PWLU: Learning Specialized Activation Functions With the Piecewise Linear Unit.

The choice of activation functions is crucial to deep neural networks. ReLU is a popular hand-designed activation function. Swish, the automatically searched activation function, outperforms ReLU on many challenging datasets. However, the search method has two main drawbacks. First, the tree-based search space is highly discrete and restricted, which is difficult to search. Second, the sample-based search method is inefficient in finding specialized activation functions for each dataset or neural architecture. To overcome these drawbacks, we propose a new activation function called Piecewise Linear Unit (PWLU), incorporating a carefully designed formulation and learning method. PWLU can learn specialized activation functions for different models, layers, or channels. Besides, we propose a non-uniform version of PWLU, which maintains sufficient flexibility but requires fewer intervals and parameters. Additionally, we generalize PWLU to three-dimensional space to define a piecewise linear surface named 2D-PWLU, which can be treated as a non-linear binary operator. Experimental results show that PWLU achieves SOTA performance on various tasks and models, and 2D-PWLU is better than element-wise addition when aggregating features from different branches. The proposed PWLU and its variation are easy to implement and efficient for inference, which can be widely applied in real-world applications.

2D PbS Nanosheets with Zigzag Edges for Efficient CO2 Photoconversion.

The rational design of transition-metal sulfide with two-dimensional (2D) structure and tunable edges on the nanoscale can effectively improve their activity for variously catalytic reactions. Herein, the 2D PbS nanosheets with abundant zigzag edges (e-PbS NS), which exhibited an excellent performance for CO2 photoconversion to CO, were constructed. The zigzag edges on the PbS NS are beneficial for exposing more active sites and promoting charge separation, thereby accelerating the kinetics process of CO2 photoreduction. This study provides a strategy to regulate structure with effective edge sites for the CO2 reduction.

In Situ Topotactic Transformation of an Interstitial Alloy for CO Electroreduction.

Electrochemical reduction of CO to value-added products holds promise for storage of energy from renewable sources. Copper can convert CO into multi-carbon (C2+ ) products during CO electroreduction. However, developing a Cu electrocatalyst with a high selectivity for CO reduction and desirable production rates for C2+ products remains challenging. Herein, highly lattice-disordered Cu3 N with abundant twin structures as a precursor electrocatalyst is examined for CO reduction. Through in situ activation during the CO reduction reaction (CORR) and concomitant release of nitrogen, the obtained metallic Cu° catalyst particles inherit the lattice dislocations present in the parent Cu3 N lattice. The de-nitrified catalyst delivers an unprecedented C2+ Faradaic efficiency of over 90% at a current density of 727 mA cm-2 in a flow cell system. Using a membrane electrode assembly (MEA) electrolyzer with a solid-state electrolyte (SSE), a 17.4 vol% ethylene stream and liquid streams with concentration of 1.45 m and 230 × 10-3 m C2+ products at the outlet of the cathode and SSE-containment layer are obtained.

Growth of boron nitride nanotubes from magnesium diboride catalysts.

The difficulty in synthesizing boron nitride nanotubes (BNNTs) in a conventional horizontal tube furnace by chemical vapor deposition (CVD) may be ascribed to the failure to identify suitable catalysts and nucleation particles. This report demonstrates that magnesium diboride (MgB2) can effectively catalyze the growth of BNNTs in such a tube furnace from various boron sources, including boron oxide (B2O3), boric acid (H3BO3), and a mixture of boron (B) and calcium oxide (CaO). This catalyst is more efficient than the possible magnesium oxide (MgO) or magnesium nitride (Mg3N2) catalysts. MgB2 efficiently catalyzes the formation of BNNTs by maintaining a liquid state and showing a dissolving capacity for B2O3 at the growth temperature, thus satisfying the criteria for the vapor-liquid-solid (VLS) mechanisms of one-dimensional nanomaterials. First-principles simulations demonstrate that B2O3 can be dissolved into the MgB2 nanoparticle. We believe that the strong catalytic behavior of MgB2 can be attributed to its robust nucleation for BNNTs and dissolubility for B2O3.

Microwave-assisted fabrication of porous hematite photoanodes for efficient solar water splitting.

Porous hematite photoanodes with wormlike networks were prepared by microwave-assisted rapid annealing of ß-FeOOH grown directly on a FTO glass. The enhanced electrochemical characteristics such as large surface area and good permeation of an electrolyte lead to the excellent photoelectrochemical performance toward the solar-driven water splitting for hydrogen production.