Spin Dynamics from Atomistic Quantum Simulations

This work establishes a unified theoretical framework using Kubo linear response theory and machine learning–assisted molecular dynamics to accurately predict spin-lattice relaxation and decoherence times of solid-state spin defects, demonstrating excellent agreement between theoretical calculations and experimental measurements for the nitrogen-vacancy center in diamond.

The Gist

Imagine a tiny, glowing defect within a diamond, like a grain of dust that functions as a microscopic quantum computer. Scientists call this the "NV center." It is special because it can store a secret (quantum information) for a long time even in heat. However, there is a problem: when the diamond heats up, the secret begins to escape, and the quantum computer stops working.

For a long time, scientists had an excellent map for how this happens in the cold, but they were at a loss when it came to predicting what happens in the heat. This article creates a new, unified map that works from room temperature up to very hot conditions.

Here is how they did it, explained with everyday analogies:

1. The Problem: The "Wobbly Table"

Imagine the NV center as a spinning top on a table.

• The Spin: The spinning top is the "quantum state."
• The Lattice: The table is the diamond crystal itself, consisting of atoms that vibrate like jelly.
• The Heat: When you heat the diamond, the "jelly" on the table begins to wobble violently.

The scientists wanted to know: How quickly does the top fall over (lose its energy) or begin to go out of sync (lose its coherence) because the table is wobbling?

2. The Old Tools versus the New Tool

Previously, scientists used two different tools to investigate this:

• Tool A (The Cold Map): Good for low temperatures, but it assumed the table was stiff and moved only in simple, predictable ways. It failed when things became hot and chaotic.
• Tool B (The Heat Estimate): Good for high temperatures, but it was often just an estimate or a rough approximation.

This article introduces a new, unified framework (based on a theory called Kubo Linear Response Theory). Think of it as a universal translator that can describe the behavior of the top, whether the table barely moves or wobbles violently. It treats energy loss and loss of synchronization as two sides of the same coin: the top tries to calm down and synchronize with the rhythm of the wobbling table.

3. The Supercomputer Simulation

To test this new map, the team had to simulate the wobbling of the diamond.

• The Challenge: To get an accurate answer, you must observe billions of atoms over a long period. Doing this with conventional supercomputers is like trying to film a hurricane with a slow-motion camera; it takes too long and costs too much.
• The Solution: They used Machine Learning (AI).
  • First, they taught an AI (a "neural network") to predict how the atoms move by learning from a few perfect but expensive computer calculations.
  • Once the AI learned the rules, it could simulate the wobbling of the diamond for nanoseconds (which is a long time in the quantum world) with incredible speed and accuracy.
  • They also taught a second AI to predict how the "top" (the spin) reacts to the wobbling table.

4. The Experiment: Verifying the Map

The team did not rely solely on the computer. They went into the lab and actually measured how long the NV center in a diamond could hold its secret at various temperatures (from 300 K to 1000 K).

The Result:
When they compared their AI-supported predictions with their real laboratory results, the numbers matched almost perfectly.

• At lower temperatures: The "top" loses its energy slowly and follows a specific pattern (like a gentle slope).
• At higher temperatures: The "top" loses its energy much faster and follows a different pattern (like a steep drop).
• The new theory correctly predicted the "transition point" (at about 500 K) where the behavior changes.

5. What They Found Out About the "Noise"

The article also breaks down why the top falls over:

• Energy Loss (T1): This happens because the top exchanges energy with the wobbling table. The AI showed that this is purely about the top jumping between different energy levels.
• Confusion (T2): This is the moment when the top gets confused and stops spinning straight. The team found that at high temperatures, the main culprit is not energy exchange, but "pure dephasing"—the table wobbles so strongly that it simply disrupts the rhythm of the top.

The Conclusion

This article provides the first complete, accurate theory explaining how quantum spins behave in hot solids. By combining a solid mathematical theory with powerful AI simulations, they proved that they can accurately predict how long a quantum system will last in heat, and this matched perfectly with real experiments. This gives scientists a reliable tool to develop better quantum sensors and computers that can function in real, warm environments.

Technical Summary

Technical Conclusion: Spin Dynamics from Atomistic Quantum Simulations

Problem Statement
Optically active solid-state spin defects, such as the nitrogen-vacancy (NV) center in diamond, are promising platforms for quantum technologies. Although record coherence times have been achieved at room temperature with isotopically pure samples, the physical processes limiting coherence at elevated temperatures remain largely ununderstood. Existing theoretical frameworks, such as the Redfield master equation and the cluster correlation expansion (CCE), are often restricted to low-temperature regimes or rely on approximations (e.g., one- and two-phonon processes, frozen lattice) that cannot capture the full anharmonic lattice dynamics required for predictions at high temperatures. Consequently, a unified theoretical framework for predicting spin relaxation times ($T_1$ and $T_2$) at high temperatures was lacking.

