Pith. sign in

REVIEW 2 major objections 7 minor 93 references

Fourier interpolation of screened Coulomb W cuts BSE k-mesh cost

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · glm-5.2

2026-07-08 22:21 UTC pith:UTY6MXTR

load-bearing objection Solid BSE implementation with dual-k-mesh interpolation; benchmarks are thorough, novelty is incremental, and the W(R) truncation assumption needs a direct convergence test but is probably fine for the systems studied. the 2 major comments →

arxiv 2607.05853 v1 pith:UTY6MXTR submitted 2026-07-07 cond-mat.mtrl-sci physics.chem-phphysics.comp-ph

Efficient Bethe-Salpeter Equation Calculations Based on Numerical Atomic Orbitals and Norm-Conserving Pseudopotentials: Dual-{boldsymbol k}-Mesh Strategy

Ziqing Guan , Yu Cao , Min-Ye Zhang , Peize Lin , Ruiyi Zhou , Xinguo Ren This is my paper
classification cond-mat.mtrl-sci physics.chem-phphysics.comp-ph
keywords boldsymbolmeshcalculationsdual-efficientatomicbasisequation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper presents an implementation of the Bethe-Salpeter equation (BSE) for optical excitation spectra using numerical atomic orbitals (NAOs) and norm-conserving pseudopotentials, integrated within the ABACUS+LibRPA software framework. The central innovation is a dual-k-mesh strategy: the computationally expensive screened Coulomb interaction W is computed on a coarse k-mesh (where GW quasiparticle energies are calculated), then Fourier-interpolated to an arbitrarily dense k-mesh for assembling and diagonalizing the BSE Hamiltonian. This interpolation is made possible by the localized resolution-of-identity (LRI) technique, which casts W into a real-space, unit-cell-indexed form W_{μν}(R) that is inherently short-ranged and can be truncated on a modest Born-von Kármán supercell. The paper validates this approach through systematic convergence tests (basis set, auxiliary basis, k-point sampling) and benchmark calculations on 28 molecules (Thiel's set) and two periodic solids (Si, MgO), demonstrating agreement with established codes at substantially reduced computational cost.

Core claim

The key finding is that the real-space screened Coulomb interaction W_{μν}(R), when expressed in the NAO+LRI framework, decays rapidly enough that it can be computed on a coarse k-mesh and Fourier-interpolated to a dense k-mesh without significant loss of accuracy in the resulting BSE absorption spectra. This separation of the k-mesh for W evaluation (coarse, affordable) from the k-mesh for BSE Hamiltonian construction (dense, converged) addresses the primary computational bottleneck in GW+BSE calculations for periodic systems. The paper demonstrates this with benchmarks showing mean absolute deviations of 32–44 meV against FHI-aims for molecular excitations, and consistent absorption peak结构

What carries the argument

The dual-k-mesh workflow: (1) DFT on both coarse and dense k-meshes; (2) GW calculation on coarse k-mesh producing quasiparticle energies and W_{μν}(R) in a small real-space supercell; (3) Fourier transform of W_{μν}(R) to dense k-mesh; (4) BSE Hamiltonian assembly and diagonalization on dense k-mesh. The LRI technique expands NAO products in atom-centered auxiliary basis functions, enabling the real-space representation. The velocity-length gauge equivalence serves as an internal diagnostic for basis-set incompleteness. Offset averaging over symmetry-reduced k-mesh shifts mitigates artifacts from high-symmetry points.

Load-bearing premise

The dual-k-mesh strategy depends on the screened Coulomb interaction W_{μν}(R) decaying fast enough in real space to be truncated on a small Born-von Kármán supercell. The paper states this truncation is safe but does not provide a quantitative analysis of truncation error versus system type, and for materials with large dielectric constants or weakly bound excitons, the range of W may be longer than assumed.

What would settle it

A material system where W_{μν}(R) decays so slowly that truncation on the Born-von Kármán supercell introduces errors comparable to or larger than the spectral features of interest, causing the interpolated BSE spectrum to diverge from a directly computed dense-k-mesh reference.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • The dual-k-mesh strategy makes converged BSE spectra for periodic systems accessible at a fraction of the uniform dense k-mesh cost, potentially enabling GW+BSE calculations for larger and more complex materials than previously feasible.
  • The real-space locality of W_{μν}(R) in the NAO+LRI framework could extend to other many-body quantities beyond BSE, such as finite-momentum exciton dispersion or nonlinear optical response, where dense k-sampling is similarly a bottleneck.
  • The systematic growth of code-to-code discrepancies with conjugation length in polyenes (up to ~200 meV for octatetraene triplet excitations) suggests that delocalized excitons stress the limits of localized basis representations and may motivate hybrid or adaptive basis strategies for extended molecular systems.
  • The offset-averaging technique for k-mesh artifacts provides a practical, parameter-free alternative to searching for optimal k-mesh shifts, generalizable to any BSE implementation suffering from high-symmetry-point artifacts.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If W_{μν}(R) is sufficiently short-ranged for Fourier interpolation, the same dual-mesh strategy could apply to dynamical (frequency-dependent) screening beyond the static approximation used here, potentially improving exciton binding energy predictions for wide-gap materials like MgO and SnO2 where the paper reports large residuals (284 meV and 102 meV vs experiment).
  • The truncation of W at large |R| may break down for materials with large dielectric constants or weakly bound, spatially extended excitons (e.g., small-gap semiconductors near a dielectric catastrophe), where the screening cloud extends beyond the Born-von Kármán supercell. A quantitative truncation-error analysis as a function of dielectric constant would clarify the boundary of applicability.
  • The finding that BSE spectra are less sensitive to auxiliary basis completeness than GW self-energies suggests that a tiered basis strategy—using more complete auxiliary bases for GW and compressed bases for BSE—could yield further computational savings without spectral degradation.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 7 minor

Summary. This manuscript presents a BSE implementation within the ABACUS+LibRPA framework using numerical atomic orbitals (NAOs) and norm-conserving pseudopotentials. The key methodological contribution is a dual-k-mesh strategy: the screened Coulomb interaction W is computed on a coarse k-mesh (inherited from the preceding G0W0 calculation), inverse-Fourier transformed to real space W_{μν}(R), and then Fourier-interpolated onto a dense k-mesh for BSE Hamiltonian assembly and diagonalization. The spatial locality of W(R) in the NAO+LRI representation makes this interpolation efficient. The paper includes convergence tests (NAO basis, auxiliary basis, k-point sampling with offset averaging), benchmarks against FHI-aims for Thiel's molecular set (28 molecules, MAD 32–44 meV), and cross-code comparisons for Si and MgO absorption spectra, plus exciton binding energies for six materials.

