Search for: All records

Environmental transmission electron microscopy probes the local structure, composition, and chemistry of materials under gas environments, while ambient-pressure X-ray photoelectron spectroscopy provides ensemble chemical and electronic structure information in gaseous conditions. Both techniques utilize similar differential pumping schemes to mitigate electron scattering by the gas phase, allowing for unique opportunities to correlate gas–surface interactions across comparable pressure ranges. Their integration has advanced the understanding of various catalytic reactions, including the water–gas-shift reaction, CO oxidation, and surface passivation dynamics. This Mini-Review discusses their methodological advancements, challenges, and potential for further integration with other in situ techniques to address complex catalytic phenomena and guide catalyst design. 

With the rapid growth of Internet of Vehicles (IoV) applications and the advancement of edge computing, resource-limited vehicles (and other IoT devices) increasingly rely on external servers to handle diverse and complex computational tasks. However, this dependence on external servers, which may be malicious or compromised, introduces significant security risks. Replication-based verifiable computing has been proposed as a solution to verify the accuracy of task results, but these approaches are vulnerable to collusion, where compromised servers return identical incorrect results to mislead the vehicle. Existing defenses against collusion either cannot ensure complete protection or become ineffective as the number of colluding servers rises. In this paper, we introduce CoVFeFE, a collusion-resilient verification framework designed to detect and mitigate collusion, even when the majority of servers are compromised. Our framework integrates a rapid detection mechanism that monitors computational conflicts, alongside a heuristic mitigation strategy that identifies and neutralizes colluding servers. Simulation results demonstrate that CoVFeFE outperforms existing solutions by successfully identifying all colluding servers, even when they constitute a majority > 50%) of the network. 

Using in situ synchrotron X-ray diffraction, we interrogate the microstructural and phase evolution of polycrystalline nickel (Ni) during redox cycling in O2, H2, and H2O environments. Oxidation in O2 promotes strong (111) texturing in both the NiO overlayer and the underlying Ni substrate. However, this crystallographic alignment is lost following reduction in H2 and subsequent reoxidation, demonstrating irreversible microstructural changes. H2 exposure leads to proton dissolution into the Ni lattice, triggering a localized phase transition from face-centered cubic (FCC) to hexagonal close-packed (HCP) Ni in hydrogen-saturated regions. In H2O-containing atmospheres, dissociative H2O adsorption produces protons that permeate the NiO layer, forming γ-NiOOH within the NiO lattice and HCP Ni beneath the NiO overlayer as protons accumulate. Kinetic analysis via the Johnson-Mehl-Avrami–Kolmogorov model uncovers distinct growth mechanisms: preoxidized Ni surfaces follow one-dimensional (1D) kinetics for NiO, γ-NiOOH, and HCP growth, whereas pristine Ni exhibits three-dimensional (3D) kinetics due to island-like nucleation and growth of NiO. These results highlight the critical interplay between H2O dissociation, hydrogen permeation, and redox-driven phase transformations, with practical implications in engineering nickel-based catalysts and hydrogen storage systems through controlled microstructural and phase evolution. 

Understanding oxide reduction is critical for advancing metal production, catalysis and energy technologies. Although carbon monoxide (CO) and hydrogen (H2) are widely used reductants, the mechanisms by which they work are often presumed to be similar, both involving lattice oxygen removal. However, because of growing interest in replacing CO with H2 to lower CO2 emissions, distinguishing gas-specific reduction pathways is critical. Yet, capturing these atomic-scale processes under reactive gas and high-temperature conditions remains challenging. Here we use environmental transmission electron microscopy, which is capable of real-time, atomic-resolution imaging of gas–solid redox reactions10,11,12,13,14,15,16, to directly visualize the gas-dependent oxide reduction dynamics in NiO. We show that CO drives surface nucleation and the growth of metallic Ni islands, leading to self-limiting surface metallization. Conversely, H2 activates a coupled surface-to-bulk transformation, where protons from dissociated H2 infiltrate the oxide lattice to promote the inward migration of surface-generated oxygen vacancies and enabling bulk metallization. By contrast, oxygen vacancies formed by CO remain confined near the surface, where they rapidly form a metallic Ni layer that inhibits further reduction. These results reveal distinct atomistic pathways for CO and H2 and provide insights that may guide metallurgical processes and catalyst design. 

Prompting has shown impressive success in enabling large pre-trained language models (LMs) to perform diverse NLP tasks, especially with only few downstream data. Automatically finding the optimal prompt for each task, however, is challenging. Most existing work resorts to tuning *soft* prompts (e.g., embeddings) which fall short of interpretability, reusability across LMs, and applicability when gradients are not accessible. *Discrete* prompts, on the other hand, are difficult to optimize, and are often created by “enumeration (e.g., paraphrasing)-then-selection” heuristics that do not explore the prompt space systematically. This paper proposes RLPrompt, an efficient discrete prompt optimization approach with reinforcement learning (RL). RLPrompt formulates a parameter-efficient policy network that generates the optimized discrete prompt after training with reward. To harness the complex and stochastic reward signals from the large LM environment, we incorporate effective reward stabilization that substantially enhances training efficiency. RLPrompt is flexibly applicable to different types of LMs, such as masked (e.g., BERT) and left-to-right models (e.g., GPTs), for both classification and generation tasks. Experiments on few-shot classification and unsupervised text style transfer show superior performance over a wide range of existing fine-tuning or prompting methods. Interestingly, the resulting optimized prompts are often ungrammatical gibberish text; and surprisingly, those gibberish prompts are transferrable between different LMs to retain significant performance, indicating that LM prompting may not follow human language patterns. 

Federated Learning (FL) under distributed concept drift is a largely unexplored area. Although concept drift is itself a well-studied phenomenon, it poses particular challenges for FL, because drifts arise staggered in time and space (across clients). Our work is the first to explicitly study data heterogeneity in both dimensions. We first demonstrate that prior solutions to drift adaptation, with their single global model, are ill-suited to staggered drifts, necessitating multiple-model solutions. We identify the problem of drift adaptation as a time-varying clustering problem, and we propose two new clustering algorithms for reacting to drifts based on local drift detection and hierarchical clustering. Empirical evaluation shows that our solutions achieve significantly higher accuracy than existing baselines, and are comparable to an idealized algorithm with oracle knowledge of the ground-truth clustering of clients to concepts at each time step.