BRENNEN

Industries / Mining and materials / Mining and quarrying / Rare earth processing

Separation

Separation turns a concentrate, a mix of chemically near-identical elements, into a pure oxide or salt of each one, usually by passing liquids through dozens or hundreds of mixing stages.

4

Milestones achieved, from 1956 to 2023.

3

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1956

    Separation by ion exchange

    At the Ames Laboratory in Iowa, Frank Spedding and J. E. Powell separated neighboring rare earths by ion exchange: the mixed elements are held on a column of resin and washed through with a chemical such as EDTA, which grips each element slightly differently, so they come out one after another. Their 1956 report describes columns 30 inches across loaded with the rare earths from 1,000 pounds of xenotime concentrate, and says the elements "are becoming commercially available at reasonable prices". Spedding founded the laboratory in 1947 and led it until 1968.

    osti.govahf.nuclearmuseum.org

    Credit

    Frank Spedding and J. E. Powell, Ames Laboratory of the US Atomic Energy Commission

  2. 1965

    Solvent extraction for color TV

    In 1965 Molybdenum Corporation of America advertised a "new process" at its Mountain Pass mine that made phosphor-grade europium oxide, the source of red in color television screens, along with cerium and lanthanum. The process was solvent extraction, in which the elements pass between a water solution and an oily liquid that pulls some of them more strongly than others. Europium-doped yttrium salts lit the colors of new TV sets that year, by the account of the Science History Institute, which holds the advertisement. By about 1995 the plant drew at least 12 separate rare earth products from one ore, by the USGS's count.

    sciencehistory.orgpubs.usgs.gov

    Credit

    Molybdenum Corporation of America (later Molycorp)

  3. 1975

    Theory replaces trial runs

    Xu Guangxian, a chemist at Peking University, took on a defense task in 1972 to separate praseodymium from neodymium, and in 1974 ran his method in an industrial trial at a Baotou rare earth plant at 99.9% purity, by his own account in a 2014 interview. In 1975 he presented a theory of cascade extraction, a way to calculate the flows and number of stages a separation needs, and in 1976 taught it for a month to about 100 engineers at the Shanghai Yuelong Chemical Plant. By 1986 Yuelong could design a cascade straight from calculation, where before it took months of trials with 200 to 300 shaken glass funnels. His theory was "widely adopted in industrial practice", according to Shanghai Jiao Tong University, and he received China's State Preeminent Science and Technology Award.

    LessonXu said the method was never patented, and that in 1980 he described it openly at an international conference to engineers from France's Rhodia plant.

    castjournals.cast.org.cnscce.sjtu.edu.cn

    Credit

    Xu Guangxian, Peking University; Shanghai Yuelong Chemical Plant

  4. 2023

    Know-how kept at home

    On 21 December 2023 China's Ministry of Commerce added the making of rare earth magnets to its catalog of technologies banned from export. Technology to extract and separate rare earths was already banned, Reuters reported, and Fastmarkets reported that the catalog covers extraction and separation, including the leaching of ionic clays. Reuters put China's share of the world's refined output at nearly 90% and wrote that Western companies, MP Materials among them, had struggled to run solvent extraction.

    LessonA method that its author says was shared freely in 1980 had become, decades later, a technology China would not export.

    mining.comfastmarkets.com

    Credit

    Ministry of Commerce of China

Watch for

  1. Heavy rare earths in Malaysia

    Lynas Rare Earths said on 16 May 2025 that a new circuit at its Malaysian plant had made dysprosium oxide to specification, making it "the world's only commercial producer of separated heavy rare earth products outside China", with terbium oxide to follow in June. Dysprosium goes into high-performance magnets for electric vehicles and wind turbines. In October 2025 an Australian producer announced plans for a further plant in Malaysia to separate terbium, dysprosium and lutetium, by the USGS's account.

    Achieved whena buyer or an outside body such as the USGS reports regular deliveries of separated dysprosium and terbium oxide from the plant.

    miningweekly.commagneticsmag.compubs.usgs.gov

    Claimed by

    Lynas Rare Earths (Amanda Lacaze, chief executive), May 2025

  2. Heavy circuit at Mountain Pass

    MP Materials said in November 2025 that it would begin commissioning a heavy rare earth separation plant at Mountain Pass in mid-2026, taking about 3,000 metric tons of feed a year, with a dysprosium and terbium circuit rated at 200 metric tons a year. In May 2026 it said commissioning would begin in the second quarter, with terbium and dysprosium production later in the year. The USGS reports that the US Department of War lent the company $150 million for the plant in August 2025.

