The Schottky Anomaly

Cool a paramagnetic salt toward absolute zero and its heat capacity does something a metal or an insulator never does: it rises out of nothing, swells into a rounded hump, and then falls away again, all within a fraction of a kelvin. There is no phase transition, no latent heat, no broken symmetry — just a smooth bump sitting where the lattice contribution has long since frozen out. This is the Schottky anomaly, and it is the unmistakable calorimetric fingerprint of a system whose particles have only a finite number of accessible energy levels. The simplest such system, a set of two-level units split by an energy gap Δ, produces the whole effect exactly. The physics is pure statistics. At temperatures far below Δ/k_B the upper level is unreachable and the system is frozen: no way to absorb heat, so C→0. At temperatures far above Δ/k_B both levels are essentially equally populated and stay that way, so adding heat barely shifts the populations and again C→0, now as a T⁻² tail. In between there is a sweet spot, near k_BT≈0.42Δ, where a small rise in temperature moves a large number of units across the gap and the system drinks heat greedily. The result is a universal curve with a peak of height C_max≈0.439 Nk_B. This article derives that curve from the two-level partition function, pins the peak location and height by direct computation, shows that the area under C/T recovers exactly the k_B ln 2 of entropy locked in the two levels, and explains why the bump is called an "anomaly" — because experimentally it appears as an unexpected hump riding on top of the ordinary lattice heat capacity. The magnetization of the same two-level paramagnet is owned elsewhere; here we own its heat capacity.

Key takeaways

A system of two-level units (gap Δ) has partition function Z = g₀ + g₁e^{−Δ/k_BT}; its heat capacity rises, peaks, and falls — the Schottky anomaly.
For equal degeneracies the curve is C/Nk_B = x²eˣ/(eˣ+1)² with x ≡ Δ/k_BT, a single universal shape independent of Δ.
The peak sits at k_BT/Δ ≈ 0.417 (x ≈ 2.40) with height C_max/Nk_B ≈ 0.439.
Low T: C ∝ (Δ/k_BT)²e^{−Δ/k_BT} freezes out exponentially (a gap signature). High T: C ∝ (Δ/k_BT)² decays as T⁻².
The area under C/T from 0 to ∞ is exactly k_B ln 2 — the entropy of the two hidden states, recovered as they thaw.
Key idea: a bounded ladder of levels can only store a finite amount of energy, so its heat capacity must vanish at both temperature extremes and bulge in between — the hump is unavoidable, not anomalous.

The core insight

A unit with just two energy levels can absorb heat only while its populations are actually shifting. That happens only in a temperature window around k_BT ∼ Δ; below it the upper level is frozen out and above it the populations are pinned near equality. A heat capacity forced to zero at both ends must peak in between — the Schottky bump, universal in shape and centred at k_BT ≈ 0.42Δ.

Where this lives today

The Schottky anomaly is a routine diagnostic in low-temperature physics and materials science. Whenever a substance is cooled into the sub-kelvin range, a heat-capacity hump betrays a set of low-lying levels — crystal-field split rare-earth ions, hyperfine-split nuclei, magnetic impurities, or the tunnelling two-level states of amorphous solids — and the peak temperature immediately fixes the splitting while the entropy under the curve counts the levels. In materials chemistry it separates a genuine magnetic ordering transition (sharp, cooperative) from a mere single-ion Schottky feature (broad, independent). The same $T^{-2}$ nuclear tail is a standard nuisance-and-tool in ultralow-temperature calorimetry, and the identical two-level partition function underlies the qubit in quantum information. The magnetization and susceptibility of the same paramagnetic units are treated under the Brillouin function; the lattice background that the hump rides on is the Debye and Einstein heat capacity; and the cooperative version, where interactions turn the smooth bump into a sharp transition, belongs to the theory of phase transitions.

The physical picture

Imagine N identical units, each of which can sit in one of two states: a ground state at energy 0 and an excited state at energy Δ. In a real crystal these might be the two orientations of a nuclear or electronic spin split by a magnetic field or a crystal field, or the two lowest levels of a paramagnetic ion. The units are independent, so the thermodynamics is fixed entirely by how a single unit shares itself between its two levels as a function of temperature, and heat capacity measures how eagerly the population shifts when you nudge the temperature. At very low temperature (k_BT ≪ Δ) essentially every unit is in the ground state. Promoting even one across the gap costs Δ, which the thermal bath cannot supply, so the excited population is exponentially tiny and does not change appreciably with a small temperature rise. No population change means no energy absorbed: the heat capacity is frozen to zero. This exponential freeze-out, C ∝ e^{−Δ/k_BT}, is the calorimetric signature of an energy gap and is the same reason a superconductor's electronic heat capacity or a semiconductor's carrier density collapses at low T.

