Hierarchically porous, ultrathin N-doped carbon nanosheets embedded with highly dispersed cobalt nanoparticles as efficient sulfur host for stable lithium-sulfur batteries

The selective hydrogenolysis of the C–O bonds of lignin-derived aryl ethers into aromatics is challenging because it is always accompanied by hydrogenation (HYD). Most metal-supported catalysts tested so far exhibit high efficiency for the C–O bond cleavage of diphenyl ether (DPE, 4-O-5 linkage in lignin) but low selectivity toward valuable aromatic compounds. Here, we report our discovery showing the feasibility of controlling the selectivity of products by tuning the catalyst support and reaction conditions. Pt/γ-Al2O3 exhibited a higher selectivity (95.7%) for aromatic products (benzene and phenol) with a nearly 100% conversion at 160 °C when 2-propanol was used as a hydrogen source. In contrast, up to 99% selectivity for saturated aromatic products (cyclohexane and cyclohexanol) and the full conversion of DPE were yielded over Pt/TiO2 at 140 °C. The H2 evolution from 2-propanol dehydrogenation confirmed that the dehydrogenation activity of 2-propanol over Pt/γ-Al2O3 was lower than that over Pt/TiO2, which effectively suppressed the deep HYD of the formed aromatics. The role of supports and reaction active sites for 2-propanol dehydrogenation was studied by first-principles calculations. Based on the experimental results and ab initio molecular dynamics simulations, the detailed mechanisms of DPE hydrogenolysis over two Pt/oxide catalysts were proposed.

Pre-trained language models have achieved state-of-the-art results in various Natural Language Processing (NLP) tasks. has shown that scaling up pre-trained language models can further exploit their enormous potential. A unified framework named ERNIE 3.0 [2] was recently proposed for pre-training large-scale knowledge enhanced models and trained a model with 10 billion parameters. ERNIE 3.0 outperformed the state-of-the-art models on various NLP tasks. In order to explore the performance of scaling up ERNIE 3.0, we train a hundred-billion-parameter model called ERNIE 3.0 Titan with up to 260 billion parameters on the PaddlePaddle [3] platform. Furthermore, we design a self-supervised adversarial loss and a controllable language modeling loss to make ERNIE 3.0 Titan generate credible and controllable texts. To reduce the computation overhead and carbon emission, we propose an online distillation framework for ERNIE 3.0 Titan, where the teacher model will teach students and train itself simultaneously. ERNIE 3.0 Titan is the largest Chinese dense pre-trained model so far. Empirical results show that the ERNIE 3.0 Titan outperforms the state-of-the-art models on 68 NLP datasets.

No abstract

While ridged, spherical, or cone superhydrophobic surfaces have been extensively utilized to explore the droplet impact dynamics and the possibility of reducing contact time, superhydrophobic surfaces with a single small pillar have received less attention. Here, we report the rebound and splashing phenomena of impact droplets on various single-pillar superhydrophobic surfaces with the pillars having smaller or equal sizes compared to the droplets. Our results indicate that the single-pillar superhydrophobic surfaces inhibit the droplet splashing compared to the flat ones, and the rebound droplets on the former sequentially exhibit three morphologies of top, bottom, and breakup rebounds with the increasing of Weber number, while those on the latter only show the (bottom) rebound. The pillar significantly enlarges the droplet spreading factor but hardly changes the droplet width. Both the relations between the maximum spreading and width factors and the Weber number on all surfaces approximately follow a classical 1/4-power law. Reduction in the contact time is observed for the rebound droplets on the single-pillar superhydrophobic surfaces, dependent on the rebound morphology. Specially, the breakup rebound nearly shortens the contact time by more than 50% with a larger pillar-to-droplet diameter ratio yielding a greater reduction. We provide scaling analyses to demonstrate that this remarkable reduction is ascribed to the decrease in the volume of each sub-droplet after breakup. Our experimental investigation and theoretical analysis provide insight into the droplet impact dynamics on single-pillar superhydrophobic surfaces.