Magnetic coupling for an intravascular blood pump

A medtech manufacturer wants to take a blood pump on a catheter into series production. The prototype measures 6 mm and runs on SmCo magnets, which need an export licence from China. The series design measures 4 mm and uses NdFeB without heavy rare earths: it transmits 1.12 mNm, 91 % of the prototype, and the signal at the Hall sensor deviates by only 3.8 %. BOMATEC overmolds the outer magnet into the impeller, qualifies the process and secures a supply without an export licence.

• 6 → 4 mm outer diameter, one third smaller
• 1.12 mNm pull-out torque, 91 % of the prototype
• 0 licences: NdFeB without heavy rare earths instead of SmCo

Representative case, combined from two BOMATEC projects and presented at MEDICA TECH FORUM 2026. Dimensions and figures come from a magnetic field simulation with assumed geometry, not from customer data. The measured value for the model check is assumed as well.

Smaller, just as strong and licence-free: each of these requirements makes the others harder. That is why they have to be solved together.

• −56 % magnet cross-section if every dimension shrinks with the diameter from 6 to 4 mm. It takes more remanence than SmCo delivers: hence NdFeB.
• −10 % flux density that a magnetized NdFeB magnet loses during overmolding. Its coercivity drops two to three times as fast with temperature as that of SmCo.
• +Dy/Tb: The usual remedy restores the coercivity and brings the export licence back.

Before we change anything, we rebuild the 6 mm prototype as a magnetic model: magnets, titanium wall, impeller hub and Hall sensor. The model gives a pull-out torque of 1.23 mNm and 37.4 mT at the sensor. Then we check it against a measurement on the real part; in this representative case the measured value is assumed.

The flushing sleeve sits in the field. That is why we specify it as non-magnetic: 316L instead of a magnetic steel, because its walls are too thin for PEEK. We check its influence with FEM.

Model check: 37.4 mT simulated, 38.6 mT measured (assumed value). Deviation −3.0 %.

We vary the inner magnet, ring thickness, length, grade, number of poles and segments. Two targets apply at the same time: pull-out torque 0.98 to 1.47 mNm and field at the sensor within ±5 %.

Simply scaled down, 0.33 mNm remain. Four segments or four poles add torque, but cost 41 % or 91 % of the sensor signal. Both targets come back mainly through a thicker outer ring: only the 316L flushing sleeve, 0.15 mm thin, covers it (prototype: 0.5 mm wall). On top of that come 7.5 instead of 5 mm length for the sensor signal and BMN-52SH, the strongest grade free of heavy rare earths.

Result: 1.12 mNm and −3.8 % at the sensor, the best of 156 candidates.

We overmold the outer magnet directly into the impeller, at our plant in Malaysia. Magnet, flushing sleeve and impeller are joined and sealed in the same shot: no adhesive, no separate seal in the blood path. The 316L flushing sleeve forms the outer wall over the Parylene-coated magnet, the plastic closes the end faces.

The flushing sleeve goes into the mold as an insert; its channels are laser-cut after overmolding. Grade and volume make up for the flux loss of about 10 % from overmolding.

New magnet material, new process: IQ, OQ and PQ run again. We document our process step, as a full PPAP on request, and support you up to the FMEA.

• Every batch tested: magnetic properties, dimensions, coating (salt spray test 96 h)
• Sampling to your specification, often 100 %: your incoming inspection can be dropped
• Traceable to the raw material batch, on-site audits welcome
• IATF 16949 with the same test equipment in Switzerland, Malaysia and China

Our plant in Malaysia injection molds, overmolds, presses and assembles, outside China. Sintered NdFeB magnets without heavy rare earths need no export licence; on request we source them from partners outside China.

• Consignment stock at the customer: eases their cash flow
• Safety stock in Switzerland, sized to demand, for up to two years

• Representative geometry, no customer dimensions. Titanium wall 0.2 mm, gaps 0.05 mm each, flushing sleeve over the ring 0.15 mm.
• Magnetostatic, without speed, temperature and eddy currents. Nominal remanence at 20 °C.
• Both magnets 2-pole, one-piece, diametrically magnetized. Overmolding loss 10 % on the NdFeB ring.
• Flushing sleeve and shaft non-magnetic and therefore not modelled, a ceramic bearing between magnet and sensor.
• Targets apply against the simulated prototype. The measured value at the sensor is assumed and serves only for the model check.
• China's export rules: as of November 2026.

Our engineers check the flushing sleeve with its channels and the eddy currents in the titanium wall with FEM before the process is qualified.

1. Treat the magnet as a design variable from the first sketch.
2. Validate the simulation on the prototype before you optimize.
3. Plan qualification and supply in the concept phase, not just before SOP.
4. Talk to us early. Then we design the magnet around your other requirements.

Why not simply stay with SmCo?

Samarium has been on China's control list since April 2025, so SmCo from China needs an export licence. The authority has up to 45 working days to decide, it often takes longer, and the rules change with politics. We have no source for SmCo outside China.

Why not NdFeB with dysprosium or terbium?

These heavy rare earths make NdFeB more temperature-resistant, but are themselves subject to China's export controls. We would have swapped one licence for the next.

Why does the coupling get longer?

At 4 mm, magnet volume is missing. A thicker ring brings most of it back, possible because only the 0.15 mm thin flushing sleeve covers it. The sensor signal also needs length: 7.5 instead of 5 mm. More than 1.5 times the prototype length was not allowed.

Why not magnetize after overmolding?

Because the outer ring is too small for that at 4 mm. It is magnetized beforehand and loses about 10 %; grade and volume make up for it.

Can this be transferred to other products?

The approach, yes. Whether a licence-free material is enough, we check for each magnet individually.

A coupling, a sensor magnet, an actuator: wherever the magnet turned out to be harder than expected. Patrick Muschak, Sales Engineer Medical & Life Sciences, looks forward to hearing from you.