Pith. sign in

REVIEW 4 major objections 3 minor 12 references

Fast and Accurate Charge Transfer Excitations via Nested Aufbau Suppressed Coupled Cluster

T0 review · 4 major / 3 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read This paper establishes that nesting a small coupled cluster treatment inside a newly derived Aufbau-suppressed second-order perturbation theory matches the accuracy of full Aufbau-suppressed coupled cluster for charge transfer excitations…

desk verdict Solid new excited-state method that delivers on its accuracy claims, with the O(1) flagged-orbital premise being the main generalization risk to watch. read the letter →

arxiv 2505.17299 v2 pith:TFACOUV6 submitted 2025-05-22 physics.chem-ph

classification physics.chem-ph
keywords chargetransferexcitationsAufbausuppressedcoupledclusterexcited-stateperturbationtheorynestedorbitalrelaxationexcitationenergiessizeconsistencyEOM-CCSDcomparison
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

This paper shows that charge-transfer excitation energies, which require both orbital relaxation and electron correlation, can be computed accurately without paying the full coupled-cluster cost. The authors derive an excited-state-specific second-order perturbation theory whose bottleneck is a non-iterative $N^{5}$ integral transformation, then nest a small Aufbau-suppressed coupled cluster (ASCC/PLASCC) calculation inside it, keeping the coupled-cluster treatment only for orbitals whose correlation is strongly affected by the excitation. On 130 valence and Rydberg states and 16 charge-transfer states, the nested method keeps typical errors below 0.1 eV and beats $N^{6}$-cost EOM-CCSD by an average of 0.25 eV for charge transfer. Because the refined orbital set is assumed to stay fixed in size as the molecule grows, the formal cost becomes non-iterative $N^{5}$ plus iterative $N^{3}$, which the authors demonstrate is enough to handle roughly 100-atom systems with explicit solvent on a single compute node.

What carries the argument

The load-bearing object is the nested ASCC/PT construction. Aufbau suppressed coupled cluster is a state-specific coupled cluster method whose exponential ansatz includes a deexcitation operator that builds post-excitation orbital relaxation into the reference. The new Aufbau-suppressed second-order perturbation theory mirrors MP2: it uses a block-diagonal zeroth-order Hamiltonian that singles out the primary hole and particle orbitals, so the coupled amplitude equations form small blocks (at most six equations for single-CSF states) and can be solved non-iteratively at O($N^{4}$), with the O($N^{5}$) integral transformation as the bottleneck. A Foster-Boys localization and orbital-matching step places ground and excited orbitals in a common local basis, and per-orbital correlation differences flag the orbitals whose correlation is strongly changed by the excitation. The iterative part of the calculation then updates only those O(1) amplitudes, which, by the connectedness of the coupled-cluster equations, keeps the iterative cost at O($N^{3}$), followed by one whole-system energy evaluation at O($N^{4}$).

What would settle it

Take a donor-bridge-acceptor molecule with a conjugated bridge and increase the bridge length while recomputing which orbitals the 0.005 eV correlation analysis flags for CC refinement; if the flagged orbital count grows with bridge length, or if the measured wall-time exponent for the iterative step grows from 3 toward 6, the central locality premise fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that the accuracy of Aufbau suppressed coupled cluster for charge-transfer excitations does not require a full-system coupled-cluster treatment: a low-order perturbation theory can decide where the coupled-cluster refinement is needed. The new perturbation theory is derived from ASCC by order analysis, with a zeroth-order Hamiltonian that keeps the amplitude equations small-block block-diagonal and non-iteratively solvable. After localizing and matching ground and excited orbitals, per-orbital correlation measures built from the PT amplitudes identify the few orbitals whose correlation changes most under the excitation. Solving the coupled-cluster residual equations only for those flagged amplitudes, freezing the rest at their PT values, and performing one final whole-system energy evaluation preserves the parent method's ~0.1 eV accuracy, improves on full ASCC in some valence and Rydberg cases, and matches CC3-quality behavior on a hydrogen-bonding charge-transfer surface. The result is a method that is more accurate than EOM-CCSD on charge transfer by 0.25 eV while having a dramatically lower asymptotic cost.

Load-bearing premise

The central assumption is that the number of orbitals whose correlation is strongly changed by a given excitation stays roughly constant as the molecule grows, so the coupled-cluster-refined orbital set does not scale with system size.

