Tailoring capping-layer composition for improved stability of mixed-halide perovskites

Abstract

Incorporating a low dimensional (LD) perovskite capping layer on top of perovskite absorber, improves the stability of perovskite solar cells (PSCs). However, in the case of mixed-halide perovskites, which can...

Machine-Learning-Assisted Screening of Interface Passivation Materials for Perovskite Solar Cells

“…Machine learning (ML) has been proposed in recent studies to assist in material screening with minimal experimental input data. Hartono et al , systematically studied the impact of various molecular features on the stability of perovskite thin films with 2D capping layers in a humid environment. A good materials screening approach was established by machine learning for the 2D perovskite layer as a humidity barrier.…”

“…A hydrogen bond donor greater than 1 represents the electronegative atoms (N, O, and F) that can form hydrogen bonds with the MA and FA cations in the perovskite. These hydrogen bond donors increase the reactivity of the ammonium salt with the 3D perovskite, 31 leading to excessive 2D-toward transformation. 58,59 The effects of concentration agree with our understanding that a low concentration means insufficient passivation and a high concentration leads to a thick passivation layer that hinders carrier transport.…”

“…Several popular regression algorithms have been used to build an effective prediction model, such as random forest (RF) regression, gradient boosting (GB) regression, support vector machine (SVM) regression, multilayer perceptron (MLP) regression, and Gaussian process (GP) regression. − ,− In addition to individual regression models, we also proposed an ensemble learning approach to integrate five individual ML regressors. The ensemble learning model combines the outputs from all regression models in order to eliminate model bias and combine the advantages of regression with different data-processing architectures (e.g., tree-based RF model and kernel-based SVM model), which reduces the possibility of overfitting in our case of a small training data set.…”

“…After the ML ensemble model was established, feature importance analysis become particularly helpful for identifying the most influential input variables. ,,, The contribution to the PCE improvement ratio is presented by the SHAP values, as shown in Figure a. The 13 input variables (i.e., 12 material descriptors and precursor concentration) are ranked based on their overall absolute impact on efficiency, regardless of the positive or negative contributions.…”

Cage Molecules Stabilize Lead Halide Perovskite Thin Films

“…Nevertheless, under the in situ testing conditions applied in this study, the POC additive method shows the potential to outperform the state-of-the-art composition engineering method that forms MA, formamidinium (FA), and cesium mixed-cation perovskites (17× stability improvement relative to MAPbI 3 ) 34 and the capping method that uses phenyltriethylammonium (PTEA) iodide to form 2D/3D heterostructured perovskites (4−6× stability improvement relative to MAPbI 3 ), as reported in our previous studies. 36,37 Additional stability tests for a list of small molecular additives and other POCs, including RCC1, RCC1-HBr, and RCC1-HBr, at low solar-cell-relevant concentrations are shown in Figures S5−7, respectively. We believe that the enormous structural, chemical and geometrical diversity of the POC family provides a rich platform for designing and optimizing additives for stabilizing perovskite thin films in the future.…”