Field-Dependent Qubit Flux Noise Simulated from Materials-Specific Disordered Exchange Interactions Between Paramagnetic Adsorbates

This paper presents a first-principles, parameter-free simulation of paramagnetic O$_2$ adsorbates on an Al$_2$O$_3$ surface that accurately reproduces experimental magnetic flux noise trends in superconducting qubits and identifies an external electric field as a viable mechanism for mitigating this noise by tuning spin-spin interactions.

The Gist

Imagine you are trying to listen to a whisper in a crowded room. That whisper is the delicate signal from a superconducting quantum computer (a super-advanced type of computer). The problem? The room is filled with people shouting, humming, and rustling their papers. In the world of quantum computers, this "noise" is called magnetic flux noise, and it's the main reason these computers make mistakes (decoherence).

For years, scientists knew this noise existed but didn't know exactly who was shouting. They suspected it came from tiny, invisible magnetic specks (defects) on the surface of the computer's materials, but they were guessing the rules of how these specks behaved.

This paper is like a team of detectives using a super-powerful microscope and a virtual reality simulator to figure out exactly who is making the noise and how to shut them up.

The Culprits: Oxygen Molecules on a Sapphire Floor

The researchers focused on a specific scenario: Aluminum quantum computers sitting on a sapphire surface. They suspected that Oxygen molecules (O₂) from the air had landed on the sapphire and were acting like tiny magnets.

Think of the sapphire surface as a dance floor. The oxygen molecules are dancers who have landed there.

• The Problem: These dancers are messy. They don't stand in neat rows; they are scattered randomly, facing different directions.
• The Noise: Because they are magnetic, they wiggle and spin. When they wiggle, they create a magnetic "static" that messes up the quantum computer's signal.

The Detective Work: No More Guessing

Previous studies tried to model this noise by making up rules (like "assume the dancers are 50% happy and 50% sad"). This paper did something different: They used real physics to calculate the rules.

1. The "First Principles" Approach: Instead of guessing, they used a supercomputer to simulate exactly how oxygen molecules stick to sapphire. They calculated the energy of every possible way an oxygen molecule could sit, tilt, or twist.
2. The "Handshake" (Exchange Interaction): When two oxygen dancers are close, they "shake hands" (interact). Sometimes they agree to spin the same way (friendly/ferromagnetic), and sometimes they fight and spin opposite ways (unfriendly/antiferromagnetic).
   • The researchers found that because the dancers are scattered randomly, the "handshakes" are a chaotic mix of strong and weak, friendly and unfriendly.
   • The Analogy: Imagine a room where some people are holding hands tightly, some are loosely touching, and some are pushing each other away. This chaotic network creates the "static" noise.

The Big Discovery: The "Freezing" Point

The team simulated what happens as the temperature changes. They found that these oxygen dancers go through a phase transition, similar to water freezing into ice, but it's a very specific type of transition called the Berezinskii-Kosterlitz-Thouless (BKT) transition.

• Below a certain temperature (around -273°C): The dancers form little groups (domains) where they all spin together.
• The Noise: Even though they are in groups, the boundaries between these groups are messy. The wiggling at these boundaries creates the magnetic noise that ruins the quantum computer.

Their simulation showed that this messy, real-world disorder perfectly matches the noise patterns scientists have measured in real labs. This confirmed that adsorbed oxygen is indeed the culprit.

The Solution: How to Quiet the Room

Once they knew the source, they looked for a way to stop the noise. They tested two "volume control" knobs:

1. The Magnetic Knob (External Magnetic Field):

   • What they did: They applied a magnetic field to the room.
   • Result: It forced all the dancers to face the same direction. When everyone is marching in lockstep, they stop wiggling chaotically. The noise dropped significantly.
   • Catch: You can't use a strong magnetic field on a quantum computer because it might break the superconductivity itself. It's like trying to quiet a room by shouting louder; it works, but it's not a good long-term solution.

