David Barry 

A Greenpeace nuclear specialist (who didn’t want to disclose is name) on rare-earth oxide pollution, radioactive residues, and why the truth only arrives after commissioning.

September 2026 · Interview conducted by video · Edited for length and clarity

This autumn, the rare-earth oxide business returns to Western Europe in earnest. At La Rochelle, Solvay’s separation line for dysprosium and terbium — the metals that keep magnets working when motors run hot — is scheduled to start up, fed in part by monazite concentrates imported from Brazil. At Lacq, Carester’s Caremag facility, a €216 million project part-financed by Japan and part-staked by an American processor, is being commissioned. In the Allier, the EMILI lithium project has cleared its public inquiry under a government that has made speed a doctrine. Every press release in the folder uses the same vocabulary: clean, circular, best available techniques.

Rare earths being what they are, the pollution question is never purely chemical — it is also radiological. So we put it to a nuclear specialist who has spent more than two decades at Greenpeace working on radioactive waste, discharge and contamination issues. The conversation ranged from mineral chemistry to the difference between a permit and a measurement — and to the one claim about “clean” rare-earth oxide production that can never be made in advance.

Q · Rare earths are marketed as the metals of the clean-energy transition. Why is a nuclear campaigner the one worried about them?

Because the physics comes free with the ore. The minerals that carry rare earths — monazite, xenotime, bastnäsite — carry thorium and uranium with them. In some monazites the thorium oxide content runs to double digits by weight. That is not contamination; that is mineralogy. When you crack those ores open with concentrated acids and separate out the oxides, the radioactivity does not disappear — it redistributes. It concentrates in pipe scales, in dusts caught on filters, in the residue streams that leave the plant as cakes and slurries.

The industry’s term of art is “NORM” — naturally occurring radioactive material. Regulators sometimes use the more honest phrase: technologically enhanced. You dig the material up dilute, and you hand it back concentrated. Once you see that, you understand why the oxide line of a rare-earth plant is a radiological-protection question as much as a chemical one — and why someone from the nuclear side of the house has opinions about it.

Q · How much of that risk is theoretical? What does the actual operating record show?

The record is long, and it is not the record of one careless country. Baotou, in Inner Mongolia, processes a large share of the world’s light rare earths; its tailings lake — kilometres across, growing for decades — has left surrounding groundwater contaminated with fluoride and heavy metals, and villages that no longer trust their wells. At Kuantan, in Malaysia, an Australian separation plant accumulated more than a million tonnes of low-level radioactive residue next to a populated coast; an IAEA peer review in 2011 recommended options that included shipping residues back to Australia; it became a national election issue, and the episode ended with regulators forcing the dirtiest processing stages out of the country. Mountain Pass, in the California desert, logged dozens of pipeline failures and leaked hundreds of thousands of gallons of radioactive wastewater on its way to corporate bankruptcy.

Europe has its own memory, which people conveniently forget: Sillamäe, in Estonia, where decades of processing left one of the continent’s largest radioactive tailings repositories standing on the shore of the Baltic. My point is not that operators are villains. My point is that this is what the physics and the chemistry do when they meet cost pressure and ordinary industrial life — everywhere, regardless of the flag on the plant.

Q · The new projects — including the French ones — insist they will be different: clean production, closed loops, zero discharge. Do you accept that?

I accept that they believe it. And I will tell you exactly how much it is worth.

The pollution risk of rare-earth oxide production is real, and it is considerable — that part is not controversial among people who have read the ledger. And even where a project is presented as fully clean — and some are presented exactly that way — the truth can only be established once the plant is running. Not in the permit file. Not in the brochure. In operation.

“Even where a project is presented as fully clean, the truth can only be established once the plant is running. The plant itself is the experiment.”

Understand the epistemics, and you understand why I say that. An environmental permit is a comparison between a model and a threshold: the company models its discharges, the regulator checks the model, the permit issues. A plant is not a model. It is a measurement — taken every day, under feedstock that varies, at scales the pilot line never saw. A demonstration unit processes kilograms; an industrial line processes tonnes per hour, and the residue streams scale with it. The monazite concentrate from one Brazilian lode is not the monazite from the next; thorium content moves with geology, batch by batch. And industrial life includes upsets: filter failures, maintenance shortcuts, the storm that exceeds the pond. None of this tells you the claims are dishonest. It tells you the claims are hypotheses — and that the experiment which tests them is called commissioning.

Q · Why can’t regulators simply verify the claims in advance?

Because verification in advance is structurally impossible — and everyone in the system quietly knows it. Assessments assume nominal operation: the plant as designed, on its best behaviour. Monitoring regimes, in many jurisdictions, are self-reported, averaged over months, and published late. Even the classification of the waste moves under your feet: residues that sit below one country’s clearance thresholds are another country’s radioactive waste, and the international standards — IAEA, Euratom — are transposed unevenly across borders.

So the honest sentence buried in every permit file is this: these are predictions, not measurements. If you want measurements, you must wait for operations — by which time the capital is sunk, the contracts are signed, and momentum has taken over the decision. That asymmetry — commitment first, verification later — is the structural problem. It is not a French problem or a Swedish problem. It is built into the industry everywhere, which is precisely why the promises arrive before the data and the data arrives after the promises.

Q · What would taking the risk seriously actually look like, concretely?

Not cancellation. Vigilance, installed before first feed rather than after first incident. The recipe is not exotic:

●Baselines before startup — multi-year pre-operational monitoring of water, sediment and biota, published, so that change can be attributed when it happens.

●Live discharge data — continuous monitoring of effluent and emissions, public by default, in near-real time, not annual averages.

●Batch-level transparency — publication of thorium and uranium assays in incoming feedstock and outgoing residues, per shipment, per campaign.

●Independent community sampling — funded standing rights for local actors to take their own samples and have them analysed independently.

●Century-scale financial assurance — decommissioning and stewardship bonds sized to the real half-life of the problem and held outside the company’s balance sheet.

●Honest categories — a facility that runs primary concentrate from abroad is a separation plant with radiological tails, whatever the word “recycling” on the gate is doing for the optics.

Every one of those six items is standard practice in some other industry, in some other country. What is missing, almost everywhere, is the political will to ask all six of a sector that is currently strategic, fashionable, and in a hurry.

Q · Last question. What should a reader — or a policymaker — take away from all this?

Take the risk seriously before the data arrives, not after. When a brochure says “fully clean production,” read it as a claim awaiting evidence, not as a fact you may quote. The plant writes its true environmental statement every single day — in its discharge logs, its residue assays, its maintenance records. The only question is whether anyone is positioned to read them.

I will grant the industry one thing, because it is true: some of these plants may indeed run clean. But that is an outcome to be demonstrated, not a premise to be granted — and the demonstration has not started yet. A ribbon-cutting is a press release. The effluent pipe is the peer review.

And to your readers in Europe specifically: you are being asked to believe that this time the chemistry will behave differently because the flag on the roof has changed. That is a testable proposition. It will be tested — at La Rochelle, at Lacq, in the Allier. What worries me is not that the experiment will run. What worries me is that the instruments are not yet in place to notice the results.