Identification of volatiles from heated tobacco biomass using direct thermogravimetric analysis—Mass spectrometry and target factor analysis

• Published in: Thermochimica acta 2018-10, Vol.668, p.132-141
• Main Authors: Davies, Ashley, Grant Nicol, James Thomas, Liu, Chuan, Tetteh, John, McAdam, Kevin
• Format: Article
• Language: English
• Subjects: Heated biomass; Mass spectrometry (MS); Multivariate analysis; Target factor analysis (TFA); Thermogravimetric analysis (TGA); Volatiles
• ISSN: 0040-6031; 1872-762X
• DOI: 10.1016/j.tca.2018.08.007

•Simulation studies successful in testing limitations of a “deep search” strategy.•Scanning factor space beyond optimum number of factors revealed correlation trends.•Target components detectable in the experimental data were distinguished.•Some targets missed if TFA were based on conventional “optimum” number of factors.•Abundant and unique mass spectra are better distinguished using TFA on TGA-MS data. Target Factor Analysis (TFA) was used to identify full spectral profiles in the presence of complex overlapping signals in mass spectrometry data. This allowed interpretation of direct thermogravimetric analysis-mass spectrometry (TGA-MS) data from two varieties of tobacco biomass, heated below 350 °C. Previous TGA-MS approaches reported in biomass research use Selected Ion Monitoring (SIM) to monitor specific ions in the evolved gases. This cannot distinguish between isobaric mass fragments, and is challenged by the complex mixture of volatiles evolved from tobacco biomass. The TFA approach instead uses the complete reference mass spectra of the target compounds. Eighteen mass spectral references were used as target compounds to test their qualitative presence, based on correlations between the target spectra and the predicted spectra from the TFA process. Both simulated and experimental data sets were used to evaluate this approach. Empirical statistical analysis recommended 7 or 8 principal components for our data, but investigating predictive capabilities beyond these levels revealed correlation trends and enhanced analytical insights. Based on this multivariate analytical strategy, which we call a “deep search”, it was found that there was a distinction between identifiable and nonidentifiable spectral groupings for the 18 target references chosen. The nonidentifiable targets persistently scored below 0.55 correlation coeffecient (R), even when the search was based on two to three times the empirically recommended number of factors. The results show that components present in a complex data matrix, generated from TGA-MS experiments on complex biomass materials could be identified by full spectral matching using TFA instead of SIM.