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

This study experimentally validates theoretical models linking control signal amplitude noise to qubit state fidelity in a 10x10 neutral atom array, demonstrating strong agreement between measured results and stochastic Schrödinger equation predictions to aid noise identification and optimal control in NISQ-era systems.

The Gist

Imagine you are trying to teach a room full of 100 identical robots to dance a perfect, synchronized routine. This is essentially what scientists are doing with quantum computers, but instead of robots, they are using tiny atoms, and instead of music, they are using invisible waves of energy (microwaves).

This paper is about a team of scientists who wanted to understand why these "atom dancers" sometimes trip up. They discovered that the music (the control signal) isn't always perfect; it has "static" or "noise" in it, just like a radio station with poor reception.

Here is the story of their experiment, broken down into simple concepts:

1. The Stage: A Grid of Atom Dancers

The scientists built a giant grid (10 by 10) of invisible "tweezers" made of laser light. Each laser trap holds a single Rubidium atom. Think of this as a dance floor with 100 specific spots where a dancer must stand.

• The Atoms: They are the qubits (the basic units of quantum information).
• The Goal: They want to make the atoms spin from one state to another perfectly, like a dancer doing a perfect 360-degree turn.

2. The Problem: The "Static" in the Music

To make the atoms dance, the scientists blast them with a microwave signal. Ideally, this signal should be smooth and steady. But in the real world, the signal has noise.

• The Analogy: Imagine trying to tell a friend to run exactly 100 meters. If you shout "Go!" but your voice cracks, shakes, or gets louder and quieter randomly, your friend might run 98 meters or 102 meters.
• In this experiment, the scientists intentionally added different types of "shaky voices" (noise) to their microwave signal to see exactly how it messed up the dance.

3. The Three Types of "Shaky Voices"

They tested three specific ways the signal could get noisy:

1. White Noise: Imagine a radio tuned between stations. It's a constant, random hiss. The signal jumps up and down unpredictably at every single moment.
2. Ornstein-Uhlenbeck Noise: Imagine a drunk person trying to walk in a straight line. They stumble, but they have a natural tendency to drift back toward the center. The noise is shaky, but it tries to "heal" itself over time.
3. Brownian Motion: Imagine a leaf floating down a river. It doesn't just wiggle; it drifts further and further away from where it started. The error builds up over time, like a snowball rolling down a hill getting bigger.

4. The Experiment: Theory vs. Reality

Before the experiment, the scientists had a mathematical recipe (a theory) that predicted exactly how much the dance would go wrong for each type of noise.

• The Prediction: "If we add this much 'hiss' (White Noise), the dancers will be 90% accurate. If we add 'drunk stumbling' (OU Noise), they will be 92% accurate."
• The Test: They ran the experiment 100 times (once for each atom) for every noise type. Because they had 100 atoms all listening to the same noisy signal at the same time, they could gather a massive amount of data very quickly. It's like testing a recipe on 100 cakes at once instead of one.

5. The Result: The Recipe Was Right!

The most exciting part of the paper is that the math matched the reality perfectly.

• When they added the "hiss," the atoms performed exactly as the math predicted.
• When they added the "drunk stumbling," the atoms performed exactly as predicted.
• When they added the "drifting leaf," the atoms performed exactly as predicted.

They even looked at the distribution of results. Instead of just saying "the average accuracy was 90%," they mapped out the whole shape of the results. They found that while the average might be good, some individual atoms did much worse, and some did much better, creating a specific "fingerprint" for each type of noise.

Why Does This Matter?

You might ask, "Why do we care if the math matches the experiment?"

1. Diagnosing the Sick Patient: Now that we know the "fingerprint" of each noise type, if a quantum computer starts making mistakes, we can look at the errors and say, "Ah! This looks like 'drunk stumbling' noise. We need to fix the part of the machine that causes that specific wobble."
2. Better Control: Knowing exactly how noise behaves allows engineers to design "noise-canceling headphones" for quantum computers. They can create control signals that are robust against these specific types of shakes.
3. Universal Truth: The math they used doesn't care if you are using atoms, electrons, or light. This means their findings help scientists working on all types of quantum computers, not just this one.

The Bottom Line

The scientists proved that they can predict exactly how "bad music" ruins a quantum dance. By intentionally adding noise and watching the atoms stumble, they validated a powerful mathematical tool. This tool will help us build better, more reliable quantum computers in the future by helping us identify and fix the specific "static" that causes them to fail.

Technical Summary

1. Problem Statement

Noise is a primary limiting factor for current Noisy Intermediate-Scale Quantum (NISQ) computers. While theoretical models exist to predict how control noise (specifically amplitude noise) affects qubit fidelity, these models often rely on assumptions that have not been rigorously validated experimentally across different noise profiles.