Significance. The dual-k-mesh strategy addresses a genuine computational bottleneck in BSE calculations for periodic systems. The implementation is cleanly specified, with careful treatment of Coulomb singularity (head/wing terms) and the truncated Coulomb operator. The use of velocity–length gauge equivalence as an internal diagnostic for basis-set incompleteness (Sec. IV.A.1) is a thoughtful contribution. The Thiel's set benchmark against FHI-aims, with identical Kohn–Sham starting points to isolate BSE-kernel differences, is well designed and yields quantitatively useful error bars. The cross-code benchmarks for Si and MgO span NAO, plane-wave, and LAPW frameworks, providing meaningful validation across fundamentally different numerical representations.

major comments (2)
  1. Sec. III.B, step 1: The statement that 'contributions from large |R| can be safely truncated' is the load-bearing assumption of the dual-k-mesh strategy, yet no independent convergence test of the real-space truncation radius for W(R) is provided. The cross-code benchmarks (Figs. 7–8, Table I) provide indirect evidence—if truncation were grossly wrong, spectra would not match—but these comparisons conflate multiple error sources (basis set, frequency grid, q→0 treatment). A brief test showing W(R) convergence with respect to the Born–von Kármán supercell size for at least one system (e.g., Si or MgO) would substantially strengthen the central claim. Alternatively, the authors should at minimum discuss the class of materials where slow decay of W(R) (large dielectric constant, small gap, weakly bound excitons) could compromise the interpolation, and note whether the coarse k-mesh used (7×
  2. Table II: Several exciton binding energies are not converged with respect to k-mesh (e.g., GaN: 74→13 meV from 11×11×7 to 21×21×14; AlN: 85 meV at 20×20×12 vs. 147 meV converged value from Ref. [9]). The MgO value (284 meV) deviates substantially from experiment (80–145 meV). While the authors are transparent about this, the claim in the abstract that benchmarks 'collectively validate the accuracy' could be read as overstating the exciton binding energy results. The authors should clarify which values are converged and which are not, perhaps by marking unconverged entries in Table II.
minor comments (7)
  1. Sec. II.C, Eq. (20): The notation k_1 and k_2 for the two k-points in the BSE Hamiltonian is introduced without explicitly stating the momentum conservation condition k_2 = k_1 + q (or similar). A brief clarifying sentence would help readers unfamiliar with the convention.
  2. Fig. 2: The workflow diagram is informative but dense. The distinction between 'Rmesh' and 'qmesh' could be made more explicit in the caption.
  3. Sec. IV.A.3: The offset-averaging weights (1/27, 8/27, 6/27, 12/27) are stated without derivation. A one-line reference to the symmetry-reduction procedure would suffice.
  4. Sec. IV.B.2, Fig. 8: The systematically lower peak intensities for MgO are attributed to 'differences in how high-lying unoccupied states and the associated transition matrix elements are represented.' This explanation is plausible but unverified. A brief comment on whether increasing the number of virtual orbitals was tested would be helpful.
  5. Table I caption: '140 excitations per category' — it would be useful to note that this is 28 molecules × 5 excitations.
  6. References [37–39] are cited as '2026' and appear to be very recent or in-press; please verify final publication details.
  7. Sec. IV.B.1: The growth of discrepancy with conjugation length in polyenes is an interesting observation. The suggested explanation (delocalized exciton sampling larger kernel volume) is reasonable but speculative; a brief qualifier would be appropriate.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive suggestions. Both major comments are well-taken. We will (1) add a dedicated W(R) real-space truncation convergence test for Si and MgO, and add a discussion of material classes where slow decay of W(R) could compromise the interpolation; (2) mark unconverged entries in Table II and soften the abstract language to avoid overstating the exciton binding energy results.

read point-by-point responses

  1. Referee: Sec. III.B, step 1: The statement that 'contributions from large |R| can be safely truncated' is the load-bearing assumption of the dual-k-mesh strategy, yet no independent convergence test of the real-space truncation radius for W(R) is provided. [...] A brief test showing W(R) convergence with respect to the Born–von Kármán supercell size for at least one system (e.g., Si or MgO) would substantially strengthen the central claim. Alternatively, the authors should at minimum discuss the class of materials where slow decay of W(R) could compromise the interpolation, and note whether the coarse k-mesh used (7×...)

    Authors: The referee is correct that the real-space truncation of W(R) is the central assumption of the dual-k-mesh strategy and that the manuscript lacks a dedicated convergence test isolating this parameter. We will address this in two ways. First, we will add a convergence test showing the BSE absorption spectrum of Si (and, if space permits, MgO) computed with varying Born–von Kármán supercell sizes for W(R), while keeping all other parameters fixed. In our preliminary checks, the Si spectrum is already well converged when the supercell extent matches the 7×7×7 coarse k-mesh used for the G0W0 calculation, consistent with the fact that the cross-code benchmarks in Figs. 7–8 show no artifacts attributable to truncation. Second, we will add an explicit discussion of the class of materials where the spatial decay of W(R) could be slow—namely, systems with large static dielectric constants, small band gaps, or weakly bound (large-radius) excitons, such as narrow-gap semiconductors and highly polarizable metals. In such cases, the Born–von Kármán supercell inherited from the coarse k-mesh may be insufficient, and the user should verify convergence by increasing the supercell size. We note that for the systems studied in this work (Si, MgO, GaN, AlN, CdS, SnO2), the coarse k-mesh of 7×7×7 (or comparable) provides a supercell large enough that W(R) is effectively converged, as evidenced by the agreement with reference codes. We will state this explicitly in the revised manuscript. revision: yes

  2. Referee: Table II: Several exciton binding energies are not converged with respect to k-mesh (e.g., GaN: 74→13 meV from 11×11×7 to 21×21×14; AlN: 85 meV at 20×20×12 vs. 147 meV converged value from Ref. [9]). The MgO value (284 meV) deviates substantially from experiment (80–145 meV). While the authors are transparent about this, the claim in the abstract that benchmarks 'collectively validate the accuracy' could be read as overstating the exciton binding energy results. The authors should clarify which values are converged and which are not, perhaps by marking unconverged entries in Table II.

