Short versionA sintering furnace has to heat a load evenly, keep it clean while it is hot, and then cool it as fast as the magnet will stand. Those three jobs pull the design in different directions.
1 · Three jobsWhat the furnace is for
A pressed magnet goes into the furnace as a fragile green compact and comes out a dense, brittle block. Sintered magnets use no binder, but the first shelf of the cycle burns off what is left of the isopropyl alcohol used as die lubricant or as the slurry the powder was milled in. After that, the furnace has three jobs. It has to heat evenly, because a block that is hotter on one side shrinks more on that side. It has to stay clean, because rare earths grab oxygen at any chance they get, so the whole cycle runs under vacuum or inert gas. And it has to cool on command.
That last one is a metallurgy problem, not a scheduling one. For NdFeB, how fast the magnet comes down after sintering and after aging shows up directly in its coercivity: faster is better, up to the point where the block cracks.1 Cooling with gas forced through the furnace is a gas quench.
Samarium-cobalt goes through the same furnaces and is fussier. After sintering it gets a solution treatment (held hot so its alloying elements dissolve evenly), a quench to freeze that state, and then a long aging with slow, controlled cooling that grows the fine cell structure its coercivity lives in. Every one of those steps is another place for an uneven furnace to show up in the magnet.
Short versionA tube furnace is a sealed quartz tube running through a small heater. It is cheap, simple and accurate. Its uniform zone is short and it cannot cool a load quickly.
2 · The tubeThe lab's furnace
Walk into any university magnet lab and you will find one. A quartz tube, sealed with a flange at each end so it can be pumped down or flushed with argon, runs straight through a heater that opens like a clamshell. The samples sit in a boat in the middle.
It is a very good furnace for what it is. The heater is simple, the controller is accurate, and the whole thing costs a small fraction of a production furnace. The catch is the uniform zone: only the middle of the heated length is actually at the set temperature. One maker's catalog lists a heated length of 440 mm with a constant-temperature zone of 150 mm in the middle of it, which is typical of the type.2 Samples that stray out of the middle are cooler than you think they are.
2 · The tubeWhy the lab can't quench
To cool a sample fast in a tube furnace you have to get the heat away from it, and the tube is in the way. The lab workarounds are all versions of moving the sample and the heater apart: pull the boat to the cold end of the tube, drop it out of a vertical furnace, or buy a furnace on rails that slides off the tube when the program ends.2
All of those work for a boat of samples weighing a few grams. None of them works for a production load, and that is the whole reason industrial furnaces look the way they do. Once the load is big enough that you can't move it out of the heat, you have to bring the cold to the load, and the rest of this page is about how.
Short versionAn industrial furnace turns the tube furnace inside out: the heater goes inside the vessel. A water-cooled steel vessel holds the vacuum, an insulated hot zone inside it holds the heat, and a hearth carries the load.
3 · Hot box, cold boxA hot box inside a cold one
In a tube furnace the heater sits outside the vessel. Industrial furnaces turn that inside out. On the outside is a steel pressure vessel with a double wall and water running between the two skins. It is a cold wall furnace: however hot the inside gets, the vessel stays near room temperature, because it has to hold a vacuum while heating and gas pressure during a quench, and steel does both well only while it is cool.
Inside the vessel is the hot zone: a box or cylinder of insulation, graphite board and felt or stacked metal heat shields, with heating elements on all four sides of its inside face. Everything hot lives in there. Power, thermocouples and the pumping line all reach it through the cold wall.
The load sits on the hearth, a set of rails on posts that carries the weight of the trays through the bottom of the hot zone. It has to be strong when it is hot and carry as little heat as possible out to the cold wall, which is a harder combination than it sounds. Every industrial furnace on the rest of this page is a variation on this arrangement; what changes is the shape of the hot zone and how the quench gas gets in and out.
Short versionThe classic industrial furnace has a round hot zone inside a round vacuum vessel, ring-shaped heating elements and three separately controlled zones. It is a sensible design that puts a square load in a round hole.
4 · The round furnaceA square load in a round hole
In the classic design the hot zone is a cylinder, lying on its side inside the cylindrical vessel, with heating elements bent into rings or ring segments around the inside wall.
The round shape makes sense for the vessel, which has to hold a vacuum and, during a quench, positive pressure. But the load is a stack of rectangular trays. Put a rectangle inside a circle and you get empty corners, and the parts of the load nearest the elements see more heat than the parts that aren't.
Makers answer with zones. The elements at the front, middle and back are wired as separate circuits so that their power can be trimmed for better uniformity; Ipsen's patents for its round TurboTreater describe exactly these front-to-rear trim zones.3
4 · The round furnaceTrimmed at one temperature
The trouble with trim zones is that the right trim depends on the temperature. The ends of a hot zone lose heat differently from the middle (the door end most of all), and the way those losses grow with temperature is not the same at each end. So a split of power that gives an even load at one temperature is wrong at another.
On a ring-element furnace the split is set by taps on the power supply. With digital taps you can set a different trim for each temperature the furnace runs at, and you have to. With analog taps you set it once, at the temperature that matters most, and live with being off everywhere else. Either way, uniformity is something you tune rather than something the furnace has.