Figure

Two-level unit: a gap Δ that fills as temperature rises

At very high temperature (k_BT ≫ Δ) the gap is negligible compared with the thermal energy, so both levels are almost equally occupied: half the units up, half down. Now the populations are pinned near 50/50 and a further temperature increase moves them only slightly (the residual imbalance scales as Δ/k_BT), so the energy the system can still absorb from a temperature rise dwindles. The heat capacity therefore falls again, this time as a slow T⁻² power law rather than an exponential. Because C must vanish at both extremes yet is positive in between, a peak is inescapable — that peak is the Schottky anomaly.

Figure

The Schottky heat-capacity bump, peak marked at (0.417, 0.439)

Everything about the curve — its universality (the same shape for every Δ once you scale the temperature by Δ/k_B), the exact peak position, and the fact that the area under it is the entropy of two states — falls out of the two-level partition function with no free parameters. That is what makes the Schottky anomaly such a clean laboratory: measure the hump, read off Δ from where it peaks, and confirm the number of levels from the entropy under it.

The core equations

A single unit has two levels; its Boltzmann statistics follow from the canonical partition function. Recall the canonical partition function machinery is owned by Canonical Partition Function; here we specialise it to two levels and extract the heat capacity.

Z_1 = g_0 + g_1\,e^{-\Delta/k_BT} \qquad (1)

where: $Z_1$ is the single-unit partition function, $g_0$ and $g_1$ are the degeneracies of the ground and excited levels, $\Delta$ is the energy gap between them, $k_B$ is Boltzmann's constant, and $T$ the temperature. For $N$ independent units the total partition function factorises, $Z = Z_1^N$ , so every extensive quantity is $N$ times the single-unit result. The mean energy per unit follows from the Boltzmann-weighted average over the two levels.

\langle E\rangle = \Delta\,\frac{g_1 e^{-\Delta/k_BT}}{g_0 + g_1 e^{-\Delta/k_BT}} = \Delta\, p_1 \qquad (2)

where: $p_1$ is the probability that a unit sits in the excited level. At low $T$ , $p_1\to 0$ ; at high $T$ , $p_1\to g_1/(g_0+g_1)$ , which is $\tfrac12$ for equal degeneracies. The energy is simply the gap times the excited-state occupation. The heat capacity is the temperature derivative of the mean energy; introducing the dimensionless variable $x\equiv\Delta/k_BT$ collapses it to a universal function.

\frac{C}{Nk_B} = \frac{g_0 g_1\,x^2\,e^{x}}{\big(g_0 e^{x} + g_1\big)^2}, \qquad x\equiv\frac{\Delta}{k_BT} \qquad (3)

where: $x\equiv\Delta/k_BT$ is the gap measured in units of the thermal energy, and the $g_0g_1$ numerator and $(g_0e^{x}+g_1)^2$ denominator carry the degeneracy dependence. For the equal-degeneracy case $g_0=g_1=1$ this reduces to the clean form used throughout, $C/Nk_B = x^2 e^{x}/(e^{x}+1)^2$ . Written out for the symmetric case $g_0=g_1=1$ , the Schottky heat capacity is the single universal curve.

\frac{C}{Nk_B} = \left(\frac{\Delta}{k_BT}\right)^{2}\frac{e^{\Delta/k_BT}}{\big(e^{\Delta/k_BT}+1\big)^{2}} = \frac{x^2 e^{x}}{(e^{x}+1)^2} \qquad (4)

where: all symbols are as above. This function is zero at both $x=0$ (high $T$ ) and $x=\infty$ (low $T$ ) and positive in between, guaranteeing a peak; its shape does not depend on $\Delta$ , only on the scaled temperature $k_BT/\Delta$ . The entropy per unit, obtained from $S=(\langle E\rangle - F)/T$ with $F=-k_BT\ln Z_1$ , interpolates between the two limits and controls the area under the curve.

