Correct application of some commonly used methods in the analysis of thermally induced processes in glass materials

Abstract

In this paper, it is pointed out to the incorrect application of the linear least squares regression method in articles dealing with the thermal characteristics of glass. On selected examples, linear least squares regression method was adequately applied to linear dependences predicted by Kissinger and Mahadevan/Moynihan equations in order to determine the activation energy of crystallization Ec and glass transition Eg of chalcogenide and oxide glasses. It is proposed to replace the roles of explanatory and response variables, i.e. to draw the dependence: 1Tx=f(ln(βTx2)), or 1Tx=f(ln(β)) where x≡c for crystallization and x≡g for glass transition. Modified equations give statistically more accurate values.The analysis of measurement uncertainty of variability parameter ΔEwas also performed in accordance with the rules of error propagation, indicating that with this parameter measurement uncertainty can make more than 50% of the value of the parameter itself, thus making conclusions and further analysis very questionable.