The elements themselves are the other weak point. A curved element is thin. It gets knocked by the load going in and out, and thermal cycling wears it out. Replacing elements is a routine line in the operating budget of a round furnace.
4 · The round furnaceJets
When the cycle calls for a quench, the vessel fills with nitrogen or argon and a blower drives it in through dozens of nozzles spaced around the hot zone wall. The gas has to find its way around the elements, around the trays and out the back. The result is very turbulent flow, and turbulence was once the point: a 1992 patent describes the standard method as forcing gas into the hot zone "from a number of different angles to create turbulence."4
Turbulence does help heat transfer. The problem is where it lands. Some nozzles end up aimed straight at a block, and that block gets a jet of cold gas on one face while the rest of it is still hot. A sintered magnet is brittle, conducts heat poorly, and expands differently along its magnetized direction than across it. One face cooling ahead of the rest builds exactly the stress that cracks it: thermal shock.
And the jets are not repeatable. Move a tray a few centimeters, load a slightly different mix of part sizes, and the jets land somewhere else. So one batch behaves unlike the next, and you can't tune that out.
4 · The round furnacePlumbing on the outside
Most classic round furnaces put the quench blower and the heat exchanger outside the vacuum vessel, joined to it by large ducts. Every flange, valve and expansion joint on that loop is another seal that has to hold a vacuum, and over years of thermal cycling some of them don't. A leak in a sintering furnace is not a nuisance. It is oxygen, delivered to a hot rare earth alloy.
That arrangement is a historical habit more than a law of nature. Before the two companies merged in 1985, Ipsen built rectangular hot zones with internal gas cooling and Abar built round hot zones with external cooling; Ipsen's answer was the TurboTreater, a round furnace with the cooling system inside the vessel.3 Round furnaces with internal quench do exist, and do well in heat treating. For magnets they mostly never got the chance, because the box arrived first.
In fairness to the round camp: the better round designs answered the jet problem with valved quadrants that let the gas come in from the top, bottom or sides on command, with no moving parts inside the vessel.5 That is real engineering. It is also a lot of hardware to fix a problem of geometry.
Short versionA box hot zone matches the shape of the load, uses heavy flat graphite elements, and keeps its top and bottom doors shut while heating. Uniformity comes from the geometry rather than from tuning.
5 · The boxA square load in a square hole
The furnace that replaced the round one is built around a rectangular hot zone. The trays are rectangular, so the hot zone is too, and each wall of heating elements faces a matching face of the load. That alone does most of the work that trim zones were trying to do.
The top and bottom walls carry large plugs, or bungs, that open and close. During heating they are shut, and the hot zone is a closed, well-insulated box that holds its heat. Heat-treating furnace patents describe the same idea as movable doors that stay closed during heating to keep the hot zone uniform and open for the quench.6
The flat walls also allow flat heating elements: thick graphite plates or bars instead of thin curved bands. They are more robust, they don't sit where the load can hit them, and they need replacing far less often.
5 · The boxTrimmed once, at the factory
Box hot zones have zones too, often more of them than round furnaces. One Italian maker's furnace puts elements on the top, bottom, both sides, front and back, as six independent zones.7 The difference is what happens after installation.
Because each wall faces a matching face of the load, the heat losses of the zones grow together as the temperature rises. The trim that makes the box uniform at one temperature is close to right at all of them. In practice the zones are trimmed once, at the factory, and left alone.
In practiceThe difference between "set it at commissioning" and "re-trim it for every recipe" sounds small on paper. On a plant floor it is the difference between a furnace you trust and a furnace you babysit.
5 · The boxThe quench: one even front
At the end of the hold, the vessel fills with gas and the bungs open. A centrifugal blower behind the hot zone pulls the hot gas out through a heat exchanger and blows the cooled gas straight up, over the top of the hot zone and around to the far side. It goes in through one bung, through the load in one direction, and out through the other.
The gas crosses the load as a single front instead of a crowd of jets. Every block sees gas from the same direction at nearly the same speed, so the temperature differences inside a block are a smooth gradient from the side facing the gas to the side away from it. A gradient like that you can plan for.
Makers differ in where they put the heat exchanger. Some wrap a round one around the blower and some stand a flat one in front of it, depending on their design and how much heat their blower will tolerate. The figure shows the second.
5 · The boxFlaps, and reversing the flow
A single direction would still leave the top of the load cooling ahead of the bottom. So the ducts between the blower and the two bungs carry flaps. Swing them over and the side that was feeding gas becomes the side taking it away: the flow through the load reverses, and the blocks that were on the leading edge are now on the trailing one.
The idea is old. Ipsen patented a swinging flap to redirect the quench stream in 1979,8 and Italian maker TAV lists cooling from above, from below, top and bottom alternately, or both together.7 When to reverse is written into the sinter and quench program, along with everything else in the cycle.
