Separation of Overlapping Processes from TGA Data and Verification by EGA

Abstract

Abstract A typical problem when analyzing the kinetics by dynamic TGA of polymer degradation consists in that different models can fit equally well, or the model does not fit the TGA data. It has been suggested that this is due to the presence of overlapping processes. High resolution TGA helps to clarify the problem in some cases, but in many other cases the problem remains unsolved. In this work, a dynamic TGA curve showing overlapping processes of degradation, typical of polymer thermal degradation, is fitted by a multiple logistic regression model. The regression function is a mixture of logistic functions where each single logistic function is weighed by a factor that represents the amount of the sample that takes place in the process represented by that function. A software program was developed to fit the data to the logistic mixture model by estimation of the parameter values. Additionally, the correlation of parameter estimates was performed by the S-Plus software. The regression was performed so that the best fitting shows the number of single decomposition processes that take place along the experiment, even if they are overlapping. Other parameters have the physical meaning of mass loss rate and position of the process in the time or temperature domain. The study of the correlations of parameter estimates gives a new insight on how each component fits a logistic model. The parameters obtained describe the kinetic of the processes in a form totally independent of the Arrhenius model. The separation of processes allows for analysis of each single process by classical methods. Although the model is mathematically supported, verification was done by comparing the results with the FTIR spectra obtained from the gases evolved along the experiment. The FTIR analysis of the gases evolved from a TGA experiment confirms that the logistic mixture model is suitable to analyze a dynamic TGA trace. It was also found that the logistic mixture model is suitable to separate single components from FTIR spectra. Some differences were found in the location of the components along the time axis, comparing DTG and FTIR data. In the FTIR, the separated peaks are broader and appear shifted to longer times. It can be attributed to the delay and mixing of components that may occur during the gas carriage from the TGA furnace to the FTIR detector.