Oscillator Strengths and Transition Dipole Moments from a Simplified Equation-of-Motion Coupled Cluster Formalism within the Frozen-Pair Approximation

This paper derives working equations for transition density matrices, dipole moments, and oscillator strengths within the EOM-frozen-pair coupled-cluster framework (EOM-fpCCSD and EOM-ptCCSD) using approximations that avoid solving $\Lambda$ equations and calculating left eigenvectors, demonstrating that these models yield improved excited-state properties compared to standard EOM-CCSD.

The Gist

Imagine you are trying to predict how a molecule will react when hit by light. In the world of chemistry, this is like trying to guess the color a new paint will be before you even mix it. To do this accurately, scientists use complex math called "Coupled Cluster" theory. It's the gold standard for accuracy, but it's also incredibly expensive and slow—like trying to solve a Rubik's cube while running a marathon.

This paper introduces a new, faster way to solve that same puzzle, specifically for molecules that are "stuck" in a difficult state (where electrons are paired up in a tricky way). Here is the breakdown of what they did, using simple analogies:

1. The Problem: The "Perfect" Recipe is Too Expensive

Standard methods (called EOM-CCSD) are like a master chef who tastes every single ingredient individually to get the perfect flavor. It works great, but it takes forever. For large molecules, this method is too slow to be useful for everyday experiments.

On the other hand, cheaper methods (like TD-DFT) are like using a food processor: fast, but sometimes they mash up the ingredients wrong, giving you a bad taste (inaccurate results), especially for complex dishes.

2. The Solution: The "Frozen Pair" Shortcut

The authors developed a new method called EOM-fpCCSD and EOM-ptCCSD.

• The Analogy: Imagine a dance floor where couples (electron pairs) are dancing. In the standard method, you have to track every single dancer's footwork perfectly. In this new "Frozen Pair" method, the authors say, "Let's lock the couples' hands together and just watch how the pairs move as a unit."
• By treating these pairs as a single, frozen unit, they can ignore a massive amount of unnecessary math. This makes the calculation much faster without losing the "perfect chef" level of accuracy.

3. The New Trick: Guessing the "Left" Side

To calculate how bright a molecule will glow (Oscillator Strengths) or how it absorbs light (Transition Dipole Moments), you usually need to solve two sides of an equation: the "Right" side (what happens) and the "Left" side (what went into it).

• The Old Way: Calculating the "Left" side is like trying to rewind a movie frame-by-frame to see exactly how the actors got into position. It's slow and computationally heavy.
• The New Way: The authors used a clever mathematical shortcut (a "matrix inverse approximation"). Instead of rewinding the movie, they looked at the final frame and used a smart guess to reconstruct the beginning.
• The Result: They avoided the heavy lifting of solving the "Left" side equations entirely, saving even more time.

4. The Test: Water and Furan

To see if their new shortcut worked, they tested it on two molecules: Water (simple) and Furan (a ring-shaped molecule often found in organic materials).

• They compared their "Frozen Pair" results against the "Gold Standard" (LR-CCSD).
• The Outcome: Their new method was almost identical to the Gold Standard. In fact, for some difficult types of excited states (where electrons are doubly excited), their method was actually better and more stable than the standard method.
• They also tested two different "maps" (orbital bases) to navigate the molecule: one standard map (HF) and one optimized map (pCCD). They found that their new method worked just as well on both maps, meaning it's very flexible.

5. The Bottom Line

The paper claims that they have successfully built a "fast lane" for calculating how molecules interact with light.

• Speed: It avoids the most expensive parts of the calculation (solving the "Left" equations and the "Lambda" equations).
• Accuracy: It produces results that are very close to the most accurate methods available today.
• Reliability: It works well even when the standard methods struggle to converge (get stuck).

In short, they found a way to get the high-quality results of a supercomputer using a much more efficient recipe, making it possible to study complex electronic materials without waiting days for the math to finish.

Technical Summary

Technical Summary: Excited State Properties from Equation-of-Motion Frozen Pair Coupled Cluster Methods

Problem Statement
Accurately describing electronic excited states and their properties, such as oscillator strengths (OSs) and transition dipole moments (TDMs), for large molecules remains a significant challenge for current quantum-chemical methods. While time-dependent density functional theory (TD-DFT) offers computational efficiency, it often suffers from functional dependence, incorrect descriptions of bond alternation in conjugated systems, and failures in modeling charge-transfer, Rydberg, and double-excitation states. Conversely, standard wave-function-based methods like Equation-of-Motion Coupled Cluster (EOM-CCSD) provide high accuracy and systematic improvability but are computationally prohibitive for routine application to large organic electronic materials due to their steep scaling. There is a critical need for methods that balance computational cost with the high accuracy of coupled-cluster theory.

