Place: CALPHAD XLVII Conference, Juriquilla, Querétaro, México
Brandon Bocklund, Department of Materials Science and Engineering, Pennsylvania State University, University Park, USA
Richard Otis, Engineering and Science Directorate, California Institute of Technology, Pasadena, CA, USA
Zi-Kui Liu, Department of Materials Science and Engineering, The Pennsylvania State University, University Park, USA
Thermodynamics is the foundation of the integrated computational materials engineering (ICME) approach for materials design and has been widely used in applications such as diffusion, precipitation, and phase field modeling through the calculation of phase diagrams (CALPHAD) method. The development of large multicomponent CALPHAD databases is limited by the time required to evaluate the constituent binary and ternary systems which grow with the square and cube of the number of elements in the database, respectively. Once developed, these databases are challenging to maintain because changing the model or parameters of a constituent unary of binary system requires updating all the higher order systems that depend on it, which limits the development and adoption of new models into existing thermodynamic descriptions. Recently, the pycalphad software package  has been released, enabling the development of new models that are represented symbolically. This work presents ESPEI, an open-source Python-based software that uses pycalphad to rapidly develop and modify databases using a combination of first-principles and experimental thermochemical and phase equilibria data. Unlike traditional database development, ESPEI uses Markov chain Monte Carlo (MCMC) to optimize and quantify the uncertainty for all model parameters simultaneously. This presentation will demonstrate the approach for dynamically generating parameters from first-principles and experimental thermochemical data and optimizing the generated parameters using MCMC for several systems, including the Cu-Mg system. Finally, a perspective on the utility of uncertainty quantification for guiding the design of thermodynamic databases will be discussed.
 Otis, R. and Liu, Z.-K. pycalphad: CALPHAD based Computational Thermodynamics in Python, J. Open Res. Softw. 5 (2017)