11.4 Kondo effect

Kondo effect fundamentals

The Kondo effect explains why certain metals with dilute magnetic impurities show an anomalous upturn in electrical resistance at low temperatures. Rather than resistance dropping monotonically as you cool a metal, it hits a minimum and then rises again. This behavior defied explanation for decades and turned out to require a fully many-body quantum mechanical treatment, making it one of the landmark problems in condensed matter theory.

The effect connects to a wide range of physics: heavy fermion behavior, quantum criticality, superconductivity in correlated systems, and the design of nanoscale quantum devices. It bridges the gap between a single quantum impurity and the macroscopic transport properties you measure in the lab.

Historical background

Jun Kondo identified the mechanism in 1964 while trying to explain the resistance minimum seen in dilute magnetic alloys (like iron dissolved in copper). Existing theories predicted resistance should just keep falling with temperature, but experiments clearly showed a minimum followed by a logarithmic rise. Kondo's perturbative calculation captured the logarithmic upturn but also predicted a divergence at zero temperature, which is unphysical. This became the famous "Kondo problem."

The resolution came in 1975 when Kenneth Wilson applied his numerical renormalization group (NRG) technique, showing that the impurity spin gets completely screened at low temperatures, forming a many-body singlet ground state. Wilson's work on this problem earned him the Nobel Prize and established renormalization group methods as essential tools in condensed matter physics.

Definition and significance

At its core, the Kondo effect describes coherent spin-flip scattering of conduction electrons off a magnetic impurity. An electron scatters off the impurity, flips both its own spin and the impurity's spin, and this process interferes quantum mechanically with subsequent scattering events. The result is a many-body resonance at the Fermi energy that grows as temperature decreases.

• Below a characteristic temperature $T_K$ (the Kondo temperature), the impurity spin becomes fully screened by a cloud of conduction electrons, forming a singlet state
• The effect demonstrates that even a single impurity can produce dramatic, non-perturbative many-body physics
• It provides the foundation for understanding quantum impurity problems more broadly, including heavy fermion compounds and quantum dot systems

Magnetic impurities in metals

When you dissolve a magnetic atom (like Fe or Co) into a non-magnetic metal host (like Cu or Au), you introduce a localized magnetic moment into a sea of itinerant conduction electrons. This sets up the essential ingredients for the Kondo effect: a local spin degree of freedom coupled to a continuum of electronic states.

Localized magnetic moments

These moments arise from partially filled d or f orbitals of transition metal or rare earth impurities. The unpaired electrons in these orbitals carry a net spin (and possibly orbital) angular momentum. In the simplest case, you can model the impurity as an $S = 1/2$ spin.

• The localized nature of d and f orbitals means the moment stays pinned at the impurity site, unlike the delocalized conduction electrons
• The impurity spin interacts with passing conduction electrons through an exchange coupling, which is the interaction that drives the Kondo effect
• Classic examples: Fe impurities in Cu, Co adatoms on Au(111) surfaces (studied extensively with STM)

Conduction electrons vs. impurities

Conduction electrons occupy delocalized Bloch states and fill up to the Fermi energy, forming a Fermi sea. The impurity introduces a localized state that hybridizes with this conduction band. This hybridization is what generates the exchange coupling between the local moment and the itinerant electrons.

The key physics unfolds as temperature drops below $T_K$:

• Conduction electrons near the Fermi level begin to correlate with the impurity spin, forming a many-body singlet
• The impurity's magnetic moment gets progressively screened, effectively disappearing from the perspective of bulk measurements
• Electrons scattering off this correlated singlet state experience enhanced scattering right at the Fermi energy, which is why resistance increases at low temperatures

Theoretical framework

Two model Hamiltonians capture the essential physics of the Kondo effect. They approach the problem from different starting points but are connected through a well-defined mathematical transformation.

s-d exchange interaction

The s-d (or Kondo) model directly describes the exchange coupling between the impurity spin $\mathbf{S}$ and the conduction electron spin density $\mathbf{s}$ at the impurity site:

$H_{sd} = J \, \mathbf{S} \cdot \mathbf{s}$

Here $J$ is the exchange coupling constant. For the Kondo effect to occur, the coupling must be antiferromagnetic ($J > 0$ in this convention), meaning the impurity and electron spins prefer to be anti-parallel.

