Multiple exciton generation (MEG) is a process that can occur in semiconductor nanocrystals, or quantum dots (QDs), whereby absorption of a photon bearing at least twice the bandgap energy produces two or more electron-hole pairs. Here, we report on photocurrent enhancement arising from MEG in lead selenide (PbSe) QD-based solar cells, as manifested by an external quantum efficiency (the spectrally resolved ratio of collected charge carriers to incident photons) that peaked at 114 ± 1% in the best device measured. The associated internal quantum efficiency (corrected for reflection and absorption losses) was 130%. We compare our results with transient absorption measurements of MEG in isolated PbSe QDs and find reasonable agreement. Our findings demonstrate that MEG charge carriers can be collected in suitably designed QD solar cells, providing ample incentive to better understand MEG within isolated and coupled QDs as a research path to enhancing the efficiency of solar light harvesting technologies.

Organic-inorganic hybrid perovskites are attracting intense research effort due to their impressive performance in solar cells. While the carrier transport parameters such as mobility and bulk carrier lifetime shows sufficient characteristics, the surface recombination, which can have major impact on the solar cell performance, has not been studied. Here we measure surface recombination dynamics in CH3NH3PbBr3 perovskite single crystals using broadband transient reflectance spectroscopy. The surface recombination velocity is found to be 3.4±0.1 × 103 cm s−1, ∼2–3 orders of magnitude lower than that in many important unpassivated semiconductors employed in solar cells. Our result suggests that the planar grain size for the perovskite thin films should be larger than ∼30 μm to avoid the influence of surface recombination on the effective carrier lifetime.

We determine the internal quantum efficiency (IQE) of the active layer of PbSe nanocrystal (NC) back-contact Schottky solar cells by combining external quantum efficiency (EQE) and total reflectance measurements with an optical model of the device stack. The model is parametrized with the complex index of refraction of each layer in the stack as calculated from ellipsometry data. Good agreement between the experimental and modeled reflectance spectra permits a quantitative estimate of the fraction of incident light absorbed by the NC films at each wavelength, thereby yielding well-constrained QE spectra for photons absorbed only by the NCs. Using a series of devices fabricated from 5.1 ( 0.4 nm diameter PbSe NCs, we show that thin NC cells achieve an EQE and an active layer IQE as high as 60 ( 5% and 80 ( 7%, respectively, while the QE of devices with NC layers thicker than about 150 nm falls, particularly in the blue, because of progressively greater light absorption in the field-free region of the films and enhanced recombination overall. Our results demonstrate that interference effects must be taken into account in order to calculate accurate optical generation profiles and IQE spectra for these thin film solar cells. The mixed modeling/experimental approach described here is a rigorous and powerful way to determine if multiple exciton generation (MEG) photocurrent is collected by devices with EQE < 100%. On the basis of the magnitudes and shapes of the IQE spectra, we conclude that the 1,2-ethanedithiol treated NC devices studied here do not produce appreciable MEG photocurrent.Light absorption in a thin film solar cell can be greatly affected by optical interference if the device is optically thin and has a highly reflective back electrode. When these conditions are met, the buildup of an optical mode structure will change the absorption spectra of the active materials relative to their spectra when free of the optical cavity formed by the device stack. Exponential, Beer's Law type absorption/ generation profiles cannot then be assumed, and accurate values of the internal quantum efficiency (IQE) of the active layers of the device can be obtained only by determining their absorption spectra as they exist within the device structure.The IQE of a solar cell is normally calculated from the ratio of its external quantum efficiency (EQE) and spectral absorptance as IQE(λ) ) EQE(λ)/[1 -F(λ) -τ(λ)], where F(λ) is the spectral reflectance and the device transmittance τ(λ) is usually zero. However, this approach can give an underestimated and misleading IQE spectrum if incident light is absorbed by the electrodes and other nonactive layers of the device in addition to the active semiconductor layer(s).We show here that the quantum efficiency of photons absorbed only within the active layer of a PbSe nanocrystal back-contact Schottky solar cell 1 can be determined by combining reflectance measurements with an optical model capable of discerning the fraction of light absorbed in each of the layers within the ...

The electronic structure and related optical properties of an emerging thin-film photovoltaic material CH3NH3PbBr3 are studied. A block-shaped α-phase CH3NH3PbBr3 single crystal with the natural ⟨100⟩ surface is synthesized solvothermally. The room-temperature dielectric function ε = ε1 + iε2 spectrum of CH3NH3PbBr3 is determined by spectroscopic ellipsometry from 0.73 to 6.45 eV. Data are modeled with a series of Tauc-Lorentz oscillators, which show the absorption edge with a strong excitonic transition at ∼2.3 eV and several above-bandgap optical structures associated with the electronic interband transitions. The energy band structure and ε data of CH3NH3PbBr3 for the CH3NH3(+) molecules oriented in the ⟨111⟩ and ⟨100⟩ directions are obtained from first-principles calculations. The overall shape of ε data shows a qualitatively good agreement with experimental results. Electronic origins of major optical structures are discussed.

Keyword spotting (KWS) plays a critical role in enabling speech-based user interactions on smart devices. Recent developments in the field of deep learning have led to wide adoption of convolutional neural networks (CNNs) in KWS systems due to their exceptional accuracy and robustness. The main challenge faced by KWS systems is the trade-off between high accuracy and low latency. Unfortunately, there has been little quantitative analysis of the actual latency of KWS models on mobile devices. This is especially concerning since conventional convolution-based KWS approaches are known to require a large number of operations to attain an adequate level of performance.In this paper, we propose a temporal convolution for real-time KWS on mobile devices. Unlike most of the 2D convolution-based KWS approaches that require a deep architecture to fully capture both low-and high-frequency domains, we exploit temporal convolutions with a compact ResNet architecture. In Google Speech Command Dataset, we achieve more than 385x speedup on Google Pixel 1 and surpass the accuracy compared to the state-of-the-art model. In addition, we release the implementation of the proposed and the baseline models including an end-to-end pipeline for training models and evaluating them on mobile devices.