The strongest public criticism of this design is fair, so here it is. Seco/Warwick argued in 2002, while selling round furnaces, that gas entering across a whole hatch can't cross a big load at uniform speed, and that much of it goes around the outside of the load instead of through it.9 That is true. It is also outweighed: a good box furnace delivers far more cooling than the magnet needs, so a predictable top-to-bottom difference is a small price, and it lets you run much higher gas pressure without breaking blocks. (The other old complaint, that reversing flaps warp and jam in big, hot systems, has been solved.10)
5 · The boxJets against a front
Here is the whole argument in two blocks. On the left, a block in a jet: one face hit hard, a cold spot you didn't plan for, and stress concentrated where the cold meets the hot. On the right, a block in a front: cooler on the side facing the gas, warmer on the far side, and the same in every block on the tray.
A sintered magnet can survive a large temperature difference if it is spread evenly across the block. What breaks it is a steep difference over a short distance, and jets make those.
5 · The boxWhat it bought me
I can put rough numbers on it from my own experience of moving a sintering process from a round furnace to a box furnace. These are ratios from one changeover, not a controlled experiment, and the figure says so.
In practiceIn the box furnace I sintered blocks twice the weight and quenched them at three times the gas pressure in half the time, without breaking them, and the variation in the knee point fell by about two orders of magnitude. I would not have believed that last number if I hadn't watched it happen.
The knee point is where a magnet's demagnetization curve turns sharply down, and how much it varies from magnet to magnet is a direct read on how evenly a batch was treated. Bigger blocks matter because every block is cut into magnets afterward, and a bigger block means fewer edges and less waste per magnet. A harder quench matters because, as the first section said, it buys coercivity.
Short versionHow the furnace is loaded matters as much as how it is built. Some designs help more than others, but none of them fixes a bad load.
6 · The loadWhat no furnace fixes for you
Everything so far assumes the gas reaches the magnets. In practice the magnets sit in stacked trays, and the green compacts usually go in closed or nearly closed steel or molybdenum boxes to protect them from the furnace atmosphere.11 The sintering box sits between the gas and the magnets. It softens the jets, and it slows the quench. How it is designed and loaded is a long subject of its own, and I'll leave it there.
The stack matters too. A 2026 simulation of the gas quench in an industrial NdFeB sintering furnace, checked against thermocouple readings, found temperature differences of more than 250 K between layers of the same load. The authors put most of that down to how the fixtures were arranged and to weak circulation inside the stack, and rearranging the fixtures in the model cut the worst difference by only about 20 K.12
That study keeps this page honest. Furnace design is not the whole story. A design that drives gas through the load in one direction makes a well-planned load work. It does not rescue a badly planned one.
Short versionA continuous furnace moves the load through a row of chambers that each do one job, with doors between them. The hot chambers never cool and the cooling chamber is built only to cool.
7 · Continuous linesA row of chambers
Everything above is a batch furnace: one chamber that heats a load, holds it, cools it, and then has to be heated again from cold for the next one. A continuous line splits the cycle up. The load goes in at one end and moves from chamber to chamber, with a sealed gate valve between each pair. One published Chinese design runs preparation, preheating and burn-off, two degassing chambers, presintering, sintering, aging and a gas-cooling chamber, each joined to the next by a valve.13
Two things follow. The hot chambers stay hot, so you stop paying to heat insulation and fixtures from cold every cycle; ULVAC, describing its multi-chamber furnaces, makes the point that the heating chamber is not cooled during the cooling step, which saves power.14 And the cooling chamber can be designed purely for cooling, with no heating elements or insulation in the gas's way.
There is a middle ground, too. Semi-continuous furnaces with two or three chambers, a heating chamber and a gas-cooling chamber moved between by a fork, get much of the energy benefit with less of the commitment.14
7 · Continuous linesSmaller loads, less flexibility
The load moving through a continuous line is usually much smaller than a batch load. ULVAC's slim continuous furnace has a uniform zone 310 mm wide, and the company credits the slim shape for uniformity in both heating and cooling.15 Less load means less distance from the outside of the load to the middle, and so less difference between them.
ULVAC says it holds more than 70% of the market for continuous lines in each of the major magnet processes: melting, sintering and aging.16 Take the exact share with salt (it is the company's own estimate), but nobody disputes that these lines are how the high-volume end of the industry runs.
What you give up is flexibility. Every chamber is tuned to one step of one recipe. Changing a recipe, or the part mix, means re-tuning the whole line, and the cost of changing over is the real price of all that efficiency. If your process still changes every month, a continuous line will punish you for it. Buy one when the process is mature.
Short versionEach design gives up something to get something else. The grid puts the trades side by side.
8 · Side by sideThe four designs, scored
The grid scores each design on a three-step scale. It is a judgment, not a measurement, and the scores are mine.
The tube gives up scale and quench for low cost and accuracy. The round furnace did the industry's work for decades and still does in plenty of plants, at the price of tuning, element wear and a quench you can't fully predict. The box accepts a predictable top-to-bottom difference in exchange for uniformity it keeps without babysitting and a hard quench that blocks survive. A continuous or semi-continuous line gives up flexibility for energy and consistency.
Which trade suits a plant depends on how many recipes it runs, how much it makes and how settled its process is. That part is a conversation, not a chart.