    Authors: We agree with this assessment. Several entries in Table II are not fully converged with respect to k-mesh sampling, and the abstract language could be misread as claiming uniform validation across all benchmark types. We will make two changes. First, we will add a column or footnote to Table II explicitly marking which exciton binding energies are converged (Si and CdS are reasonably converged at 21×21×21) and which are not (GaN, AlN, MgO, and SnO2 still show significant k-mesh dependence). Second, we will revise the abstract to distinguish more carefully between the absorption spectrum benchmarks (where cross-code agreement is strong) and the exciton binding energies (where convergence is incomplete for several materials). Specifically, we will change 'collectively validate the accuracy of the present implementation' to language that acknowledges that the absorption spectra and molecular benchmarks validate the implementation, while the exciton binding energies are presented as a systematic study of k-mesh convergence behavior rather than as fully converged benchmark values. We believe this accurately reflects the content of the paper without overstating the results. revision: yes

Circularity Check

0 steps flagged

No significant circularity found; the derivation is standard BSE theory re-implemented and validated against external codes.

full rationale

The paper's central claim — that the dual-k-mesh strategy yields reliable BSE spectra — is validated against external benchmarks (FHI-aims, BerkeleyGW, VASP, Exciting, experimental data), not against the authors' own prior results. The theoretical formulation (Eqs. 1–25) is standard many-body perturbation theory following Strinati, Onida et al., and Rohlfing & Louie. The paper explicitly states it 'essentially follows those presented in Ref. [29]' (Zhou et al., where Ren is a co-author), but this is a re-implementation in a different code framework (ABACUS+LibRPA vs. FHI-aims), not a claim whose truth depends on Ref. [29] being correct. The dual-k-mesh concept itself is attributed to prior external work (Rohlfing & Louie [5], Kammerlander [26], Gillet [27], Alliati [28]). Self-citations (Refs. [35–39, 43–48]) provide computational infrastructure (GW, LRI, RPA implementations) but are not invoked to prove the BSE results. No parameter is fitted to data and then presented as a prediction. No uniqueness theorem is invoked. The W(R) truncation assumption is a physical property of screening, not a definition that forces the conclusion. The residual concern (truncation error not independently isolated) is a correctness/generality risk, not a circularity issue. Score 1 reflects minor self-citation for infrastructure that is not load-bearing for the central claim.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 0 invented entities

No new physical entities, particles, forces, or dimensions are introduced. The paper implements established theory (GW+BSE) in an existing software framework. All axioms are standard domain assumptions common to BSE implementations. The free parameters are computational settings (broadening, orbital truncation, basis set, PCA threshold) rather than physically fitted constants. The real-space truncation cutoff for W(R) is the most concerning parameter because its value is not explicitly stated and its convergence is not independently verified.

free parameters (5)
  • Lorentzian broadening η = 0.1-0.3 eV (system-dependent)
    Not a fitted parameter in the fitting sense, but a manually chosen broadening parameter that affects spectral appearance. Values vary by system: 0.3 eV for benzene, 0.2 eV for GaN, 0.1 eV for Si, 0.15 eV for Si benchmark, 0.3 eV for MgO.
  • Number of occupied/virtual orbitals = 4-8 occupied, 4-8 virtual (system-dependent)
    Manually chosen truncation of the electron-hole pair space. Not converged to completeness for all systems; e.g., 4 occ/4 virt for Si and MgO, 6 occ/8 virt for GaN.
  • PCA threshold for auxiliary basis compression = 10^{-3} to 10^{-5}
    Controls the number of auxiliary basis functions retained. Shown to be converged at 10^{-3} for BSE (Fig. 4), but the threshold itself is a user-chosen parameter.
  • Real-space truncation cutoff for W(R) = Not explicitly stated
    The paper states W(R) is truncated on a 'small Born-von Kármán supercell' but does not specify the cutoff radius or verify convergence with respect to it independently.
  • NAO basis set (DZP/TZDP/QZTP) = TZDP adopted for benchmarks
    Basis set choice is a user parameter. TZDP chosen based on velocity-length convergence diagnostic (Fig. 3).
axioms (5)
  • domain assumption Static approximation: W(ω) is evaluated at ω=0, neglecting frequency dependence of the screened interaction in the BSE kernel.
    Stated in Sec. II.B: 'we adopt the static approximation, i.e., we neglect the frequency dependence of W(ω) and evaluate it at ω=0.' This is standard in most BSE implementations but is a known source of error for exciton binding energies.
  • domain assumption G0W0 quasiparticle energies provide an adequate starting point for BSE.
    The BSE Hamiltonian (Eq. 20) uses G0W0 quasiparticle energy differences. The paper acknowledges systematic underestimation of band gaps (e.g., Si: 1.07 vs 1.23 eV; MgO: 7.18 vs 7.98 eV), causing redshift in spectra. Self-consistent GW is noted as future work.
  • domain assumption The Tamm-Dancoff approximation (TDA) is valid for the systems studied.
    Both TDA and full BSE are implemented, but for Si the spectra are 'almost indistinguishable' (Sec. IV.B.2). The validity of TDA is system-dependent and assumed for the benchmark systems.
  • domain assumption The LRI expansion (Eq. 26) faithfully represents products of NAOs in the auxiliary basis.
    The accuracy of the LRI technique is assumed and verified empirically (Fig. 4 for GaN), but no formal error bound is provided.
  • domain assumption The real-space screened interaction W_{μν}(R) is short-ranged enough for truncation on a small supercell.
    Invoked in Sec. III.B: 'This is effective because W(R) decays rapidly in real space; contributions from large |R| can be safely truncated.' No quantitative verification of truncation error is provided.

