Rare earth separation and processing (REE oxides and metals)

Executive summary

Rare-earth element (REE) processing is not primarily a mining problem; it is a separation problem. The 15 lanthanides plus scandium and yttrium have nearly identical chemistry because they differ mainly in the inner 4f electron shell. Cheisson and Schelter describe this as “Mendeleev’s bane” and note that adjacent lanthanides are extraordinarily difficult to separate, with solvent-extraction separation factors often near 1.5–2.5 for adjacent elements (Cheisson & Schelter 2019, DOI 10.1126/science.aau7628).

Economically, the key products are individual rare-earth oxides (REOs) such as neodymium oxide (Nd₂O₃), praseodymium oxide (Pr₆O₁₁), dysprosium oxide (Dy₂O₃), and terbium oxide (Tb₄O₇), and then reduced metals/alloys such as Nd-Pr metal for NdFeB permanent magnets. USGS data show that in 2023 China supplied roughly 68% of mined REE and an estimated 85% or more of refined/separated output (USGS Mineral Commodity Summaries 2024, Rare Earths, p. 132). The bottleneck is separation capacity more than ore availability.

This note focuses on:

• the physical chemistry that makes REE separation hard;
• conventional flowsheets: roasting, leaching, impurity removal, solvent extraction (SX), ion exchange (IX), precipitation, calcination, and metal reduction;
• granted patents and peer-reviewed literature;
• worked calculations for SX stage counts;
• business-model/unit economics, customers, and public-company exposure.

The core technical problem: adjacent-lanthanide separation

REE ores contain multiple lanthanides together. After mining and beneficiation, a mixed REE concentrate is produced. The industrial task is to split that mixed concentrate into saleable individual oxides at high purity, typically 99–99.99% REO for magnet, catalyst, and phosphor applications.

The difficulty is that adjacent lanthanides have:

• identical +3 oxidation state in most aqueous/acidic processing;
• very similar ionic radii, shrinking only gradually across the series because of the lanthanide contraction (Cheisson & Schelter 2019, DOI 10.1126/science.aau7628);
• similar hydration energies and complexation behavior.

The quantitative consequence is a low separation factor,

\[

\beta_{A/B}=\frac{D_A}{D_B},

\]

where \(D_A\) and \(D_B\) are distribution coefficients of elements A and B between organic and aqueous phases. For adjacent light lanthanides such as neodymium/praseodymium, \(\beta\) is commonly around 1.4–1.6 in industrial extractant systems such as EHEHPA/P507 in chloride media (Xie et al. 2014, Minerals Engineering 56:10–28, DOI 10.1016/j.mineng.2013.10.021). This is small; a separation factor near 1 means the two elements move almost identically.

The practical result is that industrial REE separation plants use hundreds of mixer-settler stages, countercurrent washing/scrubbing, reflux, and organic-phase recycle. A standard Nd-Pr separation is not a single extraction; it is a bank of stages with scrub and strip circuits.

Feedstocks and conventional flowsheet

The main feedstocks are:

• bastnaesite (REFCO₃), the primary US/Mountain Pass and Chinese Bayan Obo mineral;
• monazite (REEPO₄), rich in light REE but containing thorium and uranium;
• xenotime (YPO₄), richer in heavy REE;
• ion-adsorption clays, important for heavy REE in southern China;
• secondary/byproduct feeds such as phosphogypsum, coal ash, recycling feeds.

USGS records the US as having significant REE resources and production from Mountain Pass, but US separated-oxide capacity remains limited relative to China (USGS 2024). MP Materials operates Mountain Pass and ships bastnaesite concentrate; the company’s 10-K describes its transition from concentrate sales to separated NdPr production, with all the associated commissioning, ramp, and financing risks (MP Materials 10-K filings, SEC EDGAR).

A generic REE processing route is:

1. Beneficiation: crushing, grinding, flotation, magnetic/electrostatic separation to produce mineral concentrate (Jordens et al. 2013, DOI 10.1016/j.mineng.2012.10.017).
2. Cracking/roasting: acid bake with concentrated H₂SO₄ or alkaline roast with NaOH/Na₂CO₃ to convert refractory bastnaesite/monazite into leachable sulfate or hydroxide salts.
3. Leaching: water or acid dissolution of REE values.
4. Impurity removal: removal of Fe, Al, Th, U, and phosphate; thorium/uranium create radioactive waste obligations especially for monazite routes (Binnemans et al. 2013, DOI 10.1016/j.jclepro.2012.12.037).
5. Primary separation: solvent extraction or ion exchange to produce individual REE streams.
6. Precipitation and calcination: REE oxalates/carbonates are precipitated and calcined to individual REOs.
7. Metal reduction: light REE metals by molten-salt electrolysis; some heavy REE metals by metallothermic reduction.

