Heavy Fermions

A heavy-fermion metal is an intermetallic conductor in which partly localized $f$-electron degrees of freedom become incorporated into exceptionally narrow, strongly renormalized low-energy electronic states. The resulting quasiparticles can have large thermodynamic and orbit-dependent effective masses, often tens to hundreds of free-electron masses and sometimes more. “Heavy” describes the low-energy electronic response, not a large bare atomic mass and not one universal scalar attached to every carrier.

The physical picture links high- and low-temperature observations. Spectroscopy and magnetism identify the local or mixed-valence $f$ states; cooling can reveal coherent bands with a large density of states and enhanced orbit-dependent masses. Comparing heat capacity, quantum oscillations and spectroscopy helps distinguish this state from unrelated causes of a large heat capacity.

Required background. Fermi Surface supplies pocket geometry and quantum-oscillation orbits. Effective Mass distinguishes dispersion, cyclotron, and thermodynamic mass inferences. Kondo Effect supplies the local-moment screening crossover, while Fermi Liquid Theory Preview supplies the quasiparticle, specific-heat, and susceptibility baseline used to test a coherent heavy metal.

Local f-Electrons and Conduction Electrons

Heavy-fermion compounds commonly contain Ce, Yb, or an actinide. A useful ionic starting point is Ce $4f^1$ or Yb $4f^{13}$, although real solids need not have an exactly integral valence. The $4f$ orbitals are spatially compact, so Coulomb repulsion and atomic spin–orbit coupling are large while direct $f$-$f$ hopping is weak. The more extended $5f$ orbitals of U and related actinides can lie closer to the boundary between localized and itinerant behavior.

Atomic spin–orbit coupling first organizes the $f$ shell into total-angular-momentum multiplets. The crystalline electric field then splits the relevant multiplet into site-symmetry representations. For ions with an odd number of electrons and time-reversal symmetry, Kramers degeneracy leaves at least a doublet. If the first excited crystal-field level lies well above the temperatures of interest, that doublet can be represented by an effective pseudospin. If several levels are thermally occupied, the degeneracy changes the Kondo scale and the magnetic anisotropy; forcing a two-level model too early can be quantitatively misleading.

Retain one local Kramers doublet and one conduction doublet for a minimal periodic Anderson model:

$$H_{\mathrm{PAM}}={}\sum_{\mathbf k,\sigma}\epsilon_{\mathbf k}c^\dagger_{\mathbf k\sigma}c_{\mathbf k\sigma} +\epsilon_f\sum_{i,\sigma}f^\dagger_{i\sigma}f_{i\sigma} +\sum_{\mathbf k,\sigma,\sigma'}\left[V_{\sigma\sigma'}(\mathbf k)c^\dagger_{\mathbf k\sigma}f_{\mathbf k\sigma'}+\mathrm{h.c.}\right] +U_f\sum_i n_{fi\uparrow}n_{fi\downarrow}.$$

The labels $\uparrow,\downarrow$ refer to the two doublet states, not necessarily pure physical spins. Here $c$ describes the broad conduction states, $f$ the correlated local states, and $V$ their symmetry-dependent mixing. The crystal field is already reflected in the selected doublet and its matrix elements. Keeping additional levels requires their energies and a corresponding multiorbital interaction, rather than adding more orbital labels to a one-orbital $U_f$ term.

Two regimes should not be conflated:

Regime	Low-energy description
Local moment	Nearly integral $f$ occupancy with costly charge excitations; Kondo exchange describes the surviving moment. Intersite order can intervene before a coherent screened metal forms.
Mixed valence	Several $f$ charge configurations participate appreciably; retain correlated orbitals and their charge fluctuations. Well-separated charge and screening scales may be absent.

In the local-moment limit, virtual charge fluctuations can be eliminated to obtain a Kondo-lattice description. That reduction is controlled only when the removed charge-excitation energies remain large compared with the retained scales. It does not make the $f$ electron physically disappear; it changes the variables used to describe its low-energy spin degree of freedom.