Kondo's original perturbative calculation (to third order in $J$) produced a contribution to resistivity proportional to $-\ln(T)$, which correctly explains the resistance upturn. But the perturbation series diverges as $T \to 0$, signaling that perturbation theory breaks down and the system flows to a strong-coupling fixed point. This breakdown is what made the problem so difficult for over a decade.

Anderson impurity model

The Anderson model is more general. Instead of putting in the exchange coupling by hand, it starts from the physical ingredients:

• An impurity energy level $\epsilon_d$
• An on-site Coulomb repulsion $U$ (the energy cost of double-occupying the impurity orbital)
• A hybridization $V$ between the impurity orbital and the conduction band

In the regime where the impurity is singly occupied (the "local moment" regime, where $\epsilon_d < E_F < \epsilon_d + U$), the Anderson model can be mapped onto the s-d model via the Schrieffer-Wolff transformation, with an effective coupling $J \propto V^2$. The Anderson model also captures mixed-valence physics and charge fluctuations that the simpler s-d model misses.

Temperature dependence

The temperature dependence of resistance is the most direct experimental signature of the Kondo effect. It reveals the competition between thermal fluctuations (which disrupt correlations) and quantum mechanical screening (which builds them).

Resistance minimum

In a normal metal, phonon scattering decreases as you lower the temperature, so resistance drops. Kondo scattering does the opposite: it contributes a term proportional to $-\ln(T)$, which grows as temperature falls. The total resistivity can be written schematically as:

$\rho(T) = \rho_0 + aT^5 - c_{\text{imp}} \ln(T)$

where $\rho_0$ is residual impurity scattering, $aT^5$ is the phonon contribution, and the logarithmic term is the Kondo contribution ($c_{\text{imp}}$ is proportional to impurity concentration).

• The minimum occurs at a temperature $T_{\min}$ where the decreasing phonon term and the increasing Kondo term balance
• For dilute magnetic alloys, $T_{\min}$ is typically a few Kelvin
• This violates Matthiessen's rule, which assumes different scattering mechanisms contribute independently and additively

Kondo temperature

The Kondo temperature $T_K$ is the fundamental energy scale of the problem. It marks the crossover from weak-coupling behavior (perturbation theory works) to strong-coupling behavior (the impurity spin is fully screened):

$T_K \approx D \exp\!\left(-\frac{1}{2J\rho_0}\right)$

where $D$ is the conduction electron bandwidth and $\rho_0$ is the density of states at the Fermi level.

• The exponential dependence means $T_K$ is extremely sensitive to the coupling $J$ and can range from millikelvins to hundreds of Kelvin depending on the system
• Below $T_K$, the impurity spin is locked into a singlet with the conduction electrons, and the system behaves as a Fermi liquid with enhanced scattering
• Above $T_K$, the impurity acts as a weakly coupled free spin

Screening of magnetic impurities

The central outcome of the Kondo effect is that conduction electrons collectively screen the impurity's magnetic moment, forming a non-magnetic singlet ground state. This screening is a genuinely many-body phenomenon involving a macroscopic number of electrons.

Formation of Kondo cloud

The Kondo cloud is the spatial region around the impurity over which conduction electrons participate in the screening. Its characteristic size is:

$\xi_K \approx \frac{\hbar v_F}{k_B T_K}$

where $v_F$ is the Fermi velocity. For systems with low $T_K$, this cloud can extend over hundreds of nanometers or more, encompassing millions of conduction electrons.

• Direct observation of the Kondo cloud remains one of the outstanding experimental challenges in the field
• Evidence for it comes from indirect probes: STM measurements on magnetic adatoms, transport through quantum dots coupled to finite-size leads, and NMR studies
• The cloud's large spatial extent explains why the Kondo effect is robust against local perturbations but sensitive to confinement in nanostructures

Spin compensation mechanism

The screening works through a coherent superposition of many-electron states near the Fermi level. Conduction electrons with spin opposite to the impurity preferentially accumulate near the impurity site, progressively compensating its moment.

