Superconductivity in Ultrasmall Metallic Grains

Abstract

We develop a theory of superconductivity in ultrasmall (nm-scale) metallic grains having a discrete electronic eigenspectrum with a mean level spacing of order of the bulk gap. The theory is based on calculating the eigenspectrum using a generalized BCS variational approach, whose applicability has been extensively demonstrated in studies of pairing correlations in nuclear physics. We discuss how conventional mean field theory breaks down with decreasing sample size, how the so-called blocking effect weakens pairing correlations in states with non-zero total spin, and how this affects the discrete eigenspectrum's behavior in a magnetic field, which favors non-zero total spin. In ultrasmall grains, spin magnetism dominates orbital magnetism, just as in thin films in a parallel field; but whereas in the latter the magnetic-field induced transition to a normal state is known to be first-order, we show that in ultrasmall grains it is softened by finite size effects. Our calculations qualitatively reproduce the magnetic-field dependent tunneling spectra for individual aluminum grains measured recently by Ralph, Black and Tinkham. We argue that previously-discussed parity effects for the odd-even ground state energy difference are presently not observable for experimental reasons, and propose an analogous parity effect for the pair-breaking energy that should be observable provided that the grain size can be controlled sufficiently well. Finally, experimental evidence is pointed out that the dominant role played by time-reversed pairs of states, well-established in bulk and in dirty superconductors, persists also in ultrasmall grains.