Pith. sign in

REVIEW 3 major objections 5 minor 80 references

A robust and efficient solver for coupled cluster equations

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper argues that a gauge-invariant Fock-commutator preconditioner, applied by Krylov solves inside a Newton iteration, makes coupled-cluster solvers robust in any orbital gauge and removes level-shift tuning.

desk verdict A sensible gauge-invariant preconditioner for Newton-Krylov CC solvers, but the efficiency claim overstates what the evidence supports. read the letter →

arxiv 2608.06669 v1 pith:2QXA4LJI submitted 2026-08-07 physics.chem-ph physics.comp-ph

Chanaka D. M. Mudiyanselage , Kangbo Li , Fabian M. Faulstich This is my paper · ORCID
classification physics.chem-phphysics.comp-ph
keywords coupled-clusterequationspreconditionedNewton–Krylovmethodgauge-invariantpreconditionerFockcommutatorfixed-pointiterationDIISaccelerationlocalcorrelationmethods
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 sets out to show that the coupled-cluster (CC) equations can be solved robustly in any orbital gauge by treating them as a nonlinear root-finding problem and preconditioning the Newton update with a gauge-invariant Fock-commutator operator $A_F$, rather than with the diagonal energy denominators used in canonical fixed-point iterations. The proposed preconditioned Newton–Krylov (PNK) framework replaces heuristic level-shifting and direct-inversion-of-the-iterative-subspace (DIIS) tuning with Krylov-based inversion of $A_F$, and the paper presents CCD-level (coupled-cluster doubles) tests on hydrogen dissociation and ethane in molecular-orbital, atomic-orbital, and randomly rotated gauges. In these tests, the method converges where fixed-point iterations diverge, and it reaches the target residual with fewer residual evaluations than carefully optimized fixed-point solvers with level-shift and DIIS. If this claim holds, it removes a major source of empirical tuning from coupled-cluster calculations and extends Newton-type solvers to local-correlation and embedding formulations that use non-canonical gauges.

What carries the argument

The load-bearing object is the gauge-invariant preconditioner $A_F$, defined by $[A_F]_{\mu\nu} = \langle\Phi_\mu|[F, X_\nu]|\Phi_0\rangle$, the commutator of the Fock operator with an excitation operator taken between the reference determinant and an excited determinant. It is the exact first-order Fock contribution to the Jacobian of the CC residual, so it acts as the correct Newton preconditioner in any orbital gauge; the canonical-gauge diagonal energy denominator is its special case. The algorithm makes $A_F^{-1}$ practical through an inner GMRES solve, then uses a Jacobian-free Newton–Krylov iteration with an adaptive forcing term for the outer linear system, so the full Jacobian is never formed.

What would settle it

A concrete experiment would be to form $A_F$ explicitly for a moderately sized molecule in several random gauges and track its smallest singular value along the PNK solution path; if any gauge drives that singular value to zero or makes the inner GMRES solves require more residual evaluations than the outer iteration saves, the robustness and efficiency claims for PNK would fail.

Watch

Extended reading notes

Core claim

The paper's central claim is that the Newton step for the coupled-cluster equations is governed, in any gauge, by the linear system $A_F \delta t = -r(t)$, where $[A_F]_{\mu\nu} = \langle\Phi_\mu|[F, X_\nu]|\Phi_0\rangle$ is the Fock-commutator approximation to the Jacobian. In the canonical molecular-orbital gauge this matrix reduces to the diagonal energy denominator $\Delta\varepsilon_\mu$, so standard fixed-point iteration is recovered as a quasi-Newton method; in a non-canonical gauge the full commutator keeps the off-diagonal Fock blocks that a diagonal approximation discards. The paper then constructs PNK by applying $A_F^{-1}$ with GMRES and using that preconditioned residual inside a Jacobian-free Newton–Krylov outer loop with an adaptive forcing term. Its reported numerical results at the CCD level show robust convergence in small-gap and gauge-transformed settings, with PNK outperforming fixed-point-plus-DIIS solvers in residual-evaluation count on the systems tested.

Load-bearing premise

The method assumes that the Fock-commutator matrix $A_F$ is nonsingular enough that GMRES can reliably apply $A_F^{-1}$ in any gauge, and that the cost of those inner solves is small enough to preserve the efficiency counted in residual evaluations.

