pith. sign in

arxiv: 2508.09113 · v1 · submitted 2025-08-12 · ❄️ cond-mat.mtrl-sci

Machine Learning Phonon Spectra for Fast and Accurate Optical Lineshapes of Defects

Mark E. Turiansky , John L. Lyons , Noam Bernstein This is my paper

Pith reviewed 2026-05-18 22:38 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci
keywords machine learning interatomic potentialsphonon spectraoptical lineshapesdefects in solidselectron-phonon couplingquantum defectsT center in siliconhybrid functionals
0
0 comments X p. Extension

The pith

Machine learning interatomic potentials trained on routine relaxation data can compute accurate phonon spectra for optical lineshapes of defects with negligible loss compared to explicit calculations.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that calculating all phonon modes in large simulation cells with hundreds of atoms creates a major computational barrier for predicting optical lineshapes of defects from first principles. Machine learning interatomic potentials overcome this barrier when fine-tuned on atomic relaxation data already generated during standard first-principles runs. This training approach proves sufficient to match explicit phonon results closely enough that derived electron-phonon coupling and optical spectra remain accurate. The resulting efficiency makes it practical to apply higher-level hybrid functionals and to examine fine details such as local vibrational mode coupling in the luminescence of the T center in silicon.

Core claim

Machine learning interatomic potentials fine-tuned on atomic relaxation data from routine first-principles calculations produce phonon spectra that enable fast computation of optical lineshapes for defects, matching explicit evaluations in large supercells with negligible accuracy loss and agreeing with experimental spectra for several defects.

What carries the argument

Machine learning interatomic potentials fine-tuned on forces and energies from atomic relaxation steps to generate phonon frequencies and modes for electron-phonon coupling calculations.

Load-bearing premise

Atomic relaxation data from standard first-principles calculations contains enough information to train machine learning interatomic potentials that reproduce the harmonic phonon spectrum needed for accurate optical lineshapes.

What would settle it

Direct comparison of the optical lineshape or phonon spectrum for a specific defect computed with the machine learning potential versus an explicit first-principles phonon calculation showing large discrepancies.

Figures

Figures reproduced from arXiv: 2508.09113 by John L. Lyons, Mark E. Turiansky, Noam Bernstein.

Figure 1
Figure 1. Figure 1: FIG. 1. The phonon density of states (left), spectral density ( [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Luminescence spectrum of [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Luminescence spectrum of N [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Luminescence spectrum of Bi [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Luminescence spectrum of the T center in Si at [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
read the original abstract

The optical properties of defects in solids produce rich physics, from gemstone coloration to single-photon emission for quantum networks. Essential to describing optical transitions is electron-phonon coupling, which can be predicted from first principles but requires computationally expensive evaluation of all phonon modes in simulation cells containing hundreds of atoms. We demonstrate that this bottleneck can be overcome using machine learning interatomic potentials with negligible accuracy loss. A key finding is that atomic relaxation data from routine first-principles calculations suffice as a dataset for fine-tuning, though additional data can further improve models. The efficiency of this approach enables studies of defect vibrational properties with high-level theory. We fine-tune to hybrid functional calculations to obtain highly accurate spectra, comparing with explicit calculations and experiments for various defects. Notably, we resolve fine details of local vibrational mode coupling in the luminescence spectrum of the T center in Si, a prominent quantum defect.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The manuscript claims that machine learning interatomic potentials (MLIPs) fine-tuned on atomic relaxation data from routine first-principles calculations can accurately reproduce phonon spectra in large defect supercells, overcoming the computational cost of explicit phonon calculations while incurring negligible accuracy loss for derived optical lineshapes. This is supported by comparisons to explicit hybrid-functional calculations and experiments across several defects, with particular emphasis on resolving fine details of local vibrational mode coupling in the luminescence spectrum of the T center in Si.

