🔬 materials science

First-principles calculation of electron-phonon spectral functions for defects using phonon interpolation

This paper introduces a phonon-interpolation method that combines localized ab initio calculations with dense $q$ -point grid sampling to accurately compute electron-phonon spectral functions and optical lineshapes for defects in wide-band-gap semiconductors, overcoming the resolution limits of standard supercell approaches as demonstrated on the nitrogen-vacancy center in diamond.

Original authors: Zoltan Santha, Gergo Thiering

Published 2026-07-24

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Zoltan Santha, Gergo Thiering

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Invisible Orchestra of a Broken Crystal

Imagine a diamond not as a sparkling gem, but as a perfectly choreographed dance floor where every atom is a dancer holding hands with its neighbors. In a perfect diamond, these dancers move in perfect unison, creating a silent, smooth rhythm. But what happens if one dancer is missing, or if a different dancer (like a nitrogen atom) steps in? This creates a "defect," a tiny glitch in the crystal's perfect rhythm. In the world of quantum physics, these defects are like tiny, glowing stars trapped inside the diamond. They are incredibly useful for future technologies, like ultra-fast computers or unhackable communication, because they can hold onto information (spin) for a long time and glow with very specific colors.

However, these glowing defects don't just sit still; they are constantly bumping into the dancing atoms around them. When the defect changes its energy state (like when it glows), it gives a little shove to the surrounding dancers. This interaction between the defect's "glow" and the crystal's "dance" is called electron-phonon coupling. Think of it like a singer (the defect) trying to hit a note while the audience (the crystal atoms) is constantly shifting their weight. If the audience is too noisy or the singer's voice is too complex, the song gets messy. To understand exactly what color the defect will glow, scientists need to map out every single way the audience can move. The problem is, the "audience" is huge, and the "dance moves" (vibrations) happen at speeds and scales that are incredibly hard to calculate with standard computer models. If the computer model is too small, it misses the slow, long waves of the dance; if it's too big, the computer crashes.

The Paper's Big Idea: A Smarter Way to Listen

In this paper, Zoltán Sántha and Gergő Thiering tackle this massive calculation problem by inventing a clever shortcut. They focus on a specific, famous defect in diamond called the negatively charged nitrogen-vacancy (NV) centre. This defect is the "singer" in our story, and the authors wanted to predict exactly how its light spectrum looks when it interacts with the diamond's vibrations.

Usually, to figure this out, scientists have to build a giant computer model of the diamond (a "supercell") containing thousands of atoms, calculate how every single atom moves, and then see how the defect shakes them. But even the biggest supercells they can build are too small to catch the very slow, long-wavelength vibrations that are crucial for the final sound of the light. It's like trying to hear the bass of a song by only listening to the first few seconds of a tiny speaker; you miss the deep rumble.

The authors' breakthrough is a method they call phonon interpolation. Instead of trying to calculate the movement of every atom in a massive, impossible-to-simulate crystal, they change the approach. They realized that when the NV defect changes its energy, the "push" it gives to the surrounding atoms is highly localized. Imagine the defect giving a sharp, quick tap to the atoms right next to it, rather than a slow, gentle wave that ripples across the whole room.

Here is the magic trick:

The Tap: They use standard, smaller computer models to calculate exactly how hard and in what direction the defect "taps" its immediate neighbors.
The Ripple: They then take that "tap" and pretend it happened in a much, much larger crystal (a "hypercell") that is too big to simulate directly.
The Math: Using a mathematical interpolation technique, they spread that single "tap" across the massive crystal to see how the vibrations ripple out. This allows them to sample millions of possible vibration modes without actually having to calculate the forces for every single atom in that giant crystal.

What They Found

When they applied this method to the NV centre in diamond, the results were striking.

First, they confirmed their hunch: the "tap" from the defect is indeed very short-ranged. They found that in a relatively small computer model (a 4×4×4 supercell, which contains about 512 carbon atoms before the defect is added), the force is already fully captured. The "tap" doesn't reach far; it dies out quickly. This means they didn't need to simulate a crystal with millions of atoms to know how the defect pushes; they just needed a small model to find the push, and then a mathematical trick to see how that push travels.

Second, they used this method to simulate a hypercell equivalent to 32×32×32 repetitions of their small model. This corresponds to a crystal with approximately 17 million atoms. This is a scale that would be impossible to simulate directly. By doing this, they recovered a smooth, continuous picture of the vibrations, rather than the jagged, sparse picture you get from small models.