Editorial extensions

If this is right

  • A CC solver can replace the diagonal energy-denominator preconditioner with $A_F$ applied via GMRES and keep the same Newton–Krylov outer loop, eliminating the gauge-dependent approximation in every orbital representation.
  • Local-correlation methods that work in atomic-orbital or other non-canonical gauges no longer need the ad hoc diagonal-Fock fix, since the same preconditioner is valid in those gauges.
  • The small-gap regime of stretched $\mathrm{H}_2$, where the fixed-point Jacobian becomes repulsive, is handled by the inexact Newton step without an optimized level shift, on the paper's evidence.
  • In the ethane tests, PNK reaches the $10^{-8}$ residual threshold in fewer residual evaluations than fixed-point with level-shift and DIIS, supporting the paper's efficiency claim when residual evaluations dominate the cost.

Reading between the lines

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

  • Editorial inference: the paper's efficiency metric counts outer residual evaluations and not the inner $A_F^{-1}$ GMRES iterations, so a wall-time comparison on larger systems is the natural next test; if inner solves dominate, the reported advantage could erode.
  • Editorial inference: the derivation of $A_F$ uses only the Fock structure of the Hamiltonian, so the same preconditioner should carry over to CCSD, higher-order truncations, and coupled-cluster response equations, though only CCD is demonstrated here.
  • Editorial inference: a useful diagnostic suggested by the paper's $2\times2$ example is to compare the spread of diagonal energy denominators with the spectral spread of the full Fock-commutator operator, predicting when the diagonal-Fock approximation will fail before running a calculation.
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 / 5 minor

Summary. The paper proposes a method for solving coupled-cluster (CC) amplitude equations called PNK: a Jacobian-free Newton-Krylov solver preconditioned by the matrix A_F with entries [A_F]_{\mu\nu} = \langle\Phi_\mu|[F, X_\nu]|\Phi_0\rangle. This preconditioner is gauge-invariant by construction, generalizing the diagonal energy-denominator preconditioner that is only valid in the canonical gauge. The authors derive working equations for CCD in an arbitrary gauge, give an explicit algorithm (Algorithm 1, Appendix F) that applies A_F^{-1} by inner GMRES, and compare PNK against fixed-point, level-shifted fixed-point plus DIIS, and unpreconditioned Newton-Krylov variants on two molecular systems: stretched H2 at CCD/cc-pVTZ and ethane in several gauges at CCD/6-31G and CCD/cc-pVTZ. The efficiency metric used throughout is the number of residual evaluations, justified by the assertion that residual evaluation dominates the computational cost. The paper claims that PNK is robust and efficient across arbitrary gauges without system-dependent level shifting, and suggests it as a promising standard solver for CC equations.

Significance. If the central claims hold, the paper would make a useful contribution to CC solver technology: replacing a gauge-dependent, ad hoc level-shifted fixed-point iteration with a preconditioned Newton-Krylov method is conceptually attractive for local-correlation and non-canonical-gauge formulations, where the standard energy-denominator preconditioner fails. The derivation of the gauge-invariant preconditioner from the commutator form of the Jacobian is standard but cleanly presented, and the authors provide explicit working equations and publicly available Julia code. They also disclose, in Appendix D, that the level-shift values used for the SFP baseline are optimized post hoc, which is honest and actually strengthens the comparison: PNK beats an SFP baseline with a shift chosen after the fact. However, the numerical evidence is limited to two small molecules at CCD level, and the efficiency claim is measured only in residual evaluations while the proposed algorithm incurs additional inner GMRES solves for A_F^{-1}, whose cost is not included.