From a Kondo Lattice to Heavy Bands

At high temperature, conduction electrons scatter incoherently from many nearly local moments. At sufficiently low temperature, some materials form a translation-coherent metallic state in which the $f$ degrees of freedom participate in Bloch-like quasiparticles. The crossover is a collective lattice phenomenon, not a copy of one isolated Kondo singlet placed independently at every site.

A two-band quasiparticle model captures the avoided crossing without pretending to derive its renormalized parameters:

$$\mathcal H_{\mathrm{qp}}(\mathbf k) = \begin{pmatrix} \epsilon_{\mathbf k} & \widetilde V_{\mathbf k}\\ \widetilde V_{\mathbf k}^{*} & \widetilde\epsilon_f \end{pmatrix}.$$

Its eigenvalues are

$$E_{\pm}(\mathbf k) = \frac{\epsilon_{\mathbf k}+\widetilde\epsilon_f}{2} \pm \frac{1}{2} \sqrt{ \left( \epsilon_{\mathbf k}-\widetilde\epsilon_f \right)^2 + 4\lvert\widetilde V_{\mathbf k}\rvert^2 }.$$

$\widetilde\epsilon_f$ and $\widetilde V_{\mathbf k}$ are low-energy parameters, generally much smaller than bare atomic scales. They may be estimated by a slave-particle mean field, dynamical mean-field calculation, or a phenomenological fit, but the displayed matrix by itself is not a microscopic proof of a heavy Fermi liquid.

For momentum-independent $\widetilde V$ and a flat $f$ level, the conduction-electron weight of either branch is

$$\begin{aligned} w_{c,\pm}(\mathbf k) &= \frac{\partial E_{\pm}} {\partial\epsilon_{\mathbf k}} \\ &= \frac{1}{2} \left[ 1 \pm \frac{ \epsilon_{\mathbf k}-\widetilde\epsilon_f }{ \sqrt{ \left( \epsilon_{\mathbf k}-\widetilde\epsilon_f \right)^2 + 4\lvert\widetilde V\rvert^2 } } \right]. \end{aligned}$$

Therefore

$$\mathbf v_{\pm}(\mathbf k) = \frac{1}{\hbar} \nabla_{\mathbf k}E_{\pm} = w_{c,\pm}(\mathbf k)\, \mathbf v_c(\mathbf k).$$

An $f$-dominated segment has small $w_c$ and a small group velocity even though it remains an electronic Bloch state. If such a segment crosses the chemical potential, its small velocity can enhance the density of states and cyclotron mass. A curvature mass also involves derivatives of the velocity and cannot be read from its magnitude alone.

Top: the two-level eigenvalues bend away from the dashed conduction band and flat $f$ level. A narrow $f$-rich segment can be heavy if the chemical potential intersects it; the plot does not fix the filling. Bottom: schematic responses illustrate that the resistivity maximum and the onset of a low-temperature Fermi-liquid regime need not coincide. The two response curves have separate arbitrary vertical scales, and their shape is not universal across compounds.

Residue is not the only mass renormalization

For a single electronic component, using energy $E$ measured from the chemical potential, the quasiparticle residue is

$$Z_{\mathbf k} = \left[ 1 - \left. \frac{\partial \operatorname{Re}\Sigma(\mathbf k,E)} {\partial E} \right|_{E=0} \right]^{-1}.$$

If the self-energy is nearly momentum independent and the Fermi surface is simple, one often writes $m^*/m_b\approx Z^{-1}$. Heavy-fermion crystals are usually multiband, anisotropic, and strongly hybridized. Momentum derivatives of $\Sigma$, orbital mixing, and band curvature then matter, so no exact material-wide identity $m^*=m_b/Z$ exists. Spectral weight, thermodynamic density of states, optical mass, penetration-depth mass, and cyclotron mass answer related but distinct questions.

