How to Balance Magnetic Energy Product, Coercivity, and Cost in High-Performance Motors

Motor designers rarely get to pick a magnet in isolation. The grade sets the remanence and the intrinsic coercivity you can work with. Geometry, leakage, and temperature then decide how much of that remanence becomes air-gap flux, how close the operating point sits to the knee, and what magnet cost the program has to defend. Push energy product up and cost often follows. Push coercivity up with bulk heavy rare earth and remanence usually drops. Hold the magnet price down and you can give away demagnetization margin. The decision is not which grade is best in a catalog. It is where, on the remanence–coercivity–cost surface, this machine still meets torque, efficiency, and durability at a cost the program can carry.

That question is sharper than a single catalog sort. Rare-earth prices move, dysprosium and terbium are costly and supply-constrained, and traction and industrial servo motors keep asking for more torque in less volume. Grain-boundary diffusion (GBD) is now a routine process route: coercivity can often be raised with a smaller remanence penalty than the same heavy rare earth dissolved through the grain. The tradeoff moves with the process route, so selection has to move with it.

The decision engineers actually face is a three-way constraint, not a single-material choice

Every motor-magnet choice sits on three constraints that pull against each other.

• Energy product is a catalog marker for torque density, not the air-gap flux itself. In a given sintered NdFeB family, a higher (BH)max grade usually has higher remanence Br. For a fixed rotor and a fixed electrical loading, more usable air-gap flux means more torque from the same stack, or the same torque from a smaller rotor — but only while the magnet stays above the knee. (BH)max is the peak of the BH curve. It is not the flux density in the gap.
• Intrinsic coercivity Hcj sets the demagnetization ceiling, together with the load line. It limits how hot the magnet can run, how much reverse field it can take from the stator in a fault or a high-current stall, and how low a permeance coefficient it can tolerate before the operating point crosses the knee. A common severe case is not cruise temperature alone. It is a short circuit or a stall while the magnet is already hot, so armature reaction adds reverse field on top of the thermal loss of Hcj. Room-temperature Hcj is not that margin.
• Cost is not the price per kilogram. Heavy rare earth dominates raw-material cost, but coating, tolerance, segmentation, and scrap on a complex geometry can move the bill as much as a grade step. A grade that shrinks the rotor, or that reaches the required Hcj with less bulk heavy rare earth, can lower system cost at a higher unit price.

These three meet at the operating point. They are not independent knobs. Buying Hcj with bulk Dy or Tb typically spends Br, so thermal margin is paid for in torque density. Buying Br back can spend Hcj, so torque density is paid for in demagnetization risk. The work is to find which constraint actually binds on this machine.

Candidate grades should be compared on the axes that decide the motor program

Default grade tables list Br, Hcj, and (BH)max. A selection table also needs the axes that move a program: demagnetization-limited temperature, heavy-rare-earth content, margin at the machine’s own worst case, and relative cost. The table is a qualitative class contrast. It is not measured data, not a supplier catalog, and not an AIC Engineering product list. It is not a basis for qualification. Confirm every number on the datasheet and the quotation for the grade you intend to buy.

Two readings follow, both qualitative. First, a GBD-treated grade is often the class to check first when hot reverse-field margin is what binds: diffusion is used to place heavy rare earth at the grain boundary, so Hcj rises with less Br loss than bulk alloying, and with less heavy rare earth in the magnet. “GBD” is not one point. Diffusion depth and the resulting Br–Hcj pair are grade-specific; read the datasheet. Second, ferrite is not a defective NdFeB grade. It is a different material family: much lower remanence, no heavy rare earth, and a lower material cost, paid for in volume and mass. Its coercivity and its worst temperature are not the NdFeB hot-fault ranking in the columns to the left. Hard ferrite typically loses remanence as it gets hot, while intrinsic coercivity rises with temperature, so a cold condition can be the harder demagnetization case. That is a reason to open the ferrite curve, not a reason to copy an NdFeB margin rule.

A low unit price that fails the real worst case is not a low cost.

Torque and cost should be reasoned from the load line, not from catalog remanence alone

When rotor geometry and cooling are defined, the magnet’s effect on torque and cost can be bounded without pretending to predict either. The relations below are first-order checks. They are not a sizing method. Close torque in electromagnetic FEA, and close cost on a quotation, before tooling.