major comments (3)
  1. [Section III and Algorithm 1 (Appendix F)] The efficiency claim is not established by the reported metric. Section III states that convergence is assessed by residual evaluations because residual construction is the dominant cost, but Algorithm 1 applies A_F^{-1} by GMRES at lines 3 and 6, and in a non-canonical gauge A_F is not diagonal (see Eq. C4). Each application of A_F to a trial vector is a four-index Fock contraction with the same polynomial scaling as a residual term, and an inner GMRES solve multiplies the per-outer-iteration cost by the number of inner iterations. Figures 4 and 5 compare PNK against SFP+DIIS and NK only in residual evaluations, so the claimed wall-time advantage is not demonstrated. The authors should either report wall-clock timings that include the inner preconditioner solves or justify that A_F^{-1} is inexpensive in the gauges tested (e.g., by diagonal dominance or by a direct factorization).
  2. [Section II, Eq. (12), and Algorithm 1] The robustness claim requires an analysis of A_F, not just an assumption that A_F^{-1} exists. The preconditioner A_F is singular whenever F has degenerate occupied-virtual pairs, as in stretched H2 at the small-gap limit discussed in Section III.A, and the paper does not provide a conditioning estimate or a regularization strategy for the inner solves. If A_F is near-singular, the inner GMRES for A_F^{-1} can stagnate, and the 'robust' claim would fail exactly in the challenging regime the paper emphasizes. Please add a conditioning analysis, report smallest singular values for the test cases, or document a practical regularization (e.g., Tikhonov or truncated SVD) and verify that the outer Newton iterations remain robust under it.
  3. [Section III and Conclusion] The numerical support is too narrow for the breadth of the abstract and conclusion. The tests are CCD on H2 and ethane only; the abstract promises 'a range of molecular systems' and the conclusion extrapolates to 'next-generation CC solvers' and local correlation methods. There are no calculations in a genuinely local-correlation setting, no larger basis sets beyond cc-pVTZ, and no demonstration that applying A_F^{-1} remains competitive at larger problem sizes. At minimum, the authors should temper the wording to match the evidence, and ideally add at least one larger or more strongly correlated system with wall-clock timings.
minor comments (5)
  1. [Introduction and throughout] There are numerous typographical errors that should be corrected: 'approache' and 'imporatant' in the introduction, 'Relieable' for 'Reliable', 'perconditioner' in Section II, and 'ad hoc' frequently missing the space before 'chosen'. A careful proofreading pass is needed.
  2. [Appendix C, after Eq. (C2)] The Wick-expansion terms in Eq. (C2) are printed as a long list of nearly identical-looking terms; the notation with repeated expression blocks is almost unreadable. It would be much clearer to present the final result (C3) and move the derivation details to a table or a supplementary derivation.
  3. [Table I and Section II] Table I classifies INK as having no preconditioner and lists 'None' for its preconditioner inversion, but the INK method still solves the linear system A_F \delta t = -r(t) by GMRES without a preconditioner. The distinction between 'preconditioner' and 'preconditioner inversion' is confusing; please clarify in the text whether the INK GMRES applies A_F directly or uses a different preconditioner.
  4. [Appendix F, Algorithm 1] The algorithm box does not define the finite-difference step \delta used at line 6, nor does the main text state how it is chosen. Since the Jacobian-free approximation is a numerical parameter that affects convergence, please give the formula or value used in the experiments.
  5. [Appendix D and Figure 6] Appendix D states that the optimal shift was chosen by minimizing the spectral radius, but the text does not explain how the spectral radius was computed (e.g., by building the Jacobian explicitly or by a power iteration). A brief sentence on this would make the comparison reproducible.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: preconditioner is a parameter-free Fock-operator commutator; only fitted baseline shifts are disclosed as impractical.

full rationale

The derivation chain is self-contained. Section II obtains the Newton update from the residual Jacobian (Eq. 7) and then uses first-order perturbation theory to approximate the Jacobian action by the Fock-operator commutator: Eq. (10) gives <Phi_mu|[F,Delta T]|Phi_0>, and Eq. (13) defines A_F from that commutator. A_F depends only on the Fock operator and the excitation manifold, not on the CC amplitudes, the convergence history, or any fitted parameter. In the canonical gauge it reduces to the energy-denominator diagonal (Eq. 11), so the advertised 'gauge-invariant formulation' is a genuine generalization of the standard denominator, not a relabeling of it. The INK and PNK solvers then solve self-contained linear systems involving A_F and the residual; no prediction is obtained by construction from a fitted quantity. The only tuned numbers in the paper are the level shifts for the SFP/SFP+DIIS baselines, and Appendix D explicitly discloses that post-hoc optimization is impractical: 'this procedure is by no means practical and is only used in this work to obtain an unbiased comparison.' Those fits are not dressed up as predictions. Self-citations in the introduction are contextual and not load-bearing for the central claim, and the cited code repository is an implementation resource. A legitimate non-circularity concern is that the efficiency metric counts only residual evaluations and omits the inner A_F^{-1} GMRES solves in Algorithm 1; that is a potential gap in the wall-time claim, but not a circular step, since the inner solves target the fixed linear system A_F x = b rather than encoding the benchmark results.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The proposed method rests on standard CC theory plus the unverified assumptions that A_F is a good nonsingular preconditioner in arbitrary gauges and that residual evaluations dominate cost. The only numerical fitting is in the baseline SFP shifts; PNK itself has two fixed algorithm constants and an unspecified finite-difference step.

