How NdFeB Rare Earth Magnets Are Made and Why Dysprosium Is the Bottleneck

An NdFeB magnet is made by separating neodymium and praseodymium from rare earth ore, reducing them to metal, alloying them with iron and boron, milling the alloy to powder, then pressing it in a magnetic field, sintering, machining, coating and magnetising it. Dysprosium is the bottleneck: hot motors need it to resist demagnetisation, and nearly all separated supply comes from one country.

That short answer hides a long chain. The US Department of Energy splits it into raw material production, separation and metal refining, magnet manufacturing, component manufacturing and end of life recovery, and finds China holding the largest share of every manufacturing stage. This guide walks the chain one step at a time, explains why a few percent of a heavy rare earth decides whether a traction motor survives a hot summer, works through the numbers for an electric car motor and an offshore wind generator, and lists what is being done to need less dysprosium. It goes one level deeper than our explainer on rare earth elements, where they go and the alternatives, which covers the 17 elements, their ores and the main substitutes. If you want the bigger picture of where these magnets end up, see how a car is made and how a smartphone is made.

Key takeaways

• The NdFeB magnet was announced in two papers in the Journal of Applied Physics on 15 March 1984: Sagawa and colleagues described sintered magnets, and Croat and colleagues described rapidly solidified (melt spun) material. Sagawa's typical specimen reached an energy product of 290 kJ/m³, about 36 MGOe.
• The US Department of Energy calls sintered NdFeB "the strongest magnets commercially available". Sintered magnets were about 93% of the NdFeB market in DOE's 2022 assessment; the rest are bonded magnets made with a plastic binder.
• A sintered magnet is roughly 30% rare earth, 69% iron and 1% boron by weight, according to the same DOE report. Rare earth metals are estimated at over 90% of the material cost.
• Dysprosium (Dy) and terbium (Tb) raise coercivity, the resistance to demagnetisation, so a magnet can work hotter. DOE's grade table runs from under 0.5% Dy for an 80 °C grade to 8.5 to 11% Dy for 200 to 220 °C grades used in traction motors.
• Grain boundary diffusion puts the heavy rare earth only where demagnetisation starts, along grain boundaries near the surface. DOE says it can halve total heavy rare earth content, and its 2023 assessment cites an industry estimate that it can remove the need for Dy entirely in medium temperature uses such as wind generators.
• China mined an estimated 270,000 of 390,000 tons of rare earth oxide in 2025 (USGS). In DOE's 2020 estimates China held about 89% of separation, 90% of metal refining and 92% of magnet alloy and magnet manufacturing, and near 100% of heavy rare earth separation.
• On 4 April 2025 China's Ministry of Commerce put samarium, gadolinium, terbium, dysprosium, lutetium, scandium and yttrium items under export licensing, naming "Dysprosium-containing NdFeB permanent magnet materials" and the terbium equivalent explicitly. An October 2025 expansion was suspended for one year in November 2025; the April controls stayed.
• USGS 2025 average prices: dysprosium oxide about $239 per kg and terbium oxide about $1,010 per kg, against about $69 for NdPr oxide.
• The software that designs these magnets and runs the furnaces, mills and presses is ordinary software with ordinary open source dependencies. That is the only part of this story where Safeguard has a role, and we keep that to one honest section near the end.

What is a neodymium iron boron magnet?

A neodymium iron boron magnet is a permanent magnet built around the tetragonal compound Nd₂Fe₁₄B. Sagawa and colleagues reported the compound in 1984 with lattice constants a = 0.880 nm and c = 1.221 nm, an approximate composition of 12 atomic percent neodymium and 6 atomic percent boron with iron as the balance, a saturation magnetisation of 1.35 T, an anisotropy constant of about 3.5 MJ/m³ and a Curie temperature of 585 K (about 312 °C). Those are the values published in 1984; treat them as the historical baseline rather than a modern datasheet.

In practice the word "neodymium" is shorthand. The DOE notes that magnet makers often use didymium, the unseparated neodymium and praseodymium mix, because the two elements are so hard to separate. Neodymium and praseodymium together are the light rare earth base. Dysprosium, terbium, gallium, cobalt, copper, aluminium, niobium and sometimes cerium appear as additions, each to tune a property.

Why is NdFeB the strongest commercial permanent magnet?

Three properties decide how useful a permanent magnet is, and the DOE lists them as coercivity (resistance to demagnetisation), maximum energy product (the magnetic energy the material can store, which depends on both coercivity and magnetisation) and maximum operating temperature. NdFeB wins on energy product. Sagawa's 1984 specimen, Nd₁₅B₈Fe₇₇, recorded a remanence (Br) of 1.23 T, an intrinsic coercivity of 960 kA/m and an energy product of 290 kJ/m³. Today, per DOE, sintered grades typically range from 35 to 52 MGOe, which is about 280 to 414 kJ/m³.

In short, iron supplies a large magnetic moment, neodymium supplies strong anisotropy (the moments prefer one crystal axis and resist being turned), and boron stabilises the structure. Sagawa's abstract notes the compound's Curie temperature was much higher than those of iron and light rare earth binary compounds. Samarium cobalt magnets keep their coercivity better at high temperature, and DOE says they suit hot applications where weight is not a concern, but NdFeB has the higher energy product at temperatures up to about 180 °C. That lets a motor or generator designer use a smaller, lighter magnet for the same job.

The weak spots are just as important. The 1984 specimen's remanence fell by 1,260 parts per million for every kelvin of temperature rise, its coercivity also falls as it warms, and the Curie temperature of around 312 °C is low compared with samarium cobalt. Rare earth metal is reactive, so uncoated NdFeB corrodes. Nearly every manufacturing step below exists to manage one of those weaknesses.

