Lecture 19: Kondo impurity model II
Overview
Subir Sachdev explains how a spin-1/2 Kondo impurity changes from Curie-law behavior at high temperature to a screened, resonant-level-like state below the Kondo temperature. He connects the Anderson impurity model to the Kondo model through the Schrieffer-Wolff transformation, derives the exponentially small scale T_K ~ D exp[-1/(rho J_K)] using renormalization-group flow, and explains the low-temperature singlet, Kondo cloud, and universal susceptibility crossover.
Key takeaways
- For an antiferromagnetic spin-1/2 Kondo impurity, renormalization-group flow makes J_K grow at low energies and generates an exponentially small scale T_K ~ D exp[-1/(rho(0)J_K)].
- The Anderson impurity model becomes a Kondo model when large U_d suppresses empty and doubly occupied impurity states; the Schrieffer-Wolff transformation generates an exchange interaction at order W^2.
- A hybridized resonant level has width gamma=pi W^2 rho(0), and its enhanced local density of states gives a susceptibility of order 1/gamma, providing a useful analogy for the Kondo response of order 1/T_K.
- At strong coupling, the impurity spin binds one conduction electron into a singlet of energy -3J_K/4, effectively excluding other electrons from the local site.
- For dilute spin-1/2 impurities, the susceptibility crossover is universal when temperature is measured in units of T_K; material-specific details primarily set the value of T_K.
- The Kondo cloud grows as the bare exchange weakens because its characteristic length is inversely related to the very small Kondo temperature.
Chapters
- At temperatures well above the exchange scale, an impurity spin S has Curie susceptibility proportional to S(S+1)/(3T).
- The impurity susceptibility is the total susceptibility minus the clean-bulk contribution, divided by the number of impurities.
- Antiferromagnetic order or spin-glass freezing can produce a downturn or kink, while ferromagnetic order produces a divergent susceptibility at the Curie temperature.
- For isolated Kondo impurities, the moment instead disappears below T_K by forming a nonlocal singlet with conduction electrons.
- Sachdev introduces a local d-level with energy epsilon_d, hybridized with conduction electrons c_k through tunneling amplitude W.
- The bath electrons have dispersion epsilon_k, and the impurity couples to the local conduction orbital c_0.
- The setup treats the impurity as a noninteracting trapped electron first, making the model solvable with Green functions.
- The analysis focuses on a level within the conduction band, where it can decay into extended states rather than form a bound state.
- The conduction density of states is defined as rho(epsilon) = sum_k delta(epsilon - epsilon_k)/V, with rho(0) evaluated at the Fermi energy.
- A d-level inside a band gap has no continuum states at its energy and can form an exponentially localized bound state.
- For a level inside the band, Sachdev approximates rho(epsilon) by the constant rho(0) near the Fermi energy when W is small.
- The bare propagators are G_c^0(k,i omega_n)=1/(i omega_n-epsilon_k) and G_d^0(i omega_n)=1/(i omega_n-epsilon_d).
- Repeated transitions between the d-level and the conduction bath form a geometric series for the full d-electron Green function.
- Summing the series gives a frequency-dependent self-energy from the momentum-summed conduction propagator.
- With a flat, symmetric density of states, the self-energy is -i gamma sgn(omega_n), where gamma = pi W^2 rho(0).
- Fermi's golden rule gives a decay rate 2 pi W^2 rho(0), so gamma is half the probability decay rate and sets the resonance width.
- The d-level spectral density is Lorentzian, with a width gamma and a peak centered near epsilon_d.
- The conduction-electron Green function gains an impurity-scattering correction proportional to W^2 G_d.
- The impurity T matrix is W^2 G_d and, in this local model, is independent of incoming and outgoing momenta.
- For a finite density of impurities, the single-impurity correction scales with impurity density rather than vanishing as 1/V.
- The Pauli susceptibility of a continuum is proportional to its density of states, so the impurity response tracks the d-level spectral density near the Fermi energy.
- When epsilon_d=0, the narrow resonance produces an impurity susceptibility of order 1/gamma.
- Sachdev uses this result to motivate the Kondo low-temperature scale: a screened spin behaves approximately like a resonance with gamma of order T_K.
- The Kondo impurity susceptibility therefore saturates at a scale of order 1/T_K rather than retaining a Curie divergence.
- An on-site interaction U_d changes the resonant-level model into the Anderson impurity model.
- The local configurations have energies 0 for an empty site, epsilon_d for one electron, and 2 epsilon_d + U_d for double occupancy.