Large and small Fermi surfaces

In a conventional, translation-invariant, spin-degenerate heavy Fermi liquid, the $f$ degrees of freedom contribute to the Luttinger count. For the spin-rotation-invariant Kondo lattice with spin-$1/2$ moments, Oshikawa (2000) established this even though the local charge states have been removed from the Hamiltonian. For primitive-cell volume $v_0$,

$$\frac{2v_0}{(2\pi)^d} \sum_{\nu} s_{\nu}V_{F,\nu} = n_c+n_f \pmod{2}.$$

$s_{\nu}=+1$ for an electron-like sheet and $-1$ for a hole-like sheet under this signed-volume convention. $n_f$ is the number of spin-$1/2$ moments per primitive cell in that Kondo model. In a periodic Anderson description with fluctuating valence, it is instead the total retained $f$-electron occupancy. Higher-spin or multichannel models require their own count; “one per moment” is not a general rule for all local multiplets. This is called a large Fermi surface; a phase in which only $n_c$ enters is called small.

Several cautions are essential. A Hall coefficient is not generally $1/(ne)$ in a compensated multiband metal. Antiferromagnetism enlarges the unit cell and folds the Brillouin zone, which can make large and small counts indistinguishable by volume alone. Lifshitz transitions can change observed frequencies without destroying Kondo entanglement. A defensible localization claim therefore combines symmetry, quantum oscillations, Hall or thermoelectric response, spectroscopy, and material-specific band calculations.

What “Large Effective Mass” Means

There is no instrument that returns one universal heavy-fermion mass. The strongest diagnosis compares complementary observables.

Electronic heat capacity

After phonon, nuclear, Schottky, and magnetic-order contributions have been separated, a Fermi liquid has

$$C_{\mathrm{el}}(T) = \gamma_{\mathrm{mol}}T + o(T),$$

with

$$\gamma_{\mathrm{mol}} = \frac{\pi^2}{3} k_{\mathrm B}^2N_{\mathrm A} N^*(E_{\mathrm F}).$$

The notation $o(T)$ means corrections smaller than linear in temperature; they need not be an analytic $T^3$ term. Here $N^*(E_{\mathrm F})$ is the renormalized quasiparticle density of states per formula unit and per energy, including both spin projections. Values of $\gamma_{\mathrm{mol}}$ in heavy-fermion compounds can exceed ordinary-metal values by two orders of magnitude. The comparison to a reliable band-structure value is more informative than comparison to the free-electron mass.

A large measured $C/T$ that continues to vary strongly as $T\to0$ may instead signal quantum-critical fluctuations, unresolved nuclear terms, disorder, or a nearby transition. Calling its finite-temperature value a Fermi-liquid $\gamma$ would then be premature.

Quantum oscillations

For a closed extremal orbit perpendicular to magnetic field,

$$F = \frac{\hbar}{2\pi e} A_{\mathrm{ext}},$$

and the positive cyclotron mass fitted from thermal damping is

$$m_c^* = \frac{\hbar^2}{2\pi} \left| \frac{\partial A(E)} {\partial E} \right|_{E=E_{\mathrm F}}.$$

The absolute value includes hole-like orbits, whose area decreases with energy. With $B$ the magnetic-field magnitude, the Lifshitz–Kosevich thermal damping factor is

$$R_T = \frac{X}{\sinh X}, \qquad X = \frac{2\pi^2k_{\mathrm B}T\,m_c^*} {\hbar eB}.$$

Fitting the temperature dependence of an oscillation amplitude yields an orbit-specific $m_c^*$. This is among the most direct heavy-mass measurements, but the heaviest sheets can be invisible because thermal, impurity, and field damping suppress their oscillations. A spectrum containing only lighter detected orbits need not account for the full heat capacity.

Spectroscopy and electrodynamics

Spectral Functions connect photoemission and tunnelling to the energy- and momentum-resolved electronic states. ARPES can reveal an avoided crossing and a narrow $f$-derived dispersion, but surface termination, matrix elements, energy resolution, and the low coherence temperature complicate extraction. Scanning tunnelling spectra can display interference between light and heavy channels rather than a direct density of states.

Optical conductivity may show a very narrow Drude response and finite-frequency spectral weight associated with hybridization. The inferred optical mass depends on the plasma frequency, the integration window, and which bands are included. Penetration depth and upper-critical-field scales provide further information in superconductors, but likewise weight the Fermi surface nonuniformly.

