Kondo Lattices

A Kondo lattice is a periodic array of localized quantum moments exchange-coupled to itinerant electrons. The same coupling that tends to screen each moment also lets conduction electrons communicate spin information between sites. The low-temperature problem is therefore collective: magnetic order, a coherent heavy Fermi liquid, a Kondo insulator, fractionalized metal, or a more complicated phase can emerge from one microscopic competition.

The phrase does not mean “many independent Kondo impurities.” Translational coherence, Fermi-surface counting, intersite exchange, filling, and lattice symmetry have no analogue in the isolated-impurity fixed point. Those ingredients are precisely what make the lattice problem difficult.

Required background. Kondo Effect supplies the single-impurity screening scale and its material signatures. RKKY Interaction supplies the conduction-electron-mediated exchange competing with screening. Heavy Fermions supplies the cross-probe diagnosis of lattice coherence, while Kondo Model Preview supplies the exchange Hamiltonian, sign convention, and screening limits used to distinguish a lattice from independent impurities.

Dense Arrays of Local Moments

Minimal Kondo-lattice model

For one symmetry-matched conduction channel and one spin-$1/2$ moment per primitive cell, write

$$\begin{aligned} H_{\mathrm{KL}} &= \sum_{\mathbf k,\sigma} \epsilon_{\mathbf k} c_{\mathbf k\sigma}^{\dagger}c_{\mathbf k\sigma} + J_K \sum_i \mathbf S_i\cdot\mathbf s_{ci}, \\ \mathbf s_{ci} &= \frac{1}{2} \sum_{\alpha,\beta} c_{i\alpha}^{\dagger} \boldsymbol{\sigma}_{\alpha\beta} c_{i\beta}. \end{aligned}$$

$c_{i\sigma}$ is the local Wannier combination that actually hybridizes with the moment, not necessarily a literal atomic orbital. With this Hamiltonian convention, $J_K>0$ is antiferromagnetic. Both spin operators are dimensionless, with spin-$1/2$ eigenvalues. The local-moment Hilbert space contains one spin-$1/2$ degree of freedom per site, with no $f$-shell charge fluctuation.

RKKY exchange is generated perturbatively by $J_K$ itself. When an additional intersite interaction is retained explicitly, the model is more precisely a Kondo–Heisenberg lattice:

$$H_{\mathrm{KH}} = H_{\mathrm{KL}} + \sum_{i<j} I_{ij}\, \mathbf S_i\cdot\mathbf S_j.$$

$I_{ij}$ may represent superexchange, direct exchange, or a phenomenological residual interaction. Adding it and also adding an RKKY interaction computed from the same conduction band can double count the same process. A model statement should say which contribution $I_{ij}$ represents.

Relation to the periodic Anderson model

The periodic Anderson model retains a correlated $f$ orbital, its Coulomb repulsion, and hybridization with conduction orbitals. The Kondo lattice follows only in a local-moment regime where virtual empty and doubly occupied $f$ states can be removed. Schrieffer–Wolff Transformation explains the block-diagonalization method; Effective Hamiltonians in Many-Body Systems derives the exchange and potential-scattering coefficients from the virtual charge paths.

The reduction loses real low-energy valence fluctuations. It becomes unreliable when the $f$ occupation is far from an integer, a charge excitation approaches the retained scales, or several atomic multiplets and conduction channels remain active. Mixed-valence compounds may still be heavy fermions, but a spin-only Kondo lattice is then not a complete microscopic model.

Which degrees of freedom are retained?

Ingredient	Why it matters
Local multiplet and conduction channels	A single spin-$1/2$ and one channel is the baseline; higher spins, non-Kramers doublets or multiple channels can change screening and order.
Hybridization and exchange symmetry	Anisotropic or momentum-dependent couplings select which states screen the moment.
Intersite interactions	Exchange range and frustration determine the competing magnetic correlations.
Filling	One moment per cell does not fix conduction density; commensurate filling can permit a Kondo insulator.
Disorder	Substitution can produce a distribution of local scales, beyond the periodic model.

