Nonequilibrium Kondo impurity: Perturbation about an exactly solvable point Kingshuk Majumdar Department of Physics and National High Magnetic Field Laboratory, University of Florida, 215 Williamson Hall, Gainesville, Florida 32611 Avraham Schiller Department of Physics and National High Magnetic Field Laboratory, University of Florida, 215 Williamson Hall, Gainesville, Florida 32611 and Department of Physics, The Ohio State University, Columbus, Ohio 43210-1106 Selman Hershfield Department of Physics and National High Magnetic Field Laboratory, University of Florida, 215 Williamson Hall, Gainesville, Florida 32611 Received 31 July 1997; revised manuscript received 14 October 1997 We perturb about an exactly solvable point for the nonequilibrium Kondo problem. In each of the three independent directions in parameter space, the differential conductance evolves smoothly as one goes away from the solvable point, and the lowest-order correction contains the logarithm of the band width, or cutoff. Perturbing towards physically realistic exchange couplings yields differential-conductance curves which more closely resemble experimental data than at the solvable point. The leading coefficient which describes the low-temperature and low-voltage scaling changes as one perturbs away from the solvable point, indicating nonuniversal behavior; however, it is restored to the solvable-point value in the limit of an infinite band width. S0163-18299800506-2 I. INTRODUCTION Recently, there has been a resurgence of interest in ex- actly solvable points for many-body problems in condensed matter physics. This is due in part to the discovery of new physical systems and in part to the discovery of new solvable points. For example, there is strong experimental evidence for the realization of clean one-dimensional interacting elec- tron systems in fractional quantum Hall systems 1,2 and quan- tum wires, 3,4 as well as for tunneling through a single mag- netic impurity. 5,6 Some of the models used to describe both of these phenomena have special points in their parameter space where a simple analytic solution can be found to the many-body problem. Some of the new solvable points which have been discovered recently are the Emery-Kivelson line for the two-channel Kondo model, 7 the g = 1 2 point for static impurity scattering in a Luttinger liquid, 8 and three new Tou- louse points for the generalized Anderson impurity model. 9 Besides being exact, one of the main advantages of solv- able points is that one can easily compute experimentally observable quantities, e.g., the susceptibility. Historically, calculations of observables for exact solutions and exactly solvable points have been done in equilibrium or linear re- sponse; however, recently there have been some solutions for nonequilibrium problems. 10,11 This paper concerns one of those solutions, namely, tunneling though a magnetic impu- rity connected to two leads. 12 The problem of tunneling through a magnetic impurity has a long history. Zero-bias anomalies associated with tun- neling through magnetic impurities were first discovered 13 in the early 1960’s. These and later experiments 14 showed the characteristic logarithmic singularities of the Kondo effect. Shortly after the original experiments, there were perturba- tive theories 15 which explained all of the qualitative features of the experiments; however, they were not able to get to the low-temperature, strong-coupling regime of the Kondo ef- fect. With the present interest in quantum dots, almost all the techniques of modern many-body physics have been applied to this problem. 16 To date, the only exact result on it beyond linear response is due to an exactly solvable point, 12 which generalizes the Toulouse 17 and Emery-Kivelson 7 solutions of the equilibrium Kondo problem. Using this solvable point for the nonequilibrium Kondo problem, a host of observables were computed: 12 electrical current, spin current, current noise, and even time-dependent response. In the case of the differential conductance, which is the most widely studied property, the solvable point shows all the qualitative features of the experiments: there is a reso- nance at zero bias—this resonance splits in an applied mag- netic field—the temperature and voltage dependences show correct Fermi-liquid behavior. Furthermore, assuming uni- versality, one can actually determine the exact scaling func- tion for the differential conductance at low temperature and voltage. 12 Although one can obtain all this information from the solvable point, it is still only one point in the parameter space. There is no reason to believe that the experimental conditions correspond to this point. Thus, a natural question to ask is what happens as one goes away from the solvable point? This is the question which we address in this paper. In particular, we perturb away from the solvable point to lowest order in all possible directions in parameter space. The ques- tions which we ask are iis the perturbation away from the solvable point smooth and nonsingular? iiDoes the quali- PHYSICAL REVIEW B 1 FEBRUARY 1998-I VOLUME 57, NUMBER 5 57 0163-1829/98/575/29919/$15.00 2991 © 1998 The American Physical Society