pith-pipeline@v1.1.0-glm · 33010 in / 4062 out tokens · 538505 ms · 2026-07-08T22:21:38.029064+00:00 · methodology

0 comments
read the original abstract

We present an efficient implementation of the Bethe--Salpeter equation (BSE) based on numerical atomic orbitals (NAOs) and norm-conserving pseudopotentials within the ABACUS+LibRPA framework. By exploiting the localized resolution-of-identity (LRI) technique, the screened Coulomb interaction is cast into a real-space, unit-cell-indexed form $W_{\mu\nu}(\boldsymbol R)$ that is inherently short-ranged and well localized. This spatial locality enables an efficient Fourier interpolation of the BSE kernel from the coarse $\boldsymbol k$-mesh used in the preceding $GW$ calculation to an arbitrarily dense $\boldsymbol k$-mesh on which the BSE Hamiltonian is assembled and diagonalized, thereby giving rise naturally to a dual-$\boldsymbol k$-mesh workflow. Building on this scheme, we systematically examine the convergence of the absorption spectra with respect to the NAO basis set, the auxiliary basis set, and the $\boldsymbol k$-point sampling. Benchmark calculations for both molecular and periodic systems collectively validate the accuracy of the present implementation and establish the dual-$\boldsymbol k$-mesh strategy as a practical and reliable approach for $GW$+BSE calculations.

Figures

Figures reproduced from arXiv: 2607.05853 by Min-Ye Zhang, Peize Lin, Ruiyi Zhou, Xinguo Ren, Yu Cao, Ziqing Guan.

Figure 1
Figure 1. Figure 1: FIG. 1. The Feynman diagram of BSE. Upper panel: The bubbles correspond to the full correlation function [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Workflow of the BSE implementation with dual- [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. BSE absorption spectra of benzene computed with three NAO basis sets (DZP, TZDP, and QZTP), shown in both the velocity and [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Convergence of the absorption spectrum with [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Convergence of the BSE absorption spectra of GaN with [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: This observation indicates that performing indepen￾dent calculations for different offsets and taking a statistically [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. BSE absorption spectra for Si with a [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: shows the corresponding benchmark for MgO, com￾pared with VASP, Exciting, and FHI-aims references from the literature, using a 14 × 14 × 14 Γ-centered mesh with 4 occupied and 4 virtual orbitals and 𝜂 = 0.3 eV. We note that the reference calculations employed slightly different sampling methods: FHI-aims [29] and VASP [68] used 15 × 15 × 15 Γ￾centered meshes, while Exciting [68] used an 11 × 11 × 11 FIG. 7… view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9. GaN exciton binding energy [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

93 extracted references · 93 canonical work pages · 1 internal anchor

  1. [1]

    2) to obtain the Kohn–Sham energies𝐸k 𝑛, eigenvectors𝑐 k 𝑠𝑛, LRI expansion coefficients𝐶 𝜇 𝑠𝑡 (R), and the Coulomb matrix𝑉 𝜇𝜈 (q)in the ABF basis

    Perform the DFT calculation within ABACUS (upper panel in Fig. 2) to obtain the Kohn–Sham energies𝐸k 𝑛, eigenvectors𝑐 k 𝑠𝑛, LRI expansion coefficients𝐶 𝜇 𝑠𝑡 (R), and the Coulomb matrix𝑉 𝜇𝜈 (q)in the ABF basis. The Coulomb matrix is first evaluated in real space as 𝑉𝜇𝜈 (R)for atom pairs within a cutoff set by the in- putparameterexx_rmesh_timestimestheNAOr...

    work page

  2. [2]

    Perform the𝐺 0𝑊0 calculation within LibRPA (mid- dle panel in Fig. 2). Starting from the non- interacting Green’s function𝐺0 𝑠𝑡 (R, 𝑖𝜏)in the real- space imaginary-time domain, the calculation proceeds through the non-interacting response function𝜒0 𝜇𝜈 and the screened Coulomb matrix𝑊𝜇𝜈 in the ABF basis, to the self-energyΣ 𝑠𝑡 in the NAO basis. These step...

    work page

  3. [3]

    Perform the BSE calculation within ABACUS (lower panel in Fig. 2). From this step onward, all computa- tions will be performed on the densek-mesh. The LRI 7 expansioncoefficients𝐶 𝜇 𝑠𝑡 (R)aretransformedfromthe NAO basis to the Kohn–Sham orbital basis to obtain e𝐶 𝜇 𝑗k2,𝑖k1 . Then, we construct the BSE Hamiltonian ma- trix via Eq. (20) using the𝐺0𝑊0 quasip...

    work page

  4. [4]

    The absorption spectrumcanbecomputedbyEq.(24)fortheTDAcase or Eq

    Diagonalize the BSE Hamiltonian𝐻BSE to obtain the exciton eigenvalues and eigenvectors. The absorption spectrumcanbecomputedbyEq.(24)fortheTDAcase or Eq. (25) for the full case. ABACUS LibRPA(LibRI) ABACUS(LibRI) INPUT, STRU, KPT 𝑉𝜇𝜈(R), 𝑉𝜇𝜈(q) 𝐶𝜇𝑠𝑡(R) 𝐸k𝑛 𝑐k𝑠𝑛 𝐺0𝑠𝑡(R, 𝑖𝜏) 𝜒0𝜇𝜈(R, 𝑖𝜏), 𝜒0𝜇𝜈(q, 𝑖𝜔) 𝑊𝜇𝜈(q, 𝑖𝜔), 𝑊𝜇𝜈(R, 𝑖𝜏) Σ𝑠𝑡(R, 𝑖𝜏)Σ𝑠𝑡(k, 𝑖𝜔) 𝐸GW𝑛k 𝑊𝜇𝜈(R, 𝑖...

    work page

  5. [5]

