PINNs-Experiments

→

Using physics-informed neural networks (PINNs) for solving atomic time-dependent Schrödinger equation (TDSE) in perturbative regime. Various experiments are available within the project framework:

• Hydrogen orbitals (solving unperturbed atomic time-independent Schrödinger equation (TISE))
• Hydrogen atom in a DC Stark setting
• Unperturbed hydrogen orbital evolution (solving unperturbed atomic TDSE)
• One-level atom in an AC Stark setting

1) Setup

Create and activate the conda environment:

conda env create -f conda.yml
conda activate pinns-experiments

2) Styling

Enforce styling with black:

python -m black src

3) Getting Started

Experiment scripts are located in examples directory. All experiment parameters can be defined by a config (.yaml) file in examples/experiment_configs, such as:

experiment_params:
  quantum_numbers:
    n: 2
    l: 1
    m: 0
  radial_extent: 30

network_params:
  backbone: "ResNet"
  input_size: 3
  output_size: 2
  num_neurons: 50
  num_blocks: 3

  num_domain: 10000
  num_boundary: 2000

  lrs: [1e-3, 1e-5, 1e-6, 1e-7]
  optimizers: ["adam", "adam", "adam", "adam"]
  epochs: [50000, 50000, 5000, 5000]
  loss_weights: [1,1,1,1,1,1,10,10,1,1,1,1]
  strict_boundary: True # loss weights and strict boundary must be compatible!

prior_params:

results_path: "Enter the path of your results folder here"

Run an example experiment

In order to run the example experiment for the hydrogen orbitals, define the path of the results directory in examples/experiment_configs/TISE_hydrogen_without_decomposition/example_resnet.yaml, e.g.:

results_path: "examples/results/TISE_hydrogen/"

Then, you can run the training script:

python TISE_hydrogen_without_decomposition --config_path ./experiment_configs/TISE_hydrogen_without_decomposition/example_resnet.yaml

The model checkpoints will be saved in the results directory given in the config file.

Evaluate your experiments

In order to evaluate the accuracy of your experiments and create beautiful plots, use the evaluation.py script in src/pinnse/. Fill the following lines in the main function:

script = "TISE_hydrogen_without_decomposition"
save_pth = "model ckpt path"
config_path = "config file path"

And go ahead and evaluate your experiments:

python evaluation.py