• This process creates the Abrikosov-Suhl resonance (also called the Kondo resonance): a sharp peak in the local density of states pinned right at the Fermi energy
• At $T = 0$, the screening is complete and the impurity behaves as a non-magnetic (unitary) scatterer
• Applying an external magnetic field splits the Kondo resonance and suppresses the effect, since the field tries to polarize the singlet state

Experimental observations

A wide range of experimental techniques have confirmed the Kondo effect and provided detailed tests of theoretical predictions, from bulk transport measurements on alloys to single-impurity spectroscopy.

Transport measurements

• Resistivity: the logarithmic upturn below $T_{\min}$ and saturation at $T \ll T_K$ are the classic signatures
• Hall effect: shows anomalous contributions from the enhanced skew scattering off Kondo impurities
• Thermopower: displays a large negative peak near $T_K$, reflecting the strong energy dependence of scattering at the Kondo resonance
• Magnetoresistance: the Kondo contribution to resistance is suppressed by magnetic fields on the scale of $k_B T_K / g\mu_B$
• Tunneling spectroscopy: in quantum dots, a zero-bias conductance peak appears when the dot is in the Kondo regime, and its width is set by $T_K$

Spectroscopic techniques

• Scanning tunneling spectroscopy (STS): the most direct single-impurity probe. The Fano lineshape observed for Co on Cu(111) is a textbook demonstration of the Kondo resonance
• Photoemission spectroscopy: reveals the Kondo resonance as a sharp feature near the Fermi level in the electronic structure of heavy fermion materials
• Neutron scattering: probes the magnetic excitation spectrum and can detect the quasielastic response associated with Kondo screening
• X-ray absorption spectroscopy: provides information on valence state and charge fluctuations in mixed-valence Kondo systems
• NMR/Knight shift: measures local magnetic susceptibility, which shows characteristic temperature dependence as the impurity moment is screened

Kondo lattice systems

When magnetic impurities form a periodic array rather than being dilute, you get a Kondo lattice. The physics becomes far richer because each impurity tries to form its own Kondo singlet with the conduction electrons, but the impurities also interact with each other magnetically.

Heavy fermion compounds

These are materials (typically Ce, Yb, or U intermetallics) where a periodic lattice of f-electron ions interacts with conduction electrons. The competition between two energy scales determines the ground state:

• Kondo screening (scale $T_K$): each f-electron moment tries to form a singlet with conduction electrons
• RKKY interaction (scale $T_{RKKY} \propto J^2 \rho_0$): an indirect magnetic coupling between f-moments mediated by conduction electrons, which favors magnetic order

When Kondo screening wins, the f-electrons hybridize coherently with the conduction band at low temperatures, forming quasiparticles with enormously enhanced effective masses (up to 1000 times the bare electron mass). This is why they're called "heavy fermions."

• Ground states include antiferromagnetic order, unconventional superconductivity, and non-Fermi liquid phases
• Examples: $\text{CeCu}_6$ (heavy Fermi liquid), $\text{UBe}_{13}$ (heavy fermion superconductor), $\text{YbRh}_2\text{Si}_2$ (near a quantum critical point)

Quantum criticality

A quantum critical point (QCP) occurs at $T = 0$ when you tune a system between two different ground states (e.g., from antiferromagnetic order to a paramagnetic heavy Fermi liquid). The tuning parameter can be pressure, magnetic field, or chemical doping.

• At the QCP, correlation lengths and timescales diverge, and Fermi liquid theory breaks down
• The influence of the QCP extends to finite temperatures, creating a "quantum critical fan" where thermodynamic and transport properties show anomalous power laws
• The classic example is $\text{CeCu}_{6-x}\text{Au}_x$, where gold doping tunes the system through an antiferromagnetic QCP near $x \approx 0.1$
• Understanding these QCPs remains an active area of research, with implications for unconventional superconductivity

Applications and implications

Quantum dots

Semiconductor quantum dots are artificial structures where you can confine a small number of electrons and couple them to metallic leads. When the dot holds an odd number of electrons (giving it a net spin), and the coupling to the leads is strong enough, a Kondo resonance forms.