    (24) are formally equivalent for bound systems, see Appendix A

    Assessing NAO basis set convergence via velocity–length equivalence In a complete basis set, the velocity and length formula- tions of the BSE spectra in Eq. (24) are formally equivalent for bound systems, see Appendix A. However, as the basis set is incomplete in practice, the two formulations may yield differentresults. Thisbehaviororiginatesfromthefact...

    work page

  6. [6]

    III, the present BSE implementation employs the localized resolution-of-identity (LRI) technique, inwhichproductsoftwoNAOsareexpandedintermsofABFs

    Auxiliary basis set convergence As discussed in Sec. III, the present BSE implementation employs the localized resolution-of-identity (LRI) technique, inwhichproductsoftwoNAOsareexpandedintermsofABFs. 8 FIG. 3. BSE absorption spectra of benzene computed with three NAO basis sets (DZP, TZDP, and QZTP), shown in both the velocity and length formulations. Lo...

    work page

  7. [7]

    For this test set, ABACUS+LibRPA results are compared directly with FHI-aims [66] for both spin chan- nels (singlet and triplet) and for both BSE variants (TDA and full BSE)

    Molecular systems: Thiel’s set To assess the accuracy of the present implementation, we benchmark against Thiel’s set [65], which contains 28 molecules. For this test set, ABACUS+LibRPA results are compared directly with FHI-aims [66] for both spin chan- nels (singlet and triplet) and for both BSE variants (TDA and full BSE). To minimize errors associated...

    work page

  8. [8]

    Periodic systems: Si and MgO We next benchmark the present implementation against es- tablishedreferencesfortwoprototypicalsystems: silicon(Si), amedium-gapcovalentsemiconductorwithweaklyboundex- citons,andmagnesiumoxide(MgO),awide-gapionicinsula- tor with strongly bound excitons. Fig. 7 compares the Si absorption spectrum computed with ABACUS+LibRPA agai...

    work page

  9. [9]

    Thisdefinitioncapturesthe energy lowering due to the attractive electron-hole interaction relative to the independent-quasiparticle picture

    Exciton binding energies Theexcitonbindingenergy𝐸 𝐵 isacentralquantitycharac- terizingthestrengthoftheelectron–holeinteractioninoptical excitations,whichisdefinedasthedifferencebetweenthemin- imum direct quasiparticle gap across the Brillouin zone and the lowest BSE excitation energy: 𝐸𝐵 =min k h 𝐸QP 𝑐k −𝐸 QP 𝑣k i −Ω BSE 1 ,(37) where𝐸 QP 𝑐k and𝐸 QP 𝑣k ar...

    work page

  10. [10]

    Onida, L

    G. Onida, L. Reining, and A. Rubio, Electronic excitations: Density-functional versus many-body Green’s-function ap- proaches, Rev. Mod. Phys.74, 601 (2002)

    work page 2002

  11. [11]

    N. T. Maitra, Charge transfer in time-dependent density func- tional theory, J. Phys.: Condens. Matter29, 423001 (2017)

    work page 2017

  12. [12]

    Rev.84, 1232 (1951)

    E.E.SalpeterandH.A.Bethe,Arelativisticequationforbound- state problems, Phys. Rev.84, 1232 (1951)

    work page 1951

  13. [13]

    Strinati, Application of the Green’s functions method to the study of the optical properties of semiconductors, Riv

    G. Strinati, Application of the Green’s functions method to the study of the optical properties of semiconductors, Riv. Nuovo Cimento11, 1 (1988)

    work page 1988

  14. [14]

    Rohlfing and S

    M. Rohlfing and S. G. Louie, Electron-hole excitations and op- ticalspectrafromfirstprinciples,Phys.Rev.B62,4927(2000)

    work page 2000

  15. [15]

    Jacquemin, I

    D. Jacquemin, I. Duchemin, and X. Blase, Benchmarking the bethe–salpeter formalism on a standard organic molecular set, 15 J. Chem. Theory Comput.11, 3290 (2015)

    work page 2015

  16. [16]

    X.Blase,I.Duchemin,D.Jacquemin,andP.-F.Loos,Thebethe– salpeterequationformalism: Fromphysicstochemistry,J.Phys. Chem. Lett.11, 7371 (2020)

    work page 2020

  17. [17]

    Okada, A

    M. Okada, A. Kutana, Y. Kureishi, Y. Kobayashi, Y. Saito, T. Saito, K. Watanabe, T. Taniguchi, S. Gupta, Y. Miyata, B. I. Yakobson, H. Shinohara, and R. Kitaura, Direct and indirect interlayer excitons in a van Der Waals heterostructure of hB- N/WS2/MoS2/hBN, ACS Nano12, 2498 (2018)

    work page 2018

  18. [18]

    A. M. Alvertis, A. Champagne, M. Del Ben, F. H. Da Jor- nada, D. Y. Qiu, M. R. Filip, and J. B. Neaton, Importance of nonuniformBrillouinzonesamplingforabinitioBethe-Salpeter equation calculations of exciton binding energies in crystalline solids, Phys. Rev. B108, 235117 (2023)

    work page 2023

  19. [19]

    V. W.-z. Yu, Y. Jin, G. Galli, and M. Govoni, GPU-accelerated solution of the Bethe–Salpeter equation for large and heteroge- neous systems, J. Chem. Theory Comput.20, 5319 (2024)

    work page 2024

  20. [20]

    N. C. Bradbury, T. Allen, M. Nguyen, K. Z. Ibrahim, and D. Neuhauser, Optimized attenuated interaction: Enabling stochastic Bethe–Salpeter spectra for large systems, J. Chem. Phys.158, 154104 (2023)

    work page 2023

  21. [21]

    Hillenbrand, J

    C. Hillenbrand, J. Li, and T. Zhu, Energy-specific Bethe– Salpeter equation implementation for efficient optical spectrum calculations, J. Chem. Phys.162, 174117 (2025)

    work page 2025

  22. [22]

    Attaccalite, M

    C. Attaccalite, M. Grüning, and A. Marini, Real-time approach to the optical properties of solids and nanostructures: Time- dependent Bethe-Salpeter equation, Phys. Rev. B84, 245110 (2011)

    work page 2011

  23. [23]