Low-temperature transport

When a heavy Fermi liquid has allowed current-relaxing electron collisions, its low-temperature resistivity can take the form

$$\rho(T) = \rho_0 + AT^2 +\cdots .$$

A quasiparticle collision rate alone does not guarantee a resistivity: momentum or current must relax, for example through lattice Umklapp processes, disorder-assisted scattering or suitable multiband transport.

The empirical Kadowaki–Woods ratio

$$R_{\mathrm{KW}} = \frac{A}{\gamma_{\mathrm{mol}}^2}$$

often groups related heavy-fermion materials more closely than ordinary metals. It is not a universal constant. Carrier density, dimensionality, Fermi-surface geometry, orbital degeneracy, current direction, and multiband short-circuiting all alter it. The ratio should be compared within a stated material class and only over a genuine common $T^2$ interval.

Coherence Temperature

“The coherence temperature” is an operational label for a crossover, and different probes need not assign the same number. The following temperatures refer to different observations:

Scale	Physical meaning and possible measurement
$T_K$	Single-ion screening scale under a specified convention; infer it from a dilute analogue, susceptibility or entropy.
$T_{\mathrm{hyb}}$ or $T^*$	Onset of a detectable hybridization-related response, such as an optical feature or Knight-shift anomaly; usually a crossover.
$T_{\mathrm{coh}}$	Coherent lattice conduction; a resistivity maximum followed by a fall or optical narrowing can locate a crossover.
$T_{\mathrm{FL}}$	Upper edge of an asymptotic Fermi-liquid regime, tested with transport and thermodynamics together.

In many Ce-based Kondo lattices, magnetic resistivity first increases on cooling because incoherent Kondo scattering strengthens, reaches a broad maximum, and then falls as coherent quasiparticles develop. Phonons, crystal-field depopulation, disorder, multiple conduction channels, or magnetic order can shift or obscure that maximum. Yb systems can show reversed pressure trends because pressure changes their valence and hybridization differently.

The onset can also be broad: spectroscopy may see hybridization before transport becomes coherent, while a true $T^2$ regime appears only much lower. Quoting a scale without its extraction rule hides this hierarchy. For example, distinguish “resistivity maximum at $18\,\mathrm K$” or “Knight-shift anomaly onset near $35\,\mathrm K$,” rather than silently calling both $T_K$.

Representative Material Families

The label covers several distinct electronic settings:

Material or family	What it illustrates
CeAl$_3$ and CeCu$_6$	Early paramagnetic heavy-fermion benchmarks with large low-temperature heat capacity. Historical single-number masses came before modern multiband resolution.
CeRu$_2$Si$_2$	Clean tetragonal system with strong anisotropy and field-tuned metamagnetic crossover. A metamagnetic crossover is not automatically a quantum critical point.
CeRhIn$_5$	Antiferromagnetism, pressure-induced superconductivity, and strong changes in de Haas–van Alphen spectra. Pressure cells restrict angular and field windows; reconstruction and breakdown scenarios must be distinguished.
YbRh$_2$Si$_2$	Very low-field antiferromagnetic suppression, Hall crossovers, and non-Fermi-liquid response. Extrapolating finite-temperature crossover widths to $T=0$ is model dependent.
CeCu$_2$Si$_2$ and CeCoIn$_5$	Bulk heavy-fermion superconductivity with strong evidence for unconventional pairing. Gap topology is material specific; CeCu$_2$Si$_2$ is not a simple universal nodal template.
UPt$_3$ and related U compounds	Multiple superconducting phases and partially itinerant $5f$ behavior. The localized-versus-itinerant split is less clean than in many Ce compounds.

No single threshold value of $\gamma$ separates these settings. Stoichiometry, disorder, strain, field orientation, and sample growth can change low-temperature behavior qualitatively.

Quantum Criticality

Pressure, magnetic field, or composition can suppress magnetic order toward zero temperature. Heavy-fermion systems are especially informative because the tuning may reorganize both the order parameter and the degree to which $f$ moments participate in the Fermi surface.