2. The Electric Knob (External Electric Field) - The Breakthrough:

   • What they did: They applied an electric field.
   • The Magic: They discovered that these oxygen molecules have a secret: their "handshake" strength changes depending on their electric charge. By applying an electric field, they could strengthen the bond between the dancers.
   • Result: When the bond is stronger, the dancers become more stable and stop wiggling as much. This reduced the noise even more effectively than the magnetic field, without breaking the superconductivity.
   • Analogy: Imagine the dancers were holding hands loosely. The electric field acted like super-glue, locking their hands together so they couldn't wiggle anymore.

Why This Matters

This paper is a game-changer because it moves from "guessing" to "knowing."

• Before: Engineers were trying to fix the noise by trial and error.
• Now: They know the noise comes from specific oxygen molecules interacting in a specific, messy way.
• The Future: Engineers can now design surfaces that prevent oxygen from sticking, or apply specific electric fields to "glue" the spins together, effectively silencing the noise and making quantum computers much more reliable.

In a nutshell: The researchers built a virtual reality of a messy dance floor of oxygen molecules, figured out exactly why they were making noise, and discovered that a simple electric field can turn that chaotic dance into a calm, silent line-up.

Technical Summary

1. Problem Statement

Superconducting quantum devices (qubits, magnetometers, dark matter detectors) suffer from magnetic flux noise, a primary source of decoherence. This noise typically exhibits a $1/f$ frequency dependence, significant cross-correlation with magnetic susceptibility, and a temperature-dependent behavior suggesting a paramagnetic origin.

• The Gap: While previous theoretical models have attributed this noise to surface defects (e.g., adsorbed oxygen or hydrogen), they often rely on generic disorder (e.g., assuming a random distribution of exchange couplings or a specific functional form). These models fail to capture the materials-specific spatial correlations and granular disorder inherent in real atomic lattices, limiting their ability to accurately predict noise magnitudes and mitigation strategies.
• The Challenge: Understanding the complex dynamics of these spin systems, which often approximate the Berezinskii-Kosterlitz-Thouless (BKT) phase transition, requires a model that incorporates realistic atomic-scale disorder, orientation-dependent interactions, and external field effects.

2. Methodology

The authors developed a first-principles-based simulation framework that eliminates free parameters and assumed functional forms for disorder. The workflow consists of three coupled stages:

A. Density Functional Theory (DFT) Calculations

• System: Paramagnetic $O_2$ molecules adsorbed on an Al-terminated $(0001)$ $Al_2O_3$ (sapphire) surface.
• Approach: Used VASP with van der Waals-corrected DFT (vdW-DF-cx) to calculate:
  • Binding Energies ($E_b$): For six stable $O_2$ orientations (grouped into two sets of three-fold degeneracy).
  • Exchange Couplings ($J$): Calculated for nearest-neighbor (NN) $O_2$ pairs in various relative orientations. The study identified 20 distinct symmetry-unique exchange values ranging from $-2.7$ meV to $+4.1$ meV.
  • Spin Anisotropy: Determined that spins prefer to lie in the plane perpendicular to the $O_2$ bond axis, approximating an XY spin model.
  • Electric Dipole Moments: Calculated the difference in electric dipole moments between ferromagnetic (FM) and antiferromagnetic (AFM) spin configurations, finding values up to $0.23$ eÅ.

B. Monte Carlo (MC) Simulations

• Configurational MC: Simulated the arrangement of $O_2$ molecules on a $20 \times 20$ lattice at $0.01$ K (representing the "frozen" disorder below 1 K). This generated realistic, disordered $O_2$ monolayers with varying coverages (50%, 75%, 100%).
• Spin MC: Using the DFT-derived exchange couplings and binding energies, the spin lattice was simulated using a Hamiltonian that includes anisotropy, external magnetic fields ($H_{ext}$), and exchange interactions.
  • Algorithm: Combined Metropolis and Wolff cluster updates to efficiently sample the spin configurations.
  • Output: Generated equilibrium spin configurations showing ferromagnetic domains separated by domain walls (often corresponding to AFM couplings or vacancies).