Significance. If the accuracy claims hold, the work removes a key bottleneck for computing electron-phonon coupling in defect systems with hundreds of atoms, enabling routine application of high-level theories such as hybrid functionals. A notable strength is the demonstration that routinely available relaxation trajectories suffice for training, together with direct end-to-end comparisons to explicit calculations and experimental spectra that validate the approach for optical lineshapes. The stress-test concern about undersampling of the quadratic region by relaxation paths does not appear to undermine the central claim, because the reported agreement in the final optical lineshapes provides an integrated test of the relevant phonon modes and couplings.

minor comments (3)
  1. Abstract: the phrase 'negligible accuracy loss' would be strengthened by a brief quantitative statement (e.g., RMS frequency error or lineshape overlap metric) that is referenced to the explicit comparisons performed later in the manuscript.
  2. Methods section (training and phonon evaluation): clarify whether the MLIP is used only for forces during relaxation or also to construct the full dynamical matrix for the supercell, and state the supercell sizes employed for both MLIP and explicit calculations.
  3. Figure captions and text describing the T-center spectrum: ensure that the specific local modes whose coupling is resolved are labeled with their frequencies and irreducible representations so that readers can directly compare to the explicit reference calculation.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive and constructive assessment of our work. The recognition that routinely available relaxation data suffice for fine-tuning MLIPs, together with the end-to-end validation against explicit hybrid-functional calculations and experimental spectra, is encouraging. We are pleased that the integrated test via optical lineshapes is viewed as sufficient to address concerns about quadratic-region sampling.

Circularity Check

0 steps flagged

No significant circularity detected in the derivation chain

full rationale

The paper trains machine learning interatomic potentials on atomic relaxation trajectories obtained from routine first-principles DFT calculations, then uses the resulting surrogate potentials to compute phonon spectra and derived optical lineshapes in large supercells. This workflow is a standard surrogate-modeling procedure in which the ML model approximates an external potential-energy surface; the phonon predictions are obtained by diagonalizing the dynamical matrix of the trained model rather than being algebraically or statistically identical to the training inputs by construction. No equations or claims in the provided text reduce a central result to a fitted parameter, a self-citation chain, or a renamed empirical pattern. Validation against explicit hybrid-functional calculations and experimental spectra supplies independent external benchmarks, rendering the derivation self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The central claim rests on the transferability of ML potentials trained primarily on relaxation data to full phonon spectra; no explicit free parameters, axioms, or invented entities are stated in the abstract.

pith-pipeline@v0.9.0 · 5686 in / 1059 out tokens · 25947 ms · 2026-05-18T22:38:15.271325+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

  • IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    We demonstrate that this bottleneck can be overcome using machine learning interatomic potentials with negligible accuracy loss. A key finding is that atomic relaxation data from routine first-principles calculations suffice as a dataset for fine-tuning

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

70 extracted references · 70 canonical work pages · 3 internal anchors

  1. [1]

    VSiVC in 4H-SiC SiC is a mature electronic material utilized for a variety of applications and has been gaining attention as a host for quantum defects [ 32]

    We shift the calculated spectrum to account for typical DFT errors, aligning the peak intensity, and normalize both spectra to the peak intensity. VSiVC in 4H-SiC SiC is a mature electronic material utilized for a variety of applications and has been gaining attention as a host for quantum defects [ 32]. One such quantum defect is the divacancy center VSi...

    work page

  2. [2]

    The experimental spectrum is normalized to approximately match the phonon sideband of the calculation

    We shift the calculated spectrum to align the ZPL. The experimental spectrum is normalized to approximately match the phonon sideband of the calculation. BiPb in CsPbBr 3 CsPbBr3 is from the family of halide perovskites, which are being actively pursued for optoelectronic devices [ 35], especially for photovoltaics where the efficiencies rival state-of-the-...

    work page

  3. [3]

    We shift the calculated spectrum to align with the peak intensity and normalize both spectra to the peak intensity. T Center in Si The T center in Si is a promising quantum defect with a telecom-wavelength spin-photon interface [ 39] and is the subject of commercial efforts to realize quantum comput- ing [ 40]. Si is an appealing host material for a variet...

    work page

  4. [4]

    When local vibrational modes are present, a Lorentzian is used in place of the Gaussian function for each local vibrational mode, following the suggestion of Ref

    and ( 8) are replaced with a Gaussian whose width varies linearly with the phonon frequency (Supplementary Information Sec- tion S2). When local vibrational modes are present, a Lorentzian is used in place of the Gaussian function for each local vibrational mode, following the suggestion of Ref

    work page

  5. [5]