\frac{S}{Nk_B} = \ln\!\big(1+e^{-x}\big) + \frac{x\,e^{-x}}{1+e^{-x}} \qquad (5)

where: again $x=\Delta/k_BT$ and $g_0=g_1=1$ . As $T\to\infty$ ( $x\to0$ ) this gives $S/Nk_B\to\ln 2$ ; as $T\to0$ ( $x\to\infty$ ) it gives $0$ . The difference, $k_B\ln 2$ per unit, is exactly the area under $C/T$ — the entropy stored in the two levels.

Derivation AThe heat capacity of a two-level system and its universal form

The goal is to obtain equation (4) directly from the partition function (1) with $g_0=g_1=1$ , showing every step from the mean energy to the final heat capacity, and to make transparent why the result depends only on the scaled variable $x=\Delta/k_BT$ and not on $\Delta$ and $T$ separately. The one subtlety is bookkeeping the chain rule: heat capacity is a derivative with respect to $T$ , but the natural variable is $x$ , and $dx/dT=-x/T$ , so a factor of $x/T$ appears and then combines with the gap to build the clean $x^2$ prefactor. Nothing is approximated; the expression is exact for a genuine two-level system.

Step 1. With $g_0=g_1=1$ , the excited-state probability from (2) is $p_1 = e^{-x}/(1+e^{-x}) = 1/(e^{x}+1)$ , the familiar two-level occupation. The mean energy per unit is $\langle E\rangle = \Delta p_1 = \Delta/(e^{x}+1)$ .

Step 2. Differentiate with respect to $T$ . Write $\langle E\rangle = \Delta/(e^{x}+1)$ and use $dx/dT = d(\Delta/k_BT)/dT = -\Delta/(k_BT^2) = -x/T$ . Then $C = d\langle E\rangle/dT = \Delta\cdot\dfrac{d}{dx}\!\left(\dfrac{1}{e^{x}+1}\right)\cdot\dfrac{dx}{dT} = \Delta\cdot\left(-\dfrac{e^{x}}{(e^{x}+1)^2}\right)\cdot\left(-\dfrac{x}{T}\right)$ .

Step 3. Collect factors: $C = \dfrac{\Delta x}{T}\dfrac{e^{x}}{(e^{x}+1)^2}$ . Since $\Delta = x k_BT$ , the prefactor $\Delta x/T = x^2 k_B$ , so $C = k_B\,x^2\,\dfrac{e^{x}}{(e^{x}+1)^2}$ , i.e. equation (4) per unit; multiplying by $N$ gives the extensive result.

Step 4. Read off the universality. The right-hand side is a function of $x=\Delta/k_BT$ alone. So the heat capacities of two salts with different gaps, plotted against $k_BT/\Delta$ , fall on the identical curve — a two-parameter family collapses to one master shape. Measuring where the hump peaks therefore delivers $\Delta$ immediately, with no fitting of the height. Verified: the companion verifier reproduces $C/Nk_B = x^2e^{x}/(e^{x}+1)^2$ symbolically from $\langle E\rangle=\Delta/(e^{x}+1)$ via $dx/dT=-x/T$ , and confirms the low- $T$ form $x^2e^{-x}$ and the high- $T$ form $x^2/4$ .

Go deeper · Derivation B · 1 minPeak location, peak height, and the entropy sum rule

Derivation BPeak location, peak height, and the entropy sum rule

Here the aim is to pin the two numbers that make the Schottky anomaly quantitative — the temperature at which the heat capacity peaks and the height there — and to prove the sum rule that the total area under $C/T$ equals $k_B\ln 2$ , the entropy of the two states. The peak is found by setting $dC/dx=0$ , which yields a transcendental equation solved numerically; the sum rule is proved cleanly by recognising that $\int C\,dT/T$ is just the entropy difference between $T=\infty$ and $T=0$ , which the partition function gives in closed form. The subtlety is that the peak equation has no elementary closed-form root, so the location must be quoted numerically — but it is a pure number, the same for every two-level system.

Step 1. Extremise $C(x)=k_B x^2 e^{x}/(e^{x}+1)^2$ . Setting $dC/dx=0$ and simplifying gives the transcendental condition $x\big(e^{x}+1\big) = 2\big(e^{x}-1\big)$ , equivalently $x = 2\tanh(x/2)\cdot\dfrac{e^{x}+1}{e^{x}+1}$ ; solving numerically gives $x_{\rm peak} = 2.3994$ .