A worked example: what the formula says about composition

Take the formula Nd₂Fe₁₄B and convert it from atoms to weight, using atomic masses of 144.24 for neodymium, 55.845 for iron and 10.81 for boron.

The pure compound is 26.7% neodymium by weight, or 11.8 atomic percent, which is where Sagawa's "about 12 at. % Nd" comes from. DOE's figure for a real sintered magnet is roughly 30% rare earth. The extra 3 or so percentage points sit in a rare earth rich phase between the grains. That grain boundary phase helps the powder sinter densely and magnetically separates neighbouring grains, and it is also why the boundary is the natural place to put dysprosium, as you will see below. The boron share matches DOE's "1% boron" exactly.

The whole process at a glance: ore to magnetised part

DOE's 2022 deep dive on rare earth permanent magnets describes the sintered route in detail. The table below follows it from the mine to a finished, magnetised part. "What comes out" is what moves to the next step.

Steps 1 to 7 are chemistry and metallurgy; steps 8 to 17 are powder metallurgy. DOE notes that the alloy can be made in one place and cast in another, but melting it twice costs more energy, and that the jet milled powder is pyrophoric, so it is hard to ship. That is one reason alloy making and magnet making tend to sit in the same plant.

From ore to oxide: how rare earths are separated

Why separation is the hard step

Every rare earth deposit contains a mixture. DOE's deep dive puts it plainly: because the rare earths are chemically similar, separating them from each other is particularly challenging, and neodymium and praseodymium are among the hardest pair. The ore also carries small amounts of uranium and thorium, which must be split off and managed as radioactive waste.

The main industrial method is solvent extraction. In DOE's description, the concentrate is cracked at high temperature with sulfuric acid or another reagent, the rare earths are dissolved, and cerium is often removed first by oxidising it to an insoluble ceric oxide, which shrinks the volume the rest of the plant has to handle. The solution then flows through trains of mixer settlers. Each unit has a mixing chamber, where an organic solvent meets the aqueous feed, and a settling chamber, where the two liquids separate by gravity. Each element moves very slightly more into one phase than its neighbour, so the separation is repeated over and over.

DOE gives the scale: trains can use hundreds of mixer settlers, and separating neodymium from praseodymium alone can take about 30. The organic phase is commonly kerosene with an organophosphorus extractant such as di(2 ethylhexyl) phosphoric acid, and oxalic acid is used to precipitate the separated rare earths. The process consumes large amounts of acid, caustic and water, and wastewater treatment and solvent purchases are major cost drivers.

Two details matter here. First, the trains separate light from heavy rare earths early and only then split individual heavy elements. DOE notes that promethium does not occur naturally, which leaves a convenient gap between the light and heavy groups. Second, when heavy rare earths are a tiny share of the feed, as in most light rare earth ores, DOE says it may not be economic to separate them into individual elements at all. A plant built for neodymium and praseodymium can therefore produce a mixed heavy concentrate it never refines.

Why many magnets use didymium instead of pure neodymium

Since splitting Nd from Pr costs another 30 or so stages, DOE says some applications do not justify it and use didymium metal, which typically has a 3 to 1 Nd to Pr ratio. If a grade needs a different ratio, separated Nd or Pr oxide is blended in. The magnet works with either element in the Nd₂Fe₁₄B lattice, so the chemistry cost is weighed against a small performance difference.

From oxide to metal: electrowinning and the ferrodysprosium step

Magnets need metal, not oxide. DOE describes two refining routes. The common one is electrowinning in a molten fluoride bath, which DOE says works on a principle similar to aluminium smelting. A typical cell uses a graphite anode and a tungsten cathode, and the electrolyte is 80 to 90% rare earth fluoride with lithium fluoride as the balance. Rare earth oxide is fed into the bath, and metal collects at the cathode. The alternative, sodium reduction of anhydrous rare earth chlorides, is a metallothermic route like the Hunter process for titanium; DOE lists its problems as making dry chloride from a salt that absorbs water, sourcing sodium metal and separating the sodium chloride by product.

For dysprosium the cell changes. DOE says ferrodysprosium is made with a consumable iron cathode, so the dysprosium alloys with the iron as it forms, giving a eutectic of about 80% Dy and 20% Fe, with dysprosium fluoride as the electrolyte. That DyFe alloy is the usual way dysprosium enters a magnet melt. China's April 2025 export notice names "Dysprosium-iron alloy" and "Terbium-dysprosium-iron alloy" among the controlled items, which shows how directly the refining step maps to trade policy.

Worked example: a 3 kA electrowinning cell

DOE's appendix sizes a cell, and the numbers show why refining is a scale business:

• A 3 kA cell needs about 260 kg of didymium fluoride to start. At a September 2021 spot price of $735 per kg, that is about $190,000 of electrolyte before any metal is made, more than the cell itself at about $110,000.
• About 2% of the bath is replaced each month.
• The cell produces about 38 tonnes of metal a year from about 44.7 tonnes of oxide. That ratio, 44.7 / 38 = 1.18 kg of oxide per kg of metal, matches the 1.18 figure DOE uses elsewhere in the report.

Check the ratio from chemistry. Nd₂O₃ is 288.48 / (288.48 + 48.00) = 85.7% neodymium by weight, so the theoretical oxide need is 1 / 0.857 = 1.17 kg per kg of metal. The real world figure is a little higher because of losses.

Now follow the metal into magnets. If an electric vehicle grade magnet carries about 20% Nd plus Pr (DOE's table gives 19 to 21.5% for EH and AH grades), one cell's 38 tonnes of NdPr metal is enough for about 38 / 0.20 = 190 tonnes of magnet. At 1.5 kg of magnet per motor, the middle of DOE's 1 to 2 kg range, that is roughly 127,000 traction motors a year from one cell, before machining losses. DOE also notes that some Chinese cells have reached 10 kA.

