IR Spectra for the EMIM-TFSI Ion Pair Using Deep Potentials.

Area of Science:

• Computational Chemistry
• Materials Science
• Spectroscopy

Background:

• Characterizing ionic liquids (ILs) is crucial but computationally intensive, especially for infrared (IR) spectra using ab initio methods.
• Accurate simulation of IL configurations, dipole moments, and IR spectra is essential for understanding their behavior.

Purpose of the Study:

• To investigate the configuration, dipole moment, and IR spectra of 1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide ([EMIM]+-[TFSI]-) using a deep potential (DP) and deep Wannier (DW) model framework.
• To evaluate the accuracy and reliability of DP and DW models against ab initio molecular dynamics (AIMD).

Main Methods:

• Integration of deep potential (DP) and deep Wannier (DW) models for molecular dynamics simulations.
• Benchmarking DP/DW simulations against ab initio molecular dynamics (AIMD) for structural, dipolar, and spectral features.
• Focus on achieving well-converged dipole distributions over extended simulation timescales (tens to hundreds of picoseconds).

Main Results:

• DP and DW models show good agreement with AIMD for dipole moment range (7-16 D, average ~10 D) and IR spectral features.
• Deep learning molecular dynamics (DW/DPMD) provides smoother, better-converged dipole distributions and accurately reproduces key vibrational bands (vS-N-S,as < vCF3 < vSO2,as).
• Classical IR spectra show discrepancies in band intensities and relative wavenumbers compared to AIMD and experimental data.

Conclusions:

• Deep learning potentials (DP and DW) are effective for simulating charged species and complex ionic interactions, outperforming classical methods.
• Achieving long simulation times with DW/DPMD is vital for enhanced spectral resolution and minimizing configuration-dependent noise.
• This framework enables advanced surrogate models for complex systems, including bulk ILs and interfaces.