[Paper Review] iGVPT2 : an interface to computational chemistry packages for anharmonic corrections to vibrational frequencies

Loïc Barnes, Baptiste Schindler+2 more|arXiv (Cornell University)|Apr 7, 2017

Advanced Chemical Physics Studies9 references3 citations

TL;DR

iGVPT2 is a free, open-source software that computes anharmonic vibrational frequency corrections using second-order vibrational perturbation theory (VPT2) and its variants (VPT2+K, DCPT2, HDCPT2), interfacing with multiple quantum chemistry packages. It enables hybrid QM//MM calculations and parallel computation of force constants, reducing real-time computation by up to ~f³ (where f is the number of normal modes), enabling efficient anharmonic IR spectrum prediction for large molecules including biomolecules.

ABSTRACT

iGVPT2 is a program for computing anharmonic corrections to vibration frequencies, based on force field expansion of the potential energy surface in normal mode coordinates. It includes second order vibrational perturbation theory (VPT2) algorithm and its derived methods (VPT2+K, DCPT2, HDCPT2). iGVPT2 is interfaced with several computation chemistry packages to compute the potential energies and dipoles derivatives. The second, third and quartic derivatives can be computed at the same level of theory but they can be also computed using different methods via one or two computational packages. iGPVT2 includes also a very fast hybrid QM//MM approach for biomolecules. It is provided free-of-charge for non-commercial research (see \url{https://sites.google.com/site/allouchear/igvpt2}).

Motivation & Objective

• To develop a flexible, efficient interface for computing anharmonic vibrational frequency corrections using VPT2 and its variants.
• To enable mixed-level quantum chemistry calculations, where harmonic frequencies are computed at a high level of theory and anharmonic corrections at a lower level.
• To support multiple computational chemistry packages (e.g., ORCA, Gaussian, Molpro, OpenMOPac) via a plugin-like interface, enhancing interoperability.
• To implement a fast hybrid QM//MM approach for large biomolecules, reducing computational cost while maintaining accuracy.
• To provide a highly parallelized framework for computing third and fourth derivatives, reducing real-time computation by up to ~f³.

Proposed method

• iGVPT2 computes anharmonic corrections using second-order vibrational perturbation theory (VPT2), including variants such as VPT2+K, DCPT2, and HDCPT2.
• It interfaces with multiple quantum chemistry packages (e.g., ORCA, Gaussian, Molpro, FireFly, GAMESS-US, DFTB+) to compute potential energy, dipole, and force derivatives.
• The software supports hybrid calculations: harmonic frequencies computed at a high-level method (e.g., B3LYP/def2-TZVPP), while cubic and quartic force constants are computed at a lower level (e.g., MMFF94 or semi-empirical methods).
• Numerical derivatives for harmonic frequencies are computed via finite difference with user-defined step size, and the process is parallelized across atoms and normal modes.
• A hybrid QM//MM approach is implemented using DFT for the QM region and MMFF94 for the rest, enabling fast and accurate anharmonic frequency prediction for large biomolecules.
• The code is parallelized to reduce real-time computation by a factor of ~f³ (f = number of normal modes), significantly accelerating third and fourth derivative calculations.

Experimental results

Research questions

• RQ1 Can anisotropic vibrational frequency corrections be efficiently computed across multiple quantum chemistry packages using a single interface?
• RQ2 Does a hybrid QM//MM approach in iGVPT2 maintain accuracy while drastically reducing computational cost for large biomolecules?
• RQ3 Can the real-time computation of third and fourth derivatives be reduced by a factor of ~f³ through parallelization?
• RQ4 How accurate are anharmonic frequencies computed with iGVPT2 compared to experimental data and other software implementations?
• RQ5 Can mixed-level calculations (e.g., high-level harmonic, low-level anharmonic corrections) yield reliable results without full high-level computation of all derivatives?

Key findings

• For the protonated dipeptide GlyGlyH+, full DFT (B3LYP/def2-TZVPP) computation of anharmonic frequencies took 109,888 CPU hours, but was reduced to 56 real-time hours using 2000 cores.
• Using the hybrid QM//MM approach (B3LYP/def2-TZVPP for harmonic, MMFF94 for anharmonic corrections), the same calculation completed in just 0.5 hours (6 seconds for third and fourth derivatives).
• The hybrid QM//MM method reproduced experimental anharmonic frequencies with high accuracy, as validated in prior work.
• Maximum deviations (MAX) across test molecules were below 1224 cm⁻¹, with mean absolute deviations (MAD) below 2.5 cm⁻¹ when compared to reference calculations.
• Root-mean-square deviations (RMSD) for anharmonic frequencies were below 4.8 cm⁻¹ when using consistent step sizes, indicating high consistency and reliability.
• The software enables accurate anharmonic frequency prediction for large systems, including biomolecules, with significant speedups through parallelization and hybrid methods.