    Local vibrational modes are identified by their inverse participation ratio (Supplementary Information Section S2). Although it is not directly needed for calcu- lating the luminescence spectrum, we define the phonon density of states as ρ(ℏω ) = 1 3N ∑ i δ(ℏω − ℏω i) , (9) where the same broadening procedure described above is used to address the delta fun...

    work page 2024

  6. [6]

    multihead

    could be viewed as a test dataset. We do not in- tend for equivalent benchmarking calculations to be done in practice when applying our approach, as that would defeat the purpose. We also do not utilize “multihead” fine-tuning [ 61]: while multihead fine-tuning can help to stabilize the model, we found that the stabilization comes at the ex- pense of better...

    work page

  7. [7]

    The MACE dynamical matrix is then used to construct the forces that would give rise to this displacement within the harmonic approximation [ 62]

    We fine-tune a MACE model to the relaxation dataset of the ground state, 30 additional configurations generated using the optimized phonon method detailed above, and 10 random displacements in the pristine supercell. The MACE dynamical matrix is then used to construct the forces that would give rise to this displacement within the harmonic approximation [ 6...

    work page

  8. [8]

    Fritsch and G

    E. Fritsch and G. R. Rossman, An update on color in gems. part 1: Introduction and colors caused by dispersed metal ions, Gems Gemol. 23, 126 (1987) . 10

    work page 1987

  9. [9]

    T. H. Maiman, Stimulated optical radiation in ruby, Na- ture 187, 493 (1960)

    work page 1960

  10. [10]

    C. E. Dreyer, A. Alkauskas, J. L. Lyons, A. Janotti, and C. G. Van de Walle, First-Principles Calculations of Point Defects for Quantum Technologies, Annu. Rev. Mater. Res. 48, 1 (2018)

    work page 2018

  11. [11]

    Wolfowicz, F

    G. Wolfowicz, F. J. Heremans, C. P. Anderson, S. Kanai, H. Seo, A. Gali, G. Galli, and D. D. Awschalom, Quan- tum guidelines for solid-state spin defects, Nat. Rev. Mater. 6, 906 (2021)

    work page 2021

  12. [12]

    M. E. Turiansky, K. Parto, G. Moody, and C. G. Van de Walle, Rational design of efficient defect-based quantum emitters, APL Photonics 9, 066117 (2024)

    work page 2024

  13. [13]

    T. E. Northup and R. Blatt, Quantum information trans- fer using photons, Nat. Photonics 8, 356 (2014)

    work page 2014

  14. [14]

    A. M. Stoneham, Theory of Defects in Solids: Electronic Structure of Defects in Insulators and Semiconductors (Clarendon Press, Oxford, 1975)

    work page 1975

  15. [15]

    Freysoldt, B

    C. Freysoldt, B. Grabowski, T. Hickel, J. Neugebauer, G. Kresse, A. Janotti, and C. G. Van de Walle, First- principles calculations for point defects in solids, Rev. Mod. Phys. 86, 253 (2014)

    work page 2014

  16. [16]

    Alkauskas, B

    A. Alkauskas, B. B. Buckley, D. D. Awschalom, and C. G. Van de Walle, First-principles theory of the luminescence lineshape for the triplet transition in diamond NV cen- tres, New J. Phys. 16, 073026 (2014)

    work page 2014

  17. [17]

    H. Wang, T. Fu, Y. Du, W. Gao, K. Huang, Z. Liu, P. Chandak, S. Liu, P. Van Katwyk, A. Deac, A. Anand- kumar, K. Bergen, C. P. Gomes, S. Ho, P. Kohli, J. Lasenby, J. Leskovec, T.-Y. Liu, A. Manrai, D. Marks, B. Ramsundar, L. Song, J. Sun, J. Tang, P. Veliˇ ckovi´ c, M. Welling, L. Zhang, C. W. Coley, Y. Bengio, and M. Zitnik, Scientific discovery in the ag...

    work page 2023

  18. [18]

    Chen and S

    C. Chen and S. P. Ong, A universal graph deep learning interatomic potential for the periodic table, Nat. Com- put. Sci. 2, 718 (2022)

    work page 2022

  19. [19]

    A foundation model for atomistic materials chemistry

    I. Batatia, P. Benner, Y. Chiang, A. M. Elena, D. P. Kov´ acs, J. Riebesell, X. R. Advincula, M. Asta, M. Avaylon, W. J. Baldwin, F. Berger, N. Bernstein, A. Bhowmik, S. M. Blau, V. C˘ arare, J. P. Darby, S. De, F. D. Pia, V. L. Deringer, R. Elijoˇ sius, Z. El-Machachi, F. Falcioni, E. Fako, A. C. Ferrari, A. Genreith-Schriever, J. George, R. E. A. Goodal...