Editorial extensions

If this is right

  • Charge-transfer excitation energies in systems with around 100 atoms and explicit solvent can be computed on a single node with typical errors below 0.1 eV, a regime previously requiring much more expensive EOM-CCSD-level calculations.
  • Nested PLASCC produces excited-state potential energy surfaces for hydrogen-bonded charge-transfer systems within about 1 kcal/mol of CC3, while preserving the correct state character where EOM-CCSD and TD-DFT mix in spurious Rydberg character.
  • The orbital-selection procedure is automatic and physically interpretable: it flags donor and acceptor regions for charge transfer and hydrogen-bond-perturbed waters for solvated excitations, without user input about where the excitation is located.
  • Freezing the PT amplitudes preserves size consistency, extensivity, and intensivity, so the nested method can be applied to larger systems without introducing size-dependent errors.
  • The same nesting pattern should extend to higher-order non-iterative corrections to the ASCC energy, potentially improving accuracy further without returning to iterative N^6 cost.

Reading between the lines

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

  • Beyond the paper's tests, the same nesting pattern should transfer to core-excitation spectroscopy, where orbital relaxation is even stronger and system-size limits are severe.
  • The 0.005 eV screening threshold is an accuracy-cost dial; an adaptive threshold tied to the magnitude of the PT correlation correction would likely make the method more robust across different basis sets and state characters.
  • The per-orbital correlation-difference map could serve as a black-box diagnostic that reports where a given excitation changes electron correlation, useful not only for screening but for interpreting charge-transfer character.
  • Combining the nesting idea with local correlation or pair-natural-orbital techniques would likely cut the O(N^5) integral-transform prefactor and reduce memory further, extending the method to even larger systems.
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

4 major / 3 minor

Summary. The manuscript introduces a second-order Aufbau-suppressed perturbation theory (PT) derived from ASCC and uses its correlation contributions to identify orbitals that are strongly affected by an excitation. A small ASCC/PLASCC calculation is then nested inside the PT, with all other amplitudes frozen at the PT level. The authors report that on 130 QUEST valence/Rydberg states and 16 CT states, nested PLASCC matches full PLASCC accuracy (CT mean unsigned error near 0.1 eV, 0.25 eV better than EOM-CCSD), with formal cost reduced to non-iterative O(N^5) plus iterative O(N^3). Three larger tests (thiopropynal chains, solvated acetone, donor-bridge-acceptor) demonstrate physically intuitive orbital selection and wall-time scaling in an initial implementation.

Significance. If the cost and accuracy claims hold, this is a valuable step toward practical high-accuracy CT calculations with explicit environments: it preserves ASCC's orbital-relaxation capabilities while removing the iterative N^6 bottleneck. The benchmark references are independent (EOM-CCSDT/LR-CC3), the SI reports sensitivity to the nesting threshold, and the large-system tests are genuinely challenging. The main reservations concern the empirical basis for the O(1) flagged-orbital premise, the post-hoc threshold choice, and the undisclosed count of excluded states; these are load-bearing for the central scaling and accuracy claims.

