Smooth Overlap of Atomic Positions (SOAP) is an algorithm used for accurately classifying and machine learning chemical environments [1,2]. For a detailed documentation, please read soapDoc.pdf in this repository.

This is a very lightweight and fast SOAP implementation for machine learning in quantum chemistry and materials physics. Once you give SOAP the .xyz in ASE format, list of positions, radius cutoff, number of basis functions and l (number of spherical harmonics), SOAP will return a numpy matrix of the power spectrum for each point. Each row corresponds to each specified point, and each column corresponds the the spectrum value.

Here is an example of the python interface:

from soaplite import getBasisFunc, get_soap_locals
from ase.build import molecule

#-------------- Define structure -----------------------------------------------
atoms = molecule("H2O")

#-------------- Define positions of desired local environments ----------------
hpos = [
    [0, 1, 2],
    [2, 3, 4]
]

#------------------ Basis function settings (rCut, N_max) ----------------------
n_max = 5
l_max = 5
r_cut = 10.0
my_alphas, my_betas = getBasisFunc(r_cut, n_max)

#--------- Get local chemical environments for each defined position -----------
x = get_soap_locals(
    atoms,
    hpos,
    my_alphas,
    my_betas,
    rCut=r_cut,
    NradBas=n_max,
    Lmax=l_max,
    crossOver=True
)

print(x)
print(x.shape)

We provide a python interface to the code with precompiled C-extension. This precompiled version should work with linux-based machines, and can be installed with:

pip install soaplite

The pip installation is still experimental for soaplite.

If you wish to use the C-libraries directly, you can compile them yourself by running

make

in the terminal after you cloned SOAPLite from github if you have the gcc compiler.

Numpy, Scipy and ASE are required for the python interface. They will be installed automatically if you use the pip installation. To install them manully, you can use the following commands:

sudo pip install numpy

sudo pip install scipy

and

sudo pip install ase

If you don't have super user access, install them by

pip install numpy --user

pip install scipy --user

and

pip install ase --user

If you wish to compile the C-extension yourself, you wil need a gcc compiler for make.

Enter the tests/ directory and run

python test_symmetry.py

This will compare the rotational and translational symmetry, and same chemical environments.

By taking the differences of the soap spectrum, we can compare the differences of the chemical environment. For example, if a point P1 gave a power spectrum S1 and at point P2 gave a spectrum S2, the difference of the chemical environment will be |S2 - S1| where || denotes the Euclidean distance. We can use this differences to classify similar/different chemical environments.

The power spectrum can also be used as an input for a neural network, kernel ridge regression or other machine learning algorithms.

• Eiaki V. Morooka - Aki78
• Marc Jäger - marchunter
• Lauri Himanen - lauri-codes

See also the list of contributors who participated in this project.

• Yu Ninomiya - Yu
• Filippo Federici - fullmetalfelix
• Yashasvi S. Ranawat - yashasvi-ranawat
• Adam Foster - suurimonster

This project is licensed under the GNU LESSER GENERAL PUBLIC LICENSE - see the LICENSE.md file for details

• [1] On representing chemical environments - Albert P. Bartók, Risi Kondor, Gábor Csányi paper
• [2] Comparing molecules and solids across structural and alchemical space - Sandip De, Albert P. Bartók, Gábor Cásnyi, and Michele Ceriotti paper