Methodology
The authors develop a unified theoretical and computational framework based on Kubo linear response theory (LRT) combined with state-of-the-art machine learning (ML) methods for molecular dynamics (MD).

1. Theoretical Framework: Within the Kubo-LRT framework, the authors derive general expressions for the spin-lattice relaxation time ($T_1$) and the decoherence time ($T_2$) as functions of temperature. These expressions link relaxation rates to the time-correlation functions of spin-lattice couplings. The formalism treats spin relaxation as the regression of conserved quantities (spin energy and transverse magnetization) under external perturbations toward equilibrium.
2. Acceleration via Machine Learning: To overcome the prohibitive computational costs of first-principles (FP) calculations required for nanosecond-long trajectories and large supercells necessary for convergence, the authors employ the following:
   • MACE ML Interatomic Potential (MLIP): Trained on spin-polarized DFT/PBE data to approximate the potential energy surface and generate MD trajectories.
   • Neural Network (NN) for ZFS: A message-passing, equivariant graph neural network trained to calculate zero-field splitting (ZFS) tensors based on local atomic environments.
3. Simulation Protocol: The authors perform nanosecond-long MD simulations of a single NV center in diamond cells containing up to 10,648 atoms over a temperature range of 400 K to 1000 K. They utilize active learning to optimize the training dataset. The ZFS time series generated from these simulations are used to evaluate the Kubo expressions for $T_1$ and $T_2$.
4. Experimental Validation: The authors measure $T_1$ for NV centers in isotopically pure diamond samples over a temperature range of 300 K to 400 K to compare with theoretical predictions.

Main Contributions

• Unified Derivation: The work derives $T_1$ and $T_2$ expressions within the Kubo-LRT framework, unifying the description of spin-lattice relaxation and decoherence as responses to external driving fields, expressed via equilibrium time-correlation functions.
• Framework for High Temperatures: The study extends the analysis of spin relaxation to high temperatures (up to 1000 K) by including all phonon processes and anharmonicity, thereby surpassing the harmonic approximations and low-temperature limits of earlier methods such as the Redfield ME and CCE.
• ML-Driven Accuracy: By combining ML potentials with NN models for ZFS tensors, the authors achieve ab-initio accuracy at a fraction of the computational cost, enabling simulations of large systems over long timescales necessary for convergence.
• Decomposition of Mechanisms: The work explicitly decomposes relaxation into secular (pure dephasing) and non-secular (population exchange) contributions, clarifying their respective roles in $T_1$ and $T_2$.

Results

• Convergence and Validation: The ML-MD approach was tested against density functional perturbation theory (DFPT) for the vibrational density of states and phonon dispersion, showing excellent agreement. The temperature dependence of the axial coefficient of the ZFS ($D$) and its derivative ($\partial D/\partial T$) also agreed well with experimental findings, despite small absolute shifts attributable to the limitations of the PBE functional.
• Relaxation Times:
  • $T_1$ (Spin-Lattice): The calculated $T_1$ values show strong temperature dependence and decrease with increasing temperature. The results agree quantitatively with experimental measurements at 400 K and 500 K. The data follow a $T^{-5}$ scaling at low temperatures ($T \leq 400$ K) and transition to a $T^{-2}$ scaling at high temperatures ($T \geq 550$ K), consistent with Raman processes involving spin one-phonon interactions.
  • $T_2$ (Decoherence): In the absence of magnetic noise, decoherence is dominated by pure dephasing (secular terms) rather than population exchange. At 500 K, the lattice-induced $T_2$ is calculated as $414 \pm 87$ $\mu$s. When magnetic noise (calculated via CCE) is included, the total $T_2$ decreases by a factor of 4 compared to the natural abundance case at 0 K.
• Mechanistic Insight: The analysis confirms that $T_1$ depends exclusively on non-secular contributions (population exchange between triplet sub-levels), while $T_2$ is influenced by both secular (pure dephasing) and non-secular terms. A Fourier decomposition reveals that low-frequency lattice vibrations significantly influence the dynamics of the ZFS tensor.

Significance
The work claims to be the first study of the coherence properties of NV centers at high temperatures combining Kubo-LRT with MD simulations. By employing ML methods to accelerate simulations, the authors successfully model systems with approximately 10,000 atoms, capturing anharmonicity and all phonon processes that are computationally inaccessible to traditional Redfield or CCE methods at these scales. The excellent agreement between numerical predictions and new experimental data validates the framework. The authors emphasize that this approach is not limited to ZFS Hamiltonians or specific defects; it is a general framework applicable to any solid-state triplet color center, molecular qubit, or system with isotopic disorder, opening a path toward the predictive design of quantum materials for high-temperature applications.