and small: a thermal runaway. That’s always a possibility when a reaction gives off waste heat while it’s running (that’s called an exothermic reaction, and some are, some aren’t — it depends on the energy balance of the bonds being broken versus the bonds being made, among other things). Heating chemical reactions almost invariably speeds them up, naturally, so the heat given off by such a reaction can make it go faster, which makes it give off even more heat, which makes it… well, now you know why it’s called a runaway reaction.

On the small scales where I’ve spent my career, the usual consequence of this is that whatever’s fitted on the top of the flask blows off, and the contents geyser out all over the fume hood. One generally doesn’t tightly seal the top of a reaction flask, not unless one knows exactly what one is doing, so there’s usually a stopper or rubber seal that gives way. I’ve walked back into my lab, looked at the floor in front of my hood, and wondered “Who on earth left a glass condenser on my floor?”, until I walked over to have a look and realized where it came from (and, um, who left it there).

But on a large scale, well, things are always different. For one thing, it’s just plain larger. There’s more energy involved. And heat transfer is a major concern on scale, because while it’s easy to cool off a 25-milliliter flask, where none of the contents are more than a centimeter from the outside wall, cooling off a 2500-gallon reactor is something else again. Needless to say, you’re not going to be able to pick it up quickly and stick it into 25,000 gallons of ice water, and even that wouldn’t do nearly as much good as you might think. The center of that reactor is a long way from the walls, and cooling those walls down can only do so much — stirring is a major concern on these scales, too.

What’s worth emphasizing is that this explosion occurred on the one hundred seventy-fifth time that T2 had run this reaction. No doubt they thought they had everything well under control — have any of you ever run the same reaction a hundred and seventy-five times in a row? But what they didn’t know was crucial: the operators had only undergraduate degrees, and the CSB report concludes that the didn’t realize that they were walking on the edge of disaster the whole time. As it turns out, the MCMT chemistry was mildly exothermic. But if the reaction got above the normal production temperature (177°C), a very exothermic side reaction kicked in. Have I mentioned that the chemistry involved was a stirred molten-sodium reaction? Yep, methylcyclopentadiene dimer, cracking to monomer, metallating with the sodium and releasing hydrogen gas. This was run in diglyme, and if the temperature went up above 199°C, the sodium would start reacting energetically with the solvent.

Experienced chemists and engineers will recognize that setup for what it is: a black-bordered invitation to disaster. Apparently the T2 chemists had experienced a few close calls in the past, without fully realizing the extent of the problem. On the morning of the explosion, the water cooling line experienced some sort of blockage, and there was (fatally) no backup cooling system in place. Ten minutes later, everything went up. In retrospect, the only thing to do when the cooling went out would have been to run for it and cover as much ground as possible in the ten minutes left, but that’s not a decision that anyone usually makes.

Leave a Reply