major comments (4)
  1. [Main text, section introducing nesting; Fig. 4 and Fig. 5] The iterative O(N^3) cost claim is derived under the premise stated just before the nesting construction: 'the number of individual orbitals that are directly involved in or strongly affected by the excitation is not likely to grow with system size.' This premise is load-bearing: if the PT correlation screening flags a growing fraction of orbitals as system size increases, the nested CC active space grows, the O(1) residual argument collapses, and the iterative cost rises toward the full ASCCSD scaling. The evidence in Fig. 4 consists of only three large systems, and the only CT scaling test (Fig. 4C) uses a saturated alkane bridge that shows saturation of the flagged set after three carbons; conjugated bridges, delocalized CT states, or many explicitly correlated solvent molecules could behave differently. The authors should either prove a bound on the flagged set for a broader class of excitations or present scaling tests over a conjugated-bridge length series and over increasing solvent-shell size; without this, the advertised non-iterative N^5/iterative N^3 scaling is not yet supported for the full CT/environment domain claimed in the abstract.
  2. [SI S4, Fig. S2; Computational Methods paragraph on thresholds] The 0.005 eV nesting threshold used for all main-text benchmark statistics was selected after examining the QUEST benchmark results (SI S4, Fig. S2). Because the same 130-state set is then used to report the accuracy improvements, the reported MUEs are in-sample estimates with no independent validation; the choice of threshold is a form of model selection on the evaluation set. The authors should report the selection procedure explicitly and provide either a held-out validation set or cross-validated accuracy, or at least quantify the sensitivity of the CT-state statistics separately from the QUEST statistics, to establish that the headline accuracy is not an artifact of threshold tuning.
  3. [SI S4] The benchmark handling of non-converged states is not fully disclosed in the main text. SI S4 states that states were entirely removed if neither of the two ASCC solutions converged, and that some states are included with only one converged solution or with energies that stalled near the convergence criterion. The number and identity of removed states are not given, so it is impossible to assess whether the reported MUEs are biased by excluding difficult CT or valence cases. The authors should list all removed/reduced states, state how many there are per method, and show that the main conclusions are unchanged when including the stalled states or applying an alternative convergence criterion.
  4. [Main text, paragraph introducing nesting; Figs. 1 and 2] The orbital flagging is based on PT correlation measures, yet the PT itself is substantially inaccurate for the target class: the text reports PT errors of ~0.5 eV for CT states and >1 eV outliers for aromatic valence states. If the PT misassigns correlation differences, it can silently fail to flag orbitals that genuinely need CC treatment, and the nested method would inherit PT-level errors without any diagnostic. The paper currently provides no validation that the PT-selected orbital set coincides with the set that a full-CC sensitivity analysis would identify, e.g., by comparing flagged orbitals against those selected by comparing full ASCCSD amplitudes or by testing a case where PT is known to have a large error. A concrete diagnostic of flagging reliability would make the accuracy claim robust.
minor comments (3)
  1. [Abstract] The phrase 'typically below 0.1 eV on average' is imprecise; the underlying statistic is the mean unsigned error, so the wording should say, for example, 'a mean unsigned error below 0.1 eV'.
  2. [Figure 3] The legend text runs together ('EOM-CCSDTDDFT/ωB97X-DNested PLASCC3') and the y-axis label in panel (b) is printed as 'potential energy (kcal/mol)excitation energy error (eV)' with no separator; these need typesetting fixes.
  3. [Results and discussion, CT benchmark paragraph] The statement that 'for all but two states, nested PLASCC closely maintains PLASCC's accuracy' should identify which two states and report their errors, since outliers are important for a benchmark claim.

Circularity Check

1 steps flagged · score 2.0 of 10

No material circularity: the central nested-CC derivation and CT benchmarks are independent; only benchmark-informed threshold selection is a mild tuning issue.

  1. fitted input called prediction [Computational Methods (orbital energy threshold) and Supporting Information S4 (threshold selection) and S1 (Hamiltonian partitioning choice)]
    "Three different energetic thresholds for the nesting procedure were tested in this study: 0.0025 eV, 0.005 eV, and 0.01 eV. The accuracy for each of the three tested thresholds for the QUEST benchmark are summarized in Figure S2. Ultimately, we decided to report the results of the 0.005 eV threshold in the main text due to its balance of computational efficiency and accuracy... In order to maintain equivalence between the perturbation theories of the ground and excited states, this partitioning scheme was also utilized for the ground state..."

    The 0.005 eV nesting threshold is selected by inspecting error statistics on the same QUEST benchmark for which the paper reports 'errors typically below 0.1 eV'; the altered ground-state zeroth-order Hamiltonian is likewise justified by improved accuracy on that benchmark (Figure S1). These are benchmark-informed model choices, so the in-sample QUEST statistics are not parameter-free predictions. The circularity is mild: the nested CC energy is not defined as the screening criterion, threshold sensitivity is disclosed, and the headline CT accuracy is evaluated on an independent 16-state CT benchmark with EOM-CCSDT/LR-CC3 references, so the central claim does not reduce by construction to the tuning.

full rationale

The core method is self-contained against external references, not against its own inputs. The Aufbau-suppressed PT is derived from the ASCC similarity-transformed Hamiltonian by an order-by-order expansion, and the nested CC step solves genuine amplitude residuals for the flagged orbital set, followed by a whole-system CC energy evaluation. Accuracy is benchmarked against independent EOM-CCSDT and LR-CC3 values, not against quantities defined by the method, and the reported improvement over EOM-CCSD is measured on CT states outside the threshold-selection set. The iterative O(N^3) cost argument follows combinatorially from the asserted O(1) flagged-orbital set and the two-body nature of the transformed Hamiltonian; that assertion is a stated physical assumption ('not likely to grow with system size') tested on three large systems, including a saturated-bridge donor-acceptor CT chain showing saturation after three bridge carbons. This is a generalization risk and a possible correctness concern for delocalized excitations or extended environments, but it is not circular. Self-citations to the authors' earlier ASCC/PLASCC work supply motivation and prior accuracy claims, but the paper re-benchmarks the methods here, so they are not load-bearing. The only mildly circular element is the disclosed selection of the 0.005 eV nesting threshold and the benchmark-informed choice of ground-state Hamiltonian partitioning in the SI; these are model-selection issues, not equivalences, and they do not force the central CT result.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The ledger is light. One numerical threshold governs which orbitals receive CC treatment and is tuned on the benchmark with disclosed sensitivity. No physical entities are invented: the method reorganizes existing correlation treatments (ASCC, MP2-style PT, orbital localization) without postulating new particles, forces, or conserved quantities. The main uncharged assumptions are the locality of excitation-affected correlations and the adequacy of second-order PT for all non-flagged correlations.

