BACKGROUND
Processes with reactive gas flows are of great technical importance for a sustainable energy and material cycle economy. In order to optimise these processes and develop new, sustainable and efficient methods, a profound understanding of the physico-chemical interactions involved is crucial. This requires non-intrusive in situ diagnostics with high temporal and spatial resolution and accuracy, with which the temperature and mole fractions of molecules can be quantitatively determined. Spontaneous Raman spectroscopy fulfils these requirements, but also poses challenges such as the low signal-to-noise ratio, the high dimensionality of the problem, the superposition of the Raman spectra of individual molecules, interference with other quantum mechanical effects and background radiation. In order to meet these challenges and enable quantitative analyses, existing evaluation procedures must be further developed and supplemented by new methods.
OBJECTIVE
The aim of the research work is to further develop conventional methods for analysing Raman spectra, such as spectral fitting and matrix inversion. These improvements should enable the robust application of the methods in new types of devices. These methods are based on spectral fingerprints, which can currently only be simulated quantum mechanically for relatively simple molecules such as N2, O2, H2, CO, CO2, H2O and N2O. A further aim of the research work is therefore to improve the derivation of these spectral fingerprints from experimental Raman spectra for molecules that cannot be simulated with sufficient accuracy. In addition, new process components based on modern machine learning methods are to be developed.
