Pith. sign in

REVIEW 3 major objections 6 minor 58 references

Real-Time Approach to the Dynamical Bethe-Salpeter Equation for Finite Systems

T0 review · 3 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that the frequency-dependent kernel of the dynamical Bethe-Salpeter equation can be built by real-time propagation and a time-domain product, avoiding frequency-grid integration and storage of the screened Coulomb…

desk verdict A real algorithmic step for dynamical BSE, but the reported shifts test a Shindo-approximated kernel, not the full-frequency BSE. read the letter →

arxiv 2608.08948 v1 pith:R72QPSWY submitted 2026-08-09 physics.chem-ph

Tucker Allen , Barry Y. Li , Daniel Neuhauser This is my paper · ORCID
classification physics.chem-ph
keywords dynamicalBethe-Salpeterequationreal-timelinearresponsetime-dependentHartreescreenedCoulombinteractionfrequency-dependentkernelTamm-Dancoffapproximationstochasticsamplingsilane
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that the frequency-dependent screened interaction entering the dynamical Bethe-Salpeter equation can be constructed by propagating the density matrix in real time and then evaluating the frequency convolution as a product in time. If correct, the dynamical BSE can be solved without constructing or storing the full frequency-dependent screened Coulomb operator, removing a major memory bottleneck and opening a path to stochastic sampling for large systems. The authors validate the formalism on silane: the static limit reproduces a standard dielectric-matrix-inversion implementation to below 1 meV, and the time-product evaluation agrees with the direct frequency convolution. The resulting dynamical correction redshifts the lowest singlet and triplet excitations, consistent with earlier dynamical-BSE studies.

What carries the argument

The load-bearing object is Eq. (17), the real-time expression for the polarization contribution to the dynamical kernel: it converts the convolution over frequency in Eq. (3) into multiplication in time by the phase factors $e^{-i(\varepsilon_b-\varepsilon_i+\Delta)t}$ and $e^{i(\varepsilon_a-\varepsilon_j+\Delta)t}$, followed by a Fourier transform. The retarded response $W^{\mathrm{pol},ij}_{ab}(t)$ is generated by time-dependent Hartree propagation: an impulsive perturbation from each occupied pair $kl$ produces an induced density matrix, which is contracted with the Coulomb tensor to give the polarization part of the screened interaction. A causal-to-time-ordered transformation puts this response on the full time axis. The kernel itself rests on the Shindo-type factorization of the four-point correlation function, Eq. (A15), which allows the two internal fermionic frequency integrals to be evaluated by contour integration so that the BSE closes.

What would settle it

Compute the dynamical excitation energies of a small molecule whose low-lying polarization mode lies near the optical excitation energy, using an unfactorized frequency-domain solution of the Bethe-Salpeter equation that retains full frequency dependence in W, and compare with the real-time time-product result. Agreement within the roughly 50 meV numerical broadening would support the Shindo factorization; a larger mismatch would show that Eq. (17) is not the full-frequency BSE kernel.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that Eq. (17) is an equivalent time-domain rendering of the frequency convolution defining the dynamical BSE kernel. The retarded screened interaction is obtained by propagating the time-dependent Hartree density matrix after impulsive perturbations built from occupied-orbital pair densities; the causal response is converted to a time-ordered interaction, multiplied by phase factors carrying the bare electron-hole transition energies, and Fourier transformed to give the full-frequency-dependent kernel without explicit frequency integration or storage of the full real-space screened Coulomb operator. The construction is validated on silane in two ways: static BSE from real-time response matches a standard dielectric-matrix-inversion implementation with mean absolute deviations below 1 meV, and the time-product evaluation of the dynamical interaction reproduces the direct frequency convolution. Dynamical screening shifts the lowest singlet excitation down by 0.12 eV and the triplet by 0.35 eV, a redshift in the same direction as earlier dynamical-BSE results.

Load-bearing premise

The load-bearing premise is the Shindo-type factorization of the four-point correlation function in Eq. (A15): if that factorization is not quantitatively accurate for molecular screening, the computed dynamical shifts are not the true full-frequency BSE shifts, and the paper gives no benchmark against a dynamical BSE solver that avoids it.

