The Kondo Effect

Why it matters

Cool almost any pure metal and its electrical resistance falls smoothly toward a small residual value set by static defects — the lattice vibrations (phonons) that scatter electrons simply freeze out. But in the 1930s experimenters found something stubbornly wrong with dilute magnetic alloys: gold with a trace of iron, or copper with a little manganese, showed resistance that fell, bottomed out, and then rose again as the temperature kept dropping. The location of this resistance minimum depended on impurity concentration, not on the pure host, so it clearly came from the magnetic atoms — yet no theory could explain why cooling would ever make a metal conduct worse.

The puzzle sat unsolved for roughly thirty years. Its resolution, by Jun Kondo in 1964, did far more than close one experimental anomaly. It exposed a genuine many-body problem: an effect that cannot be understood by treating one electron at a time, because the answer depends on how the entire Fermi sea reorganizes itself around a single spin. The mathematics that finally tamed it — the renormalization group, a running coupling that grows at low energy — turned out to be the same structure that describes asymptotic freedom in the strong nuclear force. The Kondo problem became a Rosetta stone linking condensed matter to particle physics, and it earned Kenneth Wilson a share of tools that contributed to his 1982 Nobel Prize.

How it works, step by step

The physics is a competition between two temperature-dependent scattering channels.

1. A leftover spin. A magnetic impurity — an iron or cerium atom with a partly filled d or f shell — carries a localized magnetic moment, a spin S⃗ that does not pair off with anything. Model it as a single spin-½ sitting in the metal.
2. Antiferromagnetic exchange. That local spin couples to the spins of the surrounding conduction electrons through an exchange interaction H_K = J·S⃗·s⃗(0), where s⃗(0) is the electron spin density at the impurity site. Crucially J > 0 (antiferromagnetic): the impurity wants its neighbours pointing the opposite way.
3. Spin-flip scattering. When an electron scatters off the impurity it can flip both spins at once — electron up, impurity down becomes electron down, impurity up. This is fundamentally different from ordinary potential scattering, because the impurity has internal states that change. It makes the scattering problem inseparable from the surrounding electron sea.
4. The logarithm. Kondo computed the scattering rate to third order in J and found the higher-order term contains a factor ln(D/k_B·T), where D is the bandwidth. The impurity resistance therefore grows as −ln(T) on cooling. Perturbation theory naïvely predicts a divergence at T = 0 — a warning that the true ground state is nothing like a free spin.
5. The screening cloud. As T drops below T_K the coupling effectively runs to strong values. The conduction electrons wrap the impurity in a spin-compensating cloud of size ξ_K ≈ ħ·v_F/(k_B·T_K) — tens to hundreds of nanometres — with net spin exactly opposite to the impurity.
6. The singlet. At T = 0 the impurity and its cloud lock into a collective singlet with zero total spin. The local moment is fully screened; the metal becomes an ordinary Fermi liquid again, resistance saturates at a finite unitary value, and the logarithmic divergence is cut off. The naïve infinity was an artefact of stopping the calculation too early.

The whole crossover is controlled by a single emergent energy scale, the Kondo temperature

k_B · T_K ≈ D · exp( −1 / (2 · J · ρ) )

where k_B = 1.381 × 10⁻²³ J/K is Boltzmann's constant, D is the conduction-band half-width (an energy, of order a few eV), J is the exchange coupling (energy), and ρ is the electron density of states at the Fermi level (states per unit energy). The essential singularity exp(−1/2Jρ) means T_K is fiercely sensitive to J — a modest change in coupling shifts T_K across orders of magnitude, from millikelvin to hundreds of kelvin.

The resistance minimum, quantitatively

Below the Debye regime the resistance of a dilute magnetic alloy is well described by adding a rising Kondo term to a falling phonon term:

ρ(T) = ρ₀ + a·T⁵ − b·ln(T)

Here ρ₀ is the temperature-independent residual resistivity from static defects, a·T⁵ is the Bloch–Grüneisen low-temperature phonon contribution (it falls as T drops), and −b·ln(T) is the Kondo term with b ∝ c_imp·(Jρ)³, proportional to impurity concentration c_imp (it rises as T drops). Setting dρ/dT = 0 gives the minimum:

5·a·T_min⁴ = b / T_min   ⟹   T_min = ( b / (5a) )^(1/5)   ∝  c_imp^(1/5)

The one-fifth power on concentration was a striking prediction: dilute the impurities tenfold and the minimum moves only by 10^(1/5) ≈ 1.6×, exactly matching experiments on Cu:Fe and Au:Fe. Below T_K the simple −ln(T) form breaks down; the full theory gives a smooth crossover to a Fermi-liquid resistivity that saturates as ρ(T) → ρ_unitary·[1 − (T/T_K)²].

