THz Pump Pulse-Driven Temporal Response of Liquid Water Probed by Machine-Learning-Accelerated Non-Equilibrium...

Area of Science:

• Physical Chemistry
• Computational Chemistry
• Spectroscopy

Background:

• Ultrafast spectroscopies reveal molecular dynamics in condensed phases.
• Ab initio molecular dynamics (AIMD) offers atomistic detail but is computationally expensive for non-equilibrium processes.
• Simulating excitation and relaxation requires capturing stochastic dynamics.

Purpose of the Study:

• To develop a computationally efficient method for simulating non-equilibrium molecular dynamics.
• To investigate the time-dependent response of liquid water to terahertz (THz) pulses.
• To understand the transient perturbation of water's hydrogen bond network and solvation environment.

Main Methods:

• Implemented a machine learning potential incorporating time-dependent electric fields.
• Retained ab initio molecular dynamics (AIMD)-level accuracy for electronic structure.
• Simulated liquid water's response to a 12.3 THz Gaussian pump pulse, generating 32 ns of trajectory data.

Main Results:

• The method accurately reproduced experimental observables like transient dichroism and birefringence.
• Observed THz pulse-driven librational excitation, energy transfer to translational and intramolecular vibrations.
• Demonstrated transient disruption and reorganization of the hydrogen-bond network.

Conclusions:

• The developed machine learning potential enables accurate simulation of non-equilibrium dynamics.
• THz pulses can transiently modulate water's solvation environment by perturbing its hydrogen-bond network.
• Established a broadly applicable framework for studying field-driven molecular processes.