Editorial extensions

If this is right

  • For systems where polarization modes lie near optical excitation energies, the static approximation can misestimate singlet and triplet energies by roughly 0.1-0.3 eV; the real-time dynamical kernel corrects this at the cost of propagating TDH equations rather than storing a frequency grid.
  • The nonlinear BSE can be solved by diagonalizing the dynamical matrix on a sparse grid of trial frequencies and locating roots by secant interpolation; for silane the matrix varies smoothly, so few sampling frequencies suffice.
  • The same real-time response supplies both the static and the dynamical kernel, so a single code path covers both approximations and can be checked against independent static implementations.
  • The time-domain form is compatible with stochastic sampling, in which random occupied and unoccupied orbital combinations carry the orbital-energy phase factors, offering a route to dynamical BSE for systems with hundreds to thousands of valence electrons.
  • The time-ordered interaction can also drive a memory-based density-matrix evolution, outlined in Appendix B, potentially yielding spectra or steady states without a frequency-domain diagonalization.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural test of the Shindo factorization would be to compare the real-time kernel with an unfactorized dynamical BSE calculation on a small molecule whose low-lying polarization mode overlaps the optical excitation; the paper does not provide such a benchmark.
  • The real-time construction may be most beneficial where memory rather than floating-point operations is the bottleneck, since no screened-Coulomb operator is stored on a frequency grid; profiling memory against system size would identify the crossover.
  • If the stochastic extension succeeds, the dynamical BSE could become available precisely for the molecules where the static approximation is most questionable, such as weakly screened systems and triplet excitations, because the phase factors factorize onto occupied and unoccupied subspaces.
  • The Appendix B memory-based evolution suggests that the same time-dependent interaction could bypass the nonlinear root search entirely by propagating a density matrix; whether that evolution is numerically stable is left open.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper presents a real-time linear-response formulation of the dynamical Bethe-Salpeter equation for finite systems. The screened interaction entering the BSE kernel is constructed by propagating time-dependent Hartree equations under impulsive pair-density perturbations, and the frequency convolution that defines the dynamical kernel is evaluated as a product in the time domain followed by Fourier transformation. This avoids explicit frequency grids and storage of the full screened Coulomb operator. The method is validated on SiH4 in two ways: the static BSE from real-time TDH response agrees with a frequency-domain dielectric-matrix inversion to below 1 meV, and the time-product evaluation of the dynamical interaction reproduces direct numerical frequency convolution. The dynamical BSE redshifts the lowest singlet by 0.12 eV and triplet by 0.35 eV, in qualitative agreement with earlier work by Rohlfing and Louie. The paper also sketches a stochastic extension and a memory-based time-evolution formulation.

Significance. If the central claims hold, the method gives a practical route to dynamical BSE calculations that avoids explicit frequency integration and storage of the screened interaction, and its compatibility with stochastic sampling is a genuine algorithmic advantage. The algebraic transformation from the frequency convolution to a time-domain product is exact and internally consistent, and the reproduction of an independent static implementation is a useful validation. However, the physical significance of the reported dynamical shifts is conditional on the Shindo-type factorization adopted in Appendix A, which is not benchmarked against a reference that avoids that approximation. The validation is limited to a single small molecule with no convergence tests, so the quantitative redshifts should be treated as preliminary.

