Crossed Andreev reflection revealed by self-consistent Keldysh-Usadel formalism

TL;DR

This paper demonstrates through self-consistent numerical solutions that crossed Andreev reflection can dominate over elastic cotunneling in superconductor-based heterostructures, with implications for quantum entanglement applications.

Contribution

It shows that self-consistent Keldysh-Usadel formalism reveals conditions under which CAR surpasses EC, contrary to previous predictions, and explores control via spin-splitting and chemical potential tuning.

Findings

CAR can dominate EC with proper interface resistance and superconductor length.

Self-consistency is essential for accurate modeling of non-local transport.

Spin-splitting enhances EC, but chemical potential tuning can favor CAR.

Abstract

Crossed Andreev reflection (CAR) is a process that creates entanglement between spatially separated electrons and holes. Such entangled pairs have potential applications in quantum information processing, and it is therefore relevant to determine how the probability for CAR can be increased. CAR competes with another non-local process called elastic cotunneling (EC), which does not create entanglement. In conventional normal metal/superconductor/normal metal heterostructures, earlier theoretical work predicted that EC dominates over CAR. Nevertheless, we show numerically that when the Keldysh-Usadel equations are solved self-consistently in the superconductor, CAR can dominate over EC. Self-consistency is necessary both for the conversion from a quasiparticle current to a supercurrent and to describe the spatial variation of the order parameter correctly. A requirement for the CAR probability to surpass the EC probability is that the inverse proximity effect is small. Otherwise, the subvoltage density of states becomes large and EC is strengthened by quasiparticles flowing through the superconductor. Therefore, CAR becomes dominant in the non-local transport with increasing interface resistance and length of the superconducting region. Our results show that even the simplest possible experimental setup with easily accessible normal metals and superconductors can provide dominant CAR by designing the experimental parameters correctly. We also find that spin-splitting in the superconductor increases the subvoltage density of states, and thus always favors EC over CAR. Finally, we tune the chemical potential in the leads such that transport is governed by electrons of one spin type. This can increase the CAR probability at finite values of the spin-splitting compared to using a spin-degenerate voltage bias, and provides a way to control the spin of the conduction electrons electrically.