Everyone would like to see something done about the world’s growing inventory of spent nuclear fuel. We at Flibe Energy have been working on this problem for many years, and recently were selected by the US Department of Energy for an award. At the heart of what we want to do is a salt that is discussed in a quartet of papers—two of which were based on research that Flibe Energy financially supported—that together sketch out something I find genuinely beautiful: a way to take used nuclear fuel apart, chemically and gently, and hand most of it back as feedstock instead of “waste.”

But first, that word. Waste.

The most expensive misnomer in nuclear energy

Open up a spent fuel pin and weigh out what’s inside. About 66% of it is uranium—barely touched, scarcely different from what went in. Another 25% is the zircaloy cladding, a perfectly good zirconium alloy whose only crime is having been near fission. Add the steel hardware and odds and ends, and you’re up past 95% of the mass before you’ve reached anything that earns the fear we attach to the whole package. The fission products—the genuinely hot, genuinely troublesome material, the stuff that actually makes a geologic repository a hard problem—come to about 2.4%. The transuranics are seven-tenths of one percent.

And we throw the whole pin into a canister and call all of it high-level waste.

That always struck me as not merely wasteful but unimaginative. The fission products are mixed into a much larger volume of benign material, and our entire strategy is to keep them mixed forever. The interesting question—the one ORNL was asking seriously in the 1960s before we collectively forgot how to ask it—is whether you can unmix them. And it turns out the key to unmixing is a salt.

Why fluoride, and why this one

Here is where I have to indulge my affection for fluoride chemistry, because the reason this works is the reason the molten-salt reactor worked, and it’s the same chemistry the whole field walked away from.

Everyone who tries to reprocess in molten salt reaches first for the chlorides—LiCl–KCl, the pyroprocessing eutectic Argonne developed for the Integral Fast Reactor. And zirconium misbehaves badly in chloride. It has three stable oxidation states there—Zr(I), Zr(II), Zr(IV)—so it won’t oxidize in a single clean step, and you can never quite say how much you’ve moved. Worse, ZrCl4 sublimes at a paltry 604 K, below the salt’s own melting point, and its solubility in LiCl–KCl tops out around 1.3 mol%. The zirconium literally wants to leave the room as vapor.

Now switch the anion. In fluoride, zirconium has exactly one oxidation state—Zr(IV), as ZrF4—-so every oxidation and every reduction is a single, four-electron, perfectly bookkeepable step. ZrF4 doesn’t sublime until 973 K. And instead of choking at one mole percent, you can dissolve thirty-seven mole percent of it. So you build the salt out of the thing you’re separating: LiF–NaF–ZrF4 at 26–37–37 mol%, a salt that melts at a friendly 436 °C and, as long as you stay under about 40 mol% ZrF4, keeps its vapor pressure politely low.

Someone gave it the name FLiNaZr, and I’m fond of it. It’s a cousin of the FLiBe that gave my company its name—and notably, it’s the beryllium-free cousin. That matters. Much of the reason the field reached for ZrF4-bearing salts in the first place is to get fluoride’s virtues without beryllium’s toxicity and the glovebox-and-permit apparatus beryllium demands. FLiNaZr is fluoride chemistry you can work with a little more freely.

Step one: strip the cladding, keep the zirconium

So you’ve got a salt that loves zirconium. The first thing you do with it is the most elegant.

Brenton Davis and Jinsuo Zhang at Virginia Tech—work Flibe Energy funded, with cladding supplied by Westinghouse—took pieces of zircaloy, made them the anode in FLiNaZr, and ran a current. The zirconium oxidizes off the cladding, crosses the salt as Zr(IV), and plates back out as zirconium metal on the cathode. The fission products clinging to the cladding stay behind. They measured the current efficiency at better than 90%—in fact the runs came in between 97.7% and 104.7%, which is to say essentially all of the charge went into moving zirconium and nothing else. They drove samples all the way to an open circuit, meaning every last bit of zircaloy in the salt had dissolved. And when they checked the recovered metal and the salt for any buildup of the other alloying elements, they found none.

A group at Toulouse—Massot, Gibilaro, Quaranta, Serp, and Chamelot, whom Flibe engaged to look at exactly this question—pushed the same idea further and made it an honest electrorefining process. Zircaloy-4 anode, graphite cathode, LiF–NaF–ZrF4 at 750 °C. They showed that zirconium dissolves at a less noble potential than the tin, iron, chromium, niobium, cobalt, and nickel it’s alloyed with, so by holding the anode potential you can dissolve the Zr selectively and leave the impurities sitting in the anode. The zirconium they plated out was 99.9% pure—no other metal detectable by EDS—at anodic and cathodic faradic efficiencies right around 98%. They could take the anode to 74% dissolution before they had to stop to keep impurities out of the bath.

Sit with what that means. The 25% of the used-fuel mass that we currently entomb as high-level waste—the cladding—comes back out as commercial-purity zirconium metal. That one step alone takes a quarter of the “waste” off the books.

