The thin-fi lm photovoltaic material Cu 2 ZnSnS x Se 4-x (CZTSSe) has drawn world-wide attention due to its outstanding performance and earth-abundant composition. Until recently, [ 1 ] stateof-the-art CZTSSe thin-fi lm solar cells were limited to 11.1% power conversion effi ciency (PCE), with these performance levels being achieved via a hydrazine slurry approach. [ 2 ] Other vacuum-and non-vacuum-based deposition techniques have also been successful in fabricating CZTSSe solar cells with PCE above 8%. [ 3,4 ] However, even record devices with PCE of 11% are still far below the physical limit, known as the ShockleyQueisser (SQ) limit, of about 31% effi ciency under terrestrial conditions. [ 5 ] For a solar cell with 1.13 eV bandgap such as the previous 11.1% champion, [ 2 ] the SQ limits for open circuit voltage ( V oc ) and short-circuit current density ( J sc ) are 820 mV and 43.4 mA cm −2 , respectively. The previous 11.1% champion only achieved a V oc of 460 mV and a J sc of 34.5 mA cm −2 , corresponding to about 56% and 79% of the SQ limit values. In order to boost J sc , an optical architecture with optimized transparent conductive oxide (TCO) and CdS thicknesses has recently been reported, leading to a new CZTSSe record PCE of 12.0% and a J sc that reaches 83% of the SQ limit. [ 1 ] Despite improvements in shortcircuit current, the V oc defi cit, equal to the difference between the bandgap and V oc , is currently the biggest hurdle preventing CZTSSe devices from achieving higher effi ciency. [ 6 ] Enhancement of V oc also directly improves device fi ll factor. [ 7 ] Although many factors can infl uence V oc in a solar cell, carrier generation and recombination near the charge-separating junction play a dominant role. Thus, in order to decrease the V oc defi cit and increase effi ciency beyond 12%, it is critical to understand junction characteristics, current collection, and recombination mechanisms in the current generation of devices.Here, an independently certifi ed world-record 12.6% PCE CZTSSe thin-fi lm solar cell is presented. The new champion device was fabricated using a recently described hydrazine pure-solution approach, targeting a Cu-poor and Zn-rich condition. [ 8 ] Secondary ion mass spectrometry (SIMS) shows that the obtained CZTSSe fi lms exhibit very low carbon and oxygen concentrations, comparable to fi lms fabricated by the more traditional hydrazine-slurry method. [ 9 ] The rheological properties of the particle-free solution, relative to the slurry process, signifi cantly improves the coating uniformity and fi lm structure and, consequently, the performance of the solar cells. [ 4,8 ] By simultaneously optimizing the TCO and CdS thicknesses to maximize photon transmission to the absorber and improving the bulk qualities of CZTSSe with the hydrazine pure-solution approach, both J sc and V oc are boosted in the 12.6% champion device. Device characteristics of the new champion cell, as deduced from current-voltage, quantum effi ciency, capacitance, and electron-beam-induced current (EBI...
We demonstrate that a fundamental performance bottleneck for hydrazine processed kesterite Cu2ZnSn(S,Se)4 (CZTSSe) solar cells with efficiencies reaching above 11% can be the formation of band-edge tail states, which quantum efficiency and photoluminescence data indicate is roughly twice as severe as in higher-performing Cu(In,Ga)(S,Se)2 devices. Low temperature time-resolved photoluminescence data suggest that the enhanced tailing arises primarily from electrostatic potential fluctuations induced by strong compensation and facilitated by a lower CZTSSe dielectric constant. We discuss the implications of the band tails for the voltage deficit in these devices.
A power conversion efficiency record of 10.1% was achieved for kesterite absorbers, using a Cu2ZnSn(Se,S)4 thin‐film solar cell made by hydrazine‐based solution processing. Key device characteristics were compiled, including light/dark J–V, quantum efficiency, temperature dependence of Voc and series resistance, photoluminescence, and capacitance spectroscopy, providing important insight into how the devices compare with high‐performance Cu(In,Ga)Se2. The record kesterite device was shown to be primarily limited by interface recombination, minority carrier lifetime, and series resistance. The new level of device performance points to the significant promise of the kesterites as an emerging and commercially interesting thin‐film technology. Copyright © 2011 John Wiley & Sons, Ltd.
In recent years, deep neural networks (DNN) have demonstrated significant business impact in large scale analysis and classification tasks such as speech recognition, visual object detection, pattern extraction, etc. Training of large DNNs, however, is universally considered as time consuming and computationally intensive task that demands datacenter-scale computational resources recruited for many days. Here we propose a concept of resistive processing unit (RPU) devices that can potentially accelerate DNN training by orders of magnitude while using much less power. The proposed RPU device can store and update the weight values locally thus minimizing data movement during training and allowing to fully exploit the locality and the parallelism of the training algorithm. We evaluate the effect of various RPU device features/non-idealities and system parameters on performance in order to derive the device and system level specifications for implementation of an accelerator chip for DNN training in a realistic CMOS-compatible technology. For large DNNs with about 1 billion weights this massively parallel RPU architecture can achieve acceleration factors of 30, 000 × compared to state-of-the-art microprocessors while providing power efficiency of 84, 000 GigaOps∕s∕W. Problems that currently require days of training on a datacenter-size cluster with thousands of machines can be addressed within hours on a single RPU accelerator. A system consisting of a cluster of RPU accelerators will be able to tackle Big Data problems with trillions of parameters that is impossible to address today like, for example, natural speech recognition and translation between all world languages, real-time analytics on large streams of business and scientific data, integration, and analysis of multimodal sensory data flows from a massive number of IoT (Internet of Things) sensors.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.