The filling must be stated. “One moment per site” does not imply one conduction electron per site, and the strong-coupling physics can differ sharply between dilute conduction filling, a generic metal, and the commensurate insulating case.

Why the Lattice Is Not Independent Impurities

Three effects appear as soon as the moments are dense.

Intersite correlations

Conduction-electron propagation between sites generates RKKY interactions at second order in $J_K$. Their sign and ordering wavevector depend on the full susceptibility matrix, not simply on nearest-neighbor distance. Direct or superexchange may reinforce or frustrate that tendency.

Coherent multiple scattering

In a periodic lattice, scattering from equivalent sites can reorganize into Bloch-like bands. The low-temperature heavy state is coherent in momentum space even though its microscopic origin includes local spin flips. A broad resistivity maximum or an optical hybridization onset therefore need not equal the impurity $T_K$; Heavy Fermions explains these distinct crossover temperatures.

Fermi-volume constraints

If the ground state is an ordinary translation-invariant Fermi liquid, a local spin-$1/2$ per primitive cell contributes to the Luttinger count. This global statement cannot be obtained by solving one impurity and multiplying its free energy by the number of sites.

Nozières emphasized an apparent exhaustion puzzle at low conduction filling: only a fraction of order $\rho_0k_{\mathrm B}T_K$ of conduction states lies in a narrow thermal shell around the Fermi energy, apparently too few to assign one independent screening electron to every moment. A screening cloud cannot be counted as one conduction electron permanently assigned to one moment. Kondo screening is an extended many-body correlation, and the lattice can develop distinct local-screening and coherence scales. For example, Burdin, Georges, and Grempel (2000) find regimes in a large-$N$ treatment with $T_{\mathrm{coh}}\ll T_K$, but the ratio is not universal.

Strong-Coupling Check

For the minimal Kondo lattice with no additional $I_{ij}$, the atomic limit gives a useful benchmark. On one site with exactly one conduction electron,

$$J_K\, \mathbf S_i\cdot\mathbf s_{ci} = \frac{J_K}{2} \left[ S_{\mathrm{tot}}(S_{\mathrm{tot}}+1) - \frac{3}{2} \right].$$

The singlet and triplet energies are

$$E_s = -\frac{3J_K}{4}, \qquad E_t = \frac{J_K}{4}.$$

At conduction filling $n_c=1$ and $J_K\gg t$, the zeroth-order ground state is a product of onsite singlets. Removing or adding one conduction electron leaves a site whose conduction spin is zero, costing $3J_K/4$ before hopping corrections. The atomic particle–hole cost is therefore $3J_K/2$, and finite hopping turns this into a dispersive Kondo-insulator problem.

Away from $n_c=1$, empty or doubly occupied conduction sites leave moments unscreened in this simple limit. Their motion and residual interactions can favor magnetism or other correlated states. “Large $J_K$” alone does not prove a paramagnetic heavy Fermi liquid at every filling.

Kondo Screening Versus RKKY Order

Let $\rho_0$ be the conduction density of states per spin at the Fermi energy and define

$$g = \rho_0J_K.$$

With the same one-loop convention used on the impurity pages, the two schematic scales behave as

$$\begin{aligned} \frac{k_{\mathrm B}T_K}{D} &\sim c_K \exp\left(-\frac{1}{g}\right), \\ \frac{k_{\mathrm B}T_{\mathrm{RKKY}}}{D} &\sim c_Rg^2. \end{aligned}$$

$D$ is an electronic cutoff. $c_K$ and $c_R$ absorb convention, band, geometry, and channel factors. The exponential Kondo scale is very small at weak exchange, while the RKKY scale is algebraic. This is why intersite magnetism is expected to dominate sufficiently weak $g$ even though the same antiferromagnetic $J_K$ generates both tendencies.

The competition is not literally “a local singlet on each site versus one fixed Néel state.” RKKY exchange can select ferromagnetism, incommensurate order, frustration, a spin glass in a disordered system, or a nonmagnetic valence-bond state. Kondo correlations can also survive dynamically inside an ordered phase.