Making the magnet: sintered and bonded routes

The sintered route, step by step

DOE describes sintered NdFeB as a powder metallurgy route. Here is what each step is for.

Melting and strip casting. The rare earth metals, iron, ferroboron and minor additions are induction melted, then poured onto the outer surface of a cooled rotating cylinder. The fast, directional cooling gives thin flakes with a small grain size. Fine, even grains at this stage make fine, even grains in the final magnet, and grain size is closely tied to coercivity.

Hydrogen decrepitation. The flakes are exposed to hydrogen, which the alloy absorbs; a 2017 paper in the journal Recycling describes the material breaking down into a hydrided main phase (Nd₂Fe₁₄BHx) and the rare earth rich grain boundary phase. The brittle alloy cracks apart. DOE describes this as a way to reduce the grain size further before milling, and it makes the next step much easier.

Jet milling. The cracked material is milled in a gas stream so that particles break by hitting each other, which DOE calls autogenous milling. The goal is powder in which most particles are single crystals. DOE stresses that the shape of the grains made in this step controls the magnet's microstructure and therefore its key properties. The powder is pyrophoric, so the mill, the transfer lines and the press all work under protective gas.

Aligning and pressing. The powder is pressed while a strong magnetic field is applied. Each single crystal particle rotates so its easy axis lines up with the field, then pressure locks it in place. This is what makes a sintered magnet anisotropic: strong along one direction and weak across it. Without alignment you would get an isotropic magnet with a much lower energy product.

Sintering. DOE gives 1,000 to 1,100 °C. At these temperatures the rare earth rich grain boundary material helps the pressed powder densify into a solid block. Grain size matters here too: the finer and more even the grains that survive sintering, the better the coercivity tends to be.

Machining. Sintered blocks are hard and brittle: Sagawa's 1984 specimen had a Vickers hardness of 600 and a flexural strength of 250 MPa. DOE says the largest material losses in the whole route occur in machining, depending on the final shape. The swarf is collected and, as discussed later, mostly recycled at the plant.

Coating. Rare earth metal oxidises, so DOE says the magnets are coated with a nickel film of 5 to 10 microns. Other coatings exist for specific environments, but nickel is the default.

Magnetising. Only after plating is the part magnetised in a high field, which DOE describes as aligning the magnetisation of the grains. Shipping, machining and assembling unmagnetised parts is far easier and safer than handling strongly magnetised ones, which is why this comes last.

The bonded route

Bonded magnets start from a different powder. Croat and colleagues showed in 1984 that rapid solidification of Nd-Fe-B and Pr-Fe-B alloys gave isotropic material with energy products above 14 MGOe (about 111 kJ/m³) and intrinsic coercivities around 15 kOe. Their microscopy found roughly spherical crystallites, consistent with single domain particles. In a bonded magnet that powder is mixed with a plastic resin and moulded. DOE says bonded magnets are favoured where complex shapes are needed, sintered magnets where conditions are harsh and hot, and that bonded magnets generally have lower energy products and tend to be brittle. That is why traction motors and direct drive generators rely on sintered magnets.

Sintered or bonded: a decision table

Why dysprosium and terbium are added

Coercivity, heat and the demagnetising field

A permanent magnet in a motor is constantly pushed by the field from the stator windings, and that push is strongest at high current and high temperature. If the opposing field exceeds the magnet's coercivity at that temperature, part of the magnet flips and the flux is lost for good. NdFeB's coercivity falls as it warms. DOE says Dy and Tb are used mainly in sintered magnets "to improve their resistance to demagnetization at temperatures above 120 degrees C".

DOE's 2023 Critical Materials Assessment adds that Dy, Tb and gallium can all raise coercivity, that Dy is the most frequently used, and that Tb and Ga are sometimes added in smaller amounts. A 2021 review in Metals states the problem directly: magnets without heavy rare earths "exhibit insufficient coercivity to withstand the demagnetization field at elevated temperatures".

The mechanism, in plain terms, is that a heavy rare earth atom in the Nd₂Fe₁₄B lattice raises the anisotropy that holds the moments in place, so a stronger reverse field is needed to start a flip. Reversal tends to start at grain surfaces and defects, which is why where the dysprosium sits matters as much as how much there is.

The grade table: how much dysprosium each temperature costs

Sintered grades are named by energy product and a temperature suffix. DOE's example: N42AH is 42 MGOe and rated to 220 °C, while N52 is 52 MGOe and rated to 80 °C. DOE's Table 2 gives approximate dysprosium content per suffix, for magnets made by conventional alloying:

Two patterns stand out. First, the dysprosium replaces neodymium and praseodymium almost one for one, because DOE's table assumes the total rare earth content stays at about 30%. Second, the step from 150 °C to 200 °C nearly doubles or more than doubles the dysprosium. DOE says traction motors typically need EH and AH grades, direct drive wind generators typically rely on SH grades, and hard disk drives and loudspeakers need magnets with very little if any dysprosium because they do not run hot. DOE adds that manufacturers can get below the table's Dy levels for a given grade with grain boundary diffusion and the dual alloy process, which blends NdPr metal with ferrodysprosium.

Grain boundary diffusion: dysprosium only where it is needed

Conventional alloying spreads dysprosium through every grain, including the grain interiors where it does little to stop reversal. Grain boundary diffusion (GBD) changes that. A finished, machined magnet is coated with a heavy rare earth source and heat treated, and the heavy rare earth travels inward along the grain boundary phase, forming a thin heavy rare earth rich shell on the outside of each grain near the surface. The interior of each grain keeps its high neodymium content and its high magnetisation.

