Both start as powder. The powder is lined up in a magnetic field, pressed into shape, and fused in a furnace. Air has to stay out the whole time, because rare-earth alloys react easily and a magnet’s properties depend on its exact composition.
The press works in an argon glovebox, with oxygen held below 1 ppm. The coils put about 1.8 T across the die, and every flake turns its easy axis along it. The punch then presses the loose powder (about 3 g/cm³) into a green block of about 4.4 g/cm³, roughly 58% of a finished magnet's density, before it is pushed out.
NdFeB
For NdFeB, the furnace does the main work. It fuses the powder into a nearly solid block and closes the pores that would let oxygen attack the magnet later. It also keeps the grains small, because coarse grains kill coercivity.
Under vacuum, the blocks are held near 1060 °C for about three hours. A thin liquid film at the grain edges pulls the grains together, and the block goes from about 58% of full density to nearly 100%. It shrinks most along the aligned axis. The hold is kept short and not too hot, because grains that sit too long at temperature grow coarse. A low-temperature anneal afterwards sets the grain-boundary phase. Times and temperatures are typical, not from one plant.
Why does grain size matter? Inside a magnet, the atoms in each region point the same way. The boundary between two regions pointing different ways is called a domain wall. You demagnetize a magnet by pushing those walls through it, and each wall flips a whole region as it goes.
Both boards are the same NdFeB with the same four flaws; only the grain size differs, and the fine board comes out stronger. A flip starts at a flaw, and a domain wall sweeps across that grain until it reaches the Nd-rich layer, which isn't magnetic and stops it. On the coarse board each flaw costs a ninth of the magnet, and together the flaws flip almost half. On the fine board the same flaws flip four specks, about 5% of the area. The grains with no flaw hold until a far bigger field. Fields are illustrative, and real fine grains are only 1–3 µm across.
In NdFeB, each grain is wrapped in a thin neodymium-rich layer that isn’t magnetic, and a wall can’t cross it. A flip that starts in one grain stays in that grain. Fine-grained NdFeB has grains of 1–3 micrometers. Grains that small rarely contain the tiny flaws where flips start, and when one does flip, it’s a speck the rest of the magnet barely notices.
SmCo
SmCo works differently. It’s heat-treated for about 20 hours at around 850°C, then cooled slowly. That leaves a fine cellular structure inside, and the copper-rich walls between the cells are obstacles put there on purpose.
To see why, picture a magnet as a stadium crowd with everyone facing the same direction. That shared direction is what makes it a magnet. Demagnetizing it means getting the crowd to turn around, and turning is contagious: once one person flips, their neighbors tend to follow, and the turn sweeps through the stands like a wave. The moving edge of that wave, where turned-around people meet forward-facing ones, is called a domain wall.
In aged SmCo, the crowd is split into small pens by railings, and the copper-rich walls are those railings. A flip can start, but the wave stops when it hits a railing. To turn the whole crowd, you’d have to push it through railing after railing. That catching is called pinning, and it’s where SmCo gets its resistance.
NdFeB has no railings. Once a wave gets going inside a grain, little stops it. Instead, every person in the crowd is stubborn: it takes an enormous shove to get even one of them to turn and start a wave. SmCo lets the flip start but won’t let it spread. NdFeB won’t let it start at all.
One push, the reverse field, rises on both stands at once. In aged SmCo a few weak fans turn early. Each turn spreads through its pen and stops at the railings, so most of the crowd keeps facing forward until the push is strong enough to carry the wave over railing after railing. NdFeB has no railings, but every fan is stubborn: nobody turns until the push is large, and then one fan goes and the wave sweeps the whole stand. SmCo lets the flip start but won't let it spread. NdFeB won't let it start. Push levels are illustrative.
So if NdFeB needs only hours in the furnace, why run SmCo’s day-long cycles?
Heat. Dysprosium or terbium help NdFeB resist demagnetization, but it still loses its magnetism about 3–4 times faster than SmCo per degree, and even its top grades max out around 250°C. Aerospace, defense, and satellite customers who need steady output from -55°C to 250°C pay for SmCo’s furnace time and its cobalt.
Both magnets weaken as they warm, but at different rates: about 0.12% per °C for NdFeB and about 0.035% per °C for SmCo. Across −55 to 250 °C the NdFeB block swings from about 109% to 72%, while the SmCo disc stays between about 103% and 92%. The gauges show only this reversible change. Most high-grade NdFeB is rated to around 200–230 °C, and above its rating it also loses strength for good. SmCo grades run to about 300–350 °C.
That furnace time isn’t cheap. Every SmCo part ties up a furnace much longer, so each furnace makes less and each kilogram costs more. That’s why SmCo has stayed a premium niche. NdFeB, with its short, high-volume cycle, dominates the world’s sintered-magnet output
Both furnaces are the same size and every tray carries the same load. The NdFeB program is modelled as a 6-hour cycle (sinter plus anneal). The SmCo program is a 30-hour cycle: sinter, solution treatment, about 20 hours of aging near 850 °C and a slow cool. Over 62 hours the left furnace turns out 10 trays and the right one 2, so each SmCo kilogram carries about five times the furnace time. Cycle lengths are typical hold times, not one plant's schedule.