free parameters (4)
  • SFP level shift epsilon = 0.38 (H2, Fig. 2), 1.57 (ethane cc-pVTZ, Fig. 4), 3.82e-1 (H2, Appendix D)
    Post-hoc optimized shift for the fixed-point baseline only. The paper admits in Appendix D that this is 'by no means practical'.
  • Eisenstat-Walker forcing parameters gamma and alpha = gamma=0.9, alpha=1.5
    Algorithm parameters for the inexact Newton stopping criterion in Appendix F1. Standard values but chosen by hand and not swept.
  • Finite-difference step delta for Jacobian-vector products = not stated
    Equation (8) and Algorithm 1 use delta without specifying its value or an automatic choice scheme. Convergence behavior can depend on it.
  • GMRES iteration cap m_max = not stated
    Algorithm 1 takes m_max as an input but the paper never gives the value used in the numerical experiments.
assumptions (5)
  • domain assumption The coupled cluster equations (3) are a well-posed nonlinear root-finding problem in the basins tested.
    Standard CC theory background. The paper does not prove existence or uniqueness for the systems considered.
  • domain assumption The Fock commutator approximates the Jacobian to first order, equation (10), so A_F is a useful preconditioner.
    The derivation of the gauge-invariant preconditioner truncates perturbation theory at first order; no error estimate is given for arbitrary gauges.
  • domain assumption A_F is nonsingular and its GMRES inversion is reliable in all gauges tested.
    Appendix F applies A_F^{-1} directly. Invertibility is asserted implicitly but not proved, and only two molecules are tested.
  • domain assumption Residual evaluations dominate runtime.
    Used as the efficiency measure in Section III. Wall-clock costs of the inner preconditioner solves are not reported.
  • standard math Finite-difference Jacobian-vector products with small delta are sufficiently accurate.
    Standard Newton-Krylov assumption but the accuracy depends on the unspecified choice of delta.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A robust and efficient solver for coupled cluster equations." pith.science (2026). https://pith.science/paper/2QXA4LJI

@misc{pith2026260806669,
  author       = {Pith},
  title        = {Pith review of: A robust and efficient solver for coupled cluster equations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2QXA4LJI}},
  note         = {Machine review of arXiv:2608.06669}
}
read the original abstract

The coupled-cluster (CC) equations are most frequently solved via fixed-point (FP) iterations. However, when formulated in a non-canonical gauge, as in local correlation CC, the FP iteration may converge slowly or even diverge. Practical fixes, such as level-shifting and a direct inversion of iterative subspace (DIIS), often improve the convergence, but remain fundamentally heuristic and gauge dependent. {\it Yang et al.}~demonstrated that preconditioned Newton--Krylov (PNK) methods provide substantial wall-time advantage for canonical CC. In this work, we generalize the preconditioner to arbitrary gauges by replacing the energy denominator with a gauge-invariant formulation. Combined with Krylov-based approximate Jacobian inversion, the resulting framework removes the need for level-shifting and yields robust and efficient convergence across various gauges and challenging chemical systems. Our numerical results indicate that PNK consistently outperforms carefully optimized FP-based approaches across a range of molecular systems, positioning the proposed PNK method as a promising new standard for solving the CC equations.

Figures

Figures reproduced from arXiv: 2608.06669 by the authors.

Figure 1
Figure 1. FIG. 1: (Left) the total spin expectation value [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2: Side-by-side comparison of the residual norm as [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 5
Figure 5. FIG. 5: Side-by-side comparison of the residual norm as [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: FIG. 4: Side-by-side comparison of the residual norm as [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 6
Figure 6. Figure 6: FIG. 6: (Left) The spectral radius of the FP Jacobian for the hydrogen molecule (H [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7: (Left) Side-by-side comparison of the residual norm as a function of the residual evaluations for FP, INK, [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

80 extracted references · 79 canonical work pages

  1. [1]

    F. Coester. Bound states of a many-particle system.Nu- clear Physics, 7:421–424, 1958

    work page 1958

  2. [2]

    Coester and H

    F. Coester and H. K¨ ummel. Time-dependent theory of scattering of nucleons by nuclei.Nuclear Physics, 9(2):225–236, 1958

    work page 1958

  3. [3]

    ˇC ´ ıˇ zek

    J. ˇC ´ ıˇ zek. On the correlation problem in atomic and molecular systems. calculation of wavefunction compo- nents in ursell-type expansion using quantum-field the- oretical methods.The Journal of Chemical Physics, 45(11):4256–4266, 1966

    work page 1966

  4. [4]

    Paldus, J

    J. Paldus, J. ˇC ´ ıˇ zek, and I. Shavitt. Correlation problems in atomic and molecular systems. iv. extended coupled- pair many-electron theory and its application to the BH3 molecule.Physical Review A, 5(1):50, 1972

    work page 1972

  5. [5]

