Extended Regression Analysis for Debye-Einstein Models Describing Low Temperature Heat Capacity Data of Solids

• Published in: Entropy (Basel, Switzerland) Vol. 26; no. 6; p. 452
• Main Authors: Gamsjäger, Ernst, Wiessner, Manfred
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI AG 26.05.2024; MDPI
• Subjects: Bayesian framework; Enthalpy; Entropy; Errors; Global optimization; Heat; Local optimization; Low temperature; Markov chain Monte Carlo (MCMC); Markov chains; Parameters; Probability; probability density distribution; Probability distribution; Regression analysis; Specific heat; Temperature; thermodynamic functions; Thermodynamics; thermodynamic functions; probability density distribution; Bayesian framework; regression analysis; Markov chain Monte Carlo (MCMC)
• ISSN: 1099-4300; 1099-4300
• DOI: 10.3390/e26060452

Heat capacity data of many crystalline solids can be described in a physically sound manner by Debye-Einstein integrals in the temperature range from 0K to 300K. The parameters of the Debye-Einstein approach are either obtained by a Markov chain Monte Carlo (MCMC) global optimization method or by a Levenberg-Marquardt (LM) local optimization routine. In the case of the MCMC approach the model parameters and the coefficients of a function describing the residuals of the measurement points are simultaneously optimized. Thereby, the Bayesian credible interval for the heat capacity function is obtained. Although both regression tools (LM and MCMC) are completely different approaches, not only the values of the Debye-Einstein parameters, but also their standard errors appear to be similar. The calculated model parameters and their associated standard errors are then used to derive the enthalpy, entropy and Gibbs energy as functions of temperature. By direct insertion of the MCMC parameters of all 4·105 computer runs the distributions of the integral quantities enthalpy, entropy and Gibbs energy are determined.