A 2006 paper in IEEE Transactions on Magnetics by Hirota, Nakamura, Minowa and Honshima, titled "Coercivity Enhancement by the Grain Boundary Diffusion Process to Nd-Fe-B Sintered Magnets", is one of the early publications of the process. The 2021 Metals review traces how diffusion sources evolved: first terbium or dysprosium metal, then heavy rare earth eutectic alloys to cut cost, then sources based on light rare earths and even non rare earth elements, with inorganic compounds, metals and alloys all now used in industry. The review calls GBD "the best route to fabricate high-coercive Nd-Fe-B magnets with low consumption of expensive HRE resources".

How much does it save? DOE's 2022 deep dive says total heavy rare earth content "can be halved" by GBD. DOE's 2023 assessment, citing an industry source, estimates that GBD can cut the dysprosium needed in medium temperature applications such as wind turbines by 100%, which is why that assessment's low case puts dysprosium in wind magnets at zero. There are limits. DOE notes that GBD is more pertinent to smaller magnets, because diffusion from the surface only reaches so deep, and that it is subject to intellectual property constraints. Diffusion works from the surface inward, so in practice larger parts may still need alloyed heavy rare earth.

Terbium: stronger effect, scarcer and pricier

Terbium does the same job as dysprosium, and it is scarcer still. USGS's 2025 average price for terbium oxide was about $1,010 per kg, more than four times dysprosium oxide at about $239. DOE's 2023 wind turbine assumptions set terbium as a fraction of dysprosium, based on observed ratios, rather than as a replacement. In GBD only small amounts per part are needed, which is one reason the 2021 Metals review lists terbium metal among the earliest diffusion sources.

Can you make high coercivity NdFeB with no heavy rare earths at all?

Research says partly. Hono and Sepehri-Amin, writing in Scripta Materialia in 2018, summarise microstructure studies showing that in conventional sintered magnets the grains are coupled through a ferromagnetic grain boundary phase, and coercivity is governed by domain wall pinning. Magnets in which the grains are magnetically decoupled can be made by adjusting composition to slightly neodymium rich and boron lean with gallium doping, and those reach higher coercivity through a different mechanism. Their paper discusses the prospect for heavy rare earth free, high coercivity, high remanence magnets. A 2011 paper from the same group on fine grained sintered magnets with high coercivity points at grain size as the other lever. DOE's 2022 report also mentions a carmaker's magnet that needed no heavy rare earths and less neodymium but was not used in its hybrids because of lower performance, which is a fair summary of where the field stands: real progress, with trade offs.

Worked examples: how much magnet and how much dysprosium

Primary sources publish material intensities as ranges, and they do not always agree. Here are the figures from the sources fetched for this guide, then three calculations.

The two DOE wind figures differ by a factor of four to five. The 2023 assessment is newer and explains that it picked median values from six studies; the 2022 range comes from a single cited source. We use the 2023 figure below and show what the older one would imply.

Example 1: an electric car traction motor

Assume a 1.5 kg magnet set in an EH grade, using DOE's grade table.

• Dysprosium at 8.5%: 1.5 × 0.085 = 0.128 kg, or about 128 g. At 11% it would be 165 g.
• Dysprosium oxide (Dy₂O₃) is 325 / 373 = 87.1% dysprosium by weight, so 128 g of metal needs about 128 / 0.871 = 147 g of oxide.
• At USGS's 2025 average of $239 per kg of oxide, that is about $35 of dysprosium oxide per motor.
• Nd plus Pr at about 20%: 1.5 × 0.20 = 0.30 kg of metal, or about 0.30 / 0.857 = 0.35 kg of oxide. At about $69 per kg of NdPr oxide, that is about $24.

So in a hot running motor made by conventional alloying, the dysprosium can cost more than all the neodymium and praseodymium together, while weighing less than half as much. That is the dysprosium problem in one line.

Now compare with DOE's 2023 range of 41 to 136 g of dysprosium per car. The top of that range is close to the conventional EH calculation; the bottom is roughly what you would expect if GBD halved it and the motor ran a little cooler. The 2011 USGS figure of 30 g per motor came from a 2009 estimate and from a time when traction motors were smaller. All three are plausible for different designs, which is why any single "grams of dysprosium per EV" number should come with a date and a source.

Example 2: a 15 MW offshore direct drive wind turbine

Using DOE's 2023 figures:

• Magnet mass: 650 kg per MW × 15 MW = 9,750 kg, just under 10 tonnes.
• Neodymium at the high intensity of 176.6 kg per MW: 2,649 kg.
• Praseodymium at 31.9 kg per MW: 479 kg.
• Dysprosium at 7.9 kg per MW: about 119 kg in the high case, zero in the low case that assumes full use of grain boundary diffusion.

If you used the older 2.7 to 3.2 tonnes per MW and the SH grade's 4.2% dysprosium from DOE's grade table, you would get 40 to 48 tonnes of magnet and 1.7 to 2.0 tonnes of dysprosium for the same turbine, more than ten times the 2023 high case. The gap shows how much the answer depends on assumptions about magnet mass and grade, and why DOE's 2023 team leaned on medians from several studies.

Wind generators run cooler than traction motors, which is why their dysprosium share is so much lower. DOE also points out that the most common wind generator, the doubly fed induction generator with a gearbox, contains no rare earth magnets at all. Permanent magnet generators are preferred offshore for lower maintenance, efficiency and lower weight at the top of the tower.

Example 3: how much flux a magnet loses as it warms

Sagawa's 1984 specimen had Br = 1.23 T and a remanence temperature coefficient of minus 1,260 ppm per kelvin (minus 0.126% per K). Warm it from 20 °C to 150 °C, a rise of 130 K:

• Fractional loss: 0.00126 × 130 = 0.164, or about 16%.
• Remanence at 150 °C: 1.23 × (1 − 0.164) ≈ 1.03 T.

