Can rare-earth-free materials match HoCu2 cooling at 4 K?

The 4 K gap is a specific-heat gap

A Gifford–McMahon cryocooler reaches liquid-helium-free temperatures by cycling helium gas through a packed bed of regenerator particles, which must absorb and release heat over each cycle. Near 4 K that job is normally done by heavy rare-earth intermetallics such as HoCu2, whose magnetic ions carry large moments. The paper frames the problem as a materials problem: find a non-rare-earth magnet with enough volumetric heat capacity in the 4–7 K window.

The earlier rare-earth-free candidate, the delafossite CuFe1-xAlxO2 (CFAO), already matched HoCu2 near 10 K, but its magnetic specific heat falls away below about 7 K, leaving it at roughly 50% of the benchmark at 4.2 K. The authors therefore looked for a second frustrated magnet whose peak sits where CFAO's declines, rather than trying to improve CFAO itself.

Manganese spins order low despite strong coupling

The candidate is MnNb2-xTaxO6 (MNTO), a columbite-structure antiferromagnet in which Mn2+ ions sit on an isosceles triangular lattice in the ab plane. Mn2+ has a half-filled 3d shell with spin S = 5/2, the largest available among transition-metal magnets, so its magnetic entropy scales with ln(2S+1); a high density of such spins per unit volume is what makes a large volumetric heat capacity possible.

The catch is that strong exchange usually pushes ordering to high temperatures, as in MnO or NiO. Here geometrical frustration intervenes: the Weiss temperature is about −24 K for all compositions, yet the end compounds order at Néel temperatures of only 4.4 K (MnNb2O6) and 5.7 K (MnTa2O6). That separation between interaction strength and ordering temperature is what places a sharp specific-heat peak right where the cryocooler needs it.

MNTO peaks near HoCu2 but not above it everywhere

Measured specific heat tells the more concrete story. Below about 8 K, MnNb2O6 and MnTa2O6 exceed the nonmagnetic metals Pb and Bi, and the maximum specific heat of MnNb2O6 below 5 K is 0.31 J K⁻¹ cm⁻³, essentially level with HoCu2 at 0.32 J K⁻¹ cm⁻³ and far above CFAO at 0.10 J K⁻¹ cm⁻³. The paper notes that near 4 K the specific heat of MnNb2O6 surpasses that of HoCu2.

The comparison is temperature-dependent, not uniform. Above 6 K the manganese compounds have lower specific heat than CFAO, and CFAO's own peak lies at higher temperature. That crossover is the reason the two materials are used together rather than as substitutes: each covers the range where the other is weak.

Tantalum substitution shifts ordering without weakening it

To check whether an intermediate composition works better, the authors synthesized the full solid solution x = 0 to 2 and measured magnetic susceptibility and specific heat. Increasing Ta content moves the Néel temperature monotonically upward, while the Curie–Weiss fit gives a Weiss temperature of about −24 K for every sample. The frustration parameter |ΘCW|/TN therefore falls as x rises, which the authors attribute to Ta substitution changing the ratio of the inter-chain exchange interactions J1 and J2 on the isosceles triangular network.

The sharpness of the peak at TN survives across the whole series, so long-range order remains robust even as the exchange balance changes. In the cooling tests, however, compositions x = 0, 0.5 and 1.0 gave similar cooling power, so the composition sweep did not by itself produce a clear winner on performance.

A three-layer bed turns two peaks into one gain

Cooling tests used a standard two-stage GM cryocooler with the second stage spanning 35 to 4 K. The packed bed was split along the temperature gradient: spherical lead particles at the hot end, and the magnetic material at the cold end. With MNTO alone at 30% of stage volume and 70% Pb, the best result was 0.088 W at 4.2 K, and all three MNTO compositions performed similarly.

The improvement came from a three-layer composite with MNTO at the coldest end, Pb at the hottest, and CFAO sandwiched between them. The best configuration, 30% MNTO (x = 1.0) with 20% CFAO and 50% Pb, delivered 0.163 W at 4.2 K and reached a minimum temperature of 2.80 K. That is about 40% above CFAO alone at 0.117 W and 73% of HoCu2 at 0.220 W, and it exceeds the 0.1 W nominal specification of the commercial cooler used. The cooldown curve from 300 K was almost the same as for the HoCu2 sample.

Simulation isolates the material contribution from packing

A cooling gain could come from better packing or particle shape rather than from the material. To separate the two, the authors ran one-dimensional thermal-fluid simulations in REGEN3.3 with idealized spherical particles of 250 µm diameter and a packing fraction of 63%, a setup calibrated so that the Pb/HoCu2 case yields 0.20 W at 4.2 K.

Under those fixed geometry assumptions, the simulation gave 0.148 W for the 30% MNTO/20% CFAO/50% Pb composite against 0.163 W measured, and it reproduced the large enhancement when MNTO is added to CFAO. The authors therefore conclude that the gain is an intrinsic material effect. The agreement is qualitative rather than exact: the model uses idealized heat-transfer coefficients and simplified one-dimensional gas flow, and the real crushed particles are irregular, which raises flow resistance. The simulation supports the mechanism; it is not a predictive optimization tool.

Packing fraction and particle shape cap the next gain

The measured samples were crushed particles of random shape, giving an actual packing fraction of about 60%. The authors estimate that nearly spherical granules could raise this above 65%, adding cooling power and reducing pressure drop. Thermal conductivity does not appear to be the bottleneck: MnNb2O6 shows 2–40 W m⁻¹ K⁻¹ over 4–20 K, an order of magnitude above HoCu2 at 0.28 W m⁻¹ K⁻¹, and the estimated thermal equilibration time is about 0.005 s for a 0.5 mm particle, short enough for each cycle.

The particle-size choice is a trade-off rather than a solved optimum. The 200–500 µm range used here is judged reasonable given the high conductivity, but the paper states that further experiments across particle sizes are needed for full optimization. Structural integrity was checked: particles remained intact without fracture, and specific-heat measurements on recovered samples matched the pre-test values.

73% of HoCu2 reframes the rare-earth question

The decision-relevant number is not that a rare-earth-free material exists, but where it now sits relative to the benchmark: 0.163 W versus 0.220 W at 4.2 K, with a 2.80 K minimum temperature and a cooldown rate matching the HoCu2 sample. For a 4 K GM cryocooler, that places an entirely rare-earth-free three-layer bed in the range of the incumbent rather than in a separate, lower-performance class.

The remaining gap is now specified in engineering terms rather than in materials terms. The authors locate the next increment in packing fraction above 65% through spherical granulation and in reduced pressure drop, not in finding a magnet with a larger specific-heat peak. Readers weighing this route should treat the 73% figure as tied to one cooler, one bed geometry and 4.2 K, and treat particle shape and packing as the variables most likely to move it.

Sources used in this answer