    Jiang, Q

    X. Jiang, Q. Zheng, Z. Lan, W. A. Saidi, X. Ren, and J. Zhao, Real-time GW-BSE investigations on spin-valley exciton dy- namics in monolayer transition metal dichalcogenide, Sci. Adv. 7, eabf3759 (2021)

    work page 2021

  24. [24]

    Y.-H.Chan,D.Y.Qiu,F.H.daJornada,andS.G.Louie,Giant exciton-enhanced shift currents and direct current conduction with subbandgap photo excitations produced by many-electron interactions, Proc. Natl. Acad. Sci. U.S.A.118, e1906938118 (2021)

    work page 2021

  25. [25]

    C.Hu,M.H.Naik,Y.-H.Chan,J.Ruan,andS.G.Louie,Light- induced shift current vortex crystals in moiré heterobilayers, Proc. Natl. Acad. Sci. U.S.A.120, e2314775120 (2023)

    work page 2023

  26. [26]

    Chang Lee, L

    V. Chang Lee, L. Yue, M. B. Gaarde, Y.-h. Chan, and D. Y. Qiu, Many-body enhancement of high-harmonic generation in monolayer MoS2, Nat. Commun.15, 6228 (2024)

    work page 2024

  27. [27]

    Š.MarekandJ.Wilhelm,Linearandnonlinearopticalproperties of molecules from real-time propagation based on the Bethe– Salpeter equation, J. Chem. Theory Comput.21, 9814 (2025)

    work page 2025

  28. [28]

    J.Ruan,Y.-H.Chan,andS.G.Louie,ExcitonEnhancedNonlin- ear Optical Responses in Monolayer h-BN and MoS2: Insight from First-Principles Exciton-State Coupling Formalism and Calculations, Nano Lett.24, 15533 (2024)

    work page 2024

  29. [29]

    J.Li,D.Golze,andW.Yang,CombiningRenormalizedSingles GW Methods with the Bethe–Salpeter Equation for Accurate NeutralExcitationEnergies,J.Chem.TheoryComput.18,6637 (2022)

    work page 2022

  30. [30]

    Urquiza, M

    L. Urquiza, M. Gatti, and F. Sottile, Connections between reso- nant inelastic X-ray scattering and complementary x-ray spec- troscopies: Probing excitons at the al k𝐿1 edges of𝛼-al 2o3, Phys. Rev. B109, 115157 (2024)

    work page 2024

  31. [31]

    Nicolaou, K

    A. Nicolaou, K. Ruotsalainen, L. Susana, V. Porée, L. G. Tizei, J.Koskelo,T.Taniguchi,K.Watanabe,A.Zobelli,andM.Gatti, Directmeasurementofthelongitudinalexcitondispersioninℎ- bn by resonant inelastic X-ray scattering, Phys. Rev. B112, 085207 (2025)

    work page 2025

  32. [32]

    Bajaj, N

    R. Bajaj, N. Venkatareddy, H. R. Krishnamurthy, and M. Jain, Symmetriesinzeroandfinitecenter-of-massmomentaexcitons, Phys. Rev. B112, 245127 (2025)

    work page 2025

  33. [33]

    Nalabothula, D

    M. Nalabothula, D. Sangalli, F. Paleari, S. Reichardt, and L. Wirtz, Symmetries of excitons, Phys. Rev. B113, 205130 (2026)

    work page 2026

  34. [34]

    Stöhler, S

    J. Stöhler, S. Blügel, and C. Friedrich, Efficient all- electron Bethe-Salpeter implementation using crystal symme- tries (2026), arXiv:2603.24860 [cond-mat.mtrl-sci]

    work page arXiv 2026

  35. [35]

    Kammerlander, S

    D. Kammerlander, S. Botti, M. A. L. Marques, A. Marini, and C. Attaccalite, Speeding up the solution of the Bethe-Salpeter equation by a double-grid method and wannier interpolation, Phys. Rev. B86, 125203 (2012)

    work page 2012

  36. [36]

    Gillet, M

    Y. Gillet, M. Giantomassi, and X. Gonze, Efficient on-the-fly interpolationtechniqueforbethe–salpetercalculationsofoptical spectra, Comput. Phys. Commun.203, 83 (2016)

    work page 2016

  37. [37]

    Chem.9, 763946 (2022)

    I.M.Alliati,D.Sangalli,andM.Grüning,Doublek-gridmethod forsolvingtheBethe-Salpeterequationvialanczosapproaches, Front. Chem.9, 763946 (2022)

    work page 2022

  38. [38]

    R. Zhou, Y. Yao, V. Blum, X. Ren, and Y. Kanai, All-Electron BSE@GW Method with Numeric Atom-Centered Orbitals for Extended Periodic Systems, J. Chem. Theory Comput.21, 291 (2025)

    work page 2025

  39. [39]

    V. Blum, R. Gehrke, F. Hanke, P. Havu, V. Havu, X. Ren, K. Reuter, and M. Scheffler, Ab initio molecular simulations with numeric atom-centered orbitals, Comput. Phys. Commun. 180, 2175 (2009)

    work page 2009

  40. [40]

    J. W. Abbott, C. Mera Acosta, A. Akkoush, A. Ambrosetti, V. Atalla, A. Bagrets, J. Behler, D. Berger, H. Bertschi, B. Bie- niek,J.Björk,V.Blum,S.Bohloul,C.L.Box,N.J.Boyer,D.S. Brambila,G.A.Bramley,K.R.Bryenton,M.Camarasa-Gómez, C. Carbogno, F. Caruso, S. Chutia, M. Ceriotti, G. Csányi, W. Dawson, F. A. Delesma, F. Della Sala, B. Delley, R. DiSta- sio, M. ...

    work page doi:10.1088/2516-1075/ae8067 2026

  41. [41]

    P. Li, X. Liu, M. Chen, P. Lin, X. Ren, L. Lin, C. Yang, and L.He,Large-scaleabinitiosimulationsbasedonsystematically improvable atomic basis, Comput. Mater. Sci.112, 503 (2016)

    work page 2016

  42. [42]