An illustration of the Doniach comparison, evaluated with $c_K=1$ and $c_R=0.2$. The scales cross near $g=0.212$, but this is neither a universal coupling nor a calculated phase boundary. Changing the prefactors changes the comparison; the equal-prefactor choice in exercise 2 has no crossing at all. The weak-coupling estimates also lose quantitative reliability as the exchange grows.

The Doniach Diagram

The standard one-axis diagram varies $g$ while holding the band, filling, dimensionality, crystal-field structure, and frustration fixed. It places an RKKY-dominated magnetic regime at weak coupling and a Kondo-screened paramagnet at stronger coupling. The crossing of the two estimated scales identifies a region where neither weak-coupling approximation is reliable.

That crossing is an organizing heuristic, not a derivation of a quantum critical point. In particular:

• nonuniversal prefactors and the physical band parameters shift the estimated crossing; a consistent change of density-of-states or exchange notation does not change either physical scale;
• with the plus-sign intersite Hamiltonian used here, the exchange mean-field instability is selected by the most negative eigenvalue of the momentum- and orbital-resolved exchange kernel; for pure RKKY this corresponds to the largest susceptibility eigenvalue, not one scalar $J_K^2\rho_0$;
• magnetic order can be first order, preempted by superconductivity, or split from Fermi-surface reconstruction;
• geometric frustration can suppress order without strengthening local Kondo screening;
• valence fluctuations can invalidate the spin-only model before the nominal crossing;
• disorder can broaden a clean transition into a distribution of local scales.

The diagram also suppresses phases in which magnetism and Kondo coherence coexist. An antiferromagnetic heavy Fermi liquid is perfectly possible: the heavy bands form first and are then reconstructed by a magnetic ordering wavevector. Conversely, magnetic order can coexist with a small Fermi surface when the moments remain localized.

Heavy Fermi Liquid

A paramagnetic heavy Fermi liquid has no ordinary symmetry-breaking order parameter. It is a Landau Fermi liquid whose charged quasiparticles incorporate the local-moment degrees of freedom and whose Fermi volume is correspondingly large.

Abrikosov-fermion representation

Represent a physical spin-$1/2$ by auxiliary fermions,

$$\mathbf S_i = \frac{1}{2} f_{i\alpha}^{\dagger} \boldsymbol{\sigma}_{\alpha\beta} f_{i\beta}, \qquad \sum_{\alpha} f_{i\alpha}^{\dagger}f_{i\alpha} = 1.$$

The constraint removes the empty and doubly occupied auxiliary states. The representation has a local gauge redundancy:

$$f_{i\alpha} \mapsto e^{i\theta_i}f_{i\alpha}.$$

The auxiliary $f$ fermion is a spinon, not the microscopic charged $f$ electron of the periodic Anderson model.

Large-N saddle

Generalize the spin index to $\alpha=1,\ldots,N$ and impose $\sum_\alpha f_{i\alpha}^{\dagger}f_{i\alpha}=Q$. A hybridization-channel decoupling introduces

$$b_i = -\frac{J_K}{N} \sum_{\alpha} \left\langle f_{i\alpha}^{\dagger} c_{i\alpha} \right\rangle.$$

For a uniform saddle, the quadratic part has the structure

$$H_{\mathrm{MF}} = \sum_{\mathbf k,\alpha} \begin{pmatrix} c_{\mathbf k\alpha}^{\dagger} & f_{\mathbf k\alpha}^{\dagger} \end{pmatrix} \begin{pmatrix} \epsilon_{\mathbf k} & b\\ b^* & \epsilon_f(\mathbf k)+\lambda \end{pmatrix} \begin{pmatrix} c_{\mathbf k\alpha}\\ f_{\mathbf k\alpha} \end{pmatrix} + \frac{NN_s\lvert b\rvert^2}{J_K} - \lambda QN_s.$$

Stationarity with respect to the local coefficient $b_i^*$ gives $0=\sum_\alpha\langle f_{i\alpha}^{\dagger}c_{i\alpha}\rangle+Nb_i/J_K$, fixing the sign and conjugation in the definition of $b_i$. Scalar density shifts from this channel decoupling are absorbed into the conduction-energy convention $\epsilon_{\mathbf k}$.