3.1 Solvent extraction: the dominant separation technology

Solvent extraction remains the industrial workhorse for light/medium REE separation. The Binnemans review states that SX is the most applied industrial technique for REE separation and recycling, but also the source of large chemical inventories, solvent loss, and aqueous waste (Binnemans et al. 2013, DOI 10.1016/j.jclepro.2012.12.037). Xie et al. review SX chemistry and process design for REE systems, including extractants such as D2EHPA, EHEHPA/P507, Cyanex 272/923, and TBP (Xie et al. 2014, DOI 10.1016/j.mineng.2013.10.021).

A typical Nd-Pr SX circuit uses an acid organophosphorus extractant (e.g., P507/EHEHPA) in a kerosene diluent, with HCl or HNO₃ as aqueous phase. The separation factor \(\beta_{\mathrm{Nd/Pr}}\) is low, so the circuit requires:

• many extraction stages;
• a scrub section to remove co-extracted less-extractable elements;
• a strip section to recover the loaded organic; and
• organic-phase saponification/regeneration.

3.2 Ion exchange

Ion exchange was the original industrial method, especially the historical Ames Laboratory process for high-purity REE separation. Ion exchange remains relevant for small-volume high-purity heavy REE separation, but it has lower throughput and higher reagent cost than modern SX (Binnemans et al. 2013, DOI 10.1016/j.jclepro.2012.12.037).

3.3 Emerging separations

Academic and patent literature describes alternatives intended to cut stages or waste:

• ionic-liquid extraction and aqueous biphasic systems;
• ion-imprinted polymers and resins;
• membrane-assisted extraction;
• selective crystallization/precipitation;
• electrochemical and supercritical CO₂ routes.

A granted U.S. patent assigned to University of Kentucky Research Foundation, U.S. 9,206,492 B2, “Recovery of rare earth elements and compounds from coal ash,” granted 2015-12-08, claims an acid-leach and solvent-extraction/oxalate-precipitation route for REE from coal ash (Google Patents US9206492B2). This is a real granted patent and therefore evidence of recorded claims, but not evidence of commercial cost or scalability. The distinction matters: a patent can specify a process without demonstrating that the process is economic at commercial scale.

Worked calculation: SX recovery and rare-earth separation stages

4.1 Extraction recovery from a leach solution

For a simple non-selective extraction of total REE from aqueous leach solution, the fraction remaining in the raffinate after \(N\) equilibrium stages is:

\[

\text{Fraction remaining} = \left(\frac{1}{1+E}\right)^N,

\]

where the extraction factor is

\[

E = D \times \frac{O}{A}.

\]

Use the following inputs:

• distribution ratio \(D = 4\) for REE under the chosen pH/extractant conditions (assumed for illustration; actual value depends on pH, extractant loading, and aqueous composition);
• organic/aqueous phase ratio \(O/A = 1\);
• target extraction recovery \(R = 99.9\%\), so fraction remaining = 0.001.

Then:

\[

E = 4 \times 1 = 4,

\]

\[

N = \frac{\ln(0.001)}{\ln(1/(1+4))}

= \frac{-6.9078}{\ln(0.2)}

= \frac{-6.9078}{-1.6094}

= 4.29.

\]

Rounding up, 5 equilibrium extraction stages are needed for 99.9% total REE recovery under these assumed conditions. This is a rough extraction calculation, not an adjacent-element separation.

4.2 Adjacent-element separation: Nd/Pr

For a countercurrent fractional separation, the minimum number of theoretical stages at total reflux can be estimated with the Fenske equation:

\[

N_{\min} = \frac{

\ln\left[

\left(\frac{x_A}{1-x_A}\right)_{\text{extract}}

\left(\frac{1-x_B}{x_B}\right)_{\text{raffinate}}

\right]

}{\ln \beta_{A/B}}.

\]

Take Nd as component A and Pr as component B. Set product purities of 99% Nd in the extract and 99% Pr in the raffinate:

\[

\left(\frac{x_{\mathrm{Nd}}}{1-x_{\mathrm{Nd}}}\right)_{\text{extract}} = 99,

\]

\[

\left(\frac{x_{\mathrm{Pr}}}{1-x_{\mathrm{Pr}}}\right)_{\text{raffinate}} = 99.

\]

The numerator is:

\[

\ln(99 \times 99) = \ln(9801) = 9.190.

\]

With \(\beta_{\mathrm{Nd/Pr}} = 1.4\),

\[

N_{\min} = \frac{9.190}{\ln(1.4)}

= \frac{9.190}{0.3365}

= 27.3.

\]

So about 27 theoretical stages are needed at total reflux for 99% mutual purity with \(\beta=1.4\). At \(\beta=1.5\), \(N_{\min}\approx 22.7\) stages; at \(\beta=1.3\), \(N_{\min}\approx 35.1\) stages. Real plants add extraction, scrub, strip, and organic-regeneration stages plus reflux beyond the theoretical minimum, so actual Nd-Pr circuits are often several tens of stages or more (Xie et al. 2014, DOI 10.1016/j.mineng.2013.10.021; Binnemans et al. 2013, DOI 10.1016/j.jclepro.2012.12.037).