That loss is reversible: cool the magnet and the flux returns, provided the magnet never met a reverse field larger than its coercivity at the hot temperature. Coercivity also drops with temperature, which is the irreversible risk dysprosium and terbium are there to manage. A motor designer has to budget for both: the weaker flux when hot, and enough coercivity margin at the hottest point under the highest current.

Where each step happens, and the export controls that matter

Mining: USGS 2025 estimates

USGS Mineral Commodity Summaries 2026 estimates world rare earth mine production at 390,000 tons of rare earth oxide equivalent in 2025, up from 380,000 tons. The leaders were China at 270,000 tons, the United States at 51,000, Australia at 29,000 and Burma at 22,000 (estimated from Chinese import data), followed by Thailand at 4,800, India at 2,900, Madagascar at 2,700 and Russia at 2,600. China also had the largest reserves at 44 million tons. USGS says the leading global end use of rare earths was magnets.

Mine output tells you little about dysprosium. The US mine at Mountain Pass produces bastnaesite, which DOE notes is not rich in dysprosium. Most separated heavy rare earths, DOE says, come from ionic clay deposits in China and Burma. USGS reports a Brazilian mine producing mixed concentrates from ionic clays with elevated terbium and dysprosium.

Midstream: separation, metal and magnets

DOE's 2022 table of geographic concentration for the sintered NdFeB chain, based on 2020 estimates, is the clearest picture of the midstream:

DOE adds that China's share of heavy rare earth separation is "near 100%". The pattern is the dysprosium problem in geographic form: mining has diversified, but each later step is more concentrated than the one before it, and the heavy rare earth step most of all. Since 2020 several projects have moved. USGS reports that in August 2025 the US Department of War lent $150 million to the Mountain Pass producer to build a heavy rare earth separation facility, that the same department lent $80 million to a recycler in Indiana and awarded $5.1 million to another in Texas to recover rare earths including terbium and dysprosium, that an Australian producer announced a heavy rare earth separation plant in Malaysia in October 2025, and that the US International Development Finance Corporation approved a $465 million loan in November 2025 to the Brazilian ionic clay producer. USGS also notes that while at least five US companies were developing heavy rare earth processing in 2025, none produced sustained commercial scale quantities that year.

Timeline: the magnet and the controls

What the April 2025 notice actually controls

MOFCOM Announcement No. 18 of 2025, issued on 4 April 2025 by the Ministry of Commerce and the General Administration of Customs, applies export licensing to items related to samarium, gadolinium, terbium, dysprosium, lutetium, scandium and yttrium. For dysprosium it lists dysprosium metal, dysprosium iron and terbium dysprosium iron alloys, dysprosium sputtering targets, "Dysprosium-containing NdFeB permanent magnet materials", dysprosium oxide and dysprosium compounds. The terbium section mirrors it, including terbium containing NdFeB permanent magnet materials. The notes say the controlled permanent magnet materials include magnets and magnetic powders. Exporters must apply for a licence and declare the control number on customs forms, and the announcement took effect on the date of issue.

For a magnet buyer, that means a magnet with any deliberate dysprosium or terbium content, including a GBD treated one, falls into a licensing regime, while an 80 °C grade with no heavy rare earth is not among the items this notice names. Neodymium and praseodymium are not on the April list. USGS reports that the April controls remained in effect at the end of 2025, with China issuing general export licences to selected exporters. The White House fact sheet of 1 November 2025 described those general licences as covering rare earths, gallium, germanium, antimony and graphite for US end users and their suppliers. Read the licensing status for your own trade lane from the current official notices rather than from a summary, since the October measures were suspended for one year from November 2025, not withdrawn.

The EU Critical Raw Materials Act

The European Commission says the Critical Raw Materials Act entered into force on 23 May 2024. It sets 2030 benchmarks along the strategic raw materials value chain: at least 10% of the EU's annual consumption from EU extraction, at least 40% from EU processing, at least 25% from recycling, and no more than 65% of annual consumption of each strategic raw material from a single third country. The Commission says the Act also sets requirements on recyclability and recycled content for permanent magnets, to encourage large scale magnet recycling. Against DOE's 92% figure for magnet manufacturing, the 65% single country benchmark gives a sense of the distance to travel.

Recycling magnets and alternatives under research

Swarf comes first

DOE says most rare earth recycling today is of swarf, the cuttings from machining. Because cutting happens before magnetisation, swarf is magnetic powder mixed with cutting fluid. It is usually recycled at the magnet plant by acid digestion and separating the iron and boron from the rare earths. DOE estimates the swarf stream at about 9% of global NdFeB magnet production in 2020. It mirrors the geography of magnet making, so it does not diversify supply.

End of life magnets

End of life recycling is small, per both DOE and USGS. DOE splits it into two steps.

1. Magnet recovery. Products are dismantled, the magnet parts sorted out, demagnetised and collected. The alternative is to shred the whole product and separate magnet material by magnetic and electrostatic separation, screening, abrasion and chemical etching to strip coatings, which risks contaminating and corroding the magnet.
2. Rare earth recovery. DOE describes three routes. Hydrometallurgy dissolves the magnet in acid and separates rare earths by selective precipitation, solvent extraction or ion exchange. Pyrometallurgy uses heat, either roasting or extraction with molten metal, salt or slag. Direct electrochemical routes, such as molten salt electrolysis and selective electrochemical leaching, are emerging. DOE notes that recycling is simpler than primary processing because there are fewer elements and the starting concentrations are an order of magnitude higher.

There is also a short loop. DOE describes hydrogen processing at atmospheric pressure that breaks a magnet into demagnetised alloy powder, which can be sintered or bonded again after the hydrogen is removed. It copes with coatings and adhesives but gives lower magnetic properties. A 2017 paper in the journal Recycling explains one reason: recycled hydrogen decrepitated powder from hard disk drive magnets carries more oxygen, concentrated in the rare earth rich grain boundary phase, which lowers density, coercivity and remanence. The authors tried a hydrocyclone to separate the fine grain boundary particles from the main phase and found the best separation after three passes.

