Research Interests:

Ab initio calculations, Potential Energy Surfaces (PESs), Neural Network Exponential fitting (NN-expnn), Molecular Vibrational Frequencies and wavefunctions using MCTDH, Quantum Dynamics, Excited state properties, Non-adiabatic dynamics for electron and charge transport phenomenons, and more!

Potential Energy Surfaces (PESs) using Neural Network Exponential Fitting: Determination of Vibrational Frequencies

1) Run scaled_onestg.m 

This takes the “raw” (unscaled) training, test, and validation sets and performs the fit. RMSE is given in correct (unscaled) energy units but fit parameters are in scaled units.

2) Run rescale_fit_params.py 

This takes scaled fit parameters and returns them to the correct units to use in MCDTH

3) Run new_genop.py

This takes the output files and turns them into the MCTDH operator file
 

Step 1

Ab initio computations

[Molpro]

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Scripts

Flowchart

input1, input2,input3,

submit.pbs

Geometry optimization

↓

input1d, input2d,

1D cut points, 2D cut points

↓

range.inp → trans_rand.f

↓

Random coordinates

↓

geninput.f

↓

rantr.geom

↓

input_rand.mpi →

 

 

 

 

 

 

 

 

 

 

Random energy data

↓

[r1,r2,r3,θ1,θ2,ϕ,V]

↓

en_zero.f → energy

↓

en_cut.f

↓

[r1,r2,r3,cosθ1,cosθ2,ϕ,V] →

 

 

 

 

 

 

 

(1D+2D+Random)

↓

 

Split in to train, test, and validation sets

↓

Step 2

PES fitting

[Matlab]

 

 

 

(train, test, and val set)

↓

script.m

Fit using NN-expnn

↓

 

LW, IW, b, and c files

↓

Step 3

MCTDH Calculations

[MCTDH]

 

HFCO.IW, HFCO.LW, HFCO.b, HFCO.c

↓

rescale_fitparams.py

↓

LW, IW, b, c, rd

↓

new_genop.py

↓

HONO.op  →

Generate operator file

↓

 

Vibrational frequency and wavefunction