$\lambda$ enforces the constraint on average, and $\epsilon_f(\mathbf k)$ can arise from a decoupled intersite exchange channel. Diagonalization produces the narrow hybridized bands discussed on Heavy Fermions.

Under the gauge redundancy,

$$b_i \mapsto e^{-i\theta_i}b_i.$$

Thus $b$ is not a gauge-invariant Landau order parameter. A nonzero saddle value is useful shorthand for a Higgs-like coherent regime in which spin and charge sectors are locked, but Elitzur’s theorem forbids interpreting $\langle b_i\rangle$ as a directly measurable local expectation value in the exact gauge theory. The heavy Fermi liquid does not thereby become a superconductor: locking the auxiliary gauge phase to the charge phase leaves the physical charge symmetry intact. Fermi volume, spectral poles, thermodynamics, and response functions are physical diagnostics.

The large-$N$ construction takes $N\to\infty$ with $Q/N$ fixed and couplings scaled so that the energy per flavor stays finite. It is not quantitatively controlled at physical $N=2$, and static mean field suppresses important gauge, magnetic, and spatial fluctuations.

Large Fermi volume

Define the signed Fermi-volume fraction per spin by

$$\nu_F = \frac{v_0}{(2\pi)^d} \sum_{\eta} s_{\eta}V_{F,\eta} \pmod{1}.$$

For $n_s$ spin-$1/2$ moments and $n_c$ conduction electrons per primitive cell,

$$\begin{aligned} \nu_{\mathrm{HFL}} &= \frac{n_c+n_s}{2} \pmod{1}, \\ \nu_{\mathrm{small}} &= \frac{n_c}{2} \pmod{1}. \end{aligned}$$

Oshikawa’s flux-insertion argument establishes the large count nonperturbatively when the Kondo lattice is an ordinary translation-invariant Fermi liquid. The result does not require assigning a static charged electron to each moment. Heavy Fermions explains the multiband measurement limitations.

At a suitable even total filling, the hybridized bands can be completely filled below a gap, producing a Kondo insulator rather than a metal. Whether that insulator is topological requires additional band inversion and symmetry information; Kondo screening alone does not decide the topological index.

Kondo Breakdown

Kondo breakdown denotes a zero-temperature loss of the heavy-Fermi-liquid organization in which the local moments contribute to the charged Fermi surface. The phrase is used in several related ways and should be qualified:

1. collapse of the large-$N$ hybridization saddle $b$;
2. disappearance of a low-energy Kondo resonance or coherence scale;
3. transition from a large to a small charged Fermi surface;
4. localization of $f$ charge in a periodic Anderson description;
5. a locally critical quantum transition with critical Kondo entanglement.

These statements coincide in some theoretical constructions but are not logically identical in every model.

Phase vocabulary

Description	Symmetry and charged Fermi surface
HFL or P$_L$	Translation-preserving heavy Fermi liquid with the large count.
AF$_L$	Antiferromagnet with heavy quasiparticles; the large count is folded into a magnetic Brillouin zone.
AF$_S$	Antiferromagnet whose carriers resemble a small conduction Fermi surface in an unfolded description.
FL*	Translation-preserving small charged Fermi surface, together with a fractionalized local-moment sector.

The $L/S$ labels refer to an unfolded conceptual count. In an antiferromagnet, unit-cell enlargement can make large and small volumes equivalent modulo filled bands. The two antiferromagnetic descriptions can even be adiabatically connected: changing Kondo correlations need not produce a phase transition inside the ordered state. A transition requires an additional distinction, such as a Fermi-surface topology change or topological order (Vojta, 2008).

