Lea Zimmermann | Searcharxiv

arXiv2024

Multimodal operando microscopy reveals that interfacial chemistry and nanoscale performance disorder dictate perovskite solar cell stability

Next-generation low-cost semiconductors such as halide perovskites exhibit optoelectronic properties dominated by nanoscale variations in their structure, composition and photophysics. While microscopy provides a proxy for ultimate device function, past works have focused on neat thin-films on insulating substrates, missing crucial information about charge extraction losses and recombination losses introduced by transport layers. Here we use a multimodal operando microscopy toolkit to measure nanoscale current-voltage curves, recombination losses and chemical composition in an array of state-of-the-art perovskite solar cells before and after extended operational stress. We apply this toolkit to the same scan areas before and after extended operation to reveal that devices with the highest performance have the lowest initial performance spatial heterogeneity - a crucial link that is missed in conventional microscopy. We find that subtle compositional engineering of the perovskite has surprising effects on local disorder and resilience to operational stress. Minimising variations in local efficiency, rather than compositional disorder, is predictive of improved performance and stability. Modulating the interfaces with different contact layers or passivation treatments can increase initial performance but can also lead to dramatic nanoscale, interface-dominated degradation even in the presence of local performance homogeneity, inducing spatially varying transport, recombination, and electrical losses. These operando measurements of full devices act as screenable diagnostic tools, uniquely unveiling the microscopic mechanistic origins of device performance losses and degradation in an array of halide perovskite devices and treatments. This information in turn reveals guidelines for future improvements to both performance and stability.

physics.chem-ph↗

arXiv2023

Free-form Flows: Make Any Architecture a Normalizing Flow

Normalizing Flows are generative models that directly maximize the likelihood. Previously, the design of normalizing flows was largely constrained by the need for analytical invertibility. We overcome this constraint by a training procedure that uses an efficient estimator for the gradient of the change of variables formula. This enables any dimension-preserving neural network to serve as a generative model through maximum likelihood training. Our approach allows placing the emphasis on tailoring inductive biases precisely to the task at hand. Specifically, we achieve excellent results in molecule generation benchmarks utilizing $E(n)$-equivariant networks. Moreover, our method is competitive in an inverse problem benchmark, while employing off-the-shelf ResNet architectures.

cs.LG↗

arXiv2023

Lifting Architectural Constraints of Injective Flows

Normalizing Flows explicitly maximize a full-dimensional likelihood on the training data. However, real data is typically only supported on a lower-dimensional manifold leading the model to expend significant compute on modeling noise. Injective Flows fix this by jointly learning a manifold and the distribution on it. So far, they have been limited by restrictive architectures and/or high computational cost. We lift both constraints by a new efficient estimator for the maximum likelihood loss, compatible with free-form bottleneck architectures. We further show that naively learning both the data manifold and the distribution on it can lead to divergent solutions, and use this insight to motivate a stable maximum likelihood training objective. We perform extensive experiments on toy, tabular and image data, demonstrating the competitive performance of the resulting model.

cs.LG↗