Connecting Vibrational Spectroscopy to Atomic Structure via Supervised Manifold Learning: Beyond Peak Analysis

Abstract

Vibrational spectroscopy is a nondestructive technique commonly used in chemical and physical analyses to determine atomic structures and associated properties. However, the evaluation and interpretation of spectroscopic profiles based on human-identifiable peaks can be difficult and convoluted. To address this challenge, we present a reliable protocol based on supervised manifold learning techniques meant to connect vibrational spectra to a variety of complex and diverse atomic structure configurations. As an illustration, we examined a large database of virtual vibrational spectroscopy profiles generated from atomistic simulations for silicon structures subjected to different stress, amorphization, and disordering states. We evaluated representative features in those spectra via various linear and nonlinear dimensionality reduction techniques and used the reduced representation of those features with decision trees to correlate them with structural information unavailable through classical human-identifiable peak analysis. We show that our trained model accurately (over 97% accuracy) and robustly (insensitive to noise) disentangles the contribution from the different material states, hence demonstrating a comprehensive decoding of spectroscopic profiles beyond classical (human-identifiable) peak analysis.

Unsupervised physics-informed disentanglement of multimodal data

“…Finally, we test PIMA on a large, synthetic multimodal materials dataset with the intent to discover shared, cross-modal information. Specifically, we consider a synthetic dataset of [55] in which molecular dynamic simulations of single crystal silicon atomic structures were performed with varying degrees of disorder, deformation (varying strains for uniaxial compression or hydrostatic compression, or no compression), and combinations of both disorder and deformation. The resulting dataset consists of 772 atomic structures, each generated by different disorder and deformation process parameters, where the vibrational density of states (VDoS) spectroscopy profile and average stress values are computed for each structure.…”

Expanding the Horizons of Machine Learning in Nanomaterials to Chiral Nanostructures

“…Vizozo et al aimed to find the connection between the vibrational density of states and deformation and defect states present in materials. [ 93 ] In their approach, the authors firstly pre-processed the data from DFT and MD measurements by unsupervised learning for data dimensionality reduction using PCA, an approach mentioned in Step #3 of our guidelines above in Section 2 . Next, they built a gradient boosting DT model that was capable of linking vibrational spectra to a variety of structural configurations.…”

Dataset of simulated vibrational density of states and X-ray diffraction profiles of mechanically deformed and disordered atomic structures in Gold, Iron, Magnesium, and Silicon

“…Traditional analysis of spectroscopy and diffraction spectra is typically performed manually via peak analysis and is prone to biases introduced by the practitioner or by the specific analysis techniques used that can lead to inconsistent results or inaccurate predictions. This dataset was created for the purpose of testing various machine-learning (ML) methods for extracting material properties from spectroscopic and diffraction data [ 1 , 2 ], with sub-objectives of testing how well ML techniques can be used to automate the analysis of these types of datasets without requiring human intervention as well as to test whether additional material properties could be extracted from these datasets via ML techniques that are typically considered inaccessible via traditional analysis methods. A portion of the Si data provided in this dataset was used for a previous publication [ 1 ], and this represents a substantially expansion of that work, providing researchers the ability to test their approaches on multiple characterization modalities and across multiple material systems.…”

“…To download the data, the user must install Globus Connect Personal, which can then be used to download the data to a personal machine. Related research article A portion of this dataset was used for the following publication [ 1 ]: Daniel Vizoso, Ghatu Subhash, Krishna Rajan, Rémi Dingreville Chemistry of Materials 2023 35 (3), 1186–1200 …”