Step 2. Convert to temperature: $x_{\rm peak}=\Delta/k_BT_{\rm peak}=2.3994$ means $k_BT_{\rm peak}/\Delta = 1/2.3994 = 0.4168$ . The peak sits at about four-tenths of the gap temperature.

Step 3. Evaluate the height: $C_{\rm max}/Nk_B = x_{\rm peak}^2 e^{x_{\rm peak}}/(e^{x_{\rm peak}}+1)^2 = 0.4392$ . This is a universal number — every equal-degeneracy two-level system peaks at the same fraction of $Nk_B$ , roughly $0.44\,Nk_B$ or $0.44R$ per mole.

Step 4. The entropy sum rule. Because $C=T\,dS/dT$ , the area $\int_0^\infty (C/T)\,dT = S(\infty)-S(0)$ . From (5), $S(\infty)=Nk_B\ln 2$ and $S(0)=0$ , so $\int_0^\infty (C/T)\,dT = Nk_B\ln 2$ exactly. Physically: as the gap thaws, the system unlocks precisely the $\ln 2$ of entropy hidden in its two states — no more, no less. This is the two-level count expressed thermodynamically. Verified: the verifier solves $dC/dx=0$ numerically to $x_{\rm peak}=2.3994$ , $k_BT_{\rm peak}/\Delta=0.4168$ , $C_{\rm max}/Nk_B=0.4392$ , and integrates $C/T$ numerically to recover $k_B\ln 2 = 0.6931\,k_B$ within tolerance.

Limiting cases

Each limit is a physical check on the universal curve and on the physics behind it.

Case	Result	Meaning
Low temperature $k_BT\ll\Delta$	$C/Nk_B \to x^2 e^{-x}$ , $x=\Delta/k_BT$	The upper level is frozen out; heat capacity dies exponentially — the calorimetric signature of an energy gap.
High temperature $k_BT\gg\Delta$	$C/Nk_B \to x^2/4 \propto T^{-2}$	Populations pinned near equal; only a small residual imbalance responds, so $C$ decays slowly as $T^{-2}$ (the "Schottky tail").
Peak	$k_BT/\Delta\approx0.417$ , $C_{\rm max}/Nk_B\approx0.439$	The one temperature where a nudge moves the most units across the gap — universal, gap-independent when scaled.
Unequal degeneracies $g_1>g_0$	peak shifts, area $=k_B\ln[(g_0+g_1)/g_0]$	More upper states raise the high- $T$ entropy plateau to $k_B\ln(g_0+g_1)$ and move the hump; the shape is no longer symmetric.

Common trapCalling the Schottky hump a phase transition

It is not one. A phase transition shows a singular heat capacity — a divergence, a discontinuity, or a sharp cusp — at a definite critical temperature, reflecting a diverging correlation length and collective behaviour. The Schottky anomaly is a smooth, rounded, analytic bump produced by independent units with no interactions and no cooperative order whatsoever; nothing is singular anywhere. The word "anomaly" is historical and refers to the fact that experimenters found an unexpected extra hump in the heat capacity at low temperature, on top of the ordinary lattice contribution, not to any anomaly in the mathematical sense. If you see a genuinely sharp feature, look for interactions (a magnetic ordering transition, say); the pure Schottky curve is broad and featureless apart from its single maximum, and its width is comparable to its peak temperature.

Worked examples

Example 1 — A paramagnetic salt with a 1 K splitting. A crystal-field or Zeeman interaction splits the ground doublet of a paramagnetic ion into two levels separated by $\Delta/k_B = 1$ K. Locate the heat-capacity peak and find its height per mole.

Step 1. The peak is universal in scaled temperature: $k_BT_{\rm peak}/\Delta = 0.4168$ . With $\Delta/k_B = 1$ K this gives $T_{\rm peak} = 0.4168\times 1\ \mathrm{K} = \mathbf{0.417\ K}$ — squarely in the sub-kelvin range reached by a helium-3 cryostat.

Step 2. The peak height is $C_{\rm max}/Nk_B = 0.4392$ . Per mole, $N=N_A$ and $Nk_B=R$ , so $C_{\rm max} = 0.4392\,R = 0.4392\times 8.314\ \mathrm{J\,K^{-1}mol^{-1}} = \mathbf{3.65\ J\,K^{-1}mol^{-1}}$ .