• The Gap: Previous studies often treated noise as static offsets or used density matrix formalisms that only predict average behavior. There is a lack of experimental verification for the full distribution of fidelities predicted by the Stochastic Schrödinger Equation (SSE) under time-dependent, colored noise.
• The Goal: To experimentally validate theoretical predictions regarding the relationship between the amplitude of control signal noise and the resulting fidelity distribution of neutral atom qubits.

2. Methodology

Experimental Platform

• System: A $10 \times 10$ array of optical tweezers trapping single Rubidium-85 ($^{85}\text{Rb}$) atoms.
• Qubit Encoding: Hyperfine clock states $|0\rangle = |F=2, m_F=0\rangle$ and $|1\rangle = |F=3, m_F=0\rangle$.
• Control: Global microwave pulses (3.035 GHz) drive the transition. The system achieves a Rabi frequency of $\approx 2\pi \times 50$ kHz.
• Noise Injection: An agile microwave source (Sinara Phaser + Quantum Machines OPX) generates artificial noise. The amplitude of the microwave drive is modulated with specific noise profiles at a time step of $\Delta t = 1 \mu\text{s}$.

Noise Profiles Tested

The experiment introduces three distinct types of classical control noise:

1. White Noise (WN): Uncorrelated noise.
2. Ornstein-Uhlenbeck (OU) Noise: Damped noise with a finite correlation time (colored noise).
3. Brownian Motion (BM): Integrated white noise (random walk).

Measurement Protocol

• Statistical Sampling: Due to the global nature of the microwave pulse, all 100 sites in the array experience the same noise realization simultaneously.
• Repetition: Each noise realization is effectively repeated $\approx 300 \times 50$ times (accounting for $\approx 0.5$ loading probability and 300 repetitions per realization) to gather high-statistics data on the fidelity distribution.
• Readout: State-selective fluorescence imaging determines the final state.
• Simulation: The team numerically integrates the Stochastic Schrödinger Equation (SSE) using a weak second-order Platen scheme to simulate the exact experimental conditions, including SPAM (State Preparation and Measurement) errors and Rabi frequency inhomogeneity.

3. Key Contributions

1. Experimental Validation of SSE Models: The paper provides the first direct experimental confirmation that the theoretical fidelity distributions derived from the Stochastic Schrödinger Equation (specifically from Ref. [14]) accurately match real-world quantum hardware behavior.
2. Full Distribution Analysis: Unlike standard benchmarking (e.g., Randomized Benchmarking) which yields a single average fidelity or decay rate, this work characterizes the entire probability distribution of fidelities. This allows for the analysis of higher moments (variance, tails) of the error distribution.
3. Architecture Agnosticism: By validating these models on neutral atoms, the authors demonstrate that the fidelity-noise relations are likely applicable to other qubit architectures (e.g., trapped ions, superconducting transmons), as the underlying physics of classical control noise is universal.
4. Noise Characterization Tool: The study establishes a method to identify specific noise types (WN, OU, BM) by analyzing the shape of the measured fidelity distribution, offering a diagnostic tool for quantum hardware.

4. Results

• Agreement with Theory: There is excellent agreement between the experimental data, numerical simulations, and analytic predictions for all three noise types.
  • White Noise: Shows a linear reduction in average fidelity over time.
  • Ornstein-Uhlenbeck: Exhibits a damping effect inherent to the noise correlation.
  • Brownian Motion: Shows an accelerating decrease in fidelity, characteristic of a random walk.
• Distribution Matching: The kernel density estimates of the experimental fidelity histograms match the simulated distributions, including the "tails" of the distribution (rare low-fidelity events).
• Higher Moments: The experiment successfully validated the prediction of higher moments (standard deviation) of the fidelity, not just the mean. This is crucial for error correction schemes that require all qubits to exceed a fidelity threshold, not just the average.
• Discrepancies: Experimental fidelities were slightly lower than theoretical predictions. The authors attribute this to unmodeled real-world factors such as:
  • Rabi frequency inhomogeneity across the array (coefficient of variation $\approx 0.14\%$).
  • SPAM errors (estimated $p_{0\to1} \approx 0.04$, $p_{1\to0} \approx 0.04$).
  • Microwave amplitude instabilities and phase noise.

5. Significance and Impact

• Benchmarking and Diagnostics: The validated models provide a "tested benchmark" for quantum hardware developers. By measuring the fidelity distribution, researchers can infer the specific noise spectrum of their control lines without needing complex spectral analysis hardware.
• Optimal Control: The results support the development of quantum optimal control protocols that are robust against specific noise types. Knowing the exact noise profile allows for the calculation of control trajectories that maximize fidelity for every individual realization, rather than just the mean.
• NISQ Era Relevance: As quantum computers scale, understanding the full distribution of errors (rather than just averages) becomes critical for error correction codes. This work provides the theoretical and experimental foundation for predicting how control noise degrades state preparation in large-scale systems.

In conclusion, this paper bridges the gap between stochastic quantum control theory and experimental reality, confirming that the Stochastic Schrödinger Equation is a powerful and accurate tool for predicting and diagnosing control noise in neutral atom quantum processors.