That study used magnets from hard disk drive scrap. DOE notes that hard disk drive magnets contain very little if any dysprosium, so recycling them returns mostly neodymium and praseodymium rather than easing the heavy rare earth shortage.

Alternatives under research

DOE's Critical Materials Innovation Hub, led by Ames National Laboratory, works on four areas: diversifying supply, developing substitutes, unlocking secondary sources and crosscutting research. DOE's 2022 deep dive describes the main substitute strategies:

DOE's electric motors research page states that "almost all hybrid and plug-in electric vehicles use rare earth permanent magnets in their traction motors", that induction motors offer high starting torque and reliability, and that the Beyond Rare Earth Magnets project led by Ames Laboratory is looking at lower cost permanent magnets, alongside work on reduced rare earth and magnet free motor designs. DOE's 2022 report adds an important economic point: experts estimate that if heavy rare earth prices doubled under a persistent disruption, the heavy rare earth content of certain NdFeB magnets could fall by about 43%. Price is itself a substitution tool.

Step by step: estimate the dysprosium in your product and track supply

You cannot change the chemistry of a magnet you buy, but you can know how exposed your product is. These steps work for an engineering or procurement team with a magnet parts list.

Step 1. Pull the latest USGS data. The chapters are public PDFs.

mkdir -p ~/ree-check && cd ~/ree-check
curl -sSLO https://pubs.usgs.gov/periodicals/mcs2026/mcs2026-rare-earths.pdf
curl -sSLO https://pubs.usgs.gov/periodicals/mcs2026/mcs2026-rare-earths-heavy.pdf
# pdftotext ships with poppler (brew install poppler, or apt install poppler-utils)
pdftotext -layout mcs2026-rare-earths-heavy.pdf - | grep -E "Terbium oxide|Dysprosium oxide"
pdftotext -layout mcs2026-rare-earths.pdf - | grep -E "NdPr|^China|World total"

The heavy rare earths chapter's price rows give 2021 to 2025 averages in dollars per kg; for 2025 they show terbium oxide at 1,010 and dysprosium oxide at 239. The main chapter gives NdPr oxide at 69 and China's mine production at 270,000 tons in both 2024 and 2025.

Step 2. List every magnet in the product. A simple CSV is enough. Record the part, the magnet mass, the grade suffix and whether the supplier uses grain boundary diffusion.

part,magnet_kg,grade,gbd
traction_motor_rotor,1.5,EH,no
pump_motor,0.08,H,no
speaker,0.02,N,no
seat_motor,0.05,M,no

Step 3. Estimate heavy rare earth content. This script uses DOE's 2022 grade table and, where GBD is declared, applies DOE's "can be halved" estimate. Replace both with your supplier's declared composition whenever you have it.

#!/usr/bin/env python3
"""Rough Dy exposure from a magnet parts list.
Grade Dy shares: DOE Rare Earth Permanent Magnets deep dive (2022), Table 2.
EH and AH use the low end of DOE's 8.5 to 11% range."""
import csv, sys

DY_SHARE = {"N": 0.005, "M": 0.014, "H": 0.028, "SH": 0.042,
            "UH": 0.065, "EH": 0.085, "AH": 0.085}
DY_IN_OXIDE = 325.0 / 373.0      # Dy2O3 is about 87.1% Dy by weight
DY2O3_USD_PER_KG = 239           # USGS MCS 2026, 2025 average

total_dy_g = 0.0
with open(sys.argv[1]) as f:
    for row in csv.DictReader(f):
        share = DY_SHARE[row["grade"].upper()]
        if row["gbd"].strip().lower() == "yes":
            share /= 2           # DOE: GBD "can halve" HRE content
        dy_g = float(row["magnet_kg"]) * share * 1000
        total_dy_g += dy_g
        print(f'{row["part"]:24s} {dy_g:8.1f} g Dy')

oxide_kg = total_dy_g / 1000 / DY_IN_OXIDE
print(f"{'TOTAL':24s} {total_dy_g:8.1f} g Dy")
print(f"Dy2O3 equivalent: {oxide_kg*1000:.0f} g, about ${oxide_kg*DY2O3_USD_PER_KG:.0f} at 2025 average")

Run it:

python3 dy_exposure.py magnets.csv

For the sample list the output is about 127.5 g for the traction motor, 2.2 g for the pump motor, 0.1 g for the speaker and 0.7 g for the seat motor: roughly 130 g of dysprosium, 150 g of oxide and $36 at the 2025 average price. The traction motor dominates, which is the usual result.

Step 4. Flag what is licensable. Any line with a non zero Dy or Tb share is a "Dysprosium-containing" or "Terbium-containing NdFeB permanent magnet material" under the April 2025 MOFCOM notice if it is exported from China. Mark those parts, record where each supplier's magnet, alloy and separated oxide come from, and ask for the export licence status in writing.

Step 5. Ask the design question. For each flagged part, ask whether the part really sees the grade's rated temperature. Dropping from EH to SH in the table cuts nominal dysprosium roughly in half; cooling the rotor or segmenting the magnet so GBD can reach through it may do the same. DOE's own example is that wind generators can be designed to run cooler and so use less heavy rare earth.

Step 6. Re run it when the data moves. USGS publishes a new edition every February. Re run step 1, update the price constant and keep the output with the product's compliance file.