    P. Lin, X. Ren, X. Liu, and L. He, Ab initio electronic struc- ture calculations based on numerical atomic orbitals: Basic fomalismsandrecentprogresses,WIREsComput.Mol.Sci.14, e1687 (2024)

    work page 2024

  43. [43]

    W. Zhou, D. Zheng, Q. Liu, D. Lu, Y. Liu, P. Lin, Y. Huang, X.Peng, J.J. Bao, C.Cai, Z.Jin, J.Wu, H. Zhang, G. Jin, Y.Ji, Z.Shen,X.Liu,L.Sun,Y.Cao,M.Sun,J.Liu,T.Chen,R.Liu, Y. Li, H. Han, X. Liang, T. Bao, Z. Deng, T. Liu, N. Chen, H.Ren,X.Zhang,Z.Liu,Y.Fu,M.Liu,Z.Li,T.Wen,Z.Tang, Y.Xu,W.Duan,X.Wang,Q.Gu,F.-Z.Dai,Q.Zheng,Y.Zhong, H. Xiang, X. Gong, J. Zha...

    work page 2025

  44. [44]

    R. Shi, P. Lin, M.-Y. Zhang, L. He, and X. Ren, Subquadratic- scaling real-space random phase approximation correlation en- ergy calculations for periodic systems with numerical atomic orbitals, Phys. Rev. B109, 035103 (2024)

    work page 2024

  45. [45]

    Shi, M.-Y

    R. Shi, M.-Y. Zhang, P. Lin, L. He, and X. Ren, LibRPA: A softwarepackageforlow-scalingfirst-principlescalculationsof random phase approximation electron correlation energy based on numerical atomic orbitals, Comput. Phys. Commun.309, 109496 (2025)

    work page 2025

  46. [46]

    22, 5770 (2026)

    M.-Y.Zhang,P.Lin,R.Shi,andX.Ren,Low-scalingGWcalcu- lationsofquasi-particleenergiesforextendedsystemswithinthe numericalatomicorbitalframework,J.Chem.TheoryComput. 22, 5770 (2026)

    work page 2026

  47. [47]

    B.Jia,M.-Y.Zhang,Z.Guan,H.Gong,andX.Ren,All-electron quasiparticleself-consistentGWformoleculesandperiodicsys- tems within the numerical atomic orbital framework, J. Chem. Phys.164, 194115 (2026)

    work page 2026

  48. [48]

    $G^0W^0$ implementation based on the pseudopotential and numerical-atomic-orbital basis-set framework: Algorithms and benchmarks

    H. Gong, M.-Y. Zhang, P. Lin, B. Jia, Z. Guan, L. He, and X. Ren,𝐺 0𝑊0 implementation based on the pseudopotential and numerical-atomic-orbital basis-set framework: Algorithms and benchmarks, arXiv:2605.11512 (2026)

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  49. [49]

    Strinati, Effects of dynamical screening on resonances at inner-shellthresholdsinsemiconductors,Phys.Rev.B29,5718 (1984)

    G. Strinati, Effects of dynamical screening on resonances at inner-shellthresholdsinsemiconductors,Phys.Rev.B29,5718 (1984)

    work page 1984

  50. [50]

    M. Shao, F. H. Da Jornada, C. Yang, J. Deslippe, and S. G. Louie, Structure preserving parallel algorithms for solving the Bethe–Salpeter eigenvalue problem, Linear Algebra Appl.488, 148 (2016)

    work page 2016

  51. [51]

    C.Penke,A.Marek,C.Vorwerk,C.Draxl,andP.Benner,High performance solution of skew-symmetric eigenvalue problems withapplicationsinsolvingtheBethe-Salpetereigenvalueprob- lem, Parallel Comput.96, 102639 (2020)

    work page 2020

  52. [52]

    P. Lin, X. Ren, and L. He, Accuracy of Localized Resolution of theIdentityinPeriodicHybridFunctionalCalculationswithNu- merical Atomic Orbitals, J. Phys. Chem. Lett.11, 3082 (2020)

    work page 2020

  53. [53]

    P. Lin, X. Ren, and L. He, Efficient Hybrid Density Func- tionalCalculationsforLargePeriodicSystemsUsingNumerical Atomic Orbitals, J. Chem. Theory Comput.17, 222 (2021)

    work page 2021

  54. [54]

    Cao, M.-Y

    Y. Cao, M.-Y. Zhang, P. Lin, M. Chen, and X. Ren, Apply- ing space-group symmetry to speed up hybrid-functional cal- culations within the framework of numerical atomic orbitals, J. Chem. Theory Comput.21, 8086 (2025)

    work page 2025

  55. [55]

    Scheffler, and V

    A.C.Ihrig,J.Wieferink,I.Y.Zhang,M.Ropo,X.Ren,P.Rinke, M. Scheffler, and V. Blum, Accurate localized resolution of identity approach for linear-scaling hybrid density functionals andformany-bodyperturbationtheory,NewJ.Phys.17,093020 (2015)

    work page 2015

  56. [56]

    S. V. Levchenko, X. Ren, J. Wieferink, R. Johanni, P. Rinke, V. Blum, and M. Scheffler, Hybrid functionals for large peri- odic systems in an all-electron, numeric atom-centered basis framework, Comput. Phys. Commun.192, 60 (2015)

    work page 2015

  57. [57]

    X. Ren, F. Merz, H. Jiang, Y. Yao, M. Rampp, H. Lederer, V. Blum, and M. Scheffler, All-electron periodic𝐺0𝑊0 imple- mentationwithnumericalatomicorbitalbasisfunctions: Algo- rithm and benchmarks, Phys. Rev. Mater.5, 013807 (2021)

    work page 2021

  58. [58]

    Spencer and A

    J. Spencer and A. Alavi, Efficient calculation of the exact ex- change energy in periodic systems using a truncated coulomb potential, Phys. Rev. B77, 193110 (2008)

    work page 2008

  59. [59]

