Adiabatic Demagnetization: Turning Off a Magnet to Reach a Thousandth of a Degree

The core idea: entropy that depends only on B/T

A paramagnet is a lattice studded with independent magnetic moments — unpaired electron spins in a salt, or nuclear spins in a metal. Each moment of angular-momentum quantum number J has 2J+1 orientations. With no field and no interactions, all orientations are equally likely, so the spin system's entropy is maximal:

• S_spin = R·ln(2J+1) per mole (its ceiling)
• For a spin-½ ion, 2J+1 = 2, so S = R·ln2 = 5.76 J·mol⁻¹·K⁻¹

Apply a field B and the Zeeman energy −μ·B splits the levels by ΔE ≈ g·μ_B·B. When the splitting rivals the thermal energy k_B·T, the low-energy (aligned) states win the Boltzmann lottery, the moments order, and the spin entropy drops. The crucial fact is that the ordering is controlled by a single dimensionless ratio:

• x = μB/(k_B·T) — the entropy of an ideal paramagnet is a function of B/T only.

So if you change B and T together such that S stays fixed, you must keep B/T fixed. That is the whole engine of ADR: hold entropy constant while you lower B, and T is forced to fall in proportion.

The two-stroke cycle: magnetize hot, demagnetize cold

ADR is a two-stroke process, geometrically a rectangle-and-slide on the S–T diagram:

• Stroke 1 — isothermal magnetization. With the salt thermally clamped to a heat bath at T_i (a pumped-helium or dilution-fridge stage), ramp the field from ~0 to B_i (say 1–3 T). The spins align, S_spin falls, and the released ordering energy — the latent heat of alignment — flows out into the bath. Entropy is dumped; the salt stays at T_i.
• Stroke 2 — adiabatic (isentropic) demagnetization. Break the thermal link (open a mechanical or superconducting heat switch) so the salt is isolated. Now slowly reduce B toward B_f. With no heat flow, total entropy is conserved. The spins want to re-randomize, but that costs entropy the isolated system cannot pay from outside — so it robs it from the lattice vibrations. The phonon (lattice) temperature plunges until the spin and lattice entropies rebalance.

For an ideal paramagnet where S = f(B/T), constant entropy means B_i/T_i = B_f/T_f, giving the clean result:

• T_f = T_i · (B_f/B_i)

Ramp from 2 T down to 10 mT starting at 1 K and, ideally, you reach 5 mK — a 200-fold drop in one stroke.

The magnetocaloric equation and where the cold comes from

The differential heart of ADR is the magnetocaloric effect. Along an adiabatic (dS = 0) path, a Maxwell relation from the free energy gives the temperature change per unit field:

• (∂T/∂B)_S = −(T/C_B)·(∂M/∂T)_B

Here C_B is the heat capacity at fixed field and M the magnetization. Because a paramagnet obeys the Curie law M = C·B/T (Pierre Curie, 1895), the magnetization falls as temperature rises, so (∂M/∂T)_B is negative. Lowering B (dB < 0) then makes dT negative — the sample cools. The energy bookkeeping is honest:

• The spins absorb energy T·dS_spin to re-disorder;
• Adiabatic isolation forces dS_total = dS_spin + dS_lattice = 0;
• So the lattice loses entropy and cools, transferring T·dS to the spins.

The cooling power per field sweep is set by how much entropy the spin bath can soak up — essentially its heat capacity. Once the spin entropy has been fully unfrozen (S back near R·ln(2J+1)), the salt has no more capacity and warms up as heat leaks in. That is why ADR is a single-shot refrigerator: it cools, holds for minutes to hours, then must be recharged.