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  20. [20]

    A. Loew, D. Sun, H.-C. Wang, S. Botti, and M. A. L. Marques, Universal machine learning interatomic poten- tials are ready for phonons, npj Comput. Mater. 11, 178 (2025)

    work page 2025

  21. [21]

    B. Deng, Y. Choi, P. Zhong, J. Riebesell, S. Anand, Z. Li, K. Jun, K. A. Persson, and G. Ceder, Systematic soften- ing in universal machine learning interatomic potentials, npj Comput. Mater. 11, 9 (2025)

    work page 2025

  22. [22]

    H. Kaur, F. D. Pia, I. Batatia, X. R. Advincula, B. X. Shi, J. Lan, G. Cs´ anyi, A. Michaelides, and V. Kapil, Data-efficient fine-tuning of foundational mod- els for first-principles quality sublimation enthalpies, Faraday Discuss. 256, 120 (2025)

    work page 2025

  23. [23]

    Thermal conductivity predictions with foundation atomistic models

    B. P´ ota, P. Ahlawat, G. Cs´ anyi, and M. Simoncelli,Ther- mal Conductivity Predictions with Foundation Atomistic Models (2024), arXiv:2408.00755 [cond-mat]

    work page arXiv 2024

  24. [24]

    Mosquera-Lois, S

    I. Mosquera-Lois, S. R. Kavanagh, A. M. Ganose, and A. Walsh, Machine-learning structural reconstructions for accelerated point defect calculations, npj Comput Mater 10, 1 (2024)

    work page 2024

  25. [25]

    Mosquera-Lois, J

    I. Mosquera-Lois, J. Klarbring, and A. Walsh, Point de- fect formation at finite temperatures with machine learn- ing force fields, Chem. Sci. 16, 8878 (2025)

    work page 2025

  26. [26]

    S. R. Kavanagh, Identifying Split Vacancy Defects with Machine-Learned Foundation Models and Electrostatics (2025), arXiv:2412.19330 [cond-mat]

    work page arXiv 2025

  27. [27]

    Linder¨ alv, N

    C. Linder¨ alv, N. ¨Osterbacka, J. Wiktor, and P. Erhart, Optical line shapes of color centers in solids from classical autocorrelation functions (2024), arXiv:2408.06908

    work page arXiv 2024

  28. [28]

    Sharma, A

    K. Sharma, A. Loew, H. Wang, F. A. Nilsson, M. A. L. Marques, and K. S. Thygesen, Accelerating point defect photo-emission calculations with machine learning inter- atomic potentials (2025), arXiv:2505.01403 [cond-mat]

    work page arXiv 2025

  29. [29]

    A. J. Cohen, P. Mori-S´ anchez, and W. Yang, Insights into Current Limitations of Density Functional Theory, Science 321, 792 (2008)

    work page 2008

  30. [30]

    Wickramaratne and J

    D. Wickramaratne and J. L. Lyons, Assessing the SCAN functional for deep defects and small polarons in wide band gap semiconductors and insulators, Phys. Rev. B 109, 245201 (2024)

    work page 2024

  31. [31]

    Razinkovas, M

    L. Razinkovas, M. W. Doherty, N. B. Manson, C. G. Van de Walle, and A. Alkauskas, Vibrational and vibronic structure of isolated point defects: The nitrogen-vacancy center in diamond, Phys. Rev. B 104, 045303 (2021)

    work page 2021

  32. [32]

    J. L. Lyons, A. Janotti, and C. G. Van de Walle, Carbon impurities and the yellow luminescence in GaN, Appl. Phys. Lett. 97, 152108 (2010)

    work page 2010

  33. [33]