A transition temperature tending continuously to zero is a starting point. Thermodynamic scaling, a shrinking Fermi-liquid regime and reproducible low-temperature phases on either side help distinguish a critical point from a crossover or weak first-order transition. Fermi-surface changes must be interpreted using the actual magnetic unit cell and scattering processes.

At least three frameworks may need comparison:

• Spin-density-wave criticality: Kondo coherence survives, and an itinerant heavy Fermi surface is reconstructed by the magnetic ordering wavevector. Critical scattering can be strongest near hot regions of the Fermi surface.
• Kondo destruction or local quantum criticality: an additional scale associated with Kondo entanglement collapses at the transition, and the zero-temperature Fermi volume changes between large and small. This is an influential active framework, not an established universal mechanism for every heavy-fermion critical point.
• Other reconstructions: valence fluctuations, multipolar order, frustration, disorder, or a Lifshitz transition can produce overlapping anomalies and must be tested material by material.

The Hall crossover in YbRh$_2$Si$_2$ and pressure-dependent quantum oscillations in CeRhIn$_5$ are prominent evidence for sharp Fermi-surface reorganization. Neither observation alone proves a universal Kondo-breakdown scenario. Hall response includes mobilities and vertex corrections; quantum-oscillation frequencies can change through symmetry breaking or band topology. Quantum Criticality continues the detailed material phenomenology, while general scaling belongs to Quantum Phase Transitions.

Unconventional Superconductivity

The 1979 discovery of superconductivity in CeCu$_2$Si$_2$ established that a metal with an enormous normal-state electronic heat capacity could form a superconducting condensate. A zero-resistance transition is not enough to prove this. Bulk superconductivity requires a thermodynamic anomaly and magnetic screening or another volume-sensitive measurement, with filamentary paths and secondary phases excluded.

To show that the heavy carriers participate, one seeks consistency among the normal-state $\gamma$, the superconducting heat-capacity anomaly, low-temperature entropy balance, penetration depth, thermal transport, upper critical fields, and spectroscopic gaps on heavy bands. No one ratio is universal in a multiband, strongly coupled material.

“Unconventional” must be split into two claims. A pairing-structure claim concerns crystal representation, parity or pseudospin structure, nodes, relative signs, and multicomponent symmetry breaking. A pairing-mechanism claim concerns which interaction generated that state. Neither claim establishes the other.

Evidence for nontrivial pairing structure may include:

• power-law low-temperature heat capacity, penetration depth, or thermal conductivity consistent with nodes;
• phase-sensitive or impurity response consistent with a sign-changing order parameter;
• multiple superconducting phases or broken rotational or time-reversal symmetry.

A spin resonance, proximity to magnetic fluctuations, or a pressure or field phase diagram can constrain candidate interactions, but none identifies the pairing glue or representation by itself.

Each structural or mechanism-sensitive observation has alternatives. Multiband gaps can mimic power laws over a limited range; disorder can lift or fill nodes; a neutron resonance supports but does not uniquely prove sign change. CeCoIn$_5$ has strong evidence for predominantly $d$-wave pairing. CeCu$_2$Si$_2$ illustrates why the field remains active: the measurements reviewed by Smidman and collaborators (2023) support a fully opened gap while several sign-changing multiband states remain viable. Heavy-fermion superconductivity is therefore a family of material-specific problems, not one pairing symmetry.

BCS Theory supplies the weak-coupling benchmark. Deviations from that benchmark should be stated as measured facts before a pairing glue or representation is assigned.

Exercises

1. Heavy velocity from an avoided crossing

Let $\Delta=\epsilon_{\mathbf k}-\widetilde\epsilon_f>0$ and assume $\Delta\gg\lvert\widetilde V\rvert$. Show that the lower hybridized branch is $f$-dominated and estimate its velocity relative to the bare conduction velocity.