- At the particle-hole-symmetric choice epsilon_d=-U_d/2, the singly occupied states are lowest in energy.
- For large U_d, charge fluctuations to the empty and doubly occupied states are suppressed, leaving a local spin degree of freedom.
- The Schrieffer-Wolff canonical transformation removes high-energy charge states from the Anderson impurity model.
- In the large-U_d, singly occupied regime, the remaining impurity spin couples to the conduction-electron spin density through J_K.
- The resulting Kondo Hamiltonian contains J_K S dot (c_0 dagger sigma c_0/2), with the impurity at the origin.
- The effective exchange arises at second order in W, analogous to the Hubbard-to-Heisenberg exchange J=4t^2/U.
- Renormalization-group calculations with spinons show that a small positive J_K grows as high-energy conduction states are integrated out.
- In Sachdev's convention, the flow has the form dJ_K/dl proportional to rho(0) J_K^2.
- The weak-coupling solution develops a pole at a finite logarithmic scale l*, signaling the breakdown of free-moment perturbation theory.
- The pole identifies the crossover into strong-coupling Kondo physics, not a literal divergence of the physical exchange.
- Reducing the bandwidth from D to D exp(-l*) gives the characteristic crossover energy T_K.
- The one-loop estimate is T_K ~ D exp[-1/(rho(0) J_K)], with numerical factors depending on convention and higher-order corrections.
- A weak bare exchange can therefore produce a Kondo temperature exponentially smaller than the conduction bandwidth.
- Sachdev compares this dimensional-transmutation structure with the emergence of confinement scales in QCD and the transition scale in BCS theory.
- As J_K becomes large, the impurity spin and the conduction electron at site 0 form a singlet with energy -3J_K/4.
- The singlet uses one conduction electron and cannot be broken without paying an energy of order J_K.
- Other conduction electrons are consequently blocked from occupying site 0, which acts like an infinite local potential or vacancy.
- The remaining electrons are effectively free in the presence of this excluded site, connecting the strong-coupling limit back to resonant-level scattering.
- For small bare J_K, the screening singlet extends over a large Kondo cloud with characteristic size of order 1/T_K in the lecture's units.
- At distances larger than the cloud scale, the screened impurity appears point-like to the remaining conduction electrons.
- For a dilute spin-1/2 impurity, the susceptibility is written as (1/4T) times a universal function of T/T_K.
- Once T_K is fixed, different metals share the same scaled crossover curve, from Curie behavior at high temperature to a finite low-temperature response.
- The Bethe ansatz gives the full universal Kondo crossover function, although Sachdev notes it is not generally available as a simple closed-form expression.
- Sachdev previews a large-M generalization from SU(2) spins to SU(M), intended to approximate the universal response.
- The large-M method is useful because it can extend from a single impurity to Kondo lattices and ultimately Hubbard-model problems.
- Comparing the approximation with the exact Bethe ansatz solution provides a check on the method before applying it to more complex systems.
Transcript excerpt
0:37 Okay, can you hear me? Yeah. All right. So, uh let's get started. Thanks for uh joining online and I'll be back in person on on Tuesday uh for the rest of the course. Um all right. So, we are I introduced the condo impurity model last time. Uh so roughly speaking it's uh let's see if I get this right yeah uh so you have a a spin sitting here uh and this is spin s and this is say the point r equals z um and in the bulk of the system it's
1:27 sitting in some you know this is typically could be some manganese ion. Uh and in the bulk of the system there are mobile electrons. Uh and this is some say copper host and we are interested in the behavior of this spin. Now uh and there's a coupling the most important coupling here is a coupling between the spin uh which is the the condo exchange coupling JK which will couple this spin with the spin of the mobile electrons. um and we are interested in you know what happens as various temperatures uh depending on the value of the energy
2:14 scale kai. So if I one very uh simple thing to plot is the spin susceptibility kai and if you plot this as a function of temperature then for temperatures much much bigger than the condo exchange um you should cut car you should get a behavior uh which is this qy law uh where it's equal to s into s + 1 / 3 divided by temperature. So that's the famous cury law from which you can even measure the value of S. And this is how you know uh many magnetic insulators were studied by just studying the high temperature behavior. You can read off
This is the start of the transcript. Save this video free for the whole transcript, searchable and time-coded.
In this course
Related notes
Summary, takeaways, and chapters were generated by AI from the video's transcript and may contain errors. The video belongs to its creator, Subir Sachdev.