The recent dramatic rise in power conversion efficiencies (PCEs) of perovskite solar cells (PSCs) has triggered intense research worldwide. However, high PCE values have often been reached with poor stability at an illuminated area of typically less than 0.1 square centimeter. We used heavily doped inorganic charge extraction layers in planar PSCs to achieve very rapid carrier extraction, even with 10- to 20-nanometer-thick layers, avoiding pinholes and eliminating local structural defects over large areas. The robust inorganic nature of the layers allowed for the fabrication of PSCs with an aperture area >1 square centimeter that have a PCE >15%, as certified by an accredited photovoltaic calibration laboratory. Hysteresis in the current-voltage characteristics was eliminated; the PSCs were stable, with >90% of the initial PCE remaining after 1000 hours of light soaking.

Although organic photovoltaic (OPV) cells have many advantages, their performance still lags far behind that of other photovoltaic platforms. A fundamental reason for their low performance is the low charge mobility of organic materials, leading to a limit on the active-layer thickness and efficient light absorption. In this work, guided by a semi-empirical model analysis and using the tandem cell strategy to overcome such issues, and taking advantage of the high diversity and easily tunable band structure of organic materials, a record and certified 17.29% power conversion efficiency for a two-terminal monolithic solution-processed tandem OPV is achieved.

Dye-sensitized solar cells (DSCs) using titanium dioxide (TiO 2 ) electrodes with different haze were investigated. It was found that the incident photon to current efficiency (IPCE) of DSCs increases with increase in the haze of the TiO 2 electrodes, especially in the near infrared wavelength region. Conversion efficiency of 11.1%, measured by a public test center, was achieved using high haze TiO 2 electrodes. This indicates that raising the haze of TiO 2 electrodes is an effective technique for improvement of conversion efficiency.

Click-through rate prediction is an essential task in industrial applications, such as online advertising. Recently deep learning based models have been proposed, which follow a similar Embed-ding&MLP paradigm. In these methods large scale sparse input features are first mapped into low dimensional embedding vectors, and then transformed into fixed-length vectors in a group-wise manner, finally concatenated together to fed into a multilayer perceptron (MLP) to learn the nonlinear relations among features. In this way, user features are compressed into a fixed-length representation vector, in regardless of what candidate ads are. The use of fixed-length vector will be a bottleneck, which brings difficulty for Embedding&MLP methods to capture user's diverse interests effectively from rich historical behaviors. In this paper, we propose a novel model: Deep Interest Network (DIN) which tackles this challenge by designing a local activation unit to adaptively learn the representation of user interests from historical behaviors with respect to a certain ad. This representation vector varies over different ads, improving the expressive ability of model greatly. Besides, we develop two techniques: mini-batch aware regularization and data adaptive activation function which can help training industrial deep networks with hundreds of millions of parameters. Experiments on two public datasets as well as an Alibaba real production dataset with over 2 billion samples demonstrate the effectiveness of proposed approaches, which achieve superior performance compared with state-of-the-art methods. DIN now has been successfully deployed in the online display advertising system in Alibaba, serving the main traffic.