    G. D. Purvis III and R. J. Bartlett. A full coupled-cluster singles and doubles model: The inclusion of disconnected triples.The Journal of Chemical Physics, 76(4):1910– 1918, 1982

    work page 1910

  6. [6]

    Paldus and X

    J. Paldus and X. Li. A critical assessment of coupled clus- ter method in quantum chemistry.Advances in Chemical Physics, 110:1–175, 1999

    work page 1999

  7. [7]

    T. D. Crawford and H. F. Schaefer III. An introduction to coupled cluster theory for computational chemists.Re- views in Computational Chemistry, 14:33–136, 2000

    work page 2000

  8. [8]

    R. J. Bartlett and M. Musia l. Coupled-cluster the- ory in quantum chemistry.Reviews of Modern Physics, 79(1):291–352, 2007

    work page 2007

Show all 80 references
  1. [9]

    M. C. Jones. Many electron theory.Physics Bulletin, 24(6):381–382, 1973

    1973

  2. [10]

    level–shifting

    V. R. Saunders and I. H. Hillier. A “level–shifting” method for converging closed shell hartree–fock wave functions.International Journal of Quantum Chemistry, 7(4):699–705, 1973

    1973

  3. [11]

    P. Pulay. Convergence acceleration of iterative sequences. the case of scf iteration.Chemical Physics Letters, 73(2):393–398, 1980

    1980

  4. [12]

    T. P. Hamilton and P. Pulay. Direct inversion in the iterative subspace (diis) optimization of open-shell, excited-state, and small multiconfiguration scf wave func- tions.The Journal of Chemical Physics, 84(10):5728– 5734, 1986

    1986

  5. [13]

    G. E. Scuseria, T. J. Lee, and H. F. Schaefer III. Accel- erating the convergence of the coupled-cluster approach: The use of the diis method.Chemical Physics Letters, 130(3):236–239, 1986

    1986

  6. [14]

    Hu and W

    X. Hu and W. Yang. Accelerating self-consistent field convergence with the augmented roothaan–hall energy function.The Journal of Chemical Physics, 132(5), 2010

    2010

  7. [15]

    K. N. Kudin, G. E. Scuseria, and E. Cances. A black- box self-consistent field convergence algorithm: One step closer.The Journal of Chemical Physics, 116(19):8255– 8261, 2002

    2002

  8. [16]

    P. Pulay. Improved scf convergence acceleration.Journal of Computational Chemistry, 3(4):556–560, 1982

    1982

  9. [17]

    Chupin, M

    M. Chupin, M. Dupuy, G. Legendre, and E. S´ er´ e. Convergence analysis of adaptive diis algorithms with application to electronic ground state calculations. ESAIM: Mathematical Modelling and Numerical Analy- sis, 55(6):2785–2825, 2021

    2021

  10. [18]

    Pulay, S

    P. Pulay, S. Sæbø, and W. Meyer. An efficient reformu- lation of the closed-shell self-consistent electron pair the- ory.The Journal of Chemical Physics, 81(4):1901–1905, 1984

    1901

  11. [19]

    W. D. Laidig, G. D. Purvis, and R. J. Bartlett. Can simple localized bond orbitals and coupled cluster meth- ods predict reliable molecular energies?The Journal of Physical Chemistry, 89(11):2161–2171, 1985

    1985

  12. [20]

    Adamowicz and R

    L. Adamowicz and R. J. Bartlett. Optimized virtual or- bital space for high-level correlated calculations.The Journal of Chemical Physics, 86(11):6314–6324, 1987

    1987

  13. [21]

    Sæbø and P

    S. Sæbø and P. Pulay. Local treatment of electron corre- lation.Annual Review of Physical Chemistry, 44(1):213– 236, 1993

    1993

  14. [22]

    Hampel and H

    C. Hampel and H. Werner. Local treatment of electron correlation in coupled cluster theory.The Journal of Chemical Physics, 104(16):6286–6297, 1996

    1996

  15. [23]

    Hetzer, P

    G. Hetzer, P. Pulay, and H. Werner. Multipole approxi- mation of distant pair energies in local MP2 calculations. 7 Chemical Physics Letters, 290(1-3):143–149, 1998

    1998

  16. [24]

    P. E. Maslen and M. Head-Gordon. Noniterative local second order Møller–Plesset theory: Convergence with local correlation space.The Journal of Chemical Physics, 109(17):7093–7099, 1998

    1998

  17. [25]

    G. E. Scuseria and P. Y. Ayala. Linear scaling coupled cluster and perturbation theories in the atomic orbital basis.The Journal of Chemical Physics, 111(18):8330– 8343, 1999

    1999

  18. [26]