C. Landau-Lifshitz-Gilbert (LLG) Dynamics

• Dynamics: Simulated the time evolution of magnetization using the LLG equation to generate spin trajectories.
• Noise Calculation: The Fourier transform of the total spin trajectory yielded the magnetic flux noise power spectral density.
• Conversion: Converted spin noise power to magnetic flux noise ($\Phi$) using analytical relations for a SQUID loop geometry.

3. Key Contributions

1. Materials-Specific Disorder: Unlike previous models, this work simulates an ensemble of specific atomic configurations where disorder arises naturally from the geometric arrangement of adsorbates, capturing spatial correlations absent in generic models.
2. Parameter-Free Prediction: The simulation relies entirely on DFT-derived parameters (exchange couplings, anisotropy, binding energies) without fitting free parameters to experimental data.
3. Identification of TLS Mechanism: The study identifies that specific weak exchange interactions ($\sim 0.016$ meV) allow $O_2$ pairs to act as Two-Level Systems (TLS) that couple to GHz resonators, potentially causing both flux noise and dielectric loss.
4. Electric Field Tuning: The discovery that external electric fields can modulate spin-spin interactions via spin-dependent electric dipole moments, offering a new mitigation strategy.

4. Key Results

A. Flux Noise Characteristics

• Frequency Dependence: The simulated noise spectrum follows a $1/f^{0.8-1.0}$ scaling in the 30 MHz to 6 GHz range, matching experimental observations in superconducting loops.
• Magnitude: For a typical SQUID loop, the calculated flux noise at 10 MHz is $\approx 6.6 \times 10^{-9} \phi_0/\sqrt{Hz}$ for 100% coverage. This aligns with experimental data, suggesting that roughly 50% surface coverage of $O_2$ explains the noise levels observed in real devices.
• Resonance: A magnetic resonance peak appears between 1–2 GHz (near 1.6 GHz), attributed to spin waves.

B. Temperature and Field Dependence

• Temperature: Flux noise increases by two orders of magnitude as temperature rises from 0.01 K to 0.5 K, consistent with experimental trends.
• BKT Transition: The spin-spin correlation functions exhibit algebraic decay characteristic of the BKT phase. The simulated transition temperature ($T_{BKT}$) is 121 mK, consistent with the cusp observed in SQUID susceptibility measurements at 55 mK.
• Cross-Correlation: The simulation reproduces the non-zero cross-correlation between magnetization and susceptibility ($S_{M\chi''}/S_{M S_{\chi''}}$), which decreases with temperature. This confirms the presence of ferromagnetic clusters, a feature missing in pure spin-glass models.

C. Mitigation Strategies

• Magnetic Fields: Applying an external magnetic field polarizes spins, reducing flux noise. A field of 0.1 T reduces noise by a factor of ~59, consistent with experimental findings.
• Electric Fields (Novel Finding): An applied electric field ($10^7$ V/cm) enhances the exchange coupling strength via spin-dependent dipole interactions.
  • This raises the BKT transition temperature from 121 mK to 276 mK.
  • It significantly suppresses flux noise (reducing it to $4.3 \times 10^{-9} \phi_0/\sqrt{Hz}$ at 10 MHz) and alters the frequency scaling to $1/f^{0.25}$, indicating stronger noise suppression at lower frequencies.

5. Significance

This work provides a microscopic, predictive understanding of magnetic flux noise in superconducting circuits. By linking the noise directly to the specific atomic arrangement and exchange interactions of adsorbed $O_2$ on sapphire, the authors:

• Validate the hypothesis that surface paramagnetic adsorbates are the primary source of flux noise.
• Demonstrate that surface engineering (e.g., modifying surface termination to alter exchange couplings) is a viable path to noise reduction.
• Propose electric field tuning as a dynamic method to suppress flux noise without the detrimental effects of strong magnetic fields on superconductivity.
• Highlight the dual role of these defects as both flux noise sources and dielectric loss TLSs, unifying two major decoherence mechanisms in quantum computing.