    Achieved whenMP reports separated dysprosium and terbium oxide shipped to buyers and an outside body confirms the output.

    sec.govfool.compubs.usgs.gov

    Claimed by

    MP Materials, November 2025 and May 2026

  3. A new extraction machine

    Ucore Rare Metals says its RapidSX solvent extraction machines made neodymium-praseodymium oxide samples, shipped to magnet makers in June 2026, and 99.9% dysprosium oxide in July, both at its demonstration plant in Kingston, Ontario. The US Department of War has funded the work, most recently with $4.6 million to add samarium and gadolinium. Commissioning of the first commercial machine, a 118-stage unit for 600 metric tons of rare earth oxides a year at its Louisiana site, is now expected in the first half of 2027.

    Achieved whenbuyers qualify the product and the Louisiana plant runs at its stated rate, confirmed by someone outside Ucore.

    metaltechnews.com

    Claimed by

    Ucore Rare Metals, June to September 2026

Open

  1. 01

    Score models against pilot data

    A public benchmark that takes published stage-by-stage measurements from pilot or plant solvent extraction cascades and scores how well the open models predict them, element by element and stage by stage.

    The brief
    Credit · unclaimed
  2. 02

    A ledger of firsts

    A public ledger of every claim since 2020 to have separated a rare earth oxide outside China: who, which element, what purity, how much, when, and what outside evidence backs it, such as a buyer's statement, a government award or audit, or a USGS report. It is updated every quarter.

    The brief
    Credit · unclaimed

Open 01 · the brief

Score models against pilot data

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public benchmark that takes published stage-by-stage measurements from pilot or plant solvent extraction cascades and scores how well the open models predict them, element by element and stage by stage.

Why it is still open

Single models have been checked against single campaigns: Kevin Lyon, Vivek Utgikar and Mitchell Greenhalgh's 2017 MATLAB model of an Idaho National Laboratory flowsheet matched steady-state profiles in laboratory mixer-settlers, and a thesis at the Bandung Institute of Technology compared software for modeling a pilot circuit at SGS's Lakefield plant in Canada. The open tools now available report no such test. PrOMMiS has mixer-settler models whose code says the extraction correlations come from one project report and that some rare earths lack "adequate data". A teaching notebook on GitHub simulates cascades for six rare earths, and REEL-QUEST and a second project are turning published extraction results into databases. A Crossref search for this brief found pilot papers with mixer-settler tests, but no shared benchmark that scores several models against several campaigns with public input files.

inl.elsevierpure.comdigilib.itb.ac.idgithub.comgithub.comgithub.comgithub.com

Built with what exists

PrOMMiS and the IDAES platform; the marimo notebook above; REEL-QUEST's extraction data; published pilot campaigns, for example a 2022 paper on mixer-settler tests of heavy rare earth separation in Separation and Purification Technology and a 2018 CIM Journal paper on running a rare earth separation pilot plant.

  1. First step

    Pick one published mixer-settler campaign that reports concentrations at each stage and reproduce it in PrOMMiS.

  2. Done looks like

    At least three published multi-stage campaigns reproduced in at least two open models, with the error for each element in each product stream posted and all input files public.

Open 02 · the brief

A ledger of firsts

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public ledger of every claim since 2020 to have separated a rare earth oxide outside China: who, which element, what purity, how much, when, and what outside evidence backs it, such as a buyer's statement, a government award or audit, or a USGS report. It is updated every quarter.

Why it is still open

The claims come from the companies themselves, such as Lynas's dysprosium and Ucore's 99.9% dysprosium. The USGS's February 2026 summary says several US companies made small quantities of heavy rare earth compounds and metals in 2025 but none made "sustained commercial-scale quantities", and it names none of them. Searches for this brief found news lists of projects and capacities, but no public record that sorts the claims by their evidence.

pubs.usgs.gov

Built with what exists

company releases and quarterly reports; US Department of War and Department of Energy award notices; USGS Mineral Commodity Summaries; a language model to pull claims out of filings, each checked against its page by a person.

  1. First step

    Enter the Lynas, MP Materials and Ucore claims listed above, then add each company's earlier "first" announcements.

  2. Done looks like

    At least 25 claims from at least 10 companies, each with its source and an evidence class (company only, buyer confirmed, government or USGS confirmed), and one quarterly update published.