The software side of a magnet

Magnets are designed in software before any powder is milled. Motor and generator designers use finite element electromagnetic solvers to check the demagnetising field at the hottest corner of each magnet, which is exactly where the choice between an SH and an EH grade gets made. The factory runs on software too: furnace and sintering profiles, jet mill and press controls, the field and pulse settings for magnetising, and the inspection systems that test flux and dimensions. Much of that software, from simulation codes and meshing libraries to the data pipelines and web dashboards that sit around the plant, is built from open source components. Safeguard does not analyse magnets, alloys, materials or factory processes, and its models are single purpose security models; its role is limited to that software and its dependencies, the same as for any other codebase.

Common mistakes

• Treating "neodymium magnet" as one material. An N52 for a loudspeaker and an AH grade for a traction motor have different chemistry, different heavy rare earth content and different export status. Always ask for the grade suffix.
• Assuming a magnet with no dysprosium on the label has none. Supplier datasheets often list grade and properties, not composition. Ask for the declared Dy and Tb content and whether GBD is used.
• Quoting one "grams of dysprosium per EV" figure. Sources fetched for this guide range from 30 g per motor (a 2009 estimate) to 41 to 136 g per car (DOE 2023) to well over 100 g for a conventionally alloyed EH grade. Give the date and the source.
• Mixing oxide and metal weights. Prices are usually quoted per kg of oxide, compositions per kg of metal. Dy₂O₃ is about 87% dysprosium and Nd₂O₃ about 86% neodymium, so the difference is more than 10%.
• Mixing MGOe and kJ/m³. 1 MGOe is about 7.96 kJ/m³. An N42 is about 334 kJ/m³, not 42.
• Reading mine share as supply share. China mined about 69% of 2025 output by USGS estimates, but held about 89% of separation and 92% of magnet making in DOE's 2020 figures, and near 100% of heavy rare earth separation.
• Assuming grain boundary diffusion removes the heavy rare earth problem. It cuts the amount, often by half, and for medium temperature parts possibly to zero, but it works best on thinner parts, is subject to patents, and a GBD treated magnet still contains controlled elements.
• Forgetting that magnets lose flux when hot. A 16% reversible loss from 20 °C to 150 °C, as in Example 3, is normal. Design margins, not just grade names, keep a motor safe.

FAQ

How are neodymium magnets made, step by step?

Rare earth ore is concentrated, cracked with acid and separated by solvent extraction into individual oxides. The oxides are reduced to metal by electrowinning. The metal is melted with iron and boron, strip cast into flakes, broken up with hydrogen and jet milled to a fine powder. The powder is aligned and pressed in a magnetic field, sintered at 1,000 to 1,100 °C, machined to shape, nickel plated and finally magnetised in a strong field.

Why is dysprosium used in neodymium magnets?

Dysprosium raises coercivity, the resistance to demagnetisation, so the magnet keeps working at high temperature and under strong opposing fields. DOE says dysprosium and terbium are used mainly in sintered magnets to improve demagnetisation resistance above 120 °C. Traction motors, which need 200 to 220 °C grades, are the biggest users.

How much dysprosium is in an electric car?

It depends on the motor design and process. DOE's 2023 assessment gives 41 to 136 g per car. A conventionally alloyed 1.5 kg EH grade magnet set would contain about 128 g. A 2011 USGS report cited 30 g per motor from a 2009 estimate. Grain boundary diffusion and cooler designs push the figure down.

What is grain boundary diffusion in NdFeB magnets?

It is a post sintering treatment. A dysprosium or terbium source is applied to the surface of a finished magnet and heat treated, so the heavy rare earth diffuses along the grain boundaries and concentrates in a thin shell around each grain, where magnetic reversal starts. DOE says it can halve heavy rare earth content; it works best on thinner parts.

What is the difference between sintered and bonded NdFeB magnets?

Sintered magnets are pressed from aligned powder and sintered into dense, anisotropic blocks with the highest energy products, typically 35 to 52 MGOe. Bonded magnets mix isotropic, rapidly solidified powder with a plastic binder and are moulded into complex shapes. DOE puts sintered magnets at about 93% of the market.

What do the letters in N42SH or N52 mean?

The number is the maximum energy product in MGOe. The suffix is the temperature class: none for 80 °C, M for 100 °C, H for 120 °C, SH for 150 °C, UH for 180 °C, EH for 200 °C and AH for 220 °C, per DOE's grade table. Higher temperature classes need more dysprosium or terbium unless the maker uses grain boundary diffusion.

Which countries make NdFeB magnets?

DOE estimated that in 2020 China made about 92% of magnet alloy and magnets, Japan about 7% and Vietnam about 1%, with small amounts elsewhere including Germany and the United States. Separation and metal refining were similarly concentrated in China.

Did China ban dysprosium exports?

No. On 4 April 2025 China placed dysprosium, terbium and five other rare earths, including dysprosium and terbium containing NdFeB magnet materials, under export licensing. Exports need a licence rather than being banned. Expanded controls announced on 9 October 2025 were suspended for one year in November 2025, and USGS reports China began issuing general licences to selected exporters.

Can neodymium magnets be recycled?

Yes, but mostly at small scale for end of life magnets. Most recycling today is of machining swarf at magnet plants, about 9% of global magnet production in 2020 per DOE. End of life routes include hydrogen processing back to powder, acid based hydrometallurgy, high temperature pyrometallurgy and emerging electrochemical methods.

Are there magnets as strong as NdFeB without rare earths?

Not yet at the same performance. DOE says no other magnets are available with an energy product similar to NdFeB. Research focuses on "gap" magnets of about 10 to 25 MGOe, such as iron nitride, iron nickel and cerium substituted NdFeB, and on motor designs that use less magnet or none.

Why software hygiene still applies to magnets

None of what follows touches a magnet or a furnace. It applies to the software that surrounds them: the simulation tools, plant software, supplier portals and dashboards that a magnet maker or a motor team builds and runs, all of which pull in open source dependencies that need patching like any other code.

