Deep Learning Approach for Raman Spectroscopy

Raman spectroscopy is a widely used technique for organic and inorganic chemical material identification. Throughout the last century, major improvements in lasers, spectrometers, detectors, and holographic optical components have uplifted Raman spectroscopy as an effective device for a variety of different applications including fundamental chemical and material research, medical diagnostics, bio-science, in-situ process monitoring and planetary investigations. Undoubtedly, mathematical data analysis has been playing a vital role to speed up the migration of Raman spectroscopy to explore different applications. It supports researchers to customize spectral interpretation and overcome the limitations of the physical components in the Raman instrument. However, large, and complex datasets, interferences from instrumentation noise and sample properties which mask the true features of samples still make Raman spectroscopy as a challenging tool. Deep learning is a powerful machine learning strategy to build exploratory and predictive models from large raw datasets and has gained more attention in chemical research over recent years. This chapter demonstrates the application of deep learning techniques for Raman signal-extraction, feature-learning and modelling complex relationships as a support to researchers to overcome the challenges in Raman based chemical analysis.

Mineralogy, Mineral Chemistry and SWIR Spectral Reflectance of Chlorite and White Mica

Hyperspectral reflectance has the potential to provide rapid and low-cost mineralogical and chemical information that can be used to vector in mineral systems. However, the spectral signature of white mica and chlorite, despite numerous studies, is not fully understood. In this study, we review the mineralogy and chemistry of different white mica and chlorite types and investigate what mineralogical and chemical changes are responsible for the apparent shifts in the shortwave infrared (SWIR) spectroscopic absorption features. We demonstrate that the spectral signature of white mica is more complex than previously documented and is influenced by the Tschermak substitution, as well as the sum of interlayer cations. We show that an increase in the interlayer deficiencies towards illite is associated with a change from steep to shallow slopes between the wavelength position of the 2200 nm feature (2200 W) and Mg, Al(VI) and Si. These changes in slope imply that white micas with different elemental chemistry may be associated with the same 2200 W values and vice versa, contrary to traditional interpretation. We recommend that traditional interpretations should only be used in true white mica with sum interlayer cations (I) > 0.95. The spectral signature of trioctahedral chlorite (clinochlore, sheridanite, chamosite and ripidolite) record similar spectral relationships to those observed in previous studies. However, dioctahedral Al-rich chlorite (sudoite, cookeite and donbassite) has a different spectral response with Mg increasing with 2250 W, which is the opposite of traditional trioctahedral chlorite spectral interpretation. In addition, it was shown that dioctahedral chlorite has a 2200 W absorption feature that may introduce erroneous spectral interpretations of white mica and chlorite mixtures. Therefore, care should be used when interpreting the spectral signature of chlorite. We recommend that spectral studies should be complemented with electron microprobe analyses on a subset of at least 30 samples to identify the type of muscovite and chlorite. This will allow the sum I of white mica to be obtained, as well as estimate the slope of 2200 W absorption trends with Mg, Al(vi), and Si. Preliminary probe data will allow more accurate spectral interpretations and allow the user to understand the limitations in their hyperspectral datasets.

Particle Size-Frequency Distributions of the OSIRIS-REx Candidate Sample Sites on Asteroid (101955) Bennu

We manually mapped particles ranging in longest axis from 0.3 cm to 95 m on (101955) Bennu for the Origins, Spectral Interpretation, Resource Identification, and Security–Regolith Explorer (OSIRIS-REx) asteroid sample return mission. This enabled the mission to identify candidate sample collection sites and shed light on the processes that have shaped the surface of this rubble-pile asteroid. Building on a global survey of particles, we used higher-resolution data from regional observations to calculate particle size-frequency distributions (PSFDs) and assess the viability of four candidate sites for sample collection (presence of unobstructed particles ≤ 2 cm). The four candidate sites have common characteristics: each is situated within a crater with a relative abundance of sampleable material. Their PSFDs, however, indicate that each site has experienced different geologic processing. The PSFD power-law slopes range from −3.0 ± 0.2 to −2.3 ± 0.1 across the four sites, based on images with a 0.01-m pixel scale. These values are consistent with, or shallower than, the global survey measurements. At one site, Osprey, the particle packing density appears to reach geometric saturation. We evaluate the uncertainty in these measurements and discuss their implications for other remotely sensed and mapped particles, and their importance to OSIRIS-REx sampling operations.

Functional Group Identification for FTIR Spectra Using Image-Based Machine Learning Models

<p>Fourier Transform Infrared Spectroscopy (FTIR) is a ubiquitous spectroscopic technique. Spectral interpretation is a time-consuming process, but it yields important information about functional groups present in compounds and in complex substances. We develop a generalizable model via a machine learning (ML) algorithm using Convolutional Neural Networks (CNNs) to identify the presence of functional groups in gas phase FTIR spectra. The ML models will reduce the amount of time required to analyze functional groups and facilitate interpretation of FTIR spectra. Through web scraping, we acquire intensity-frequency data from 8728 gas phase organic molecules within the NIST spectral database and transform the data into images. We successfully train models for 15 of the most common organic functional groups, which we then determine via identification from previously untrained spectra. These models serve to expand the application of FTIR measurements for facile analysis of organic samples. Our approach was done such that we have broad functional group models that inference in tandem to provide full interpretation of a spectrum. We present the first implementation of ML using image-based CNNs for predicting functional groups from a spectroscopic method.</p>

Functional Group Identification for FTIR Spectra Using Image-Based Machine Learning Models

<p>Fourier Transform Infrared Spectroscopy (FTIR) is a ubiquitous spectroscopic technique. Spectral interpretation is a time-consuming process, but it yields important information about functional groups present in compounds and in complex substances. We develop a generalizable model via a machine learning (ML) algorithm using Convolutional Neural Networks (CNNs) to identify the presence of functional groups in gas phase FTIR spectra. The ML models will reduce the amount of time required to analyze functional groups and facilitate interpretation of FTIR spectra. Through web scraping, we acquire intensity-frequency data from 8728 gas phase organic molecules within the NIST spectral database and transform the data into images. We successfully train models for 15 of the most common organic functional groups, which we then determine via identification from previously untrained spectra. These models serve to expand the application of FTIR measurements for facile analysis of organic samples. Our approach was done such that we have broad functional group models that inference in tandem to provide full interpretation of a spectrum. We present the first implementation of ML using image-based CNNs for predicting functional groups from a spectroscopic method.</p>