Anomalous Loss Reduction Below Two-Level System Saturation in Aluminum Superconducting Resonators

TL;DR

This paper investigates aluminum superconducting resonators and finds an unexpected decrease in loss at low temperatures and powers, explained by TLS response bandwidth reduction, with a model combining discrete TLS and GTM fitting the data.

Contribution

It introduces a combined model of discrete TLS ensemble and generalized tunneling model to explain anomalous loss reduction in superconducting resonators.

Findings

Loss decreases with lowering temperature below TLS saturation.

Loss follows a logarithmic power dependence at higher powers.

The combined model accurately fits the experimental data.

Abstract

Superconducting resonators are widely used in many applications such as qubit readout for quantum computing, and kinetic inductance detectors. These resonators are susceptible to numerous loss and noise mechanisms, especially the dissipation due to two-level systems (TLS) which become the dominant source of loss in the few-photon and low temperature regime. In this study, capacitively-coupled aluminum half-wavelength coplanar waveguide resonators are investigated. Surprisingly, the loss of the resonators was observed to decrease with a lowering temperature at low excitation powers and temperatures below the TLS saturation. This behavior is attributed to the reduction of the TLS resonant response bandwidth with decreasing temperature and power to below the detuning between the TLS and the resonant photon frequency in a discrete ensemble of TLS. When response bandwidths of TLS are smaller than their detunings from the resonance, the resonant response and thus the loss is reduced. At higher excitation powers, the loss follows a logarithmic power dependence, consistent with predictions from the generalized tunneling model (GTM). A model combining the discrete TLS ensemble with the GTM is proposed and matches the temperature and power dependence of the measured internal loss of the resonator with reasonable parameters.