Nuclear power is kinetically trapped

Note: The views expressed here are my own bad ideas and do not reflect the views of Powerhouse.

Diamond and graphite are cool because they are both 100% made up of carbon atoms, and yet are totally different in almost every characteristic. Thermodynamically, graphite is the most stable allotrope of carbon, but diamonds are incredibly stable¹ because they are “kinetically trapped.” Why am I talking about this esoteric chemistry thing? Because it relates to a toy model I use in my brain to think about all sorts of business and I want to test it out on you. A short digression…

In chemistry, a reaction coordinate is this idea that chemical reactions follow a thermochemical potential² as a molecule transforms from (say) diamond, to (say) graphite. This is like Chemistry 101, but I’m almost certainly getting it wrong because I never properly learned chemistry. The model looks like the figure³ below, where everyone wants to “roll downhill” but is often blocked by an activation barrier. In this case, the activation barrier for diamond is the incredible bond strength between all those carbon atoms (and pressure? I dunno). So, if you heat this diamond up enough, you can get over the activation barrier and zoom downhill to the thermodynamically stable configuration, graphite.⁴ 

So how does this relate to nuclear power? Well, this toy model is nice because it explains why some ideas succeed and others fail. You might want to deploy the superior technology (graphite), but you’re kinetically trapped in the current state (diamond). The good news is that all sorts of tools exist to get you out of the well. The simplest one is money (energy in this metaphor). You take money, that gives you enough energy to make your product more affordable, you get over the activation barrier, you win. Then there’s more tricky stuff like finding a niche customer with a high willingness to pay that lets you grow with your customers. I like to think of that as an intermediate state or going around the barrier. Or maybe you benefit from some kind of policy (say a feed-in tariff) that lowers the barrier to adoption (we might call this a catalyst or an enzyme, which is a catalyst made out of proteins). The goal here is to get to the superior configuration. And you know it’s superior because graphite is thermodynamically favored — nature wants graphite.

That’s great, you might say, but what do I do with all that? Well I have two thoughts here. One relates to learning cycles (kinetics) and the other relates to the competition (intermediate states).

The Transition State

I sort of feel like there’s a bifurcated situation⁵ where we either end up with generation being dominated by variable wind and solar or dominated by firm power (SMR/geothermal/fusion/etc.). It just doesn’t seem like the variable renewables and firm power are as friendly as we’d like. I’ve read the Jesse Jenkins work on how flexible firm power can complement variable power, but I have to admit that I don’t grok the theory. Like, there’s no denying that cheap wind and solar are robbing nukes and other firm power sources of revenue. Maybe they’re pushing that revenue into high cost hours like the evening ramp or calm winter nights, but is it all commensurate? Is an equivalent of money on offer in those high cost hours to make up for the super low-cost hours? If not, the revenue those plants can earn will become a moving target that gets lower and lower as more renewables are deployed.

So now we’re headed towards a transition state that is uniquely suited to storage, transmission, and flexible natural gas. Batteries and LDES love that cheap power, and each one will be tailored for daily, multi-day, or seasonal discharge cycles. Even if nuclear might ultimately have a lower cost, it doesn’t today and that is putting them further and further behind. Meanwhile, the market for battery storage is in maximum rocket-ship mode, although it remains to be seen if/when a market emerges for LDES. Similarly, increasing wind and solar will create huge spatial price disparities that incentivize transmission (or maybe storage). And flexible gas generators with low capex will hang about for those peak hours where there isn’t enough storage and transmission (hopefully we can add some CCUS). 

The Kinetics

The learning cycles of nuclear power development are among the worst I’ve ever encountered. When I was at Alta Devices⁶ we would obsess over learning cycles. The fact that it took weeks to learn the result of an experiment meant we only got like 20 experiments per year! The key reason why SpaceX succeeded is that it was willing to fail, which turned multi-year learning cycles into multi-month learning cycles. Wind turbines, photovoltaics, batteries, and electrochemical cells all have a multi-week learning cycle, at worst. And of course, software has a multi-day (if not sub-hourly) learning cycle which causes no end of envy in us hardware nerds.⁷

Jamie Beard described this when explaining why funding for one geothermal project was not enough: “You need two or three or five projects before you’re optimized and scalable.” That’s hard for geothermal, but maybe unacceptable when the potential outcome of a bad project is a nuclear meltdown. A few weeks ago Tim Hade at Scale Microgrids noted that his policy was “move fast and never break things” If he’s being that careful with just some PV modules and batteries, what does that mean about nuclear power?

Reasons I’m wrong

A few ideas, although I’m sure folks have lots of ideas for why I’m wrong.

Folks today are pretty concerned that wind and solar might stall out, maybe for lack of land or interconnection. This is going – here’s an Ezra Klein quote from the Princeton Net Zero America project. “A plausible path to decarbonization, modeled by researchers at Princeton, sees wind and solar using up to 590,000 square kilometers — which is roughly equal to the land mass of Connecticut, Illinois, Indiana, Kentucky, Massachusetts, Ohio, Rhode Island and Tennessee put together.” I tend to be less worried about this but energy transitions have stalled before!

As all nuclear engineers should know, superheated water is also a kinetically trapped state. You can heat water up over 100C and if it’s pure enough, it just won’t boil. But then you poke it and… fwoosh.⁸ Maybe that’s what can happen with nukes (but in a good way!). France now hopes to build up to 14 new reactors by 2050, China is planning to build 150 nuclear reactors by 2035, and competition between Westinghouse and KHNP is heating up.

Interestingly, there’s a flavor of this reaction coordinate theory that applies to protein folding, Anfinsen’s dogma, also known as the thermodynamic hypothesis. The claim is that proteins will naturally fold into the lowest free energy or most thermodynamically stable configuration (e.g. diamond will become graphite). I like the analogy of our energy system as a really complicated protein trying to figure out how to fold itself properly.⁹ Note there’s a key qualifier about kinetics though: “the path in the free energy surface from the unfolded to the folded state must be reasonably smooth or, in other words, that the folding of the chain must not involve highly complex changes in the shape.” Going from high renewables to nukes sure feels like a complex change in shape!

What to do?

In order to move quickly, nuclear needs to be able to make mistakes. This is when folks bring up small modular reactors, but I’m not sure they’re thinking small enough. They need to be able to blow stuff up like SpaceX. But no matter how small you go, I think we’re always going to be (rightly) unhappy about fission reactors blowing up, so it seems like fusion is (surprisingly) the best option. On that front, I kind of like what Avalanche Energy Designs is working on. It seems small enough that they can tolerate a few failures, but what do I know? 

I want to qualify all of this by saying that I really do see nuclear power as key to a longer term vision to achieve energy abundance and become an interstellar species. I’m just not confident it’ll get us there in time to meet our climate goals. A lot of super smart people are working on nuclear power and I’m hopeful they’ll figure it out, but I’m not counting on it for the climate. 

Notes

1. Great memes in that one.
2. A.K.A. “energy”.
3. Not to be confused with the Hype Cycle, popularized by Shayle Kann.
4. I’m assuming there’s no oxygen involved here.
5. Some people call this path dependency.
6. For more on Alta, see this amazing post mortem by Ben Lenail.
7. The short learning cycles of software has meant that all sorts of problems that might be better solved by hardware got solved by software first because it just got there faster. Which is mostly okay IMO.
8. You can also supercool water (wow the Veritasium guy looks so young).
9. AlphaFold has made good progress solving proteins, but I don’t think they’ve solved energy systems yet.