major comments (3)
  1. [II.A and Appendix A, Eq. (A15)] The central physical claim—that the method goes beyond the static screening approximation with full-frequency dependence in the screened interaction—rests on the Shindo-type factorization in Eq. (A15), but this factorization is not validated. The agreement between the time-product and frequency-convolution routes reported in Section III only confirms that Eq. (17) correctly implements Eq. (3); it does not test the approximation in Eq. (3) itself. Since the dynamical shifts in Table I are presented as physical predictions, a comparison against a dynamical BSE solver that avoids the Shindo factorization, using the same LDA+scissor input, is needed to support the attribution of these shifts to full-frequency screening.
  2. [III, Table I] The comparison with Ref. [13] cannot serve as a quantitative validation because the reference uses GW quasiparticle states and a plasmon-pole model, as the authors acknowledge in the text. The statement that dynamical screening redshifts the excitations is supported, but the magnitudes (0.12 eV singlet, 0.35 eV triplet) are not benchmarked. Please either provide a same-input reference calculation or explicitly label these magnitudes as model-dependent within the Shindo approximation.
  3. [III, Eqs. (13)-(19)] The reported numerical results have no convergence analysis. The choices N_c=24, dt=0.05 au, T_max=20 fs, gamma~100 meV, and eta~50 meV are stated, but their effect on the dynamical shifts is not quantified. Because the shifts are of order 0.1-0.35 eV, the numerical error from Gaussian damping, finite propagation time, and the frequency grid used for secant interpolation could be a substantial fraction of the effect. Please report convergence with respect to T_max (or gamma), eta, basis size, and number of frequency samples.
minor comments (6)
  1. [Eqs. (14) and (17)] The retarded response in Eq. (14) uses Gaussian damping e^{-gamma^2 t^2/2}, while the time-domain convolution in Eq. (17) introduces e^{-eta|t|}. The relationship between gamma and eta, and how both appear in the final time-ordered W(t) entering Eq. (17), should be stated explicitly.
  2. [Eq. (20)] Eq. (20) writes a relation for the full W(t), whereas Eq. (17) uses the polarization part W_pol(t), with the bare Coulomb term added separately in Eq. (18). Please clarify whether Eq. (20) includes the instantaneous delta-function term and how the delta function is handled in the stochastic sampling context.
  3. [Section III] The secant root-finding procedure is described in words but the frequency grid spacing and the number of sampled frequencies are not given. This information is needed to assess the numerical stability of the reported roots.
  4. [References] References [52] and [53] are identical (Romaniello, Guyot, and Reining, J. Chem. Phys. 131, 154111 (2009)); the duplicate should be removed or replaced.
  5. [Table I caption] The column layout of Table I is confusing: the columns labeled 'Static', 'Time Product', and 'Frequency Convolution' under the present work, followed by 'Static' and 'Dynamic' under Ref. [13], are not clearly separated. The caption should explicitly indicate which columns refer to which implementation.
  6. [Eq. (21)] The stochastic notation, particularly the distinction between overlined and double-overlined random states, is introduced abruptly. A brief definition of the sampling distribution and the normalization of the random states would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Eq. (17) is an exact time-domain reformulation of Eq. (3), and no fitted parameter is renamed as a prediction.

full rationale

The derivation chain is self-contained. Eq. (17) is an exact Fourier/convolution identity relative to Eq. (3): the paper explicitly states that the orbital-energy phase factors originate from the energy denominators in Eq. (3), so the time-product evaluation is mathematically equivalent to the frequency convolution. This is an algebraic reformulation, not an output-dependent fit or a self-referential derivation. The paper also explicitly labels Eq. (3) as a Shindo-like approximation and derives it in Appendix A through the stated factorization (A15), rather than importing the kernel as a fitted or externally imposed result. The screened interaction W(omega) is obtained from time-dependent Hartree/RPA response, an independent model, and no parameter is fitted to the reported excitation energies. The only external fit, the scissor shift Delta, is fixed to the GW gap of Ref. [13], not to the BSE excitation energies that are the outputs. The validation that the time product reproduces the frequency convolution is tautological in the sense that both evaluate the same kernel, but it is a numerical consistency check, not a circular claim. Self-citations appear only in describing the causal-to-time-ordered transformation and in speculative future stochastic extensions; they are not load-bearing for the central derivation. The absence of a non-Shindo dynamical BSE benchmark is a validation gap and a correctness risk, but it is not circularity under the criteria used here.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The central claim rests on standard DFT/TDH inputs, three documented physical approximations (RPA screening, TDA, Shindo factorization), and a fitted scissor shift. The only fitted physical parameter is the scissor; the numerical damping and broadening are convergence parameters. No new entities are introduced.