• The Kondo temperature can be tuned in situ by adjusting gate voltages, something impossible with real atomic impurities
• This tunability makes quantum dots ideal platforms for studying Kondo physics quantitatively, including non-equilibrium effects (finite bias voltage)
• The Kondo resonance enhances conductance through the dot at zero bias, reaching the unitary limit of $2e^2/h$ for a single spin-degenerate channel
• GaAs/AlGaAs heterostructure dots were among the first systems to demonstrate this controllable Kondo effect

Nanostructures and devices

The Kondo effect appears across a variety of nanoscale systems:

• Single-molecule transistors: molecules like $\text{C}_{60}$ bridging electrodes show Kondo resonances when they carry a net spin
• Carbon nanotubes: exhibit Kondo physics with tunable coupling strength, and can realize multi-channel Kondo effects due to orbital degeneracy
• Magnetic adatoms on surfaces: STM studies of individual atoms (Co, Fe, Mn on metal surfaces) reveal the spatial and spectral structure of the Kondo resonance
• Kondo insulators: bulk materials like $\text{SmB}_6$ where Kondo hybridization opens a gap in the entire band structure. These are now actively studied as candidates for topological insulators
• Potential applications include Kondo-based spin filters and components for quantum information processing

Beyond the Kondo effect

Multi-channel Kondo effect

In the standard (single-channel) Kondo effect, one channel of conduction electrons screens the impurity spin, and you get a Fermi liquid ground state. If the impurity couples symmetrically to $k > 1$ independent channels, the physics changes qualitatively.

• For $k = 2$ channels and an $S = 1/2$ impurity, the two channels "overscreen" the spin, leading to a non-Fermi liquid ground state with a residual entropy of $\frac{1}{2} k_B \ln 2$ per impurity
• Thermodynamic quantities show fractional power-law temperature dependences instead of the integer power laws of Fermi liquid theory
• Experimental realizations are challenging but have been achieved in mesoscopic devices (charge two-channel Kondo effect in quantum dots) and proposed in certain rare earth systems like $\text{Pr}_x\text{La}_{1-x}\text{Ag}_2$ (quadrupolar Kondo effect)

Non-Fermi liquid behavior

Standard Fermi liquid theory predicts specific heat linear in $T$, resistivity going as $T^2$, and a temperature-independent Pauli susceptibility at low $T$. Non-Fermi liquid (NFL) behavior refers to systematic deviations from these predictions.

• NFL behavior often appears near quantum critical points in heavy fermion systems, where the Kondo and RKKY scales are comparable
• Typical signatures: specific heat diverging as $-T \ln T$, resistivity linear in $T$, and Curie-like divergences in susceptibility
• Systems like $\text{CeCu}_{5.9}\text{Au}_{0.1}$ and $\text{YbRh}_2\text{Si}_2$ near their QCPs are canonical examples
• The theoretical description of NFL behavior remains an open problem, with competing scenarios including local quantum criticality and spin-density-wave criticality

Numerical methods

Analytical approaches to the Kondo problem break down in the strong-coupling regime, making numerical methods essential. Two techniques have been particularly important.

Numerical renormalization group

Wilson's NRG was the method that originally solved the Kondo problem. The key idea is to exploit the logarithmic energy structure of the problem:

1. Discretize the conduction band on a logarithmic grid (energies $D\Lambda^{-n}$, where $\Lambda > 1$ is a discretization parameter)
2. Map the discretized Hamiltonian onto a semi-infinite chain, with the impurity at one end
3. Diagonalize the chain iteratively, adding one site at a time and keeping only the lowest-energy states at each step
4. Extract thermodynamic quantities, spectral functions, and fixed-point structure from the flow of energy levels

NRG captures the full crossover from weak to strong coupling and gives highly accurate results for thermodynamic and spectral properties. Modern extensions include time-dependent NRG for non-equilibrium dynamics and multi-orbital/multi-channel generalizations.

Quantum Monte Carlo simulations

QMC methods provide a complementary approach, particularly useful for more complex impurity models and lattice problems.

• The Hirsch-Fye algorithm was the workhorse for the Anderson impurity model for many years, based on discretizing imaginary time and using a Hubbard-Stratonovich transformation
• Continuous-time QMC (CT-QMC) methods have largely replaced it, eliminating discretization errors and offering better scaling. Two main flavors exist: CT-HYB (expansion in hybridization) and CT-INT (expansion in interaction)
• QMC is the standard impurity solver within dynamical mean-field theory (DMFT), which maps lattice problems onto self-consistent impurity models
• The main limitation is the sign problem, which makes certain parameter regimes (e.g., frustrated multi-orbital systems, real-time dynamics) exponentially expensive to simulate