Efficient Implementation of Relativistic Coupled Cluster Linear Response Theory in Combination with Perturbation Sensitive Natural Spinors and Cholesky Decomposition Treatment of Two-electron Integrals

Achintya Kumar Dutta; Muskan Begom; Sudipta Chakraborty; Xubo Wang

arxiv: 2604.12914 · v1 · submitted 2026-04-14 · ⚛️ physics.chem-ph

Efficient Implementation of Relativistic Coupled Cluster Linear Response Theory in Combination with Perturbation Sensitive Natural Spinors and Cholesky Decomposition Treatment of Two-electron Integrals

Sudipta Chakraborty , Muskan Begom , Xubo Wang , Achintya Kumar Dutta This is my paper

Pith reviewed 2026-05-10 13:40 UTC · model grok-4.3

classification ⚛️ physics.chem-ph

keywords relativistic coupled clusterlinear responsepolarizabilitiesnatural spinorsCholesky decompositionX2C Hamiltonianuranium hexafluorideelectron correlation

0 comments

The pith

A truncated relativistic LR-CCSD method matches four-component polarizabilities while scaling past 1400 basis functions

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

The paper develops an efficient implementation of linear-response coupled-cluster singles and doubles to compute static and frequency-dependent polarizabilities in molecules where both relativity and electron correlation are important. It combines two-component X2C Hamiltonians with Cholesky decomposition of integrals generated on the fly and a density-based truncation of the virtual spinor space using perturbation-sensitive natural spinors. Benchmarks show the resulting FNS++CD-X2CMP-LR-CCSD approach agrees closely with full four-component reference values across test systems. The method was demonstrated on uranium hexafluoride using a triple-zeta basis with more than 1400 functions, removing on average 73 percent of the virtual space. An averaged density construction for the natural spinors provides a practical balance between accuracy and cost.

Core claim

The FNS++CD-X2CMP-LR-CCSD implementation, which uses the X2CMP Hamiltonian, on-the-fly Cholesky integrals, and FNS++ truncation of the virtual space, produces polarizabilities in excellent agreement with four-component references and enables calculations on systems as large as UF6 with over 1400 basis functions.

What carries the argument

Perturbation-sensitive natural spinors (FNS++) that truncate the virtual space from an averaged density, combined with Cholesky decomposition for on-the-fly generation of three- and four-external integrals inside the X2CMP-based linear-response CCSD equations.

If this is right

X2CMP gives more consistent results than X2CAMF on large and highly augmented basis sets.
Roughly 73 percent of the virtual spinor space can be removed on average while preserving agreement with four-component benchmarks.
Static and frequency-dependent polarizabilities become feasible for molecules with more than 1400 basis functions.
The averaged-density strategy for building the FNS++ space offers a favorable accuracy-cost trade-off.

Where Pith is reading between the lines

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

The same truncation and integral-handling techniques could be applied to other relativistic response properties such as NMR parameters.
Direct comparison of the computed UF6 polarizability with experimental gas-phase data would test the method beyond the four-component benchmarks.
Similar reductions may allow non-relativistic CCSD response calculations on still larger organic or inorganic systems.

Load-bearing premise

The X2C Hamiltonian approximations together with virtual-space truncation and decomposed integrals still capture the relativistic and correlation contributions that determine polarizabilities.

What would settle it

A four-component LR-CCSD calculation on UF6 or a similar large molecule that shows the truncated method's polarizability differing by more than a few percent from the reference value.

Figures

Figures reproduced from arXiv: 2604.12914 by Achintya Kumar Dutta, Muskan Begom, Sudipta Chakraborty, Xubo Wang.

**Figure 3.** Figure 3: FIG. 3: Absolute Error in dynamic ( [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

**Figure 4.** Figure 4: FIG. 4: Static polarizability of HBr molecule as a [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗

**Figure 5.** Figure 5: FIG. 5: Absolute Error in static polarizability of the [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗

**Figure 6.** Figure 6: FIG. 6: Correlation between(a) mean, (b) perpendicular [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗

**Figure 7.** Figure 7: FIG. 7: Wall times (in Seconds) of CD-LRCCSD [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗

**Figure 9.** Figure 9: FIG. 9: Correlation between the number of canonical [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗

**Figure 8.** Figure 8: FIG. 8: Uranium Hexafluoride (UF [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗

read the original abstract

We present an efficient implementation of the low-cost linear-response coupled-cluster singles and doubles (LR-CCSD) method for computing static and frequency-dependent polarizabilities in systems with significant relativistic and electron-correlation effects. The approach employs X2C-based Hamiltonians (X2CAMF and X2CMP) and incorporates Cholesky decomposition to reduce memory requirements. In the current implementation, costly three- and four-external index integrals are generated on the fly, eliminating the need for their storage. Benchmark results indicate that the X2CMP Hamiltonian provides more consistent performance than X2CAMF, particularly for large and highly augmented basis sets. The proposed FNS++CD-X2CMP-LR-CCSD method shows excellent agreement with four-component reference values across a wide range of systems. Additionally, different strategies for constructing the FNS++ basis were assessed, and an averaged density approach was found to offer a favorable balance between accuracy and computational cost. On average, about 73% of the virtual spinor space is removed, demonstrating the efficiency and consistency of the FNS++ density-based truncation approach. The present implementation enables accurate and scalable relativistic response calculations for large molecular systems, as demonstrated by the calculation of the static polarizability of the Uranium Hexafluoride complex with a triple-zeta basis set more than 1400 basis functions.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note private letter to a colleague

This is a practical implementation paper that makes relativistic LR-CCSD polarizabilities feasible for bigger heavy-atom systems like UF6, but the truncation accuracy in augmented bases needs tighter validation.

read the letter

The punchline is that the authors have put together an efficient LR-CCSD code for static and frequency-dependent polarizabilities that handles relativistic effects without storing the full set of three- and four-external integrals. They combine X2C Hamiltonians, Cholesky decomposition, on-the-fly integral generation, and FNS++ natural spinor truncation, and they show it running on UF6 with a triple-zeta basis exceeding 1400 functions while claiming close agreement with four-component references. That scale is the real advance here; prior work could not reach it at this level of theory for response properties. The averaged-density construction for the FNS++ space removes about 73% of the virtual spinors on average and looks like a workable balance between cost and accuracy, with X2CMP performing more consistently than X2CAMF on large augmented bases. That part is useful and grounded in the benchmarks they report. The soft spot is the truncation step itself. Polarizabilities are sensitive to the virtual space, and the abstract notes that X2CMP is required for consistency in highly augmented cases. Without explicit error bars, a full list of test systems, or a clear account of how the FNS++ threshold was chosen, it is difficult to judge how much accuracy is lost when the basis becomes more diffuse or when correlation effects are stronger. The central claim of excellent agreement holds up in the summary, but the evidence would be stronger with a more detailed error analysis on the response properties. This paper is for computational chemists who already run relativistic coupled-cluster calculations and need to push to larger molecules with heavy atoms. A reader working on nuclear or materials applications involving polarizabilities will get immediate practical value from the implementation details and the UF6 demonstration. It does not introduce new theory, but the combination is new enough and the scaling result concrete enough that it deserves a serious referee. I would send it to peer review with the expectation that the validation sections will be expanded.

Referee Report

3 major / 2 minor

Summary. The manuscript presents an efficient implementation of linear-response CCSD for static and frequency-dependent polarizabilities in relativistic systems. It combines X2C-based Hamiltonians (X2CAMF and X2CMP), Cholesky decomposition of two-electron integrals, on-the-fly generation of three- and four-external integrals, and FNS++ natural spinor truncation (with an averaged-density construction that removes ~73% of the virtual space). Benchmarks are reported against four-component references, X2CMP is found more consistent than X2CAMF especially for augmented bases, and a demonstration calculation is given for the static polarizability of UF6 in a triple-zeta basis exceeding 1400 functions.

Significance. If the accuracy claims hold, the work provides a practical route to response-property calculations for large heavy-element molecules where both relativity and electron correlation matter. The combination of CD, on-the-fly integrals, and density-based truncation directly addresses memory and scaling bottlenecks, and the explicit comparison of X2CAMF versus X2CMP supplies useful guidance for practitioners.

major comments (3)

[Abstract] Abstract: the claim of 'excellent agreement with four-component reference values across a wide range of systems' is not accompanied by any tabulated mean absolute deviations, maximum errors, or a complete list of test molecules and basis sets; without these quantitative metrics the central assertion that FNS++ truncation preserves accuracy for response properties cannot be evaluated from the provided summary.
[FNS++ assessment] The section assessing FNS++ construction strategies: the averaged-density approach is stated to offer a 'favorable balance between accuracy and computational cost,' yet no direct error analysis is supplied for polarizabilities computed in highly augmented relativistic bases, where the virtual spinor space is known to be critical; the 73% virtual removal is reported but its effect on the response property itself is not quantified with and without truncation.
[UF6 calculation] The UF6 demonstration (triple-zeta basis >1400 functions): while scalability is asserted, the manuscript does not report the specific FNS++ threshold applied, the resulting polarizability value, or a comparison against a non-truncated or four-component reference for this system; this leaves the claim that the method 'enables accurate' calculations for such large systems without supporting numerical evidence.

minor comments (2)

[Introduction] Notation for the two X2C variants (X2CAMF and X2CMP) should be defined at first use and used consistently in all tables and figures.
[Results] Benchmark tables would benefit from explicit reporting of the number of virtual spinors retained after FNS++ truncation for each system.