    Sch¨ utz

    M. Sch¨ utz. Low-order scaling local electron corre- lation methods. III. linear scaling local perturbative triples correction (T).The Journal of Chemical Physics, 113(22):9986–10001, 2000

    2000

  19. [27]

    Sch¨ utz and H

    M. Sch¨ utz and H. Werner. Local perturbative triples correction (T) with linear cost scaling.Chemical Physics Letters, 318(4-5):370–378, 2000

    2000

  20. [28]

    Flocke and R

    N. Flocke and R. J. Bartlett. A natural linear scal- ing coupled-cluster method.The Journal of Chemical Physics, 121(22):10935–10944, 2004

    2004

  21. [29]

    Neese, F

    F. Neese, F. Wennmohs, and A. Hansen. Efficient and accurate local approximations to coupled-electron pair approaches: An attempt to revive the pair natural orbital method.The Journal of Chemical Physics, 130(11), 2009

    2009

  22. [30]

    Riplinger and F

    C. Riplinger and F. Neese. An efficient and near linear scaling pair natural orbital based local coupled cluster method.The Journal of Chemical Physics, 138(3), 2013

    2013

  23. [31]

    Pinski, C

    P. Pinski, C. Riplinger, E. F. Valeev, and F. Neese. SparseMaps–a systematic infrastructure for reduced- scaling electronic structure methods. I. an efficient and simple linear scaling local MP2 method that uses an in- termediate basis of pair natural orbitals.The Journal of C...

    2015

  24. [32]

    Pavoˇ sevi´ c, P

    F. Pavoˇ sevi´ c, P. Pinski, C. Riplinger, F. Neese, and E. F. Valeev. SparseMaps–a systematic infrastructure for reduced-scaling electronic structure methods. IV. linear- scaling second-order explicitly correlated energy with pair natural orbitals.The Journal of Chemical Phys...

    2016

  25. [33]

    Saitow, U

    M. Saitow, U. Becker, C. Riplinger, E. F. Valeev, and F. Neese. A new near-linear scaling, efficient and accu- rate, open-shell domain-based local pair natural orbital coupled cluster singles and doubles theory.The Journal of Chemical Physics, 146(16), 2017

    2017

  26. [34]

    Y. Guo, C. Riplinger, U. Becker, D. G. Liakos, Y. Mi- nenkov, L. Cavallo, and F. Neese. Communication: An improved linear scaling perturbative triples correction for the domain based local pair-natural orbital based sin- gles and doubles coupled cluster method [DLPNO-CCSD (T)]...

    2018

  27. [35]

    T. P. ˇZivkovi´ c and H. J. Monkhorst. Analytic connection between configuration–interaction and coupled-cluster solutions.Journal of Mathematical Physics, 19(5):1007– 1022, 1978

    1978

  28. [36]

    Adams, K

    B. Adams, K. Jankowski, and J. Paldus. Symmetry- adapted coupled-pair approach to the many-electron cor- relation problem. II. Application to the be atom.Physical Review A, 24(5):2316, 1981

    1981

  29. [37]

    Piecuch, S

    P. Piecuch, S. Zarrabian, J. Paldus, and J. ˇC ´ ıˇ zek. Coupled-cluster approaches with an approximate account of triexcitations and the optimized-inner-projection tech- nique. II. Coupled-cluster results for cyclic-polyene model systems.Physical Review B, 42(6):3351, 1990

    1990

  30. [38]

    Paldus, P

    J. Paldus, P. Piecuch, L. Pylypow, and B. Jeziorski. Ap- plication of Hilbert-space coupled-cluster theory to sim- ple (H2)2 model systems: Planar models.Physical Review A, 47(4):2738, 1993

    1993

  31. [39]

    Piecuch and L

    P. Piecuch and L. Adamowicz. Solving the single- reference coupled-cluster equations involving highly ex- cited clusters in quasidegenerate situations.The Journal of Chemical Physics, 100(8):5857–5869, 1994

    1994

  32. [40]

    Kowalski and K

    K. Kowalski and K. Jankowski. Towards complete solu- tions to systems of nonlinear equations of many-electron theories.Physical Review Letters, 81(6):1195, 1998

    1998

  33. [41]

    Piecuch and K

    P. Piecuch and K. Kowalski. In search of the relationship between multiple solutions characterizing coupled-cluster theories. In J. Leszczynski, editor,Computational Chem- istry, Reviews of Current Trends, pages 1–104. World Sci- entific, 2000

    2000

  34. [42]

    F. M. Faulstich and A. Laestadius. Homotopy contin- uation methods for coupled-cluster theory in quantum chemistry.Molecular Physics, page e2258599, 2023

