Simulation of chemical dynamics in high dimensionality with low-rank tensor-train Chebyshev (TT-Chebyshev) and function-train Chebyshev (FT-Chebyshev) quantum dynamics

Micheline B. Soley, Paul Bergold, Alex A. Gorodetsky, Victor S. Batista, J. Chem. Theory Comput. 2022, 18, 1, 25-36.

The TT-Chebyshev program requires the following:

• MATLAB
• H-Tucker Tensor Toolbox
• TT-Toolbox

The C FT-Chebyshev program requires the following:

• C compiler
• CMake
• Compressed Continuous Computation (C3) library (which requires BLAS, LAPACK, and optionally SWIG-python)

Note the C version of the program reproduces the results in the paper.

The Python functional tensor-train Chebyshev (FTTC) code requires the following:

• Miniconda
• Compressed Continuous Computation (C3) library

Comparison of TT- and FT-Chebyshev results with the provided scripts requires:

• Gnuplot

To run the TT-Chebyshev code in folder TensorTrain, run tt_CHEBSOFT.m in MATLAB. Data files can then be converted to Gnuplot format by running GnuplotConversion.m in MATLAB.

To run the C FT-Chebyshev code in folder FunctionTrain, compile with commands:

cmake .
make

and run the program as follows:

./ft_cheb

Troubleshooting details are included in the readme.txt file in the FT-Chebyshev folder.

To run the Python FT-Chebyshev code in folder FTTCPython use the command:

python fttc.py

Troubleshooting details are included in the readme.txt file in the Python FT-Chebyshev folder.

TT- and C FT-Chebyshev results can be compared by moving the TT- and C FT-Chebyshev data files to folder TensorTrainData and FunctionTrainData in folder Comparisons and using the following commands:

gnuplot wavefunctionsnapshots.gpt
gnuplot autocorrelationcompare.gpt

Additional documentation is included in the CodeDocumentation.pdf file in the main folder.