free parameters (3)
  • Scissor correction Delta = 7.18 eV
    Applied to all bare LDA transition energies to match the 13.79 eV GW gap reported in Ref. [13]; the BSE matrix Eq. (1) and the phase factors in Eq. (17) depend on it, so all reported excitation energies shift with this choice.
  • Gaussian damping gamma = 3/T_max ~ 100 meV
    Introduced in Eq. (14) to damp the finite-time retarded response before Fourier transform; chosen by hand and affects the frequency resolution of W(omega) and hence the dynamical kernel.
  • Numerical broadening eta = ~50 meV
    Broadening in the frequency denominators of Eqs. (3) and (17); chosen by hand; affects root finding and can mask or smooth near-degenerate poles.
assumptions (6)
  • domain assumption The screened interaction W(omega) is obtained at the time-dependent Hartree/RPA level via Eq. (11).
    The time-dependent Hartree Hamiltonian with only the Hartree potential and no exchange-correlation kernel is used for the polarization; this is a standard but approximate screening model.
  • domain assumption The Tamm-Dancoff approximation (TDA), neglecting the resonant-antiresonant coupling block, is used for the BSE.
    The paper states that both the software implementation and results employ the TDA; the paper notes the off-diagonal block is excluded because of numerical divergences.
  • domain assumption The Shindo-like factorization, Eq. (A15), is accurate enough for the dynamical BSE.
    The factorization is required to close the BSE when W depends on frequency; if it fails, the kernel in Eq. (A20) would not represent the full dynamical BSE.
  • domain assumption LDA-DFT orbitals with a rigid scissor sufficiently approximate GW quasiparticle states for the BSE.
    Input states are taken from LDA with a scissor Delta fitted to the Ref. [13] GW gap, rather than from an actual GW calculation.
  • domain assumption The finite-grid plane-wave LDA ground state is converged sufficiently for BSE excitations.
    The 16^3 grid with 0.6 Bohr spacing is used without a convergence study; static agreement with the dielectric-matrix implementation is evidence but not a full convergence test.
  • standard math Fourier and convolution identities, plus contour integration, used in Appendix A are valid.
    The time-product Eq. (17) and the internal-frequency integrals in Eq. (A19) rely on standard complex analysis and Fourier transforms.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Real-Time Approach to the Dynamical Bethe-Salpeter Equation for Finite Systems." pith.science (2026). https://pith.science/paper/R72QPSWY

@misc{pith2026260808948,
  author       = {Pith},
  title        = {Pith review of: Real-Time Approach to the Dynamical Bethe-Salpeter Equation for Finite Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R72QPSWY}},
  note         = {Machine review of arXiv:2608.08948}
}
abstract

We present a real-time linear-response approach to solving the dynamical Bethe-Salpeter equation (BSE). The polarization part of the screened interaction is obtained from time-dependent Hartree propagation of the one-particle density matrix within an orbital basis-set representation. The frequency convolution defining the dynamical kernel is evaluated as a product in time, avoiding numerical integration and storage of the full screened Coulomb operator. The resulting nonlinear eigenvalue problem is then solved directly, going beyond the static screening approximation with full-frequency dependence in the screened interaction. While the deterministic scaling remains $\mathcal{O}(N^6)$, the time propagation formulation is readily compatible with grid-based stochastic sampling methods, which will open the possibility for dynamical BSE calculations of very large systems.

Figures

Figures reproduced from arXiv: 2608.08948 by the authors.

Figure 1
Figure 1. FIG. 1. Static BSE excitation energies of SiH [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 48 canonical work pages

  1. [13]

    Rohlfing and S

    M. Rohlfing and S. G. Louie, Phys. Rev. B62, 4927 (2000)

    2000

  2. [1]

    C. A. Ullrich, APL Computational Physics1, 020901 (2025)

    2025

  3. [2]

    N. T. Maitra, F. Zhang, R. J. Cave, and K. Burke, The Journal of Chemical Physics120, 5932 (2004)

    work page 2004

  4. [3]

    B. G. Levine, C. Ko, J. Quenneville, and T. J. Mart ´Inez, Molecular Physics104, 1039 (2006)

    work page 2006

  5. [4]

    Romaniello, D

    P. Romaniello, D. Sangalli, J. A. Berger, F. Sottile, L. G. Molinari, L. Reining, and G. Onida, The Journal of Chemical Physics130(2009)

    work page 2009

  6. [5]

    Sangalli, P

    D. Sangalli, P. Romaniello, G. Onida, and A. Marini, The Journal of Chemical Physics134 (2011)

    work page 2011

  7. [6]

    Zhang, S

    D. Zhang, S. N. Steinmann, and W. Yang, The Journal of Chemical Physics139, 154109 (2013)

    work page 2013

  8. [7]

    Rebolini and J

    E. Rebolini and J. Toulouse, The Journal of Chemical Physics144, 094107 (2016)

    work page 2016

Show all 58 references
  1. [8]

    Lacombe and N

    L. Lacombe and N. T. Maitra, npj Computational Materials9(2023)

    2023

  2. [9]