    M.Azizi,J.Wilhelm,D.Golze,M.Giantomassi,R.L.Panadés- Barrueta, F. A. Delesma, A. Buccheri, A. Gulans, P. Rinke, C. Draxl, and X. Gonze, Time-frequency component of the GreenX library: Minimax grids for efficient RPA and GW cal- culations, J. Open Source Softw.8, 5570 (2023)

    work page 2023

  60. [60]

    Kaltak, J

    M. Kaltak, J. Klimeš, and G. Kresse, Low Scaling Algorithms for the Random Phase Approximation: Imaginary Time and Laplace Transformations, J. Chem. Theory Comput.10, 2498 (2014)

    work page 2014

  61. [61]

    P.Liu,M.Kaltak,J.Klimeš,andG.Kresse,CubicscalingGW: Towardsfastquasiparticlecalculations,Phys.Rev.B94,165109 (2016)

    work page 2016

  62. [62]

    LibRPA repository:https://github.com/Srlive1201/ LibRPA(2026)

    work page 2026

  63. [63]

    LibRI repository:https://github.com/abacusmodeling/ LibRI(2026)

    work page 2026

  64. [64]

    Lehtola, A review on non-relativistic, fully numerical elec- tronic structure calculations on atoms and diatomic molecules, Int

    S. Lehtola, A review on non-relativistic, fully numerical elec- tronic structure calculations on atoms and diatomic molecules, Int. J. Quantum Chem.119, e25968 (2019)

    work page 2019

  65. [65]

    H. Peng, S. Yang, H. Jiang, H. Weng, and X. Ren, Basis-Set- Error-FreeRandom-PhaseApproximationCorrelationEnergies forAtomsBasedontheSternheimerEquation,J.Chem.Theory Comput. 10.1021/acs.jctc.3c00668 (2023)

    work page doi:10.1021/acs.jctc.3c00668 2023

  66. [66]

    Peng and X

    H. Peng and X. Ren, Ab initio correlated calculations with- out finite-basis-set error: Numerically precise all-electron random-phase-approximation correlation energies for diatomic molecules, Phys. Rev. A112, 062819 (2025)

    work page 2025

  67. [67]

    Chen, G.-C

    M. Chen, G.-C. Guo, and L. He, Systematically improvable optimized atomic basis sets for ab initio calculations, J. Phys.: Condens. Matter22, 445501 (2010)

    work page 2010

  68. [68]

    P.Lin,X.Ren,andL.He,Strategyforconstructingcompactnu- merical atomic orbital basis sets by incorporating the gradients of reference wavefunctions, Phys. Rev. B103, 235131 (2021)

    work page 2021

  69. [69]

    A.Jain,S.P.Ong,G.Hautier, etal.,Commentary: TheMaterials Project: Amaterialsgenomeapproachtoacceleratingmaterials innovation, APL Mater.1, 011002 (2013)

    work page 2013

  70. [70]

    Laskowski, N

    R. Laskowski, N. E. Christensen, G. Santi, and C. Ambrosch- Draxl, Ab initio calculations of excitons in GaN, Phys. Rev. B 72, 035204 (2005)

    work page 2005

  71. [71]

    Struct.1, 037001 (2019)

    C.Vorwerk,B.Aurich,C.Cocchi,andC.Draxl,Bethe–Salpeter equationforabsorptionandscatteringspectroscopy: Implemen- tation in the exciting code, Electron. Struct.1, 037001 (2019)

    work page 2019

  72. [72]

    T.Sander,E.Maggio,andG.Kresse,BeyondtheTamm-Dancoff approximationforextendedsystemsusingexactdiagonalization, Phys. Rev. B92, 045209 (2015)

    work page 2015

  73. [73]

    D.E.AspnesandA.A.Studna,Dielectricfunctionsandoptical parameters of Si, Ge, GaP, GaAs, GaSb, InP, InAs, and InSb 17 from 1.5 to 6.0 eV, Phys. Rev. B27, 985 (1983)

    work page 1983

  74. [74]

    Schreiber, M

    M. Schreiber, M. R. Silva-Junior, S. P. A. Sauer, and W. Thiel, Benchmarks for electronically excited states: CASPT2, CC2, CCSD, and CC3, J. Chem. Phys.128, 134110 (2008)

    work page 2008

  75. [75]

    C. Liu, J. Kloppenburg, Y. Yao, X. Ren, H. Appel, Y. Kanai, and V.Blum, All-electron ab initioBethe-Salpeter equation ap- proach to neutral excitations in molecules with numeric atom- centered orbitals, J. Chem. Phys.152, 044105 (2020)

    work page 2020

  76. [76]

    Deslippe, G

    J. Deslippe, G. Samsonidze, D. A. Strubbe, M. Jain, M. L. Co- hen, and S. G. Louie, Berkeleygw: A massively parallel com- puterpackageforthecalculationofthequasiparticleandoptical propertiesofmaterialsandnanostructures,Comput.Phys.Com- mun.183, 1269 (2012)

    work page 2012

  77. [77]

    Begum, M

    V. Begum, M. E. Gruner, C. Vorwerk, C. Draxl, and R. Pentcheva, Theoretical description of optical and x-ray ab- sorption spectra of MgO including many-body effects, Phys. Rev. B103, 195128 (2021)

    work page 2021

  78. [78]

    D. M. Roessler and W. C. Walker, Electronic Spectrum and Ultraviolet Optical Properties of Crystalline MgO, Phys. Rev. 159, 733 (1967)

    work page 1967

  79. [79]

    Scholz, Y

    R.A.R.Leute,M.Feneberg,R.Sauer,K.Thonke,S.B.Thapa, F. Scholz, Y. Taniyasu, and M. Kasu, Photoluminescence of highly excited AlN: Biexcitons and exciton-exciton scattering, Appl. Phys. Lett.95, 031903 (2009)

    work page 2009

  80. [80]

    J. Li, K. B. Nam, M. L. Nakarmi, J. Y. Lin, H. X. Jiang, P. Car- rier, and S.-H. Wei, Band structure and fundamental optical transitions in wurtzite AlN, Appl. Phys. Lett.83, 5163 (2003)

    work page 2003

Showing first 80 references.