4.3 Sensitivity of Nd/Pr theoretical stages to \(\beta\)

This calculation shows why small apparent improvements in \(\beta\) are commercially valuable: moving \(\beta_{\mathrm{Nd/Pr}}\) from 1.4 to 1.5 reduces the theoretical minimum stage count by about 17%.

Parameter table

Patents and IP: what a granted patent does and does not prove

A granted patent is a legal document of claimed invention, not a pilot-plant data sheet, not a cost study, and not proof of commercial viability. The coal-ash patent above, U.S. 9,206,492 B2, shows that a US institution has recorded claims for REE recovery from coal ash (Google Patents US9206492B2). That supports the existence of disclosed process chemistry and an assignee’s effort to protect it. It does not establish that the process can compete with bastnaesite/monazite or Chinese SX economics.

The same rule applies to company R&D claims. MP Materials, for example, markets Mountain Pass as the only operating rare-earth mine and separation facility in the Western Hemisphere, but its SEC filings contain risk-factor language about limited separated-production history, ramp risks, supply-chain dependence, and the need to achieve commercial-scale quality and output (MP Materials 10-K, SEC EDGAR). This is not a contradiction in the legal sense, but it is a material gap between promotional disclosure and filed risk disclosure: marketing emphasizes capability, while the 10-K conditions that capability on successful commissioning and continued financing.

Energy Fuels similarly markets its White Mesa Mill rare-earth separation capabilities, while its SEC filings describe a uranium/vanadium company adding a new REE separation business with associated operating and regulatory risk (Energy Fuels 10-K, SEC EDGAR).

Environmental, regulatory, and strategic constraints

REE separation has substantial environmental liabilities:

• acid or alkaline baking consumes large reagent volumes;
• monazite processing can mobilize thorium and uranium, creating radioactive waste management requirements (Binnemans et al. 2013, DOI 10.1016/j.jclepro.2012.12.037);
• SX can emit volatile organic solvents and generate large aqueous raffinate streams;
• fluoride and phosphate species in bastnaesite/monazite require careful handling.

These liabilities are a regulatory barrier to entry and a cost driver, not only a technical issue. USGS notes that the United States imports the majority of its separated REE compounds and magnets, which is a supply-chain vulnerability (USGS MCS 2024 Rare Earths).

Unit economics / business model

Who monetizes this technology:

• Miners and concentrate producers sell mixed REO concentrate or carbonates. Their economics depend on contained REO grade, basket price, and radionuclide penalties.
• Separators/processors sell individual separated oxides at a premium over mixed concentrate. They capture value from the separation margin: the difference between the basket price of individual NdPr/Tb/Dy oxides and the mixed-concentrate feed cost.
• Metal/alloy producers convert oxides into metal and NdFeB alloy, capturing further value before magnet makers.
• Some firms monetize toll separation or offtake rather than spot sales; MP’s filings disclose offtake/concentrate relationships and its separation ramp (MP Materials 10-K, SEC EDGAR).

Margins are not stable across the value chain. The pure upstream concentrate business is commodity-exposed; the separated NdPr oxide business has a higher margin but also higher capital and operating intensity. Because the key adjacent-element separation factors are low, the cost structure is dominated by:

• number of SX stages;
• reagent and solvent losses;
• acid/caustic consumption;
• waste management, especially Th/U removal in monazite routes (Binnemans et al. 2013, DOI 10.1016/j.jclepro.2012.12.037);
• energy for metal reduction.

I do not present a single “typical margin” number here because margin varies by feedstock, product purity, and commodity cycle, and any single figure without a specific filing would be fabricated. Investors should read MP Materials and Energy Fuels 10-K filings for current revenue, cost, and margin line items. Any margin figure I could insert from memory would require needs primary source validation.

The value capture is not symmetrical. In a tight separated-oxide market, the separator captures the scarce margin; in a glut, mixed-concentrate producers may still generate positive cash flow but separation projects may not justify capital. This is a conventional refining-spread business model.

Disconfirming evidence

The thesis implied by this note is that REE separation is a high-barrier, stage-intensive, strategically scarce processing bottleneck, and that US/rest-of-world separation projects are economically relevant but technically and financially constrained.

That thesis would be weakened or disproved if:

1. China materially relaxes rare-earth export/processing controls and allows abundant, cheap separated oxide exports without quota or license constraints.
2. A non-SX technology demonstrates adjacent-element separation factors above ~3.0 at pilot/commercial scale with lower waste and cost; that would reduce the stage-count barrier and potentially commoditize separation.
3. MP Materials and Energy Fuels 10-K disclosures show repeated delays, lower NdPr purity, higher unit costs, or inability to achieve commercial-scale production, directly undermining the domestic-separation thesis (MP Materials 10-K; Energy Fuels 10-K).
4. Permanent-magnet demand shifts structurally to non-REE motor designs or substitutes at scale, reducing NdPr/Dy/Tb price support.
5. Large and lower-cost REE byproduct recovery emerges from non-REE metal/phosphate operations at scale, eroding the margin of dedicated bastnaesite/monazite separation projects.