Methodology
This work derives working equations for transition density matrices within the Equation-of-Motion Frozen-Pair Coupled Cluster (EOM-fpCC) framework, specifically focusing on the EOM-fpCCSD and EOM-ptCCSD models. The approach builds upon the pair Coupled Cluster Doubles (pCCD) ansatz, which restricts the cluster operator to the seniority-zero sector (electron-pair excitations), and extends it to excited states.

Key methodological features include:

• Ground State Reference: The methods utilize a pCCD ground-state wavefunction. In the EOM-ptCCSD model, the cluster operator is partitioned into internal (pCCD pair) and external (broken-pair) components. In the EOM-fpCCSD model, the projection equations are enforced for each excitation manifold, effectively decoupling seniority-zero and seniority-non-zero sectors.
• Approximate Transition Moments: To avoid the computationally expensive solution of the coupled-cluster $\Lambda$ equations and the explicit calculation of left EOM eigenvectors, the authors employ two approximations:
  1. Expectation Value Approximation: The left eigenvector operator $\hat{\Lambda}$ is approximated by the cluster operator $\hat{T}$ ($\hat{\Lambda}^\dagger \approx \hat{T}$).
  2. Matrix Inverse Approximation: The left eigenvectors ($L^\dagger$) are approximated using the right eigenvectors ($R$) via the relation $L^\dagger \approx (R^\dagger R)^{-1}R$.
• Orbital Basis: Calculations were performed using both canonical Hartree–Fock (HF) orbitals and natural pCCD orbitals to assess the dependence of excited-state properties on the orbital basis choice.

Key Contributions

• Derivation of Working Equations: The authors present the formal derivation for transition dipole moments and oscillator strengths within the EOM-fpCCSD and EOM-ptCCSD formalisms.
• Computational Efficiency: The proposed scheme eliminates the need for solving $\Lambda$ equations and calculating left eigenvectors explicitly, significantly reducing computational overhead compared to standard linear-response (LR) approaches.
• Benchmarking: The study benchmarks the newly developed models against the standard LR-CCSD method for water and furan across multiple basis sets (cc-pVDZ, cc-pVTZ, cc-pVQZ, and aug-cc-pVTZ).

Results

• Accuracy: The results demonstrate that EOM-fpCCSD and EOM-ptCCSD models provide a description of excited-state properties that is improved compared to the standard EOM-CCSD variant. Specifically, EOM-fpCCSD yields the smallest percentage errors relative to LR-CCSD reference data for both dipole strengths (DS) and oscillator strengths (OS).
• Orbital Independence: The choice of orbital basis (canonical HF vs. natural pCCD) was found to have a negligible influence on the excited-state description within the EOM-fpCCSD and EOM-ptCCSD models. This contrasts with simpler LR-pCCD models and suggests these methods are robust even when the HF reference is unstable.
• Error Metrics: Overall errors in TDMs and OSs generally remained below 5% in most cases. The methods showed no significant dependence on basis set size or the inclusion of diffuse functions.
• Size-Intensivity: The authors note that, by design, the approximate excited-state properties are not size-intensive due to disconnected terms in the right transition density matrices. However, for the small molecules investigated (water and furan), this size-intensivity problem was negligible, as LR-CCSD and standard EOM-CCSD yielded identical properties.

Significance and Claims
The paper claims that the recently introduced single-reference coupled-cluster approaches (EOM-fpCCSD and EOM-ptCCSD) can be reliably extended to study excited-state spectra and properties of complex electronic structures. The work demonstrates that these models offer a favorable trade-off between cost and accuracy, retaining the systematic improvability of coupled-cluster theory while significantly reducing computational scaling. The ability to use natural pCCD orbitals without loss of accuracy opens the door to modeling excited-state properties in complex systems where the standard HF reference may be unreliable. The authors conclude that these methods are particularly suitable for describing states with substantial double-excitation character and charge-transfer states, which are often challenging for other methods.