Step two: let the salt eat the fuel

Now the part that made me grin.

Once the cladding is gone and you’re left with the UO₂ fuel sitting in a bath that’s already 37 mol% ZrF4, you don’t need to reach for the harsh fluorinating gases to dissolve it. The ZrF4 is already a fluorinating agent. Zirconium forms an extraordinarily stable oxide, and ZrF4 is more than happy to trade its fluorine for somebody else’s oxygen:

> ZrF4 + UO₂ → ZrO₂ + UF4 (ΔG° = –35.8 kJ at 873 K)

The reaction runs downhill on its own. Davis and Zhang measured the rates—for UO₂, and for the lanthanide and yttrium oxides that stand in for fission products (Y₂O₃ was actually the fastest)—across 773 to 923 K, and showed the salt converts the oxide fuel into dissolved fluorides without a molecule of fluorine gas. For the stubborn transuranic oxides like NpO₂ and AmO₂ that won’t go on their own, a little zirconium metal in the bath acts as a reducing agent and pulls them along too.

And here’s the economy of it. The classical fluoride-volatility route to UF₆—the process ORNL pioneered, the process that is the chemical heart of any fluoride-fueled reactor—wants you to drive uranium all the way from UO2 to UF6 with expensive, ferociously corrosive F2 or NF3. But if the cheap ZrF4 has already fluorinated UO₂ to UF4, then your aggressive gas only has to add the last two fluorines, UF4 → UF6, and the volatile uranium hexafluoride boils right out of the salt. The plutonium, the rest of the transuranics, and the bulk of the fission products stay behind in the melt, where reductive extraction or electrochemistry can sort them—and where the transuranics become fuel for the next reactor rather than a disposal problem.

This is the old Oak Ridge fluoride-volatility instinct, modernized and made gentler by letting a benign salt do the heavy lifting. It is exactly the kind of institutional knowledge I worry we let slip away when the molten-salt program was shut down.

Knowing the salt itself

None of this is hand-waving chemistry; people have done the hard fundamental work of understanding FLiNaZr as a material. Mathieu Salanne and colleagues, back in 2009, married first-principles simulation to real conductivity measurements and showed how ZrF4 stitches the melt together—fluorine ions bridging between zirconiums, forming fluorozirconate complexes and chains that slow the melt down as you add more ZrF4. Rajni Chahal, Shubhojit Banerjee, and Stephen Lam at UMass Lowell carried that further with ab-initio molecular dynamics, mapping the [ZrF6]2-, [ZrF7]3-, and [ZrF8]4- complexes, the network formation, the densities and heat capacities and viscosities—the boring-sounding numbers that are the difference between a chalkboard idea and a real flowsheet you can pump through a pipe.

We are, in other words, no longer guessing about this salt. We know its structure, its transport, its thermophysics, and now—from the Virginia Tech and Toulouse work—its electrochemistry and its reaction kinetics against the actual oxides in used fuel.

Key takeaways

String the four papers together and you get a single, coherent process, and it’s hard for me not to be excited by it:

1. Declad the fuel electrochemically and recover the zirconium as 99.9%-pure metal—a quarter of the “waste,” gone.
2. Dissolve the oxide fuel by letting the ZrF4 already in the salt fluorinate it, cheaply, with no fluorine gas.
3. Volatilize the uranium as UF6 with only a whisper of NF3 or F2 to finish the job.
4. Leave the transuranics and fission products in the salt, separable, with the actinides bound for the next reactor instead of a canister.

What’s left to actually call waste at the end is close to that 2.4% we started with—the fission products, and only the fission products. Everything else is product.

This is the part of nuclear energy I keep coming back to. The reactor gets the magazine covers, but the chemistry—the quiet work of fluoride salts and electrode potentials and Gibbs free energies—is where the fuel cycle is won or lost. It’s the part Oak Ridge understood deeply and the part the rest of the industry spent fifty years forgetting. Watching a salt like FLiNaZr take a used fuel pin apart, piece by piece, and hand most of it back as something useful, feels less like a new invention than like remembering something we always knew how to do.

We just had to pick the right salt.

—

Sources: B. Davis & J. Zhang, “Electrochemical decladding of used nuclear fuel,” Progress in Nuclear Energy 173 (2024) 105256. B. Davis & J. Zhang, “Fluorination of UO2, La2O3, and Y2O3 using ZrF4,” Journal of Nuclear Materials 592 (2024) 154971. R. Chahal, S. Banerjee & S. T. Lam, “Short- to Intermediate-Range Structure, Transport, and Thermophysical Properties of LiF–NaF–ZrF4 Molten Salts,” Frontiers in Physics 10 (2022) 830468. M. Salanne et al., “Transport in molten LiF–NaF–ZrF4 mixtures,” Journal of Fluorine Chemistry 130 (2009) 61–66. L. Massot et al., “Zirconium Metal Recovery from Zircaloy in Molten Fluorides,” Materials Sciences and Applications 15 (2024) 559–571.