    2023

  35. [43]

    Kowalski and P

    K. Kowalski and P. Piecuch. Complete set of solutions of multireference coupled-cluster equations: The state- universal formalism.Physical Review A, 61(5):052506, 2000

    2000

  36. [44]

    Szak´ acs and P

    P. Szak´ acs and P. Surj´ an. Stability conditions for the cou- pled cluster equations.International Journal of Quantum Chemistry, 108(12):2043–2052, 2008

    2008

  37. [45]

    Schneider

    R. Schneider. Analysis of the projected coupled cluster method in electronic structure calculation.Numerische Mathematik, 113(3):433–471, 2009

    2009

  38. [46]

    Rohwedder

    T. Rohwedder. The continuous coupled cluster formulation for the electronic Schr¨ odinger equation. ESAIM: Mathematical Modelling and Numerical Analy- sis, 47(2):421–447, 2013

    2013

  39. [47]

    Rohwedder and R

    T. Rohwedder and R. Schneider. Error estimates for the coupled cluster method.ESAIM: Mathematical Mod- elling and Numerical Analysis, 47(6):1553–1582, 2013

    2013

  40. [48]

    F. M. Faulstich, H. E. Kristiansen, M. A. Csirik, S. Kvaal, T. B. Pedersen, and A. Laestadius. S-diagnostic–an a posteriori error assessment for single-reference coupled- cluster methods.The Journal of Physical Chemistry A, 127(43):9106–9120, 2023

    2023

  41. [49]

    Laestadius and S

    A. Laestadius and S. Kvaal. Analysis of the extended coupled-cluster method in quantum chemistry.SIAM Journal on Numerical Analysis, 56(2):660–683, 2018

    2018

  42. [50]

    Laestadius and F

    A. Laestadius and F. M. Faulstich. The coupled- cluster formalism–a mathematical perspective.Molecular Physics, 117(17):2362–2373, 2019

    2019

  43. [51]

    F. M. Faulstich, A. Laestadius, ¨O. Legeza, R. Schneider, and S. Kvaal. Analysis of the tailored coupled-cluster method in quantum chemistry.SIAM Journal on Nu- merical Analysis, 57(6):2579–2607, 2019

    2019

  44. [52]

    M. A. Csirik and A. Laestadius. Coupled-cluster theory revisited – part I: Discretization.ESAIM: Mathematical Modelling and Numerical Analysis, 57(2):645–670, 2023

    2023

  45. [53]

    M. A. Csirik and A. Laestadius. Coupled-cluster the- ory revisited – part II: Analysis of the single-reference coupled-cluster equations.ESAIM: Mathematical Mod- elling and Numerical Analysis, 57(2):545–583, 2023

    2023

  46. [54]

    Hassan, Y

    M. Hassan, Y. Maday, and Y. Wang. Analysis of the sin- gle reference coupled cluster method for electronic struc- ture calculations: the full-coupled cluster equations.Nu- merische Mathematik, 155(1-2):121–173, 2023

    2023

  47. [55]

    Hassan, Y

    M. Hassan, Y. Maday, and Y. Wang. Analysis of the sin- gle reference coupled cluster method for electronic struc- ture calculations: The discrete coupled cluster equations. 8 arXiv:2311.00637, 2023

    2023

  48. [56]

    F. M. Faulstich and M. Oster. Coupled cluster the- ory: Toward an algebraic geometry formulation.SIAM Journal on Applied Algebra and Geometry, 8(1):138–188, 2024

    2024

  49. [57]

    F. M. Faulstich, B. Sturmfels, and S. Sverrisd´ ottir. Al- gebraic varieties in quantum chemistry.Foundations of Computational Mathematics, pages 1–32, 2024

    2024

  50. [58]

    F. M. Faulstich. Recent mathematical advances in cou- pled cluster theory.International Journal of Quantum Chemistry, 124(13):e27437, 2024

    2024

  51. [59]

    Mih´ alka and J

    Z. Mih´ alka and J. Noga. Exploring alternative ap- proaches to improve the convergence pattern in solving the coupled-cluster equations.Molecular Physics, 121(11- 12):e2140084, 2023

    2023

  52. [60]

    F. M. Faulstich, V. Galgano, E. Neuhaus, and I. Portakal. On the coupled cluster doubles truncation variety of four electrons.arXiv:2602.16580, 2026

    2026

  53. [61]

    F. M. Faulstich, Y. Khoo, and K. Li. Augmented la- grangian method for coupled-cluster.arXiv:2403.16381, 2024

    2024 arXiv

  54. [62]