    Hedin, Phys

    L. Hedin, Phys. Rev.139, A796 (1965)

    1965

  3. [10]

    M. S. Hybertsen and S. G. Louie, Physical Review Letters55, 1418 (1985)

    1985

  4. [11]

    Strinati, La Rivista del Nuovo Cimento11, 1 (1988)

    G. Strinati, La Rivista del Nuovo Cimento11, 1 (1988)

    1988

  5. [12]

    Blase, I

    X. Blase, I. Duchemin, D. Jacquemin, and P.-F. Loos, The Journal of Physical Chemistry Letters11, 7371 (2020)

    2020

  6. [14]

    Onida, L

    G. Onida, L. Reining, and A. Rubio, Rev. Mod. Phys.74, 601 (2002)

    2002

  7. [15]

    Jacquemin, I

    D. Jacquemin, I. Duchemin, and X. Blase, The Journal of Physical Chemistry Letters8, 1524 (2017)

    2017

  8. [16]

    Li and V

    J. Li and V. Olevano, Journal of Photochemistry and Photobiology B: Biology232, 112475 (2022)

    2022

  9. [17]

    Rocca, Y

    D. Rocca, Y. Ping, R. Gebauer, and G. Galli, Physical Review B85, 045116 (2012)

    2012

  10. [18]

    Jacquemin, I

    D. Jacquemin, I. Duchemin, and X. Blase, Journal of Chemical Theory and Computation11, 3290 (2015)

    2015

  11. [19]

    X. Gui, C. Holzer, and W. Klopper, Journal of Chemical Theory and Computation14, 2127 19 (2018)

    2018

  12. [20]

    F¨ orster and L

    A. F¨ orster and L. Visscher, Journal of Chemical Theory and Computation18, 6779 (2022)

    2022

  13. [21]

    Marini and R

    A. Marini and R. Del Sole, Physical Review Letters91, 176402 (2003)

    2003

  14. [22]

    Bechstedt,Many-Body Approach to Electronic Excitations: Concepts and Applications (Springer Berlin Heidelberg, 2015)

    F. Bechstedt,Many-Body Approach to Electronic Excitations: Concepts and Applications (Springer Berlin Heidelberg, 2015)

    2015

  15. [23]

    Y. Ma, M. Rohlfing, and C. Molteni, Physical Review B80, 241405 (2009)

    2009

  16. [24]

    Loos and X

    P.-F. Loos and X. Blase, The Journal of Chemical Physics153, 114120 (2020)

    2020

  17. [25]

    Zhang, J

    X. Zhang, J. A. Leveillee, and A. Schleife, Physical Review B107, 235205 (2023)

    2023

  18. [26]

    M. Wen, G. Harsha, and D. Zgid, Journal of Chemical Theory and Computation (2026)

    2026

  19. [27]

    All-electron dynamical Bethe-Salpeter Equa- tion for extended systems with atom-centered orbital basis set,

    R. Zhou, S. Liu, J. Xu, Y. Yao, and Y. Kanai, “All-electron dynamical Bethe-Salpeter Equa- tion for extended systems with atom-centered orbital basis set,” (2026), arXiv:2606.08350 [physics.chem-ph]

    2026 arXiv

  20. [28]

    S. J. Bintrim and T. C. Berkelbach, The Journal of Chemical Physics156(2022)

    2022

  21. [29]

    Kurzweil and R

    Y. Kurzweil and R. Baer, Physical Review B77(2008)

    2008

  22. [30]

    K. J. Daas, S. Crisostomo, and K. Burke, Phys. Rev. Lett.137, 028002 (2026)

    2026

  23. [31]

    Casanova and A

    D. Casanova and A. I. Krylov, Physical Chemistry Chemical Physics22, 4326 (2020)

    2020

  24. [32]

    W. Park, J. Shen, S. Lee, P. Piecuch, M. Filatov, and C. H. Choi, The Journal of Physical Chemistry Letters12, 9720 (2021)

    2021

  25. [33]

    D. B. Dar and N. T. Maitra, The Journal of Physical Chemistry Letters16, 703 (2025)

    2025

  26. [34]

    Authier and P.-F

    J. Authier and P.-F. Loos, The Journal of Chemical Physics153, 184105 (2020)

    2020

  27. [35]