What sets the floor: the internal field b

The idealized T_f = T_i·(B_f/B_i) predicts absolute zero as B_f → 0 — which the Third Law forbids. The escape hatch is that real spins are never truly independent. Even with the external magnet off, each moment feels an internal field b from dipole–dipole coupling, exchange, and hyperfine interactions. The correct expression replaces B with an effective field:

• T_f = T_i · √(B_f² + b²) / √(B_i² + b²)

As B_f → 0 the temperature saturates at T_f ≈ T_i·b/√(B_i²+b²), not zero. The internal field b is the entire game in choosing a salt:

• Cerium magnesium nitrate (CMN) has b of only a few millitesla and orders magnetically near 2 mK — long the coldest electronic paramagnet, used as a primary thermometer.
• Chromium/iron ammonium alums have larger b (tens of mT) and floor out near 3–15 mK.
• Gadolinium gallium garnet (GGG) and gadolinium lithium fluoride (GLF) are dense, high-J solids favored for continuous space-borne ADR because they carry large entropy per volume.

Below the ordering temperature the spins freeze on their own and the magnetocaloric handle is gone — that internal field is the physical wall between electron ADR and true absolute zero.

Nuclear demagnetization: the same trick, 2000× deeper

To beat the millikelvin floor, swap electron spins for nuclear spins. A nuclear magnetic moment is of order the nuclear magneton μ_N = 5.05×10⁻²⁷ J/T — about 1/1836 of the Bohr magneton — so nuclear dipole–dipole coupling, and hence the internal field b, is smaller by a similar factor. Nuclei stay paramagnetic to nanokelvin spin temperatures.

The price is that the tiny moment demands enormous B/T to polarize: μ_N·B/k_B ≈ 0.37 mK per tesla, so an 8-tesla field only imposes x ~ 1 at about 3 mK. Nuclear demag therefore starts from a dilution refrigerator pre-cool near 10 mK, ramps a metal (typically copper or PrNi₅) in 6–9 T, isolates it, and demagnetizes toward a few mT:

• Helsinki's Low Temperature Laboratory drove rhodium nuclear spins to a record spin temperature of about 100–280 pK — the coldest temperature ever recorded in matter (copper nuclei, the classic first-stage material, reach the tens-of-nanokelvin range).
• The conduction electrons and lattice, coupled to the nuclei by the Korringa mechanism, follow to the low-microkelvin range, cooling attached samples for studies of superfluid ³He and nuclear magnetism.

PrNi₅ is special: hyperfine enhancement multiplies the effective nuclear moment, giving big cooling power per mole and letting a single stage reach ~0.4 mK from a modest field.

Where it earns its keep — and where it's headed

ADR is not merely a laboratory curiosity; it is the workhorse cooler where cryogens are inconvenient:

• Space cryogenics. Because ADR needs only a magnet and a heat switch — no gravity-dependent liquid — it cooled the microcalorimeter detectors on Japan's Hitomi and XRISM X-ray observatories to ~50 mK in orbit, where energy resolution of a few eV demands a rock-steady sub-100-mK bath.
• Continuous ADR. Chaining two or more stages with alternating heat switches yields a cyclic cooler that holds a detector at 50–100 mK indefinitely, replacing the single-shot limitation.
• Quantum hardware. ADR and nuclear-demag stages provide the ultra-low base temperatures for superconducting-qubit and topological-matter experiments below what a dilution fridge alone delivers.
• Room-temperature magnetic refrigeration. The same magnetocaloric effect near the Curie point of gadolinium (T_C ≈ 293 K, giving roughly 2 K of cooling per tesla) is being engineered into solid-state, gas-free refrigerators — a green alternative to vapor-compression.

From Giauque's 0.25 K pill (Nobel Prize, 1949) to rhodium nuclei at 100 pK, every step is the same accounting: keep entropy fixed, lower the field, and let the spins extract the cold.

Electron-spin vs nuclear-spin adiabatic demagnetization: same physics, moments ~2000× smaller.

Property	Electron ADR (paramagnetic salt)	Nuclear demagnetization
Magnetic moment	~1 μ_B = 9.27×10⁻²⁴ J/T	~1 μ_N = 5.05×10⁻²⁷ J/T
Starting temperature	1–4 K (pumped He)	~10 mK (dilution fridge)
Field used	1–3 T	6–9 T
Final temperature	1–10 mK (salt-limited)	1 μK down to ~100 pK
Limiting scale	Internal field b ~ few mT	Nuclear ordering / dipolar b
Working substance	CMN, Fe/Cr alum, GGG	Copper, PrNi₅ nuclei