Their calculations revealed a few key details about the NV centre's light:

The Main Beat: The strongest interaction happens at an energy of 63 meV. This creates a distinct "sideband" in the light spectrum, a series of peaks that appear next to the main color of the light.
The Low Rumble: At very low energies, the vibrations follow a predictable, linear pattern, which their method captured perfectly.
The Fine Details: They were able to see tiny, sharp structures in the spectrum caused by specific features in the diamond's vibration patterns (called van Hove singularities), which smaller models completely missed.

When they compared their simulated light spectrum to real-world experiments, the match was excellent for emission (when the defect glows). The simulated curve perfectly reproduced the main peaks and the smaller, wiggly details seen in lab experiments at low temperatures.

However, they noted a slight hiccup with absorption (when the defect soaks up light). The simulation using the ground-state vibrations didn't match the experiment as perfectly. They suspect this is because the excited state of the NV centre is "Jahn-Teller active," meaning it's a bit unstable and wants to twist its shape in complex ways that their current simplified math doesn't fully capture. When they tried using the vibrations of the excited state instead, the match improved, suggesting that the "twist" is important for absorption but less so for emission.

The Bottom Line

This paper doesn't claim to have solved every mystery of diamond defects, but it provides a powerful new tool. It shows that you don't need to brute-force your way through a simulation of 17 million atoms to get accurate results. Instead, by understanding that the defect's influence is local, you can use a small, manageable calculation to find the "source" of the vibration and then mathematically expand it to a massive scale.

The authors demonstrated that this "phonon interpolation" method allows scientists to see the fine details of how defects interact with light, bridging the gap between small, manageable computer models and the vast, continuous reality of a crystal. It's a bit like being able to hear the full, deep sound of a symphony orchestra by listening to just the conductor's baton and knowing exactly how the music should ripple through the hall. For researchers trying to build quantum technologies, this means they can now predict the behavior of these tiny quantum stars with much greater precision and less computing power.

Technical Summary: First-Principles Calculation of Electron-Phonon Spectral Functions for Defects Using Phonon Interpolation

Problem Statement
Point defects in wide-band-gap semiconductors, such as the negatively charged nitrogen-vacancy (NV) center in diamond, exhibit optical spectra heavily influenced by electron-phonon coupling. Accurate first-principles prediction of these spectra requires calculating the phonon-resolved electron-phonon coupling density (Huang-Rhys factors) and the resulting optical line-shape functions. However, direct ab initio calculations are severely limited by the finite size of accessible supercells. A supercell containing $N$ ions yields only $3N-3$ vibrational modes, resulting in a sparse sampling of the phonon spectrum. This sparsity fails to capture the long-wavelength acoustic continuum (crucial for low-energy behavior) and the fine structures near van Hove singularities in the optical phonon regime, both of which are essential for reproducing accurate experimental line shapes. While previous approaches have utilized embedding strategies or machine-learned potentials to access larger systems, they often involve complex model construction or rely on the transferability of trained potentials.

Methodology
The authors propose a phonon-interpolation method that decouples the local defect physics from the sampling of the vibrational continuum. The core of the method relies on the observation that for charge-conserving optical transitions between localized defect states in a gapped host, the transition-induced force is strongly localized around the defect.

Transition-Force Reconstruction: Instead of using the relaxed displacement field of a finite supercell (which represents a periodic array of defects), the method reconstructs the localized transition-force source ( $F^{tr}$ ). This is derived from the difference in Born-Oppenheimer forces between the initial and final electronic states at the same nuclear configuration. The force is calculated using density functional theory (DFT) within a finite supercell (e.g., $4\times4\times4$ ).
Locality and Convergence: The transition force is shown to decay exponentially with distance from the defect. Once the supercell is large enough to suppress periodic image overlap of the defect orbitals, the transition force converges rapidly. This allows the force field obtained from a manageable ab initio supercell to serve as an accurate source term for an isolated defect.
Hypercell Interpolation: The method constructs a "hypercell" by repeating the defect-containing supercell $M_1 \times M_2 \times M_3$ times. The force-constant matrices (dynamical matrices) are interpolated from the ab initio supercell to this larger hypercell. By diagonalizing these interpolated dynamical matrices on dense $q$ -point grids, the method generates a dense set of vibrational modes (phonons) for the hypercell.
Spectral Function Calculation: The localized transition force is coupled to this dense vibrational continuum. The partial Huang-Rhys factors ( $S_{q,\nu}$ ) are calculated by projecting the transition force onto the interpolated phonon eigenvectors. This allows for the calculation of smooth, continuous Huang-Rhys spectral densities $S(\hbar\omega)$ and optical line-shape functions ( $L_{em}$ and $L_{abs}$ ) with ultrafine spectral resolution, effectively simulating systems with millions of atoms (e.g., $32\times32\times32$ hypercells $\approx$ 17 million atoms) without performing direct ab initio calculations on such scales.