Solution

For the lower branch,

$$w_{c,-} = \frac{1}{2} \left[ 1 - \frac{\Delta} {\sqrt{\Delta^2+4\lvert\widetilde V\rvert^2}} \right].$$

Expanding the square root,

$$\sqrt{\Delta^2+4\lvert\widetilde V\rvert^2} \simeq \Delta \left( 1+ \frac{2\lvert\widetilde V\rvert^2}{\Delta^2} \right),$$

so

$$w_{c,-} \simeq \frac{\lvert\widetilde V\rvert^2}{\Delta^2}.$$

Therefore

$$\frac{\lvert\mathbf v_-\rvert} {\lvert\mathbf v_c\rvert} \simeq \frac{\lvert\widetilde V\rvert^2}{\Delta^2} \ll1.$$

The branch remains a coherent mixture, but its small conduction weight makes it nearly flat in this simplified model. At the exact avoided crossing, $w_{c,\pm}=1/2$; the extreme slowing occurs on the $f$-rich side, not exactly at the crossing.

2. What can be inferred from heat capacity?

A compound has $\gamma_{\mathrm{exp}}=800\,\mathrm{mJ\,mol^{-1}K^{-2}}$, while a noninteracting band calculation gives $\gamma_{\mathrm{band}}=8\,\mathrm{mJ\,mol^{-1}K^{-2}}$. What is the robust inference, and what stronger statement is unjustified?

Solution

The ratio is

$$\frac{\gamma_{\mathrm{exp}}} {\gamma_{\mathrm{band}}} = 100.$$

If both values refer to the same formula unit and the low-temperature state is a Fermi liquid, the total quasiparticle density of states is enhanced by roughly a factor of $100$ relative to the calculation. It is unjustified to conclude that every carrier has $m^*=100m_e$ or even $m^*=100m_b$. Different sheets contribute different densities of states, the calculated band masses are not all $m_e$, and momentum-dependent self-energy and hybridization break a one-mass description. Quantum oscillations or momentum-resolved spectroscopy are needed to distribute the enhancement among sheets.

3. Thermal damping of a heavy orbit

Using

$$X \simeq 14.69\, \frac{m_c^*}{m_e} \frac{T[\mathrm K]}{B[\mathrm T]},$$

estimate $R_T=X/\sinh X$ for $m_c^*=8m_e$ and $B=15\,\mathrm T$ at (a) $T=0.50\,\mathrm K$ and (b) $T=0.10\,\mathrm K$.

Solution

At $0.50\,\mathrm K$,

$$X \simeq 14.69 \frac{8(0.50)}{15} \simeq 3.92,$$

so

$$R_T \simeq \frac{3.92}{\sinh(3.92)} \simeq 0.16.$$

At $0.10\,\mathrm K$,

$$X \simeq 0.78, \qquad R_T \simeq \frac{0.78}{\sinh(0.78)} \simeq 0.90.$$

Cooling by a factor of five changes the thermal damping dramatically. This is why very low temperatures and high fields are needed to observe heavy orbits, and why undetected sheets cannot automatically be declared absent.

4. Distinguish screening from lattice coherence

A material shows impurity-like susceptibility scaling near $50\,\mathrm K$, an optical hybridization feature below $35\,\mathrm K$, a broad resistivity maximum at $18\,\mathrm K$, and stable $\rho=\rho_0+AT^2$ only below $1.2\,\mathrm K$. Assign the four scales and explain whether their separation is contradictory.

Solution

A defensible assignment is

$$T_K\sim50\,\mathrm K, \qquad T_{\mathrm{hyb}}\sim35\,\mathrm K, \qquad T_{\mathrm{coh}}\sim18\,\mathrm K, \qquad T_{\mathrm{FL}}\sim1.2\,\mathrm K.$$

The numbers are operational and need not coincide. Local screening correlations can become appreciable before a momentum-coherent lattice response is visible; spectroscopy can detect hybridization before transport loses incoherent scattering; and asymptotic Fermi-liquid behavior can require a much lower temperature. The separation is physical information, not an inconsistency.

5. Large versus small Fermi volume

Consider a translation-invariant, spin-degenerate metal with conduction density $n_c=0.6$ per primitive cell and one spin-$1/2$ local moment per cell. Ignoring filled bands, what fractions of the Brillouin-zone volume are enclosed per spin in small- and large-Fermi-surface descriptions?

