Vibrational Coupled Cluster Computations in Polyspherical Coordinates with the Exact Analytical Kinetic Energy Operator

Emil Lund Klinting, David Lauvergnat, Ove ChristiansenJournal of Chemical Theory and Computation 2020 · DOI

TL;DR

This paper introduces polyspherical coordinates for calculating molecular vibrations, offering a more efficient and accurate approach than traditional rectilinear coordinates, especially for complex molecular motions. The authors demonstrate that this new method accelerates convergence in vibrational calculations and handles challenging molecules like hydrogen peroxide effectively.

What Problem Does This Paper Address?

Calculating the vibrational behavior of molecules is crucial for understanding their properties and reactions. This involves solving the nuclear Schrödinger equation, which requires defining coordinates to describe molecular geometry and the kinetic energy operator. Traditional methods often use rectilinear coordinates (like normal coordinates), which can become computationally expensive and less accurate for molecules with large-amplitude motions or complex internal dynamics. This paper addresses the challenge of efficiently and accurately calculating anharmonic vibrational states for such molecules by introducing and testing a curvilinear coordinate system, polyspherical coordinates (PsCs).

Methods

The core of this work lies in the application of polyspherical coordinates (PsCs) to vibrational coupled cluster (VCC) theory. PsCs are a type of curvilinear coordinate that naturally describe internal molecular motion, including angles and distances. The authors utilize the TANA program to generate analytical expressions for the nuclear kinetic energy operator (KEO) in PsCs. They also employ the Adaptive Density-Guided Approach (ADGA) for constructing potential energy surfaces (PESs) and the Vibrational Coupled Cluster (VCC) model for calculating vibrational excitation energies. The key innovation is the integration of PsCs into these established computational frameworks, allowing for a more natural and efficient representation of molecular vibrations. They compare their results with calculations using traditional rectilinear coordinates: normal coordinates (NCs) and hybrid optimized and localized coordinates (HOLCs).

Datasets

Not discussed in the paper. The paper focuses on theoretical methodology and its application to specific molecules (formaldehyde and hydrogen peroxide), rather than using pre-existing datasets for training.

Key Insights and Ideas

  • PsCs offer a more natural description of molecular motion: Unlike rectilinear coordinates, PsCs inherently capture the angular and distance relationships within a molecule, which is particularly beneficial for describing large-amplitude motions and complex internal dynamics.
  • Accelerated convergence with PsCs: The use of PsCs leads to a faster convergence of vibrational calculations, meaning fewer computational steps are needed to reach accurate results, especially when using truncated methods.
  • Reduced computational cost: By improving convergence, PsCs can significantly reduce the computational effort required for accurate vibrational calculations, making it feasible to study larger and more complex molecules.
  • Effective for challenging molecules: The study demonstrates that PsCs are highly effective for molecules with non-trivial internal motions, such as the torsional motion in hydrogen peroxide, where rectilinear coordinates can struggle.
  • ADGA and VCC work well with PsCs: The integration of PsCs with established methods like ADGA for PES construction and VCC for vibrational structure calculations proves to be a powerful combination.

Key Results

  • Formaldehyde: Calculations showed that PsC parametrization consistently required the fewest single-point calculations (SPCs) to converge the potential energy surfaces (PESs) across various mode combination (MC) levels. For example, to converge a 2M PES, PsCs required 1639 SPCs, compared to 3172 for NCs and 3845 for HOLCs.
  • Vibrational Excitation Energies (Formaldehyde): The fundamental and first overtone excitation energies calculated using PsCs were very similar to those obtained with NCs and HOLCs, with mean absolute deviations (MADs) of less than 1 cm⁻¹. This indicates that PsCs achieve comparable accuracy with significantly less computational effort.
  • Hydrogen Peroxide: The study focused solely on PsC parametrization for hydrogen peroxide due to its complex internal motion. The results showed that PsCs could accurately describe the ground state torsional splitting, even at low levels of approximation.
  • Convergence: Figures and tables demonstrate that PsCs lead to faster convergence of VCC calculations towards the full vibrational configuration interaction (FVCI) limit compared to rectilinear coordinates. For instance, achieving a MAD of less than 1 cm⁻¹ for fundamental excitation energies required a VCC[4] model with a 3M PES using PsCs, whereas NCs required a 4M PES.

Limitations and Future Work

  • Computational Cost for Large Systems: While PsCs offer advantages, the exact analytical nuclear KEO can become computationally demanding for larger systems due to high levels of mode-coupling. The authors suggest that future research could explore systematic approximations to the KEO.
  • Electronic Structure Methods: For hydrogen peroxide, the authors note that their calculations lacked some electronic corrections. They suggest repeating calculations with more sophisticated electronic structure methods to further ascertain the accuracy of their approach.
  • Periodic B-splines: The use of non-periodic B-splines for representing vibrational wave functions was a limitation for hydrogen peroxide's internal motion. Future work could incorporate periodic B-splines for a more accurate representation.
  • Larger Molecules and Full Dimensionality: The authors propose applying PsCs to medium to large-sized molecules in full dimensionality, either independently or in comparison with other coordinate sets like FALCON.
  • Frozen Mode Approximation and Subsystems: Exploring the frozen mode approximation and the ability to define subsystems for PsC generation and KEO calculation are suggested as logical extensions.

Summary updated

Abstract

We present the first use of curvilinear vibrational coordinates, specifically polyspherical coordinates, in combination with vibrational coupled cluster theory. The polyspherical coordinates are used in the context of both the adaptive density-guided approach to potential energy surface construction and in the subsequent vibrational coupled cluster calculations of anharmonic vibrational states. Results obtained based on the polyspherical coordinate parametrization are compared to results obtained with the use of rectilinear vibrational coordinates, namely, normal coordinates and hybrid optimized and localized coordinates for the formaldehyde molecule. This comparison is carried out with the full vibrational configuration interaction model, using the respective fully coupled potential energy surfaces and untruncated kinetic energy operators. The polyspherical coordinates are shown to facilitate an acceleration of convergence for truncated methods when compared to the use of normal coordinates. We furthermore report on calculations on the hydrogen peroxide molecule in the polyspherical coordinate parametrization. The polyspherical vibrational coordinates are shown to perform very well, even for truncated methods, especially when considering the difficulty that rectilinear vibrational coordinates can exhibit in treating complicated internal molecular motion.
BibTeX
@article{klinting2020vibrational,
  author = {Emil Lund Klinting and David Lauvergnat and Ove Christiansen},
  title = {{Vibrational Coupled Cluster Computations in Polyspherical Coordinates with the Exact Analytical Kinetic Energy Operator}},
  journal = {Journal of Chemical Theory and Computation},
  year = {2020},
  doi = {10.1021/acs.jctc.0c00261},
}
Figure