Step 3. Check the entropy: integrating the measured $C/T$ from 0 up through the peak and out to a few kelvin must give $R\ln 2 = 5.76\ \mathrm{J\,K^{-1}mol^{-1}}$ . Recovering exactly $R\ln 2$ confirms there are two levels (a doublet); recovering $R\ln 3$ would betray a hidden triplet. The takeaway: the peak temperature reads off $\Delta$ , and the area reads off the number of levels.

Example 2 — A nuclear Schottky tail above the peak. Nuclear spins in a metal are split by hyperfine or applied fields by only $\Delta/k_B\sim$ millikelvin. Where is the hump, and what is seen in an ordinary cryostat that only reaches, say, 10 mK?

Step 1. With $\Delta/k_B = 1$ mK, the peak is at $T_{\rm peak}=0.417$ mK — far below what most cryostats reach. So the hump itself is invisible in a 10 mK experiment.

Step 2. At 10 mK one is on the high-temperature side, $x=\Delta/k_BT = 1/10 = 0.1$ , deep in the $T^{-2}$ tail. Here $C/Nk_B \approx x^2/4 = (0.1)^2/4 = 2.5\times10^{-3}$ , and as temperature is lowered the nuclear contribution grows as $1/T^2$ . This rising $C\propto T^{-2}$ nuclear tail, appearing beneath the falling electronic and lattice terms, is exactly what one sees experimentally and is the everyday form the Schottky anomaly takes in low-temperature calorimetry.

Step 3. Sanity check the number: the exact value at $x=0.1$ is $C/Nk_B = (0.1)^2 e^{0.1}/(e^{0.1}+1)^2 = 2.494\times10^{-3}$ , agreeing with the $x^2/4$ estimate to better than 1%. The takeaway: even when the peak is out of reach, its $T^{-2}$ tail is a reliable, quantitative fingerprint of a small splitting.

Origins · 1 mintwo-level system and how this was discovered

two-level system

the simplest quantum ladder · A finite ladder of energy levels is the minimal model behind the Schottky anomaly, first analysed by Walter Schottky

The two-level system — a unit with exactly two accessible energy states separated by a gap — is the simplest nontrivial quantum system and the workhorse behind the Schottky anomaly. Walter Schottky (1886–1976), the German physicist better known for the Schottky barrier and the shot-noise formula, pointed out in 1922 that a system with a few low-lying levels must contribute a characteristic hump to the specific heat, distinct from the lattice's monotonic rise. The two-level idealisation captures Zeeman-split spins, crystal-field doublets, tunnelling states in glasses, and nuclear-spin systems; the same partition function $Z=1+e^{-\Delta/k_BT}$ recurs across statistical physics, quantum optics, and information theory (as the qubit). Its heat capacity — vanishing at both temperature extremes and peaking near $k_BT\approx0.42\Delta$ — is universal, which is why the Schottky curve is a template rather than a case study.

History

1922

— Walter Schottky notes that systems with a few low-lying energy levels add a characteristic hump to the specific heat (Physikalische Zeitschrift)

1930

s — low-temperature calorimetry of paramagnetic salts reveals sub-kelvin heat-capacity humps matching the two-level curve

1934

— Giauque and collaborators use the entropy under such humps to characterise magnetic salts for adiabatic demagnetization

1950

s — nuclear Schottky anomalies (millikelvin splittings) observed as T⁻² tails in the heat capacity of metals

1972

— the tunnelling two-level-system model of glasses (Anderson–Halperin–Varma; Phillips) reuses the same Schottky physics for amorphous solids

Going further

References. Reif, Fundamentals of Statistical and Thermal Physics, §7 (two-level systems) · Kittel & Kroemer, Thermal Physics, Ch. 2–3 · Ashcroft & Mermin, Solid State Physics, Ch. 23 (magnetic heat capacity) · Gopal, Specific Heats at Low Temperatures (Schottky anomalies in solids) · Schottky, Physikalische Zeitschrift 23, 448 (1922).