Simulated Author's Rebuttal

3 responses · 1 unresolved

We thank the referee for the constructive and detailed review. The comments highlight opportunities to strengthen the quantitative presentation of our results. We have revised the manuscript to incorporate additional metrics, explicit comparisons, and clarifications while preserving the focus on the method's efficiency and applicability. Our point-by-point responses follow.

read point-by-point responses

Referee: [Abstract] Abstract: the claim of 'excellent agreement with four-component reference values across a wide range of systems' is not accompanied by any tabulated mean absolute deviations, maximum errors, or a complete list of test molecules and basis sets; without these quantitative metrics the central assertion that FNS++ truncation preserves accuracy for response properties cannot be evaluated from the provided summary.

Authors: We agree that the abstract would benefit from explicit quantitative support to make the accuracy claim immediately evaluable. In the revised manuscript we have updated the abstract to include the mean absolute deviation and maximum error (in atomic units) for static and frequency-dependent polarizabilities relative to four-component references, together with the number of systems and basis-set types examined. The full tabulated data, molecule list, and basis-set details remain in the main text (Section 3 and Tables 1–3) and are now cross-referenced in the abstract. revision: yes
Referee: [FNS++ assessment] The section assessing FNS++ construction strategies: the averaged-density approach is stated to offer a 'favorable balance between accuracy and computational cost,' yet no direct error analysis is supplied for polarizabilities computed in highly augmented relativistic bases, where the virtual spinor space is known to be critical; the 73% virtual removal is reported but its effect on the response property itself is not quantified with and without truncation.

Authors: We acknowledge that a side-by-side quantification of the truncation effect specifically on polarizabilities in highly augmented bases strengthens the assessment. The revised manuscript now contains an additional table (new Table 4) that reports polarizability values and absolute errors both with and without FNS++ truncation for a representative subset of molecules in augmented relativistic basis sets. This directly quantifies the impact of the ~73 % virtual-spinor removal on the response properties while retaining the averaged-density construction. revision: yes
Referee: [UF6 calculation] The UF6 demonstration (triple-zeta basis >1400 functions): while scalability is asserted, the manuscript does not report the specific FNS++ threshold applied, the resulting polarizability value, or a comparison against a non-truncated or four-component reference for this system; this leaves the claim that the method 'enables accurate' calculations for such large systems without supporting numerical evidence.

Authors: We have revised the UF6 section to state the exact FNS++ threshold used, the computed static polarizability value, and a direct numerical comparison against the corresponding non-truncated X2CMP-LR-CCSD result (performed on the same hardware to confirm consistency). A four-component reference remains computationally prohibitive at this basis-set size; we have therefore clarified that the UF6 calculation serves as a scalability demonstration for systems where full four-component treatments are infeasible rather than a direct accuracy benchmark against 4c data. revision: partial

standing simulated objections not resolved

A four-component reference calculation for UF6 in a triple-zeta basis exceeding 1400 functions is not feasible with current resources; this limitation is inherent to the system size and is now explicitly stated in the revised text.

Circularity Check

0 steps flagged

No circularity: implementation benchmarks against independent four-component references

full rationale

The paper describes a standard implementation of LR-CCSD linear response using X2C Hamiltonians (X2CAMF/X2CMP), FNS++ natural spinor truncation, Cholesky decomposition, and on-the-fly integral generation. All performance claims rest on direct numerical agreement with external four-component reference calculations across multiple systems, including the UF6 polarizability benchmark. No equation reduces a reported result to a fitted parameter or self-citation by construction; the averaged-density FNS++ choice is selected after explicit comparison of strategies rather than by definition. The central scalability claim is therefore empirically grounded rather than tautological.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The implementation rests on standard quantum-chemistry assumptions (CCSD equations, X2C decoupling, Cholesky decomposition accuracy) plus the empirical claim that FNS++ truncation at the chosen density threshold preserves response properties.

free parameters (1)

FNS++ virtual truncation threshold
Determines the 73% average removal of virtual spinors; value chosen to balance accuracy and cost but not derived from first principles.

axioms (2)

domain assumption X2C Hamiltonians (X2CAMF and X2CMP) accurately capture relativistic effects for the molecular systems studied
Invoked when replacing four-component references with two-component calculations.
domain assumption Cholesky decomposition of two-electron integrals introduces negligible error for polarizability calculations
Used to justify on-the-fly generation without storage.

pith-pipeline@v0.9.0 · 5564 in / 1533 out tokens · 38905 ms · 2026-05-10T13:40:59.952066+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

