Raman microscopy studies of carbon particles from diesel particulate matter (DPM) and coal dust

Diesel Particulate Matter (DPM) and coal dust samples were characterised using Raman microscopy, X-ray Photoelectron Spectroscopy (XPS), Diffuse Reflectance Infrared Fourier Transform Spectroscopy (DRIFTS), Thermo-Gravimetric Analysis (TGA), X-ray Fluorescence (XRF) spectrometry and Scanning electron Microscopy. The sp2/sp3 carbon bonding ratios for DPM and coal dust were determined as 6.1 and 0.7, respectively, from XPS. Principal Component Analysis (PCA) was successfully implemented as a tool for distinguishing between the very similar DPM and coal dust Raman spectra, with over 99% of the variance contained in the first principal component. DPM and coal dust mixtures with known compositions were produced. Raman instrumental parameters were systematically optimised by varying the objective lenses, acquisition times and laser powers, to improve spectral and obtain the most reproducible integrated spectral areas. A rotation stage was developed and employed to spin the specimens during analysis, resulting in a larger sampling area. This resulted in a more representative sampling regime for the heterogeneous specimens and a considerable improvement in the reproducibility of integrated spectral areas. The error in the integrated spectral areas of 10 replicate spectra of different mixtures ranged from 5-22% before implementation of the rotating stage and was notably reduced to 2 -6% due to the action of spinning. Raman spectra of mixtures were used to construct a Partial Least Squares (PLS) model. The R2 values for the DPM and coal dust were 0.865 and 0.763, respectively. The differential bum-off of volatile organics during the Raman analysis due to localised heating from the laser hindered the ability to gain highly reproducible spectra and thus markedly affected the PLS model. A method development stage aimed at improving the R2 values was applied to the samples. This involved heat-treating the specimens to 625°C in an inert nitrogen atmosphere, before the Raman analysis. The resultant PLS model, after heat-treatment, dramatically improved the R2 values such that the DPM and coal dust were 0.974 and 0.907, respectively. This model was used to predict the composition of a test sample with known amounts of DPM and coal dust. The concentrations predicted by the model were 166 ± 3.9pg for the DPM and 68 ± 7.8jxg for the coal dust. The model slightly overestimated the amount of DPM present in the sample but gave a large underestimation of the coal dust content. The diagnostics of the model were investigated and recommendations for the improvement of future models were given.