Air-gap flux depends on remanence and on the load line. Catalog Br is the closed-circuit flux density, not the gap flux density. On a straight demagnetization line, with a dimensionless SI permeance coefficient defined as $P_c=|B_m/({\mu }_0{H}_m)|$,

Here $B_m$ and $B_r$ are flux densities, and $P_c$ and the recoil permeability ${\mu }_{\mathrm{r}\mathrm{e}\mathrm{c}}$ are dimensionless, so the expression is dimensionally consistent. For sintered NdFeB, ${\mu }_{\mathrm{r}\mathrm{e}\mathrm{c}}$ is near unity, which is why $P_c$ matters: a thinner magnet, a larger effective gap, or more leakage lowers $P_c$ and pulls $B_m$ down the curve. This $B_m$ is inside the magnet. Gap density is lower again by the magnet-to-gap area ratio and by leakage. Do not put $B_m$ in a torque estimate and call it $B_g$.

A permeance coefficient quoted in gauss per oersted is a cgs convention. It is not the SI $P_c$ above. Do not mix them.

The linear expression is already invalid once the operating point reaches the knee. That is exactly the demagnetization case. Do not use it to sign off margin. Thickness is still a real lever: it moves $P_c$ without adding rare earth, and it has to be chosen with the grade, not after it.

Torque follows gap flux only inside a stated regime. In the current-limited range, before the iron saturates, torque scales roughly with air-gap flux, and that flux scales with Br only along the load line you actually have. A modest loss of Br can be traded for a useful gain in Hcj if the load line and the thermal margin hold. The scaling misses saturation, winding factor, and pole count. Cutting magnet volume to save cost can cut flux, and therefore torque, by more than the mass change suggests.

Cost per unit torque is not the magnet price list. No closed cost-per-torque formula is used here. Heavy rare earth is already inside material price if the quotation is for a finished grade; adding it again double-counts. Coating, segmentation, tolerance, and scrap sit in the same numerator. The denominator is delivered torque at the required duty, including whether the grade shrinks the rotor or eases cooling. Compare that outcome, not the price per kilogram.

The workable window is set by the worst-case point and by margins you can actually hold

For a high-performance sintered-NdFeB motor, a usable selection window is procedural, not a universal set of factors.

1. Qualify against the worst demagnetization case, not the nominal point. Fix the maximum magnet temperature and the reverse field in the fault or stall you actually have to survive. Then require the knee of the intrinsic curve, at that temperature, to stay clear of the operating point. Room-temperature Hcj does not make that check.
2. Hold a margin that covers the scatter you really have. Hcj varies by lot, magnet temperature is uncertain, and fault current is not a single number. How much clearance that demands is a project decision. It is not a fixed multiplier that can be copied from one machine to another. Extra margin that the fault and the thermal model never use is heavy rare earth you are paying for.
3. Hold magnet thickness on the load line before you buy a higher grade. Reducing thickness lowers $P_c$ and walks the operating point toward the knee. There is no universal thickness-to-gap ratio at which risk suddenly appears. A magnet that is thin relative to the gap and the leakage path can demagnetize even if the grade name looks conservative. Recover load line and grade together.
4. Look at GBD grades when hot coercivity is the binding constraint. Supplier literature on grain-boundary diffusion describes a smaller Br penalty, and less heavy rare earth, than bulk Dy/Tb addition for a similar coercivity gain. Treat that as a reason to request the Br–Hcj pair, not as a guaranteed offset. Confirm both numbers, and the recommended temperature limit, on the grade you will qualify.
5. Check the hot operating point in FEA, and in hardware if the risk justifies it. Behavior on the knee is nonlinear. A paper load line will not catch a local reverse field at a magnet corner, or a magnet that runs hotter than the winding sensor.

The right clearance depends on temperature uncertainty and on how severe the electrical fault is. Copying another program’s factor does not transfer that physics.

This tradeoff fails in known ways, and those limits should bound the selection

The balance fails in repeatable ways. The list is limited to modes that follow from the load line, the NdFeB temperature coefficients, and ordinary manufacturing cost. It is not a field-failure survey.