Why a symmetric small Fermi surface is nontrivial

With one spin-$1/2$ moment per primitive cell and unbroken spin rotation and translation, a unique, fully gapped, short-range-entangled local-moment background is forbidden by Lieb–Schultz–Mattis–Oshikawa–Hastings constraints. If translation is preserved and the charged Fermi surface counts only $n_c$, the missing local-moment contribution must be carried by additional low-energy or topological structure.

A fractionalized Fermi liquid, FL*, supplies that structure: conduction electrons form a small charged Fermi surface while the local moments form a spin liquid with neutral spinons and an emergent gauge field. For example, a $\mathbb Z_2$ spin liquid can have gapped neutral excitations; a $U(1)$ construction may instead have a spinon Fermi surface. The generalized Luttinger relation accounts for the fractionalized sector (Senthil, Sachdev, and Vojta, 2003). FL* is a theoretically consistent phase, not a synonym for any experimentally observed small Fermi surface.

Possible transition routes

The one-axis Doniach picture suggests a direct magnetic-to-HFL transition, but an extended phase space permits several routes:

• Itinerant spin-density-wave route: HFL $\to$ AF$_L$. Kondo coherence survives; the heavy Fermi surface is folded and reconstructed by magnetic order.
• Direct Kondo-destruction route: AF$_S$ $\to$ HFL. Magnetic order disappears as the Kondo scale collapses, so order-parameter and Fermi-surface criticality coincide.
• Two-step magnetic route: AF$_S$ $\to$ AF$_L$ $\to$ HFL. In a model with a genuine distinction between the two antiferromagnets, their transition precedes the disappearance of magnetic order. Other models have only a crossover between these antiferromagnetic limits.
• Fractionalized route: AF$_S$ $\to$ FL* $\to$ HFL, or a direct FL* $\to$ HFL transition when frustration suppresses order.

Whether a direct continuous transition exists depends on dimension, symmetry, gauge dynamics, Fermi-surface geometry, and the microscopic model. Large-$N$, extended dynamical mean-field theory, and special sign-problem-free lattice models demonstrate particular routes; none proves a universal phase diagram for all heavy-fermion materials.

Testing a proposed breakdown

Look for a common vanishing scale in Fermi-surface reconstruction, loss of coherent spectral weight, spin dynamics and thermodynamics, while accounting for magnetic folding. A Hall crossover or changed oscillation frequency alone is insufficient: Lifshitz transitions, valence changes and ordinary antiferromagnetic reconstruction can produce similar signatures. Nor does loss of zero-energy coherence require all finite-frequency Kondo correlations to disappear.

Some locally critical theories predict frequency-over-temperature scaling of the form

$$\chi''(\mathbf Q,\omega,T) = \chi_*\left(\frac{T_*}{T}\right)^{\alpha} \Phi_{\mathbf Q} \left( \frac{\hbar\omega}{k_{\mathrm B}T} \right).$$

Here $\chi_*$ and $T_*$ set nonuniversal susceptibility and temperature units. The exponent and scaling function are framework and material dependent. Observing an approximate collapse over a finite window does not, by itself, identify the microscopic fixed point.

Calculating the Lattice State

Strong-coupling expansions test the onsite-singlet limit. Large-$N$ methods reveal coherent bands and gauge structure but can miss strong corrections at physical $N=2$. Dynamical mean-field theory treats local dynamics self-consistently; ordinary single-site DMFT omits nonlocal correlations, while extended DMFT includes a dynamical intersite environment within its own approximation. Sign-free quantum Monte Carlo and tensor networks provide complementary checks where their sampling or geometry permits.

Dimension matters. The gapless paramagnetic metal studied in one dimension by Shibata, Tsvelik, and Ueda (1997) is a Luttinger liquid, not a higher-dimensional heavy Fermi liquid. Other fillings and couplings can produce insulating or ferromagnetic states. Compare methods in a shared parameter and dimensional regime before comparing their phase diagrams.