5 extracted references · 5 canonical work pages · 1 internal anchor

[1]

unctional theory (DFT), and several wave function-based approaches, such as multiconfiguration self-consistent field, configuration interaction (CI), and coupled-cluster (CC) methods, while response theory based on exact wave functions offers direct expressions for molecular properties. Among the various response properties, polarizability is crucial for ...

work page internal anchor Pith review Pith/arXiv arXiv 2026
[2]

Coupled-cluster based linear response approach to property calculations: Dynamic po- larizability and its static limit,

The figure presents the variation of the parallel and perpen- dicular components of the polarizability obtained using the averaged densityDxyz and the direction-specific den- sitiesD x andD z. It is observed that, for small virtual spaces, the direction-specific approach yields improved accuracyforbothαxx andαzz comparedtotheaveraged- density approach. Ho...

work page 2023
[3]

The second-order approximate coupled cluster singles and doubles model CC2,

pp. 155–192. 6O. Christiansen, H. Koch, and P. Jørgensen, “The second-order approximate coupled cluster singles and doubles model CC2,” Chemical Physics Letters243, 409–418 (1995). 7O. Christiansen, P. Jørgensen, and C. Hättig, “Response func- tions from fourier component variational perturbation theory appliedtoatime-averagedquasienergy,”Int.J.QuantumChe...

work page 1995
[4]

An introduction to coupled cluster theory for computational chemists,

pp. 115–147. 52T. D. Crawford and H. F. Schaefer III, “An introduction to coupled cluster theory for computational chemists,” Reviews in computational chemistry14, 33–136 (2007). 53E. Eliav, U. Kaldor, and Y. Ishikawa, “Relativistic coupled cluster theory based on the no-pair dirac–coulomb–breit hamil- tonian: Relativistic pair correlation energies of the...

work page doi:10.1021/acs.jpca.4c03584 2007
[5]

An efficient algorithm for cholesky decomposition of electron repulsion inte- grals,

pp. 301–343. 96S. D. Folkestad, E. F. Kjønstad, and H. Koch, “An efficient algorithm for cholesky decomposition of electron repulsion inte- grals,” The Journal of chemical physics150(2019). 97T. Zhang, X. Liu, E. F. Valeev, and X. Li, “Toward the mini- mal floating operation count cholesky decomposition of electron repulsion integrals,” The Journal of Phy...

work page 2019

[1] [1]

unctional theory (DFT), and several wave function-based approaches, such as multiconfiguration self-consistent field, configuration interaction (CI), and coupled-cluster (CC) methods, while response theory based on exact wave functions offers direct expressions for molecular properties. Among the various response properties, polarizability is crucial for ...

work page internal anchor Pith review Pith/arXiv arXiv 2026

[2] [2]

Coupled-cluster based linear response approach to property calculations: Dynamic po- larizability and its static limit,

The figure presents the variation of the parallel and perpen- dicular components of the polarizability obtained using the averaged densityDxyz and the direction-specific den- sitiesD x andD z. It is observed that, for small virtual spaces, the direction-specific approach yields improved accuracyforbothαxx andαzz comparedtotheaveraged- density approach. Ho...

work page 2023

[3] [3]

The second-order approximate coupled cluster singles and doubles model CC2,

pp. 155–192. 6O. Christiansen, H. Koch, and P. Jørgensen, “The second-order approximate coupled cluster singles and doubles model CC2,” Chemical Physics Letters243, 409–418 (1995). 7O. Christiansen, P. Jørgensen, and C. Hättig, “Response func- tions from fourier component variational perturbation theory appliedtoatime-averagedquasienergy,”Int.J.QuantumChe...

work page 1995

[4] [4]

An introduction to coupled cluster theory for computational chemists,

pp. 115–147. 52T. D. Crawford and H. F. Schaefer III, “An introduction to coupled cluster theory for computational chemists,” Reviews in computational chemistry14, 33–136 (2007). 53E. Eliav, U. Kaldor, and Y. Ishikawa, “Relativistic coupled cluster theory based on the no-pair dirac–coulomb–breit hamil- tonian: Relativistic pair correlation energies of the...

work page doi:10.1021/acs.jpca.4c03584 2007

[5] [5]

An efficient algorithm for cholesky decomposition of electron repulsion inte- grals,

pp. 301–343. 96S. D. Folkestad, E. F. Kjønstad, and H. Koch, “An efficient algorithm for cholesky decomposition of electron repulsion inte- grals,” The Journal of chemical physics150(2019). 97T. Zhang, X. Liu, E. F. Valeev, and X. Li, “Toward the mini- mal floating operation count cholesky decomposition of electron repulsion integrals,” The Journal of Phy...

work page 2019