Convolutional Neural Networks for Challenges in Automated Nuclide Identification

Improvements in Radio-Isotope IDentification (RIID) algorithms have seen a resurgence in interest with the increased accessibility of machine learning models. Convolutional Neural Network (CNN)-based models have been developed to identify arbitrary mixtures of unstable nuclides from gamma spectra. In service of this, methods for the simulation and pre-processing of training data were also developed. The implementation of 1D multi-class, multi-label CNNs demonstrated good generalisation to real spectra with poor statistics and significant gain shifts. It is also shown that even basic CNN architectures prove reliable for RIID under the challenging conditions of heavy shielding and close source geometries, and may be extended to generalised solutions for pragmatic RIID.

Raman Laser Spectrometer: Application to 12C/13C Isotope Identification in CH4 and CO2 Greenhouse Gases

A compact Raman laser gas spectrometer is developed. It comprises a high-power green laser at 532.123 nm as an excitation source and a specially designed gas cell with an internal volume of less than 0.6 cm3 that can withstand gas pressures up to 100 atm. The resolution of the spectrometer is ~1 cm−1. The Raman spectra of chemically pure isotopically enriched carbon dioxide (12CO2, 13CO2) and methane (12CH4, 13CH4) gases are studied. The expected limit of detection (LOD) is less than 100 ppm for the isotopologues of CO2 and less than 25 ppm for those of CH4 (at a gas pressure of 50 atm.), making the developed spectrometer promising for studying the sources of emissions of greenhouse gases by resolving their isotopologue composition. We also show the suitability of the spectrometer for Raman spectroscopy of human exhalation.

Isotope Identification Mechanisms Enabled by Swept-Wavelength Raman Spectroscopy

Swept-wavelength Raman signatures have been measured for isotopic variants of polyethylene, acetic acid, and potassium sulfates. The swept-wavelength measurements produce two-dimensional Raman signatures which enable identification techniques based on changes in Raman peak amplitudes as a function of wavelength. In addition to the typical Raman peak energy shifts, which results from the change in isotope mass, three wavelength dependent mechanisms for isotope identification have been identified. Changes in the shape of the Raman signal, the presence and absence of Raman peaks over specific wavelength ranges, and changes in absorption of the Raman signal were observed as a result of isotopic substitution. These features provide additional specificity in the isotopic Raman signatures which suggests swept-wavelength Raman signatures will facilitate the identification of isotopes in complex and dirty mixtures. Measurements in the visible range suggest that the identification mechanisms are primarily evident in the ultraviolet, or resonance Raman, region.