Preliminary tests to optimize compound specific isotopic methods for source attribution of levoglucosan and PAHs in airborne particulate matter

This thesis develops and critically evaluates compound-specific stable carbon isotope analysis workflows for levoglucosan and selected polycyclic aromatic hydrocarbons in airborne particulate matter, with the aim of improving source attribution in complex atmospheric matrices. The analytical strategy combines GC–MS screening and method optimization with GC–C–IRMS measurements and integrates meteorological information and air mass classification to interpret the results obtained for real samples. For levoglucosan, extraction, purification, and derivatization steps were assessed through recovery and repeatability experiments. Derivatization to triacetyllevoglucosan was investigated by varying temperature, reaction time, and pyridine volume, showing that reaction temperature had the clearest effect on derivatization yield. Tests performed with reference materials demonstrated that the workflow could achieve low derivatization bias variability. These results indicate that acetylation is a promising derivatization strategy for levoglucosan CSIA, although its quantitative robustness still requires further validation through reagent isotope characterization and mass-balance correction. Instead, for ambient PM samples, levoglucosan CSIA was frequently affected by coelution and high background. The tested deconvolution approach did not provide a demonstrable improvement over standard integration within the available dataset. Therefore, levoglucosan δ13C results were interpreted conservatively, recognizing that the uncertainty cannot be robustly estimated under the current analytical conditions. To better quantify the uncertainty, the influence of the interferent peaks should be better characterized. For PAHs, extraction and cleanup protocols achieved acceptable recoveries for low molecular weight compounds in test samples. High molecular weight compounds were more difficult to measure and showed higher dispersion. However, PAH CSIA in real samples was not feasible because target analytes did not reach the required intensity for GC–C–IRMS, indicating that higher on column mass and deeper purification are necessary for routine application. The results also show that mild purification based on silica gel columns, even when applied to relatively clean matrices such as PM extracts, is not sufficient to remove matrix-related interferences. Coelution with non-target compounds and elevated chromatographic baselines prevented reliable PAH δ13C determination, even when the presence of target compounds was confirmed by GC–MS, most often at low concentration levels. 5 Meteorological analysis was implemented on the continuous time series across the sampling campaign rather than on short regime windows, because meteorology is hourly while isotope data are daily or discrete, and the number of isotope observations per window is often limited. At the daily scale, no consistent dependence of levoglucosan δ13C or of the difference of levoglucosan δ13C on corresponding days PM2.5 and PM10 samples (Δδ) on wind speed or wind direction was observed, suggesting no significant change of levoglucosan origin with changing meteorology. Overall, this study shows that CSIA performance in real PM samples is primarily controlled by chromatographic peak resolution and isotope biases due to contamination and fractionation processes. It also establishes the basis for a dedicated method for levoglucosan δ13C measurement in PM samples by identifying acetylation as a promising derivatization route and by defining the main analytical challenges that must be overcome before routine application. Practical pathways to improve robustness include enhanced cleanup, increased sampled mass, routine use of certified isotope reference materials, reagent isotope characterization, and a more systematic assessment of how co-eluting peaks and baseline interferences affect the reliability of δ13C values.