Fix it without waiting for upstream

Whatever the vulnerability, most of the risk sits in two gaps. The first is the time between a fix being published and that fix actually running in your production builds. The second, and worse, is the time before any fix exists at all, while a maintainer works out a patch and a release schedule. Waiting through both gaps is a habit, not a requirement. Safeguard gives teams three ways to close them.

1. Autonomous remediation: pull requests, not tickets

When a fixed version exists, Safeguard's Griffin models open a tested, policy gated pull request that moves the affected dependency to the safe version. You can run it on one repository or in bulk across every project that carries the vulnerable version. Remediation covers eight categories, including dependencies, licenses, security posture and findings. Reachability analysis ranks the vulnerable code paths your application actually calls first, so the first pull requests you merge are the ones that matter most. Autonomous remediation ships switched off, with every category set to Manual, so your team decides how much runs on its own and what still needs a human approval.

2. Gold Open Source: the safe version in seconds, free

gold.safeguard.sh is free and needs no login. Search any CVE or package to see affected and fixed versions, CVSS, EPSS, CISA KEV status and a recommended safe version, or use the "Am I affected?" checker. The same data is available through a free read API, README status badges and RSS threat feeds. The safeguard-sh/gold-scan-action GitHub Action can fail a build on a known exploited vulnerability, on any known vulnerability, or never, so a version you have just removed cannot quietly come back in next week's merge.

3. Gold Registry: start from components that are already fixed

registry.safeguard.sh provides hardened, drop in replacements for upstream packages and container images. Every Gold artifact is held continuously to zero exploitable critical or high CVEs, zero malware and a completed license check, with SLSA L3 or higher provenance, a Sigstore signature and a public evidence page you can hand to an auditor. An artifact loses its Gold badge if it regresses. Teams can point their own package directory at the registry so hardened equivalents resolve without changing how developers install anything.

When upstream has no fix yet, or will not ship one, Safeguard first works to get the fix accepted upstream. If the issue is severe or upstream will not move, Safeguard builds a fork, fixes the vulnerability there and publishes a hardened Gold build, so you are not stuck on a maintainer's release schedule. You can also request a hardened, compatibility tested build of a specific package. This fork and patch flow is in early access today, so ask which packages it covers before you plan around it.

One honest limit: when the vulnerable code lives in a vendor appliance, a hosted service or an operating system runtime you do not build yourself, the fix still has to come from that vendor or distribution. The steps above apply to the open source dependencies and container images you build and ship, which is where most teams carry the largest and least tracked share of their exposure.

See how Safeguard would help

None of these demos involves a magnet. They show the software side of this guide: finding a vulnerable open source component in code you build, fixing it with a person approving each step, and running scans on hardware you control. To try the CLI, download it from https://safeguard.sh/download and read the setup guide at https://docs.safeguard.sh/docs/cli .

The first demo is a real Safeguard CLI session. Asked to fix the lodash CVE in a repo, it reads the finding and advisory, asks before each command, edits package.json to lodash 4.17.21, runs npm install and npm test with approval, prints a summary and pushes nothing, then shows /permissions, /model and /scan. The same flow applies to a supplier portal, a plant dashboard or the build tooling around a motor design team's simulation scripts.

Click through it at your own pace, or open it full screen.

The second demo sets up the Safeguard runner on your own machine, approves it with its device code, then scans from the terminal. Factory software often lives on separate plant networks, and a runner you host yourself keeps the scan close to that code.

Click through it at your own pace, or open it full screen.

The third demo connects Claude to the Safeguard MCP server with OAuth and chooses which tools Claude can use. For an engineering team, that means asking plain questions about which projects carry a given component version, with the assistant limited to the tools you picked.

Click through it at your own pace, or open it full screen.

Quick check

Sources

• USGS Mineral Commodity Summaries 2026: Rare Earths
• USGS Mineral Commodity Summaries 2026: Rare Earths (Heavy)
• USGS Scientific Investigations Report 2011-5094: Rare Earth Elements, End Use and Recyclability
• US DOE: Rare Earth Permanent Magnets, Supply Chain Deep Dive Assessment (February 2022)
• US DOE: Critical Materials Assessment (July 2023)
• US DOE: Electric Motors Research and Development
• Ames National Laboratory: Critical Materials Innovation Hub
• MOFCOM: Announcement No. 18 of 2025 on export control of some medium and heavy rare earth related items
• The White House: Fact sheet on the economic and trade deal with China (1 November 2025)
• European Commission: Critical Raw Materials Act
• European Commission press release IP/24/2748: Critical Raw Materials Act enters into force
• Sagawa et al., New material for permanent magnets on a base of Nd and Fe, Journal of Applied Physics 55, 2083 (1984)
• Croat et al., Pr-Fe and Nd-Fe-based materials: a new class of high-performance permanent magnets, Journal of Applied Physics 55, 2078 (1984)
• Hirota et al., Coercivity Enhancement by the Grain Boundary Diffusion Process to Nd-Fe-B Sintered Magnets, IEEE Transactions on Magnetics (2006)
• He et al., Grain Boundary Diffusion Sources and Their Coating Methods for Nd-Fe-B Permanent Magnets, Metals 11, 1434 (2021)
• Hono and Sepehri-Amin, Prospect for HRE-free high coercivity Nd-Fe-B permanent magnets, Scripta Materialia (2018)
• Sepehri-Amin et al., Microstructure of fine-grained Nd-Fe-B sintered magnets with high coercivity, Scripta Materialia (2011)
• Awais et al., Hydrocyclone Separation of Hydrogen Decrepitated NdFeB, Recycling 2, 22 (2017)

Last reviewed: 23 October 2026. Export licensing, prices and capacity figures change; check the linked official sources for the latest status.