    F. M. Faulstich and S. Sverrisd´ ottir. Algebraic geometry for spin-adapted coupled cluster theory. arXiv:2601.16646, 2026

    2026

  55. [63]

    Sverrisd´ ottir and F

    S. Sverrisd´ ottir and F. M. Faulstich. Exploring ground and excited states via single reference coupled-cluster theory and algebraic geometry.Journal of Chemical The- ory and Computation, 20(19):8517–8528, 2024

    2024

  56. [64]

    Sverrisd´ ottir

    S. Sverrisd´ ottir. Algebraic varieties in second quantiza- tion.SIAM Journal on Applied Algebra and Geometry, 10(2):490–517, 2026

    2026

  57. [65]

    D. A. Matthews and J. F. Stanton. Accelerating the con- vergence of higher-order coupled cluster methods.The Journal of Chemical Physics, 143(20), 2015

    2015

  58. [66]

    E. F. Kjønstad, S. D. Folkestad, and H. Koch. Ac- celerated multimodel newton-type algorithms for faster convergence of ground and excited state coupled clus- ter equations.The Journal of Chemical Physics, 153(1), 2020

    2020

  59. [67]

    D. A. Matthews. Accelerating the convergence of higher-order coupled-cluster methods II: Coupled-cluster λequations and dynamic damping.Molecular Physics, 118(19-20):e1757774, 2020

    2020

  60. [68]

    C. Yang, J. Brabec, L. Veis, D. B. Williams-Young, and K. Kowalski. Solving coupled cluster equations by the Newton Krylov method.Frontiers in Chemistry, 8:590184, 2020

    2020

  61. [69]

    P. N. Brown and Y. Saad. Convergence theory of nonlin- ear Newton–Krylov algorithms.SIAM Journal on Opti- mization, 4(2):297–330, 1994

    1994

  62. [70]

    D. A. Knoll and D. E. Keyes. Jacobian-free Newton– Krylov methods: a survey of approaches and applica- tions.Journal of Computational Physics, 193(2):357–397, 2004

    2004

  63. [71]

    Saad and M

    Y. Saad and M. H. Schultz. Gmres: A generalized min- imal residual algorithm for solving nonsymmetric lin- ear systems.SIAM Journal on Scientific and Statistical Computing, 7(3):856–869, 1986

    1986

  64. [72]

    J. Hubbard. The description of collective motions in terms of many-body perturbation theory.Proceedings of the Royal Society A, 240(1223):539–560, 1957

    1957

  65. [73]

    N. M. Hugenholtz. Perturbation approach to the Fermi gas model of heavy nuclei.Physica, 23(1-5):533–545, 1957

    1957

  66. [74]

    Shavitt and R

    I. Shavitt and R. J. Bartlett.Many-Body Methods in Chemistry and Physics: MBPT and Coupled-Cluster Theory. Cambridge Molecular Science. Cambridge Uni- versity Press, 2009

    2009

  67. [75]

    Helgaker, P

    T. Helgaker, P. Jorgensen, and J. Olsen.Molecular electronic-structure theory. John Wiley & Sons, 2013

    2013

  68. [76]

    C. D. M. Mudiyanselage. ARCC.jl, 2026

    2026

  69. [77]

    S. C. Eisenstat and H. F. Walker. Choosing the forcing terms in an inexact Newton method.SIAM Journal on Scientific Computing, 17(1):16–32, 1996. 9 Appendix A: CCD Equations in an arbitrary gauge

    1996

  70. [78]

    Molecular orbital representation We first summarize the CCD equations in the molecular orbital (MO) basis following Ref. [74]. Letfdenote the Fock matrix,vthe electron-repulsion integrals (ERIs), andw ab ij = 2v ab ij −v ba ij the anti-symmetrized ERIs. We define the intermedi...

  71. [79]

    [25] and define the linear transformationU, which maps the MO basis onto the desired gauge

    Random gauge representation To derive the CCD equations in a random gauge, we follow Ref. [25] and define the linear transformationU, which maps the MO basis onto the desired gauge. Applying this transformation to the Fock matrix gives ¯f=U f UT ,(A6) where ¯fdenotes the Fock ...

  72. [80]

    We adopt the Eisenstat–Walker forcing term Ref

    Adaptive F orcing In Algorithm 1, we employ adaptive forcing to avoid oversolving the linearized problem within the GMRES loop. We adopt the Eisenstat–Walker forcing term Ref. [77], defined as ηk =γ ∥r(θ(k))∥ ∥r(θ(k−1))∥ α ,(F1) where the inner GMRES iteration terminates once ...

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.