Molecular Vibration Analysis Skill

Goal

Calculate the vibrational frequencies ($\nu$), normal modes, and zero-point energy (ZPE) of non-periodic (finite) systems — molecules, clusters, and adsorbates — within the harmonic approximation using Machine Learning Interatomic Potentials (MLIPs).

[!IMPORTANT] This skill is for molecules and finite systems only. For periodic crystals, use the phonon skill instead.

Background

In the harmonic approximation, the potential energy surface near a local minimum is approximated as $V \approx V_0 + \frac{1}{2} \sum_{ij} H_{ij} \Delta r_i \Delta r_j$, where $H_{ij} = \frac{\partial^2 V}{\partial r_i \partial r_j}$ is the Hessian (force constant) matrix. Diagonalizing the mass-weighted Hessian yields $3N$ eigenvalues: for a nonlinear molecule, $3N-6$ are real vibrational modes (and $3N-5$ for linear molecules), while the remaining eigenvalues correspond to translational and rotational degrees of freedom (near zero).

1. Prerequisites

• An MLIP wrapper must be available (MACEWrapper, MatGLWrapper, or FAIRCHEMWrapper).
• ASE must be installed in the relevant conda environment.
• The input structure must be a molecule or cluster (non-periodic). Periodic systems should use mat-phonon.