No abstract

Developing an efficient catalyst for H 2 generation from hydrolysis of ammonia borane (AB) in a controllable and sustainable way is a prerequisite to implementing H 2 as an alternative energy vector. To meet this requirement, the present work designs and constructs a potential Ag@Pd core−shell catalyst, in which the plasmonic Ag nanocube core acts as a light absorber and the ultrathin Pd shell as the actively catalytic site. First, the core−shell structure of the prepared Ag@Pd samples is characterized and confirmed by an electron microscope and energy dispersive spectroscopy. Then their catalytic performances for AB hydrolysis are investigated and compared with Ag nanocubes under the light on/off condition and at different temperatures. The results show that the H 2 generation rate on Ag@Pd is greatly enhanced, which is found to benefit from the reduced activation energy under light illumination. Further investigation of wavelength-dependent performance verifies the plasmondriven nature of the Ag@Pd catalysts. The subsequent optical simulation using the finite element method indicates that the absorbed light energy is instead preferential to dissipate into the nonplasmonic but catalytic Pd site when the Pd shell coats on the plasmonic Ag nanocube. This leads to the concentration of energetic charge carriers in the outer Pd shell and thus the much higher catalytic efficiency of Ag@Pd core−shell catalyst than its single compartment for breaking down the N−B bond of AB to produce H 2 gas.

A high loading of Mn(II)-metalated porphyrin was achievable in a 2D porphyrin-based Mn-MOF induced by an ionic liquid. The excellent stability, sufficient redox potential, atomically dispersed porphyrin Mn(II) sites, desired...

No abstract

Transformer architectures rely on explicit position encodings in order to preserve a notion of word order. In this paper, we argue that existing work does not fully utilize position information. For example, the initial proposal of a sinusoid embedding is fixed and not learnable. In this paper, we first review absolute position embeddings and existing methods for relative position embeddings. We then propose new techniques that encourage increased interaction between query, key and relative position embeddings in the self-attention mechanism. Our most promising approach is a generalization of the absolute position embedding, improving results on SQuAD1.1 compared to previous position embeddings approaches. In addition, we address the inductive property of whether a position embedding can be robust enough to handle long sequences. We demonstrate empirically that our relative position embedding method is reasonably generalized and robust from the inductive perspective. Finally, we show that our proposed method can be adopted as a near drop-in replacement for improving the accuracy of large models with a small computational budget.