Key Contributions

Formulation: The paper formulates a phonon-interpolation procedure that directly connects to the harmonic Huang-Rhys theory of optical line shapes. It replaces the displacement field with a localized transition-force source, enabling the use of interpolated dynamical matrices to sample the vibrational continuum.
Efficiency: The method avoids the prohibitive cost of direct ab initio calculations on massive supercells while retaining the accuracy of ab initio descriptions for local defect physics. It is computationally more accessible than embedding methods that require constructing large embedded models and managing cut-off force-constant matrices.
Implementation: The authors provide a publicly available Python software package (pysideband) that interfaces with standard ab initio codes (via Phonopy) to post-process force constants and extrapolate vibrational properties.

Results
The method was benchmarked on the NV $^-$ center in diamond, specifically the optical transition between the $^3A_2$ ground state and the $^3E$ excited state.

Force Localization: Analysis of the transition force in supercells ranging from $2\times2\times2$ to $4\times4\times4$ confirmed strong localization. Convergence metrics (participation number, RMS localization radius) stabilized in the $4\times4\times4$ supercell, validating the use of this size as a source for the isolated defect force.
Spectral Density: Interpolation to a $32\times32\times32$ $32 \times 32 \times 32$ hypercell yielded smooth Huang-Rhys spectral densities. The results showed:
- A dominant coupling band centered near 63 meV.
- Correct linear scaling ( $S(\hbar\omega) \propto \hbar\omega$ ) in the low-energy acoustic region, avoiding unphysical divergences.
- Resolution of fine structures associated with van Hove singularities near the optical phonon band edge, which are typically missed in sparse supercell calculations.
Line Shapes:
- Emission: The calculated emission spectrum (using ground-state phonons) showed excellent agreement with low-temperature experimental data, reproducing the zero-phonon line, the 63 meV phonon replicas, and the sideband envelope.
- Absorption: Calculations using ground-state phonons showed discrepancies with experimental absorption profiles, particularly in the first phonon peak. However, recalculating absorption using excited-state phonons significantly improved the agreement, recovering the experimentally observed flat, weakly split structure of the main sideband. This highlights the sensitivity of absorption to the Jahn-Teller active nature of the $^3E$ state.
Convergence: The total Huang-Rhys factor ( $S_{tot}$ ) was shown to converge systematically with hypercell size following a power-law approach, allowing for the extrapolation of infinite-system limits.

Significance and Claims
The paper claims that the transition-force interpolation method provides an efficient and systematically improvable route to converged electron-phonon spectral functions for localized defects. It addresses the specific limitation of sparse vibrational sampling in standard ab initio defect calculations without requiring prohibitively large direct calculations.

The authors emphasize that their approach:

Retains the ab initio description of local defect relaxation while overcoming vibrational sampling limitations.
Is distinct from and complementary to machine-learned potential approaches, as its precision is controlled by the underlying first-principles force source rather than potential training errors.
Is simpler to deploy than embedding methods, as it does not require constructing large embedded models or managing complex cut-off matrices.

The work acknowledges limitations, noting that the current implementation relies on the parallel-mode, equal-frequency approximation and harmonic treatment. Consequently, while it successfully describes emission, the absorption spectrum requires careful consideration of the excited-state surface and Jahn-Teller activity, suggesting that full Duschinsky mode mixing and explicit vibronic coupling treatments would be necessary for a quantitatively complete description of systems where the final-state potential-energy surface plays a dominant role.

The Invisible Orchestra of a Broken Crystal

The Paper's Big Idea: A Smarter Way to Listen

What They Found

The Bottom Line

Technical Summary: First-Principles Calculation of Electron-Phonon Spectral Functions for Defects Using Phonon Interpolation

More like this