Exercises

1. Atomic-limit Kondo insulator

At $t=0$, $J_K>0$, and $n_c=1$, determine the ground-state energy for $N_s$ sites. Then compute the cost of creating one empty and one doubly occupied conduction site while keeping the total conduction number fixed.

Solution

Each site forms a local singlet with energy $-3J_K/4$, so

$$E_0 = -\frac{3J_K}{4}N_s.$$

An empty conduction site has $\mathbf s_{ci}=0$, so its exchange energy is zero. A doubly occupied single conduction orbital is a spin singlet and also has $\mathbf s_{ci}=0$. Relative to two onsite Kondo singlets, the particle–hole pair therefore costs

$$\Delta_{\mathrm{ph}} = 2 \left( \frac{3J_K}{4} \right) = \frac{3J_K}{2}.$$

Finite hopping broadens the particle and hole states and reduces the gap. The result is an atomic-limit benchmark, not the exact finite-bandwidth Kondo-insulator gap.

2. Doniach-scale estimate

Set $c_K=c_R=1$. Compare $k_{\mathrm B}T_K/D=e^{-1/g}$ and $k_{\mathrm B}T_{\mathrm{RKKY}}/D=g^2$ at (a) $g=0.08$ and (b) $g=0.18$. Which tendency is larger in each case?

Solution

For $g=0.08$,

$$\begin{aligned} \frac{k_{\mathrm B}T_K}{D} &= e^{-12.5} \simeq 3.73\times10^{-6}, \\ \frac{k_{\mathrm B}T_{\mathrm{RKKY}}}{D} &= 6.4\times10^{-3}. \end{aligned}$$

The RKKY estimate is larger.

For $g=0.18$,

$$\begin{aligned} \frac{k_{\mathrm B}T_K}{D} &= e^{-1/0.18} \simeq 3.87\times10^{-3}, \\ \frac{k_{\mathrm B}T_{\mathrm{RKKY}}}{D} &= 3.24\times10^{-2}. \end{aligned}$$

The RKKY estimate is still larger, although the Kondo scale has increased much more rapidly. With the arbitrarily fixed $c_K=c_R=1$, there is actually no crossing: $e^{-1/g}/g^2$ has maximum $4e^{-2}<1$ at $g=1/2$. A schematic Doniach crossing requires different prefactors or extensions beyond these weak-coupling estimates; it is not a quantitative prediction of $g_c$.

3. Gauge charge of the hybridization field

Under $f_{i\alpha}\mapsto e^{i\theta_i}f_{i\alpha}$, determine how $b_i$ must transform so that $b_i c_{i\alpha}^{\dagger}f_{i\alpha}+\mathrm{h.c.}$ is invariant. Why is $\langle b_i\rangle$ not an ordinary order parameter?

Solution

The bilinear transforms as

$$c_{i\alpha}^{\dagger}f_{i\alpha} \mapsto e^{i\theta_i} c_{i\alpha}^{\dagger}f_{i\alpha}.$$

Therefore

$$b_i \mapsto e^{-i\theta_i}b_i.$$

$b_i$ carries the auxiliary gauge charge, so its phase can be changed by a local redundancy. Elitzur’s theorem forbids a gauge-noninvariant local expectation value in the exact theory. A nonzero mean-field saddle diagnoses a Higgs-like coherent regime, while physical evidence must be expressed through gauge-invariant spectra, Fermi volume, thermodynamics, or response.

4. Can the small-Fermi-surface paramagnet be ordinary?

Suppose one spin-$1/2$ moment occupies each primitive cell, translation and spin rotation are unbroken, and the charged Fermi surface counts only $n_c$. Can the remaining moment sector be a unique, fully gapped, short-range-entangled paramagnet?

Solution

Not under those assumptions. The half-odd-integer moment per primitive cell carries a Lieb–Schultz–Mattis–Oshikawa–Hastings anomaly. A unique, symmetric, fully gapped, short-range-entangled state cannot absorb it.

