Manuel Cordova^1*, Edgar A. Engel², Artur Stefaniuk¹, Federico Paruzzo¹, Albert Hofstetter¹, Michele Ceriotti^3*, Lyndon Emsley^1*

1 Institut des Sciences et Ingénierie Chimiques, École Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland

2 Theory of Condensed Matter Group, Cavendish Laboratory, University of Cambridge, J. J. Thomson Avenue, Cambridge CB3 0HE, U.K.

3 Institut des matériaux, École Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland

* Corresponding authors emails: manuel.cordova@epfl.ch, michele.ceriotti@epfl.ch, lyndon.emsley@epfl.ch

DOI10.24435/materialscloud:a9-4n [version v1]

Publication date: Nov 11, 2022

How to cite this record

Manuel Cordova, Edgar A. Engel, Artur Stefaniuk, Federico Paruzzo, Albert Hofstetter, Michele Ceriotti, Lyndon Emsley, A machine learning model of chemical shifts for chemically and structurally diverse molecular solids, Materials Cloud Archive 2022.147 (2022), https://doi.org/10.24435/materialscloud:a9-4n

Description

Nuclear magnetic resonance (NMR) chemical shifts are a direct probe of local atomic environments and can be used to determine the structure of solid materials. However, the substantial computational cost required to predict accurate chemical shifts is a key bottleneck for NMR crystallography. We recently introduced ShiftML, a machine-learning model of chemical shifts in molecular solids, trained on minimum-energy geometries of materials composed of C, H, N, O, and S that provides rapid chemical shift predictions with density functional theory (DFT) accuracy. Here, we extend the capabilities of ShiftML to predict chemical shifts for both finite temperature structures and more chemically diverse compounds, while retaining the same speed and accuracy. For a benchmark set of 13 molecular solids, we find a root-mean-squared error of 0.47 ppm with respect to experiment for 1H shift predictions (compared to 0.35 ppm for explicit DFT calculations), while reducing the computational cost by over four orders of magnitude.

Materials Cloud sections using this data

Files

File name	Size	Description
ShiftML2.zip MD5md5:12d0872eb54908dc6ea6244895b4da0e	55.3 KiB	ShiftML2 Python package
Experimental_benchmark_nmr.zip MD5md5:68520acd7410ffdd60627679501b2c38	461.7 KiB	GIPAW NMR computations performed on a set of crystal structures used to compare ShiftML2 with experimental chemical shifts
final_training_test_sets.zip MD5md5:7fda429b5eadf7378ec1903ca19cd278	115.3 MiB	Final set of training and test environments for each nucleus
initial_training_test_sets.zip MD5md5:5bc1ddc62a1fb532e8707b723a5de6ab	39.0 MiB	Initial set of training and test environments for each nucleus
relax_vs_md.zip MD5md5:a4c9f521cf6352704e9fd477b7767967	6.3 MiB	Snapshots from AIMD simulation used to compare the accuracy of the model in relaxed and distorted structures
ShiftML_poly.zip MD5md5:28b1c98148fef74cd3991adf97f3f55e	537.2 KiB	Set of candidate structures with GIPAW NMR computations for three molecular solids
training_scripts.zip MD5md5:2a952934a684939ef148beb13116a1da	2.8 MiB	Python notebook scripts used to train the model
ShiftML_v2_model_H.pk MD5md5:0155e4f88b00a52c7ee7555012961149	3.5 GiB	ShiftML model (kernel) for hydrogen
ShiftML_v2_model_C.pk MD5md5:915ddef4e10c74b1d92215315500a289	3.3 GiB	ShiftML model (kernel) for carbon
ShiftML_v2_model_N.pk MD5md5:2f0fce6d788c6b7c07d917454dcff6a5	3.4 GiB	ShiftML model (kernel) for nitrogen
ShiftML_v2_model_O.pk MD5md5:421a80c90cdd54110dcbe96b94cb3ab5	3.5 GiB	ShiftML model (kernel) for oxygen
ShiftML_v2_model_F.pk MD5md5:71a8c7dc1b010f76857e2cc41bcf2e16	1.5 GiB	ShiftML model (kernel) for fluorine
ShiftML_v2_model_S.pk MD5md5:8373a1dcc7065aec7dbc91d2f6cb17ab	1.0 GiB	ShiftML model (kernel) for sulfur
ShiftML_v2_model_P.pk MD5md5:6f1c6a6cdbb60de78ddb9cac0edf3ec0	362.4 MiB	ShiftML model (kernel) for phosphorus
ShiftML_v2_model_Cl.pk MD5md5:d92aefad57270115a756fd3e301a8de5	944.0 MiB	ShiftML model (kernel) for chlorine
ShiftML_v2_model_Na.pk MD5md5:7fd44392f47b46730e22d1adf5576dd7	51.7 MiB	ShiftML model (kernel) for sodium
ShiftML_v2_model_Ca.pk MD5md5:0be6d72103c591f8239164f922fbbad4	25.3 MiB	ShiftML model (kernel) for calcium
ShiftML_v2_model_K.pk MD5md5:1d9e1ae13b2e8c97b82ff16cc216bfe3	44.5 MiB	ShiftML model (kernel) for potassium
ShiftML_v2_model_Mg.pk MD5md5:407b0ee50683d1296e6d9480413cc165	11.4 MiB	ShiftML model (kernel) for magnesium
README.txt MD5md5:c4bcccd488166a2df257f652412574aa	1.2 KiB	File descriptions and instructions for package installation

License

Files and data are licensed under the terms of the following license: Creative Commons Attribution 4.0 International.
Metadata, except for email addresses, are licensed under the Creative Commons Attribution Share-Alike 4.0 International license.

External references

Keywords

machine learning MARVEL/DD1 SNSF