• Irreversible demagnetization on the knee. If the worst case is a terminal short or a stall at high magnet temperature, a high-Br, low-Hcj grade can lose flux permanently. Cooling does not restore irreversible loss. This is the usual way a price-driven NdFeB choice fails in service. A linear load-line formula will not show it.
• Magnet temperature taken from the wrong sensor. Winding temperature is not magnet temperature. Depending on loss split and the cooling path, the magnet can run hotter or cooler than the copper. An optimistic thermal model produces an aggressive grade. Where the margin is tight, measure magnet temperature, or bound it, rather than inferring it from the winding.
• Thickness cut to save cost, grade left unchanged. A thinner magnet lowers the permeance coefficient and spends demagnetization margin even when the grade code is the same. Geometry and grade are one decision.
• A low-remanence substitute forced into an NdFeB envelope. Ferrite is a legitimate cost class when volume and mass can grow. It is not a drop-in for a volume-limited or mass-limited rotor: the remanence gap is large, and the missing flux does not appear because the grade is cheaper. Do not apply the NdFeB hot-fault margin to ferrite. Its limiting temperature is often different, and cold demagnetization has to be checked on the ferrite curve.
• Heavy rare earth added “for safety.” Hcj the machine never uses is cost without a torque or durability return, and it ties the bill to the tightest part of the rare-earth supply.
• Coating and tolerance left out of the comparison. Corrosion protection and the tolerance on a segmented or ring geometry can erase the saving from a cheaper grade. A simple geometry in a slightly higher grade can cost less than a hard-to-hold shape in a cheaper one.
• Signing the grade off on room-temperature Hcj. NdFeB intrinsic coercivity falls as temperature rises. A grade that looks comfortable at 20 °C can be on the knee at the real magnet temperature. The datasheet curve at temperature, on the real load line, is the selection document.

This window assumes a sintered NdFeB rotor whose worst magnet temperature and worst reverse field can be stated. If either is still unknown, grade selection is early. The same numerical window does not transfer to a material family with a different demagnetization curve or different temperature coefficients.

AIC Engineering supports this selection window with circuit design, prototypes, and inspection

The AIC Engineering team works with motor designers on the selection this article describes: grade, geometry, and thermal margin set against the worst-case operating point, not against the nominal point. Within that scope, the work includes:

• Magnetic-circuit and structure design together, so load line, magnet thickness, and grade are chosen as one set rather than in sequence.
• Special motor magnet assemblies — multi-pole rings, radially oriented rings, Halbach arrays, and linear-motor magnet sets — where the flux path can be shaped to the machine. Whether a given geometry reduces heavy-rare-earth content is a result to calculate, not a property of the assembly type.
• Rapid prototyping in 3–7 days, so candidate grades and geometries can be in hand before tooling is committed. That interval is prototype lead time. It is not a hot-fault or thermal-soak test.
• Permanent-magnet quality inspection, so incoming magnetic properties and consistency can be checked against the grade that was qualified. Inspection confirms the received material. It does not by itself create the Hcj margin; the margin still has to be designed and, where needed, tested at temperature.

Where production volume and qualification timing require it, global supply and regional delivery support can be aligned with the program schedule.

Next steps start from the worst-case operating point, not from the catalog

1. State the worst-case point first — maximum magnet temperature, reverse field under the fault or stall you must survive, and the torque-density target — before opening a grade list.
2. Place candidate grades on the axes above, and compare total magnet cost against delivered torque, including coating, scrap, rotor size, and cooling. Do not stop at price per kilogram.
3. Set the demagnetization clearance from your own temperature uncertainty and fault current, then check it in FEA on the intrinsic curve at temperature. Add a hot-state prototype test when the margin is thin or the thermal model is weak.
4. Contact AIC Engineering for custom magnetic-circuit design and rapid prototyping — visit https://www.aicmagnetics.com for a free consultation and tailored engineering solutions. Bring the worst-case operating point. The team will help find the grade, geometry, and margin that balance energy product, coercivity, and cost for your machine.

References

1. Coey, J.M.D., Magnetism and Magnetic Materials, Cambridge University Press — remanence, coercivity, energy product, and the contrast between rare-earth and ferrite permanent magnets.
2. Parker, R.J., Advances in Permanent Magnetism, Wiley — magnetic-circuit design and load-line methods.
3. Magnet-supplier technical handbooks on sintered NdFeB grades and grain-boundary diffusion — the source for any specific Br–Hcj pair or recommended temperature limit. The class table in this article is not a substitute.
4. Standard permanent-magnet synchronous-machine design texts on torque constant, magnetic loading, and demagnetization. They are background, not a qualification record for a grade.