Worked example — locating the minimum

Take a copper film with iron impurities, T_K ≈ 30 K. Suppose the phonon coefficient gives a·T⁵ and the Kondo coefficient gives −b·ln(T) with b/a = 4.0 × 10⁶ K⁵ (typical order of magnitude for a few parts per million of Fe). Then

T_min = ( b / (5a) )^(1/5) = ( 4.0e6 / 5 )^(1/5) K = (8.0e5)^(1/5) ≈ 15 K

So the resistance bottoms out near 15 K and rises below it — comfortably above absolute zero and within easy reach of a helium cryostat, which is why the effect was found so early. Now halve the iron concentration: b drops by 2×, and T_min shifts to (4.0e5)^(1/5) ≈ 13 K, a change of only ~1.15×. The weak concentration dependence is the fingerprint that distinguishes the Kondo minimum from an impurity phase transition, which would scale far more strongly.

Free spin versus screened singlet

Property	Above T_K (free-moment regime)	Below T_K (Kondo regime)
Impurity spin	Nearly free local moment	Screened into many-body singlet, S_total = 0
Magnetic susceptibility χ	Curie law, χ ∝ 1/T (diverges)	Pauli-like, χ → constant ∝ 1/(k_B·T_K)
Impurity resistivity	Rises as −ln(T)	Saturates at unitary limit, ρ ∝ 1 − (T/T_K)²
Impurity entropy	k_B·ln 2 (two spin states)	0 (unique singlet ground state)
Effective coupling J	Weak (perturbative)	Strong (RG flows to strong coupling)
Analog in QCD	Asymptotic freedom (weak at high E)	Confinement (strong at low E)

Kondo temperatures across systems

System	Impurity / local moment	Approx. T_K
Au:Fe (dilute alloy)	Fe in gold	~0.3 K
Cu:Fe (dilute alloy)	Fe in copper	~30 K
Cu:Mn (dilute alloy)	Mn in copper	~0.01 K (nearly free spin)
CeCu₆ (heavy fermion)	Ce 4f lattice	~5 K
CeAl₃ (heavy fermion)	Ce 4f lattice	~4 K
Semiconductor quantum dot	Single trapped electron	~1 K (tunable by gate)
Single Co atom on Au(111) (STM)	Cobalt 3d moment	~70 K

The asymptotic-freedom analog

Anderson's "poor man's scaling" (1970) and Wilson's numerical renormalization group (1975) both frame the Kondo problem as a running coupling. As you integrate out high-energy electronic states — effectively lowering the temperature or energy scale — the dimensionless coupling g = Jρ obeys a scaling equation whose leading term is dg/d(ln D) = −2g². Because the sign drives g to grow as D shrinks, the coupling that looks weak at high temperature becomes strong at low temperature, and the flow is cut off only when the singlet forms at k_B·T_K.

This is precisely the logic of asymptotic freedom in quantum chromodynamics, where the strong coupling is small at high energy and grows toward confinement at low energy. The Kondo model is the simplest solvable system exhibiting that behaviour, which is why it is a standard pedagogical gateway to the renormalization group. Wilson's numerical solution in 1975, and the exact Bethe-ansatz solutions found by Andrei and independently by Wiegmann around 1980, confirmed the singlet ground state and the universal crossover controlled by the single scale T_K.

Common misconceptions

• "The resistance rises because scattering off impurities increases." Static impurities give a temperature-independent ρ₀. The rise is specifically from spin-flip scattering off a magnetic moment; nonmagnetic impurities show no Kondo minimum.
• "The resistance diverges at T = 0." Only the naïve perturbation series diverges. The true ground state is a screened singlet and the resistivity saturates at a finite unitary limit — the log is cut off at T_K.
• "J is ferromagnetic." The Kondo effect requires antiferromagnetic exchange (J > 0). Ferromagnetic coupling scales the other way, to weak coupling, and gives no singlet or resistance upturn.
• "It's a phase transition at T_K." T_K is a crossover scale, not a critical temperature. There is no symmetry breaking and no singularity in thermodynamic functions — properties evolve smoothly through T_K.
• "The screening cloud is tiny." Its length ξ_K ≈ ħ·v_F/(k_B·T_K) can reach hundreds of nanometres for small T_K — mesoscopic, not atomic.
• "It only happens in messy alloys." A single-electron quantum dot, a single adatom under an STM tip, or a dense lattice of rare-earth ions all display Kondo physics — the ingredients are just a localized spin and a Fermi sea.

From dirty alloys to designer atoms

The modern face of the Kondo effect is the quantum dot. Trap an odd number of electrons on a nanoscale island and the topmost unpaired electron acts as the impurity spin; the leads on either side are the Fermi sea. Below T_K the dot, which would otherwise block current by Coulomb blockade, develops a sharp zero-bias conductance peak that climbs to the unitary limit G = 2e²/h ≈ 77.5 μS — the dot becomes perfectly transparent. This tunable, single-impurity Kondo effect was observed by Goldhaber-Gordon and colleagues in 1998, turning a decades-old alloy anomaly into a controllable quantum device. The same physics underlies heavy-fermion metals, where a whole lattice of Kondo ions dresses the electrons into quasiparticles with effective masses hundreds of times the bare value.