Getting back to the question of the oil-cooled nuclear reactor in Pinawa, Manitoba: my assignment was to get a new monitoring system installed on the leak detection system. The fifty-seven pressure tubes of the reactor were each surrounded by a larger tube, and these outer tubes were continuously purged with CO2 gas. The leak detection system was based on monitoring the purge gas for trace hydrocarbons, which would indicate that a pressure tube was leaking. To pinpoint the location of any possible leaks, the 57 tubes were separately routed through a system of solenoid valves to a monitoring station.
To the great chagrin of the reactor operators, the detection system indicated all kinds of leaks! At least a quarter of the channels showed hydrocarbon readings well into the tens of parts per million. This was a very serious matter.
Until someone got the bright idea of just letting the multiplexer valve sit on a single channel for a while. It turned out that after ten minutes the reading would go back down to zero. You could clearly see it on the strip chart recorder which left a telltale line of ink, one after another, for each channel in sequence. The high readings were obviously some kind of instrumentation glitch, a "surge" they called it. The true reading was the number showing after ten minutes purging a single channel. Problem solved.
And this was where things sat when I was given the assignment. I was most definitely not expected to get into the question of explaining the "surges": the system was working just fine. All we needed was some more modern equipment. Strip chart recorders were after all very very 1960's: this was the 80's and we were converting to computers for all our data monitoring. And one of the major benefits of computer data logging would be to get rid of those annoying surge readings. You would just program the computer to switch channels, wait ten minutes, and then log the reading only when it had had a chance to "settle down".
I still don't know what made me think of it, but it occurred to me that I might be able to explain the surges. What if there was some chemical reaction taking place in the sampling lines whereby hydrocarbons were being broken down to some other lighter species which were then going through the detector without showing their presence? Then, when the valve switched to the next sampling line, some unsampled gas would still be left sitting in the last tube. With 57 channels at 10 minutes each, it would be almost six hours before that tube would be sampled again. Maybe that was enough time for the reaction to reverse and the products be converted back into methane. That would explain the surge, and it would explain why the surge disappeared after ten minutes: once the fresh gas reached the detector, the remaining hydrocarbons would have been washed out of the system.
But what kind of reaction could be responsible for this wierd behavior. Taking methane as an example of a typical hydrocarbon, I wrote out:
CH4 + CO2 => ???
I did some trial and error and came up a couple of possible reactions. The one that seemed most interesting was:
CH4 + CO2 => 2CO + 2H2
If you haven't done chemistry for a while, you might want to note that the left and right hand sides of this equation each have two carbons, two oxygens, and four hydrogens. So it is indeed a balanced chemical equation. The question is: does this reaction actually take place?
There's a way to tell if a reaction is expected to take place or not, and it's something you learn in first year chemistry. It's called the Gibbs Free Energy and its a formula that combines the enthalpy and entropy of a reaction into a combined measure of spontaneity. In short, all things being equal, if the Gibbs Free Energy is negative, the reaction should go. If it's positive, then it won't.
Let's ignore for a moment just what is the physical meaning of enthalpy and entropy: the fact is you can look them up in the chemical handbooks and add them up. It's not that hard and what you find is that the Gibbs Free Energy for the reaction in question is decidedly positive. So my theory was wrong: thermodynamically speaking, the reaction shouldn't occur.
Was that correct? Maybe I'd made a mistake in the calculation. I went over it again and got the same result. Then, I noticed something: in the formula for Gibbs Free Energy, the entropy term is multiplied by the temperature. Out of habit I had used STP (Standard Temperature and Pressure) conditions, but of course the reactor ran at a coolant temperature of 300 degrees Celsius. Maybe this would make a difference?
Sure enough, it did. For this reaction, the internal energy of the molecules (the enthalpy) was definitely higher on the right hand side, which inhibited the reaction: but the entropy contribution was in the other direction; since it was multiplied by temperature, the reaction became more favorable the hotter it ran. This makes sense because entropy is a measure of disorder, and the products consist of four molecules while the reactants are only two. So at sufficiently high temperatures the entropy should dominate and the reaction procede.
I quickly redid the figures, and once again I was disappointed. Even with the corrected temperature, the additional contribution of the entropy term was still not sufficient to tilt the balance from positive to negative. The reaction was still a "no go."
And yet: the reaction does take place, and I was able to prove it by directly measuring carbon monoxide in a freely running sample line! The explanation of this mystery will follow in my next blog post.