free parameters (2)
  • Nesting orbital energy threshold = 0.005 eV; 0.0025 and 0.01 eV tested
    Orbitals whose PT correlation measure changes by more than this threshold between ground and excited states are refined at CC level. Main-text value chosen for 'balance of computational efficiency and accuracy' after seeing benchmark results (SI S4), so reported accuracy is for a benchmark-tuned choice.
  • ESMF reference CSF singular value cutoff = 0.2
    CSFs with singular value above 0.2 define the excited state reference (Section IV). Standard practice from prior ESMF work; not fitted to targets here.
assumptions (4)
  • domain assumption Only O(1) orbitals are strongly affected by an excitation, independent of system size.
    Load-bearing for the iterative O(N^3) cost; demonstrated by saturation of the flagged set in the donor-bridge-acceptor test and by localized flags in solvated acetone, but not proven for delocalized or multi-CSF excitations.
  • domain assumption Second-order Aufbau suppressed PT is adequate for all correlations not strongly affected by the excitation.
    The bare PT has MUEs of about 0.4-0.5 eV with outliers beyond 1 eV (Figures 1-2), so the screening must identify relevant orbitals even where PT energies are inaccurate; this is shown empirically, not analyzed in detail.
  • domain assumption ESMF provides a qualitatively correct orbital-relaxed single-CSF excited reference.
    Carried over from same-group prior work (refs 60-62); nested accuracy depends on it, and convergence failures are disclosed in SI S4.
  • standard math The PT amplitude equations decouple into small, system-size-independent blocks in a semicanonical basis.
    SI S2 argues the zeroth-order Hamiltonian's off-diagonal elements couple only within the small primary hole/particle space (largest blocks of 6 equations in the single-CSF case), enabling non-iterative O(N^4) solution; the algebra is summarized but not fully reproduced in the main text.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fast and Accurate Charge Transfer Excitations via Nested Aufbau Suppressed Coupled Cluster." pith.science (2026). https://pith.science/paper/TFACOUV6

@misc{pith2026250517299,
  author       = {Pith},
  title        = {Pith review of: Fast and Accurate Charge Transfer Excitations via Nested Aufbau Suppressed Coupled Cluster},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TFACOUV6}},
  note         = {Machine review of arXiv:2505.17299}
}
abstract

Modeling charge transfer well can require treating post-excitation orbital relaxations and handling medium to large molecules in realistic environments. By combining a state-specific correlation treatment with such orbital relaxations, Aufbau suppressed coupled cluster has proven accurate for charge transfer, but, like many coupled cluster methods, it struggles with large system sizes. We derive a low-cost Aufbau suppressed second order perturbation theory and show that, by nesting a small coupled cluster treatment inside of it, computational cost and scaling are reduced while accuracy is maintained. Formal asymptotic costs are dropped from iterative $N^6$ to non-iterative $N^5$ plus iterative $N^3$, and we test an initial implementation that can handle about 100 atoms and 800 orbitals on a single computational node. Charge transfer excitation energy errors are typically below 0.1 eV on average, with an average 0.25 eV improvement over $N^6$-cost equation of motion coupled cluster with singles and doubles.

Figures

Figures reproduced from arXiv: 2505.17299 by the authors.

Figure 1
Figure 1. FIG. 1. Excitation energy error distributions for various methods on 130 valence and Rydberg single-CSF singlet states from [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Mean unsigned errors (MUEs) for various methods [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: FIG. 5. Wall time measured for Aufbau Suppressed PT (top), [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 4
Figure 4. Figure 4: FIG. 4. A gas phase, simple valence excitation (A), a sim [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

12 extracted references · 12 canonical work pages

  1. [5]

    For an explicit solvation example, we examine in Fig- ure 4B the n → π∗ transition in solvated acetone. The explicit waters play a non-innocent role, as the hydro- gen bonding that initially stabilizes the ground state is disrupted when one of the hydrogen-bond-accepting lone pair electrons is transferred into the π∗ orbital by the excitation. Crucially, ...