Solution

For an electron-like Fermi sea, the per-spin occupied fraction is half the counted density. A small Fermi surface counts only conduction electrons:

$$f_{\mathrm{small}} = \frac{n_c}{2} = 0.30.$$

A conventional heavy Fermi liquid counts the local moment as one additional electron per cell:

$$f_{\mathrm{large}} = \frac{n_c+1}{2} = 0.80.$$

The difference is one half of a Brillouin zone per spin. If antiferromagnetism doubles the real-space unit cell, Brillouin-zone folding can make these counts equivalent modulo filled bands, so volume alone may no longer distinguish the phases.

Related Topics

• Kondo Lattices develops the dense-moment model, screening competition and candidate breakdown phases.
• Quantum Criticality compares endpoint and scaling tests across materials.
• Unconventional Superconductivity explains pairing symmetry and the measurements that distinguish it from pairing mechanism.
• Heat Capacity and Thermodynamics and Magnetic Susceptibility explain background subtraction, entropy and response fits.

References

• K. Andres, J. E. Graebner, and H. R. Ott, “4f-virtual-bound-state formation in CeAl$_3$ at low temperatures,” Physical Review Letters 35, 1779–1782 (1975).
• S. Doniach, “The Kondo lattice and weak antiferromagnetism,” Physica B+C 91, 231–234 (1977).
• P. Gegenwart, Q. Si, and F. Steglich, “Quantum criticality in heavy-fermion metals,” Nature Physics 4, 186–197 (2008).
• A. C. Hewson, The Kondo Problem to Heavy Fermions (Cambridge University Press, 1993).
• H. Hu, L. Chen, and Q. Si, “Quantum critical metals and loss of quasiparticles,” Nature Physics 20, 1863–1873 (2024).
• K. Kadowaki and S. B. Woods, “Universal relationship of the resistivity and specific heat in heavy-fermion compounds,” Solid State Communications 58, 507–509 (1986).
• M. Oshikawa, “Topological approach to Luttinger’s theorem and the Fermi surface of a Kondo lattice,” Physical Review Letters 84, 3370–3373 (2000).
• S. Paschen, T. Lühmann, S. Wirth, et al., “Hall-effect evolution across a heavy-fermion quantum critical point,” Nature 432, 881–885 (2004).
• C. Pfleiderer, “Superconducting phases of $f$-electron compounds,” Reviews of Modern Physics 81, 1551–1624 (2009).
• H. Shishido, R. Settai, H. Harima, and Y. Ōnuki, “A drastic change of the Fermi surface at a critical pressure in CeRhIn$_5$: dHvA study under pressure,” Journal of the Physical Society of Japan 74, 1103–1106 (2005).
• M. Smidman, O. Stockert, E. M. Nica, et al., “Colloquium: Unconventional fully gapped superconductivity in the heavy-fermion metal CeCu$_2$Si$_2$,” Reviews of Modern Physics 95, 031002 (2023).
• F. Steglich, J. Aarts, C. D. Bredl, et al., “Superconductivity in the presence of strong Pauli paramagnetism: CeCu$_2$Si$_2$,” Physical Review Letters 43, 1892–1896 (1979).
• G. R. Stewart, “Heavy-fermion systems,” Reviews of Modern Physics 56, 755–787 (1984).
• O. Stockert, J. Arndt, E. Faulhaber, et al., “Magnetically driven superconductivity in CeCu$_2$Si$_2$,” Nature Physics 7, 119–124 (2011).
• H. v. Löhneysen, A. Rosch, M. Vojta, and P. Wölfle, “Fermi-liquid instabilities at magnetic quantum phase transitions,” Reviews of Modern Physics 79, 1015–1075 (2007).
• S. Wirth and F. Steglich, “Exploring heavy fermions from macroscopic to microscopic length scales,” Nature Reviews Materials 1, 16051 (2016).
• Y.-f. Yang, Z. Fisk, H.-O. Lee, J. D. Thompson, and D. Pines, “Scaling the Kondo lattice,” Nature 454, 611–613 (2008).