    Mackoit-Sinkeviˇ cien˙ e, M

    M. Mackoit-Sinkeviˇ cien˙ e, M. Maciaszek, C. G. Van de Walle, and A. Alkauskas, Carbon dimer defect as a source of the 4.1 eV luminescence in hexagonal boron nitride, Appl. Phys. Lett. 115, 212101 (2019)

    work page 2019

  34. [34]

    Batatia, D

    I. Batatia, D. P. Kovacs, G. Simm, C. Ortner, and G. Csanyi, MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields, in Advances in Neural Information Processing Systems, Vol. 35 (2022) pp. 11423–11436

    work page 2022

  35. [35]

    Batatia, S

    I. Batatia, S. Batzner, D. P. Kov´ acs, A. Musaelian, G. N. C. Simm, R. Drautz, C. Ortner, B. Kozinsky, and G. Cs´ anyi, The design space of E(3)-equivariant atom- centred interatomic potentials, Nat. Mach. Intell. 7, 56 (2025)

    work page 2025

  36. [36]

    Janotti and C

    A. Janotti and C. G. Van de Walle, Fundamentals of zinc oxide as a semiconductor, Rep. Prog. Phys. 72, 126501 (2009)

    work page 2009

  37. [37]

    J. L. Lyons, A. Janotti, and C. G. Van de Walle, Why ni- trogen cannot lead to p-type conductivity in ZnO, Appl. Phys. Lett. 95, 252105 (2009)

    work page 2009

  38. [38]

    M. C. Tarun, M. Z. Iqbal, and M. D. McCluskey, Nitrogen is a deep acceptor in ZnO, AIP Adv. 1, 022105 (2011) . 11

    work page 2011

  39. [39]

    N. T. Son, C. P. Anderson, A. Bourassa, K. C. Miao, C. Babin, M. Widmann, M. Niethammer, J. Ul Hassan, N. Morioka, I. G. Ivanov, F. Kaiser, J. Wrachtrup, and D. D. Awschalom, Developing silicon carbide for quan- tum spintronics, Appl. Phys. Lett. 116, 190501 (2020)

    work page 2020

  40. [40]

    D. J. Christle, P. V. Klimov, C. F. de las Casas, K. Sz´ asz, V. Iv´ ady, V. Jokubavicius, J. Ul Hassan, M. Syv¨ aj¨ arvi, W. F. Koehl, T. Ohshima, N. T. Son, E. Janz´ en, ´A. Gali, and D. D. Awschalom, Isolated Spin Qubits in SiC with a High-Fidelity Infrared Spin-to-Photon Interface, Phys. Rev. X 7, 021046 (2017)

    work page 2017

  41. [41]

    Y. Jin, M. Govoni, G. Wolfowicz, S. E. Sullivan, F. J. Heremans, D. D. Awschalom, and G. Galli, Photolumi- nescence spectra of point defects in semiconductors: Val- idation of first-principles calculations, Phys. Rev. Mater. 5, 084603 (2021)

    work page 2021

  42. [42]

    A. Dey, J. Ye, A. De, E. Debroye, S. K. Ha, E. Bladt, A. S. Kshirsagar, Z. Wang, J. Yin, Y. Wang, L. N. Quan, F. Yan, M. Gao, X. Li, J. Shamsi, T. Debnath, M. Cao, M. A. Scheel, S. Kumar, J. A. Steele, M. Ger- hard, L. Chouhan, K. Xu, X.-g. Wu, Y. Li, Y. Zhang, A. Dutta, C. Han, I. Vincon, A. L. Rogach, A. Nag, A. Samanta, B. A. Korgel, C.-J. Shih, D. R. ...

    work page 2021

  43. [43]

    L. Pan, Z. Liu, C. Welton, V. V. Klepov, J. A. Pe- ters, M. C. De Siena, A. Benadia, I. Pandey, A. Miceli, D. Y. Chung, G. N. M. Reddy, B. W. Wessels, and M. G. Kanatzidis, Ultrahigh-flux x-ray detection by a solution- grown perovskite cspbbr 3 single-crystal semiconductor detector, Adv. Mater. 35, 2211840 (2023)

    work page 2023

  44. [44]