  2. [153]

    F.; Bartlett, R

    8Stanton, J. F.; Bartlett, R. J. The equation of motion coupled- cluster method. A systematic biorthogonal approach to molecu- lar excitation energies, transition probabilities, and excited state properties. The Journal of chemical physics 1993, 98, 7029–

  3. [300]

    K.; Gagliardi, L.; Pedersen, T

    38Aquilante, F.; Todorova, T. K.; Gagliardi, L.; Pedersen, T. B.; Roos, B. O. Systematic truncation of the virtual space in mul- ticonfigurational perturbation theory. The Journal of chemical physics 2009, 131, 034113. 39Segarra-Mart ´ ı, J.; Garavelli, M.; Aquilante, F. Multiconfigura- tional second-order perturbation theory with frozen natural or- bital...

  4. [462]

    L.; Domcke, W

    10Sobolewski, A. L.; Domcke, W. Ab initio study of the excited-state coupled electron–proton-transfer process in the 2- aminopyridine dimer. Chemical Physics 2003, 294, 73–83. 11Dreuw, A.; Weisman, J. L.; Head-Gordon, M. Long-range charge- transfer excited states in time-dependent density functional the- ory require non-local exchange. The Journal of chem...

  5. [997]

    Time-dependent density functional theory: Past, present, and future

    5Burke, K.; Werschnik, J.; Gross, E. Time-dependent density functional theory: Past, present, and future. The Journal of chemical physics 2005, 123, 062206. 6Casida, M. E.; Huix-Rotllant, M. Progress in time-dependent density-functional theory. Annual review of physical chemistry 2012, 63, 287–323. 7Rowe, D. Equations-of-motion method and the extended she...

  6. [1474]

    49Martin, R. L. Natural transition orbitals. The Journal of chem- ical physics 2003, 118, 4775–4777. 50Myhre, R. H.; S´ anchez de Mer´ as, A. M.; Koch, H. Multi-level coupled cluster theory. The Journal of chemical physics2014, 141, 224105. 51Nooijen, M.; Bartlett, R. J. Similarity transformed equation-of- motion coupled-cluster theory: Details, examples,...

  7. [1540]

    S.; Neuscamman, E

    62Hardikar, T. S.; Neuscamman, E. A self-consistent field formula- tion of excited state mean field theory. The Journal of chemical physics 2020, 153, 164108. 63Rubensson, E. H.; Sa lek, P. Systematic sparse matrix error con- trol for linear scaling electronic structure calculations. Journal of computational chemistry2005, 26, 1628–1637. 64Koch, H.; S´ an...

  8. [2009]

    Combining coupled cluster and perturbation theory

    24Nooijen, M. Combining coupled cluster and perturbation theory. The Journal of Chemical Physics1999, 111, 10815–10826. 25Bochevarov, A. D.; Sherrill, C. D. Hybrid correlation mod- els based on active-space partitioning: Correcting second-order Møller–Plesset perturbation theory for bond-breaking reactions. The Journal of Chemical Physics2005, 122, 234110...

Show all 12 references
  1. [2025]

    Modeling charge transfer well can require treating post-excitation orbital relaxations and handling medium to large molecules in realistic environments. By combining a state-specific correlation treatment with such orbital relaxations, Aufbau suppressed coupled cluster has pro...

  2. [2395]

    B.; Schirmer, J

    30Trofimov, A. B.; Schirmer, J. An efficient polarization propaga- tor approach to valence electron excitation spectra. Journal of Physics B: Atomic, Molecular and Optical Physics 1995, 28,

  3. [3784]

    40Kumar, A.; Crawford, T. D. Frozen virtual natural orbitals for coupled-cluster linear-response theory. The Journal of Physical 9 Chemistry A 2017, 121, 708–716. 41Mukhopadhyay, T.; Jangid, B.; Dutta, A. K. State-specific frozen natural orbital for reduced-cost algebraic diag...

  4. [7039]

    9Krylov, A. I. Equation-of-motion coupled-cluster methods for open-shell and electronically excited species: The hitchhiker’s guide to Fock space. Annu. Rev. Phys. Chem.2008, 59, 433–

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.