Regression Machine Learning Models Used to Predict DFT-Computed NMR Parameters of Zeolites

Abstract

Machine learning approaches can drastically decrease the computational time for the predictions of spectroscopic properties in materials, while preserving the quality of the computational approaches. We studied the performance of kernel-ridge regression (KRR) and gradient boosting regressor (GBR) models trained on the isotropic shielding values, computed with density-functional theory (DFT), in a series of different known zeolites containing out-of-frame metal cations or fluorine anion and organic structure-directing cations. The smooth overlap of atomic position descriptors were computed from the DFT-optimised Cartesian coordinates of each atoms in the zeolite crystal cells. The use of these descriptors as inputs in both machine learning regression methods led to the prediction of the DFT isotropic shielding values with mean errors within 0.6 ppm. The results showed that the GBR model scales better than the KRR model.

The need for operando modelling of 27Al NMR in zeolites: the effect of temperature, topology and water

“…More accurate shielding (tensor) predictions can be achieved using, e.g. , Kernel Ridge Regression 54,55 or (equivariant) neural networks. 56,57 However, these ML methods use high-dimensional descriptors, making a physical interpretation of the structure– σ correlation difficult.…”

Impact of noise on inverse design: the case of NMR spectra matching

“…31 and commonly being used in learning NMR properties from quantum chemical calculations. 37,[42][43][44][45][46] We use a Laplacian kernel and the local atomic Faber-Christensen-Huang-Lilienfeld (FCHL19 (ref. 13 C and 1 H spectra matching.…”

Theory and computation of nuclear shielding

“…7 Miscelaneous examples, including machine learning of nuclear magnetic shielding modeling Current density, nuclear shielding and magnetizability in a very intriguing molecule of infinitene, resembling the infinity symbol or stretched coronene, were calculated and analyzed by Summa et al 146 Kutateladze et al 147 reported on machine learning-augmented density functional theory (DU8ML) nuclear magnetic resonance computations for high-throughput solution structure validation. Mineva et al 148 used regression machine learning models to predict DFT-calculated NMR parameters of zeolites. Costa et al 149 applied DP4þ and ANN-PRA methodology to determine the relative configuration of a natural product (alpha-bisabol).…”