    Brittman, B

    S. Brittman, B. T. Jones, M. H. Stewart, B. A. March- eschi, P. D. Cunningham, N. V. Proscia, C. T. Ellis, J. P. Murphy, K. Sendgikoski, T. H. Brintlinger, and J. L. Lyons, Near-infrared photoluminescence from bismuth, a deep defect in cesium lead bromide perovskite, Appl. Phys. Lett. 126, 032103 (2025)

    work page 2025

  45. [45]

    Zhang, M

    X. Zhang, M. E. Turiansky, and C. G. Van de Walle, Cor- rectly Assessing Defect Tolerance in Halide Perovskites, J. Phys. Chem. C 124, 6022 (2020)

    work page 2020

  46. [46]

    Bergeron, C

    L. Bergeron, C. Chartrand, A. T. K. Kurkjian, K. J. Morse, H. Riemann, N. V. Abrosimov, P. Becker, H.-J. Pohl, M. L. W. Thewalt, and S. Simmons, Silicon-integrated telecommunications photon-spin inter- face, PRX Quantum 1, 020301 (2020)

    work page 2020

  47. [47]

    Afzal, M

    F. Afzal, M. Akhlaghi, S. J. Beale, O. Bedroya, K. Bell, L. Bergeron, K. Bonsma-Fisher, P. Bychkova, Z. M. E. Chaisson, C. Chartrand, C. Clear, A. Dar- cie, A. DeAbreu, C. DeLisle, L. A. Duncan, C. D. Smith, J. Dunn, A. Ebrahimi, N. Evetts, D. F. Pin- heiro, P. Fuentes, T. Georgiou, B. Guha, R. Haenel, D. Higginbottom, D. M. Jackson, N. Jahed, A. Khorshi-...

    work page 2024

  48. [48]

    Karaiskaj, M

    D. Karaiskaj, M. L. W. Thewalt, T. Ruf, M. Cardona, H.-J. Pohl, G. G. Deviatych, P. G. Sennikov, and H. Rie- mann, Photoluminescence of Isotopically Purified Silicon: How Sharp are Bound Exciton Transitions?, Phys. Rev. Lett. 86, 6010 (2001)

    work page 2001

  49. [49]

    Dhaliah, Y

    D. Dhaliah, Y. Xiong, A. Sipahigil, S. M. Griffin, and G. Hautier, First-principles study of the T center in sili- con, Phys. Rev. Mater. 6, L053201 (2022)

    work page 2022

  50. [50]

    Komsa, T

    H.-P. Komsa, T. T. Rantala, and A. Pasquarello, Finite- size supercell correction schemes for charged defect cal- culations, Phys. Rev. B 86, 045112 (2012)

    work page 2012

  51. [51]

    Zhong, D

    P. Zhong, D. Kim, D. S. King, and B. Cheng, Ma- chine learning interatomic potential can infer electrical response (2025), arXiv:2504.05169 [cond-mat]

    work page arXiv 2025

  52. [52]

    Falletta, A

    S. Falletta, A. Cepellotti, A. Johansson, C. W. Tan, M. L . Descoteaux, A. Musaelian, C. J. Owen, and B. Kozinsky, Unified differentiable learning of electric response, Nat. Commun. 16, 4031 (2025)

    work page 2025

  53. [53]

    S. Fang, M. Geiger, J. G. Checkelsky, and T. Smidt, Phonon predictions with E(3)-equivariant graph neural networks (2024), arXiv:2403.11347 [cond-mat]

    work page arXiv 2024

  54. [54]

    Kaltak, J

    M. Kaltak, J. Klimeˇ s, and G. Kresse, Cubic scaling algorithm for the random phase approximation: Self- interstitials and vacancies in Si, Phys. Rev. B 90, 054115 (2014)

    work page 2014

  55. [55]

    Y. Jin, V. W.-z. Yu, M. Govoni, A. C. Xu, and G. Galli, Excited State Properties of Point Defects in Semiconduc- tors and Insulators Investigated with Time-Dependent Density Functional Theory, J. Chem. Theory Comput. 19, 8689 (2023)

    work page 2023

  56. [56]

