Fidelity Relations in an Array of Neutral Atom Qubits -- Experimental Validation of Control Noise

TL;DR

The study experimentally validates SSE-based fidelity-noise relations for amplitude noise on a neutral-atom qubit array, demonstrating that measured fidelity distributions under white, Ornstein-Uhlenbeck, and Brownian noise closely match analytic SSE predictions and simulations. Using a 10×10 Rubidium-85 tweezer platform with global microwave control, the authors inject controlled amplitude noise, characterize noise-induced fidelity degradation, and quantify SPAM and Rabi-inhomogeneity contributions. The results establish a robust benchmark for predicting and diagnosing control-noise effects in NISQ devices, with implications for noise-aware quantum control strategies across architectures. The approach provides a pathway to use fidelity distributions for noise identification and to guide the development of optimized control protocols in realistic quantum processors.

Abstract

Noise is a hindering factor for current-era quantum computers. In this study, we experimentally validate the theoretical relationships between amplitude noise of the control signal and qubit state fidelity. The experiment comprises a 10x10 site optical tweezer array stochastically loaded with single rubidium-85 atoms. A global microwave field is used to manipulate the state of the hyperfine qubits. With precise control of the time-dependent amplitude of the microwave drive, we apply control signals featuring artificial noise. We systematically analyze the impact of various noise profiles on the fidelity distribution of the quantum states. The measured fidelities are compared against theoretical predictions made using the stochastic Schrödinger equation. Our results show a good agreement between the experimentally measured and theoretically predicted results. This validation is consequential, as the model provides critical information on noise identification and optimal control protocols in NISQ-era quantum systems.

Figures (15)

• Figure 1: Conceptual representation of the experiment. Using a microwave antenna, a control signal with superimposed noise is sent to an array of neutral atom qubits. The state evolution under the applied noise leads to a reduction of the final state fidelity, as depicted by the pink arrow.
• Figure 2: Simplified diagram of the microwave drive circuit. The artificial noise is introduced as a time varying amplitude of the 200MHz produced by a Quantum Machines Operator-X module (QM OPX+). This signal is upconverted to 3.035GHz to be resonant with the atomic transition, by mixing with a 2.8 GHz base frequency from a Sinara Phaser, controlled by Artiq.
• Figure 3: Qubit fidelity for various Ornstein-Uhlenbeck noise strengths. Average final fidelities after 200µs pulse over 75 realizations, 100 sites, and 300 measurements per realization. Analytic results (circles) with 1$\sigma$ deviation over the possible realizations (dashed lines), experimental (triangles) and simulation (squares) for varying noise strengths $\gamma$ with error bars indicating 1$\sigma$ over the realizations. Ornstein-Uhlenbeck noise with 200 time steps and $\kappa=5\cdot 10^3$ s$^{-1}$.
• Figure 4: Noise evolution over time for 3 types of noise. Average fidelities over 75 noisy realizations, 100 sites, and 300 measurements per realization for equivalent 200µs pulses in experiment (triangles) and simulation (squares). Analytic results (black) together with their $1\sigma$ standard deviation over 75 realizations (dotted lines). a) white noise (WN) with $\gamma=6$ s$^{-1/2}$, b) Ornstein-Uhlenbeck noise (OU) with $\gamma=6$ s$^{-1/2}$ and $\kappa=5\cdot 10^3$ s$^{-1}$, and c) Brownian motion (BM) with $\gamma=0.04$ s$^{-1/2}$. Error bars indicating 1$\sigma$ over the realizations
• Figure 5: Qubit fidelity standard deviation for white noise as a function of time. Average final fidelities after 200µs pulse over 75 realizations, 100 sites, and 300 measurements per realization. Analytic results (circles) with 1$\sigma$ deviation over the possible realizations (dashed lines), experimental (triangles) and simulation (squares) with error bars indicating 1$\sigma$ over the realizations. White noise with 200 time steps and $\gamma=6$ s$^{-1/2}$.