Being a reactive hazards person, I try to keep up on the reports posted by the US Chemical Safety Board (CSB). In my view, the CSB does exemplary work in root cause analysis of what are often very complex events leading to disaster. I wholeheartedly recommend that people in the process side of chemistry peruse the many reports and videos posted on the CSB website.
The development of any technology in the real world involves what I refer to as
“the discovery of new failure modes”.
While it is possible to anticipate many kinds of failure modes, it often happens that plant operations will present the opportunity to line up the planets in a particular way that was left out of the failure analysis.
A recent account from the CSB is the report on the T2 Laboratories accident in late 2007 in Jacksonville, FL. This accident killed 4 employees and injured 32 in many of the adjacent businesses. The explosive yield was estimated by the CSB investigators to be equivalent to 1450 lbs of TNT.
What is most instructive about this incident is the extent to which the thermokinetic behaviour was unknown to the owner/operators. This accident illustrates that thermal decomposition modes leading to runaway can happen despite a large number of successful runs.
I won’t go into too much detail since the report itself should be read by those interested in such things. But the upshot is that the reactor contents (MeCp dimer, Na, and diglyme) accelerated to a temperature that lead to the exothermic reaction of sodium metal and solvent diglyme. The reaction contents accelerated, raising the temperature and pressure to the rupture disk yield pressure of 400 psi. However, the acceleration was far too energetic for this safety device. The vessel exploded, hurling fire and fragments off the site. Just prior to the explosion, the owner/engineer directed the operators to leave the control room, saying prophetically, “there is going to be a fire”.
While the owners did perform some process development and did have the used vessel professionally inspected, what was left out was a study of the aptitude of the reaction to self-heat into a runaway condition. The company rightly anticipated the exothermicity of the sodium reaction with MeCp monomer and in fact, relied on the exotherm to raise the rxn temperature to a level where the economics would be more favorable. But what nobody at T2 anticipated was the runaway potential of the reaction of the sodium with the diglyme. No doubt they thought that the cooling jacket would prevent temperature excursions leading to a runaway.
The various glymes are often chosen as reaction solvents owing to their diether character as well as their high boiling points. Troublesome compounds or reactions requiring a polar solvent can be dissolved at high temperature and reacted in this high boiler. In certain cases, reactions can be run in a glyme and the product conveniently distilled out of the reaction mixture. Perhaps this is what they were doing in the MCMT process, I don’t know. This level of detail was not provided in the report.
Lamentations on Chemistry 
(MeCp dimer, Na, and diglyme) and HYDROGEN byproduct when making MeCp(-). Hydrogen can leach carbon out of steel at high temps as methane. Hydrogen embrittlement is a bad thing overall. Hydrogen in air is unforgiving. They should have had a 2X containment sump above the reactor. “8^>)
Yeah, there were all kinds of problems with the situation, not the least of which was hubris.
I read the report last Sunday. Monday morning, I come into work, the operators are adding bricks of sodium into a 500 gal glass lined reactor with molten “starting material” to make the sodium salt (+ hydrogen). There must have been too much air in the reactor, because—boom. Fortunately, the 6″ manway was only clamped loosely with C-clamps, this blew open and relieved the pressure (along with an impressive spout of fire). No injuries, but some re-thinking (grounding the contents, oxygen monitor in the reactor, adding the sodium as a liquid via a pump to minimize reactor opening, etc…). We were very lucky….
How very interesting. Glad to hear that there were no injuries. Did they add the Na under N2 purge? Solids addition to a reactor is always problematic. We use purged solids-addition funnels for Mg. Not sure what we use with Na. I’ll find out.
yes the reactor was purged…but only a 1/4″ N2 line for a 500 gal reactor (!). I mentioned this to the engineer, so maybe they will change it. They are going with a 2 valve purged addition chamber for the solid Na. Also, adding less Na per addition.
I have a few hair raising stories about Na that I have to keep to myself. It really is a treacherous metal. A brick of Na is really just brick of electrons lookin’ for trouble.
The purged addition chamber is a great idea, even if it does take a bit longer to use.
What about “no backup cooling” doesn’t make one shiver? (Handloading Na blocks doesn’t sound all that fun or smart, either.)
I would have figured (such as BuLi/THF) that the behavior of alkali metals with ether solvents might have made someone at T2 think about that as a possible failure mode. Is that reasonable?
The process was a batch reaction- all of the reagents and all of the potential energy were in the vessel. The exotherm was managed with addition of water to a jacket that relied largely on the heat of water evaporation to carry away the dQ as steam. Procedures defining threshold temperatures calling for more water constitute an administrative control. In earlier runs, the temperature was maintained with the existing engineering controls in the water feed system.
But the question that was not taken seriously enough was, what if the cooling system fails in the off condition? The absence of a backup system or even a better primary cooling system was the result of choices made by the owners. They felt they had a one-pot process they could control with minimal capital investment. They were wrong.
The deal with sodium is this- you have to handle it in some way. Na bricks can be handled and charged into a reactor safely, if the contingencies are thought out. Fact is, Na is a geat source of easily available and cheap electrons for covalent bond making. A retired friend of mine used to make commercial quantities of Ph4Si using PhNa and SiCl4. It was used as a water repellant for canvas. They quit only when the market dried up.
T2 was clever in that they cracked the MeCp dimer in the presence of Na, which promptly deprotonated the MeCp and generated H2 which was vented, avoiding accumulation of MeCp monomer.
T2 was not clever in failing to recognize how Na and diglyme represented a large accumulation of energy. I have no doubt the owners could have talked for hours about how they had the system under control, all the while failing to have recognized the thermally accessable failure mode just around the corner.
The choice of a 400 psi rupture disk is puzzling. As suggested in the report, a 75 psi disk might have prevented the runaway by venting energy at a lower temperature. I know from experience that decompositions having a high temperature onset may be predisposed to rapid decomposition/deflagration.
There is always a bit of hubris in these kind of disasters.