An open-source environmental chamber for materials-stability testing using an optical proxy

Abstract

This study is motivated by the desire to disseminate a low-cost, high-precision, high-throughput environmental chamber to test materials and devices under elevated humidity, temperature, and light. This paper documents the...

Vision-driven Autocharacterization of Perovskite Semiconductors

“…Furthermore, the process of band gap curve fitting using inclination angles developed by Escobedo-Morales et al [4] produces numerical inconsistencies and is removed in our implementation in favor of a simpler and more robust version of RMSE minimization. Keesey & Tiihonen et al [7] develop a procedure of optically extracting a figure of merit to quantify the stability of perovskites but this methodology requires the manual segmentation of samples one by one, which becomes infeasible for a human researcher when the number of samples increases significantly. In this paper, we build from these prior works by demonstrating the performance of three autocharacterization methods that are scalable to arbitrarily many samples using a computer vision-driven approach.…”

“…In the perovskite semiconductor system, the performance metrics of band gap and stability are largely governed by the chemical composition of the material [23,7]. However, since the 80 samples are synthesized using the high-throughput method described in Section 2.1, the exact composition of each unique deposit (the value of x) is not known a priori.…”

“…To determine stability, the perovskite semiconductors were subject to an accelerated degradation chamber, and the three-channel RGB approximation of the color change of the sample was recorded as a proxy to sample stability. [7] The chamber, pictured in Figure 5, maintains a uniform temperature, illumination, and humidity level around the samples, and obtains real-time images of the perovskite semiconductors every time step ∆t for a duration of time T resulting in Ω(∆t) images. [7] The vision segmentation algorithm described in Algorithm 1 is applied to the first droplet image, Ω, to output the segmented locations and reflectance data, Φ, of each spatiallynon uniform sample across the duration of the experiment with a fixed camera position.…”

High-Throughput Multi-Scale Experiments to Probe the Stability of Metal Halide Perovskites

“…For the optical images, conventional image processing and arithmetical algorithms would be enough. Figure 2a shows the different workflows for handling experimental results obtained from MHP thin-film samples and single-crystal samples, as proposed by Keesey et al 3 For thin-film samples, the sum and average values of overall pixels are calculated for all three channels separately, while for single-crystal samples, a preprocessing operation would automatically segment the region of interest (ROI) and calculate the average RGB values on the ROI only. For a pure α-phase MHP sample with dark black color, all the RGB channels would have equally low values, or a relatively larger R channel value for Br-rich MHP.…”

“…An electric stage enables automatic sample changing during the test. (b) The high-dimension time-series data processing with simple discrete differential algorithms or ML methods. , The top figure was reproduced with permission from ref . Copyright 2022 Royal Society of Chemistry.…”