Bayesim: A tool for adaptive grid model fitting with Bayesian inference

Abstract

Bayesian inference is a widely used and powerful analytical technique in fields such as astronomy and particle physics but has historically been underutilized in some other disciplines including semiconductor devices. In this work, we introduce bayesim, a Python package that utilizes adaptive grid sampling to efficiently generate a probability distribution over multiple input parameters to a forward model using a collection of experimental measurements. We discuss the implementation choices made in the code, showcase two examples in photovoltaics, and discuss general prerequisites for the approach to apply to other systems. Program summaryProgram Title: BayesimProgram Files doi:http://dx.doi.org/10.17632/fch5m6p3nn.1Licensing provisions: GNU General Public License 2Programming language: Python 3Supplementary material: uncertainty_figure.pngNature of problem: Many parameters in forward numerical models, e.g. for photovoltaic (PV) device behavior, are difficult to measure via direct experiment. In addition, in early-stage PV materials (and many other systems of interest), there are many unknown parameters and a desire to know their values in a small amount of time to enable high-throughput materials screening, making the time investment for direct measurement prohibitive. However, measurement of electrical properties of devices is comparatively easy, cheap, and automatable.Solution method: We employ Bayesian inference to “invert” the device models using the high-throughput experimental data, running the model forward with many combinations of parameters and generating a probability distribution over the inputs. This enables fitting of a number of parameters limited only by the quantity of experimental data and computational power available. The code is available open-source as a Python package and includes features such as adaptive grid fitting, model uncertainty calculation, and a variety of visualization options.Additional comments: The code is freely available on Github (https://github.com/pv-lab/bayesim) and thoroughly documented online ( https://pv-lab.github.io/bayesim/_build/html/index.html).