    Loos and P

    P.-F. Loos and P. Romaniello, The Journal of Chemical Physics156, 164101 (2022)

    2022

  28. [36]

    H. F. Wilson, F. m. c. Gygi, and G. Galli, Phys. Rev. B78, 113303 (2008)

    2008

  29. [37]

    X. Ren, P. Rinke, V. Blum, J. Wieferink, A. Tkatchenko, A. Sanfilippo, K. Reuter, and M. Scheffler, New Journal of Physics14, 053020 (2012)

    2012

  30. [38]

    Neuhauser, Y

    D. Neuhauser, Y. Gao, C. Arntsen, C. Karshenas, E. Rabani, and R. Baer, Phys. Rev. Lett. 113, 076402 (2014)

    2014

  31. [39]

    Duchemin and X

    I. Duchemin and X. Blase, Journal of Chemical Theory and Computation17, 2383 (2021)

    2021

  32. [40]

    C.-N. Yeh, S. Iskakov, D. Zgid, and E. Gull, Physical Review B106, 235104 (2022)

    2022

  33. [41]

    Rabani, R

    E. Rabani, R. Baer, and D. Neuhauser, Physical Review B91, 235302 (2015)

    2015

  34. [42]

    N. C. Bradbury, T. Allen, M. Nguyen, K. Z. Ibrahim, and D. Neuhauser, The Journal of 20 Chemical Physics158, 154104 (2023)

    2023

  35. [43]

    Hillenbrand, J

    C. Hillenbrand, J. Li, and T. Zhu, The Journal of Chemical Physics162, 174117 (2025)

    2025

  36. [44]

    Vlˇ cek, W

    V. Vlˇ cek, W. Li, R. Baer, E. Rabani, and D. Neuhauser, Physical Review B98, 075107 (2018)

    2018

  37. [45]

    N. C. Bradbury, M. Nguyen, J. R. Caram, and D. Neuhauser, The Journal of Chemical Physics 157, 031104 (2022)

    2022

  38. [46]

    N. C. Bradbury, B. Y. Li, T. Allen, J. R. Caram, and D. Neuhauser, The Journal of Chemical Physics161, 141101 (2024)

    2024

  39. [47]

    Allen, B

    T. Allen, B. Y. Li, N. C. Bradbury, and D. Neuhauser, Journal of Chemical Theory and Computation , acs.jctc.6c00440 (2026)

    2026

  40. [48]

    J. P. Perdew and Y. Wang, Phys. Rev. B45, 13244 (1992)

    1992

  41. [49]

    A. L. Fetter and J. D. Walecka,Quantum Theory of Many-Particle Systems, Dover Books on Physics (Dover Publications, Newburyport, 2012)

    2012

  42. [50]

    Troullier and J

    N. Troullier and J. L. Martins, Phys. Rev. B43, 1993 (1991)

    1991

  43. [51]

    G. J. Martyna and M. E. Tuckerman, The Journal of Chemical Physics110, 2810 (1999)

    1999

  44. [53]

    Romaniello, S

    P. Romaniello, S. Guyot, and L. Reining, The Journal of Chemical Physics131, 154111 (2009)

    2009

  45. [54]

    Neuhauser, E

    D. Neuhauser, E. Rabani, and R. Baer, Journal of Chemical Theory and Computation9, 24 (2012)

    2012

  46. [55]

    R. Baer, D. Neuhauser, and E. Rabani, Phys. Rev. Lett.111, 106402 (2013)

    2013

  47. [56]

    N. C. Bradbury, T. Allen, M. Nguyen, and D. Neuhauser, Journal of Chemical Theory and Computation19, 9239 (2023)

    2023

  48. [57]

    Y. Gao, D. Neuhauser, R. Baer, and E. Rabani, The Journal of Chemical Physics142, 034106 (2015)

    2015

  49. [58]

    Allen, M

    T. Allen, M. Nguyen, and D. Neuhauser, The Journal of Chemical Physics161, 114116 (2024)

    2024

  50. [59]

    P. S. Thomas, M. Nguyen, D. Bazile, T. Allen, B. Y. Li, W. Li, M. Del Ben, J. Deslippe, and D. Neuhauser, Journal of Chemical Theory and Computation22, 3960 (2026). 21

    2026

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.