At least one assumption must change. The moment sector may break translation or spin symmetry, remain gapless, or possess topological order. FL* supplies a fractionalized moment sector. It may have gapped neutral excitations and topological order, or gapless spinons and gauge fluctuations; both are not required simultaneously. This additional structure accounts for the missing local-moment contribution while the charged Fermi surface remains small.

5. Breakdown or ordinary reconstruction?

A field-tuned antiferromagnet shows a jump in one quantum-oscillation frequency at the critical field, but neutron diffraction also shows that the magnetic unit cell disappears there. No independent hybridization scale is resolved. Does this establish Kondo breakdown? Name three further tests.

Solution

No. Removing the magnetic supercell unfolds the Brillouin zone and can change oscillation frequencies even if Kondo coherence survives. A simultaneous Lifshitz transition is another alternative.

Useful further tests include:

1. compare calculated AF$_L$ and AF$_S$ orbit sets, including magnetic breakdown and field-dependent visibility;
2. track a Hall or thermoelectric crossover to lower temperature and test whether its width extrapolates to zero;
3. look for collapse of a low-energy hybridization or quasiparticle-residue scale in spectroscopy;
4. test for local dynamical criticality and a common $E^*$ scale in thermodynamics;
5. search for an intermediate AF$_L$ or AF$_S$ phase under a second tuning parameter.

Kondo breakdown is a joint statement about entanglement, Fermi count, and low-energy scales, not a synonym for any reconstructed orbit.

Related Topics

• Fractionalization explains excitations and gauge structure in a spin liquid.
• Quantum Criticality compares tuning and scaling in materials.
• Non-Fermi Liquids distinguishes mechanisms that invalidate quasiparticle descriptions.
• Mean-Field Theory explains saddle points, self-consistency and fluctuation checks.

References

• S. Burdin, A. Georges, and D. R. Grempel, “Coherence scale of the Kondo lattice,” Physical Review Letters 85, 1048–1051 (2000).
• P. Coleman, “1/$N$ expansion for the Kondo lattice,” Physical Review B 28, 5255–5262 (1983).
• P. Coleman, C. Pépin, Q. Si, and R. Ramazashvili, “How do Fermi liquids get heavy and die?” Journal of Physics: Condensed Matter 13, R723–R738 (2001).
• 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).
• P. Nozières, “Impuretés magnétiques et effet Kondo,” Annales de Physique 10, 19–35 (1985).
• P. Nozières, “Some comments on Kondo lattices and the Mott transition,” European Physical Journal B 6, 447–457 (1998).
• 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 and Q. Si, “Quantum phases driven by strong correlations,” Nature Reviews Physics 3, 9–26 (2021).
• N. Read and D. M. Newns, “On the solution of the Coqblin–Schrieffer Hamiltonian by the large-$N$ expansion technique,” Journal of Physics C: Solid State Physics 16, 3273–3295 (1983).
• T. Senthil, S. Sachdev, and M. Vojta, “Fractionalized Fermi liquids,” Physical Review Letters 90, 216403 (2003).
• T. Senthil, M. Vojta, and S. Sachdev, “Weak magnetism and non-Fermi liquids near heavy-fermion critical points,” Physical Review B 69, 035111 (2004).
• N. Shibata, A. Tsvelik, and K. Ueda, “One-dimensional Kondo lattice model as a Tomonaga–Luttinger liquid,” Physical Review B 56, 330–334 (1997).
• Q. Si, S. Rabello, K. Ingersent, and J. L. Smith, “Locally critical quantum phase transitions in strongly correlated metals,” Nature 413, 804–808 (2001).
• Q. Si and F. Steglich, “Heavy fermions and quantum phase transitions,” Science 329, 1161–1166 (2010).
• Q. Si and S. Paschen, “Quantum phase transitions in heavy fermion metals and Kondo insulators,” physica status solidi (b) 250, 425–438 (2013).
• M. Vojta, “From itinerant to local-moment antiferromagnetism in Kondo lattices: Adiabatic continuity versus quantum phase transitions,” Physical Review B 78, 125109 (2008).
• 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).