Search for: All records

We present a method for designing spectrally- selective optoelectronic films with a finite absorption bandwidth. We demonstrate the process by designing a film composed of lead sulfide colloidal quantum dots (PbS-CQDs). Designs incorporate the patterning of absorbing PbS-CQD films into photonic crystal- like slabs which couple incident light into leaky modes within the plane of the absorbing films, modulating the absorption spectrum. Computational times required to calculate optical spectra are drastically decreased by implementing the Fourier Modal Method. Furthermore, a supervised machine-learning-based inverse design methodology is presented which allows tailoring of the PbS-CQD film optical properties for use in a variety of photovoltaic applications, such as tandem cells in which spectral tailoring can enable current-matching flexibility. 

The inverse design of photovoltaic 2D photonic crystals using machine learning will be presented. The technique bypasses calculation of photonic bandstructure in favor of directly computing designer-friendly properties such as spectral transmission. 

Abstract The morphology, chemical composition, and electronic uniformity of thin‐film solution‐processed optoelectronics are believed to greatly affect device performance. Although scanning probe microscopies can address variations on the micrometer scale, the field of view is still limited to well under the typical device area, as well as the size of extrinsic defects introduced during fabrication. Herein, a micrometer‐resolution 2D characterization method with millimeter‐scale field of view is demonstrated, which simultaneously collects photoluminescence spectra, photocurrent transients, and photovoltage transients. This high‐resolution morphology mapping is used to quantify the distribution and strength of the local optoelectronic property variations in colloidal quantum dot solar cells due to film defects, physical damage, and contaminants across nearly the entire test device area, and the extent to which these variations account for overall performance losses. It is found that macroscopic defects have effects that are confined to their localized areas, rarely prove fatal for device performance, and are largely not responsible for device shunting. Moreover, quantitative analysis based on statistical partitioning methods of such data is used to show how defect identification can be automated while identifying variations in underlying properties such as mobilities and recombination strengths and the mechanisms by which they govern device behavior.