    Muechler, D

    L. Muechler, D. I. Badrtdinov, A. Hampel, J. Cano, M. R¨ osner, and C. E. Dreyer, Quantum embedding meth- ods for correlated excited states of point defects: Case studies and challenges, Phys. Rev. B 105, 235104 (2022)

    work page 2022

  57. [57]

    Kresse and J

    G. Kresse and J. Furthm¨ uller, Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set, Phys. Rev. B 54, 11169 (1996)

    work page 1996

  58. [58]

    Kresse and J

    G. Kresse and J. Furthm¨ uller, Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set, Comput. Mater. Sci. 6, 15 (1996)

    work page 1996

  59. [59]

    J. Heyd, G. E. Scuseria, and M. Ernzerhof, Hybrid func- tionals based on a screened Coulomb potential, J. Chem. Phys. 118, 8207 (2003)

    work page 2003

  60. [60]

    Hybrid functionals based on a screened Coulomb poten- tial

    J. Heyd, G. E. Scuseria, and M. Ernzerhof, Erratum: “Hybrid functionals based on a screened Coulomb poten- tial” [J. Chem. Phys. 118, 8207 (2003)], J. Chem. Phys. 124, 219906 (2006)

    work page 2003

  61. [61]

    Lax, The Franck-Condon Principle and Its Applica- tion to Crystals, J

    M. Lax, The Franck-Condon Principle and Its Applica- tion to Crystals, J. Chem. Phys. 20, 1752 (1952)

    work page 1952

  62. [62]

    Kubo and Y

    R. Kubo and Y. Toyozawa, The Method of Generating Function Applied to Radiative and Non-Radiative Tran- sitions of a Trapped Electron in a Crystal, Prog. Theor. Phys. 12, 805 (1954)

    work page 1954

  63. [63]

    Silkinis, M

    R. Silkinis, M. Maciaszek, V. ˇZalandauskas, M. E. Ba- then, L. Vines, A. Alkauskas, and L. Razinkovas, Optical 12 lineshapes of the C center in silicon from ab initio calcu- lations: Interplay of localized modes and bulk phonons, Phys. Rev. B 111, 125136 (2025)

    work page 2025

  64. [64]

    Open Materials 2024 (OMat24) Inorganic Materials Dataset and Models

    L. Barroso-Luque, M. Shuaibi, X. Fu, B. M. Wood, M. Dzamba, M. Gao, A. Rizvi, C. L. Zitnick, and Z. W. Ulissi, Open Materials 2024 (OMat24) Inorganic Materi- als Dataset and Models (2024), arXiv:2410.12771 [cond- mat]

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  65. [65]

    J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized Gradient Approximation Made Simple, Phys. Rev. Lett. 77, 3865 (1996)

    work page 1996

  66. [66]

    V. I. Anisimov, J. Zaanen, and O. K. Andersen, Band theory and Mott insulators: Hubbard U instead of Stoner I, Phys. Rev. B 44, 943 (1991)

    work page 1991

  67. [67]

    Batatia and J

    I. Batatia and J. Hart, (private communication)

    work page

  68. [68]

    MACE - Machine Learning Force Fields Documenta- tion, https://mace-docs.readthedocs.io/en/latest/, Accessed: 2025-07-31

    work page 2025

  69. [70]

    M. E. Turiansky, lineshape tools, online (2025). Supplementary Information for Machine Learning Phonon Spectra for Fast and Accurate Optical Lineshapes of Defects Mark E. Turiansky, ∗ John L. Lyons, and Noam Bernstein US Naval Research Laboratory, 4555 Overlook Avenue SW, Washi ngton, DC 20375, USA (Dated: September 11, 2025) S1. COMPUT A TIONAL DET AILS ...

    work page 2025

  70. [71]

    Approximate Excited-State Potential Energy Surfaces for Defects in Solids

    is used. B. F oundation Model We first assess the performance of the baseline foundation model for pre dicting luminescence spectra. With the foundation model, we compute F0 and find a value of 1.52 eV/ ˚ A for C N in GaN, 1.00 eV/ ˚ A for the NV center in diamond, and 0.97 eV/ ˚ A for the carbon dimer in hBN (see table S2). Unsurprisingly, the foundation m...

    work page internal anchor Pith review Pith/arXiv arXiv 1996