Provenance. A verifier should reproduce: (1) the symbolic reduction $C/Nk_B = x^2e^{x}/(e^{x}+1)^2$ from $\langle E\rangle=\Delta/(e^{x}+1)$ ; (2) the peak $x_{\rm peak}=2.3994$ , $k_BT_{\rm peak}/\Delta=0.4168$ , $C_{\rm max}/Nk_B=0.4392$ ; (3) the low- $T$ form $x^2e^{-x}$ and high- $T$ form $x^2/4$ (checked at $x=0.1$ : $2.49\times10^{-3}$ ); (4) the entropy limits $S(\infty)=k_B\ln 2$ , $S(0)=0$ and the sum rule $\int_0^\infty (C/T)dT = k_B\ln 2$ ; (5) the degenerate case $g_1/g_0=2$ (shifted peak, plateau $k_B\ln 3$ ).

Builds on: Canonical Partition Function · The Two-Level Paramagnet and the Brillouin Function · Heat Capacity and the Fluctuation–Response Relation. Leads to: Negative Temperature · The Debye Model of Phonons · Nuclear Schottky anomalies in solids.

Why is it called an "anomaly" if it is completely ordinary statistics?

Purely for historical reasons. Early low-temperature experimenters expected the heat capacity to fall smoothly toward zero (lattice as T³, electrons as T), and were surprised to find an extra hump appearing at very low temperature. They labelled the unexpected feature an "anomaly." Theoretically there is nothing anomalous about it at all: it is the exact heat capacity of a system with a finite number of levels, as inevitable as the freeze-out of any gapped degree of freedom. The name stuck even after the physics became textbook-standard.

Why must the heat capacity vanish at both high and low temperature?

Heat capacity measures how much energy the system absorbs per unit temperature rise, which is proportional to how fast its populations shift. At low T essentially all units are in the ground state and stay there (promotion costs Δ, unavailable), so nothing shifts and C→0. At high T the populations are already saturated near their infinite-temperature values (50/50 for equal degeneracies) and barely move with further heating, so again almost no energy is absorbed and C→0. A finite ladder can hold only a bounded amount of energy, so once that ceiling is approached the heat capacity must decline — the bump between the two zeros is forced.

How is this related to the magnetization of the same paramagnet?

They are two windows on the identical two-level system. The Brillouin-function treatment computes the magnetization — how the population imbalance responds to a magnetic field — while the Schottky anomaly computes the heat capacity — how the energy stored in that imbalance responds to temperature. Both come from the same partition function $Z=1+e^{-\Delta/k_BT}$ (with Δ set by the Zeeman splitting $g\mu_B B$ for a spin in a field). The magnetization saturates as the field aligns the moments; the heat capacity peaks when the thermal energy matches the splitting. Measuring both on one sample overdetermines Δ and is a standard consistency check.

What changes if the two levels have different degeneracies?

The high-temperature entropy plateau rises from $k_B\ln 2$ to $k_B\ln(g_0+g_1)$ , because more states become available, and the area under $C/T$ becomes $k_B\ln[(g_0+g_1)/g_0]$ . The peak also shifts: with a more degenerate upper level ( $g_1>g_0$ ) the excited states are entropically favoured, so the population climbs at lower temperature and the hump moves down and changes height and symmetry. The functional form is $C/Nk_B = x^2\,(g_1/g_0)e^{x}/(e^{x}+g_1/g_0)^2$ up to the definition of $x$ ; setting $g_1/g_0=1$ recovers the symmetric curve.

Can I use the Schottky curve to measure an energy splitting?

Yes — that is its main practical use. Measure the heat capacity through the hump, read the peak temperature $T_{\rm peak}$ , and the splitting follows immediately from $\Delta = k_BT_{\rm peak}/0.4168 \approx 2.40\,k_BT_{\rm peak}$ . Then integrate $C/T$ over the hump; the area confirms the number of levels ( $R\ln 2$ for a doublet, $R\ln 3$ for a triplet, and so on). This is routine for crystal-field levels in rare-earth compounds and for hyperfine and Zeeman splittings, and needs no adjustable fitting of the peak height because that height is universal.

Does the T⁻² tail ever cause practical trouble?

Often. In ultralow-temperature experiments the nuclear-spin Schottky tail, $C\propto T^{-2}$ , can dominate the sample's heat capacity below a few tens of millikelvin, enormously increasing the heat that must be removed to cool further and lengthening thermal time constants. It is simultaneously a nuisance (it fights the cooling) and a resource (the nuclear spins are the working substance of nuclear demagnetization refrigeration, which exploits exactly this level system to reach microkelvin temperatures). Recognising the $1/T^2$ signature is essential to separating the nuclear contribution from electronic and lattice terms.