RESUMEN
The vulnerability of modern neural networks to random noise and deliberate attacks has raised concerns about their robustness, particularly as they are increasingly utilized in safety- and security-critical applications. Although recent research efforts were made to enhance robustness through retraining with adversarial examples or employing data augmentation techniques, a comprehensive investigation into the effects of training data perturbations on model robustness remains lacking. This paper presents the first extensive empirical study investigating the influence of data perturbations during model retraining. The experimental analysis focuses on both random and adversarial robustness, following established practices in the field of robustness analysis. Various types of perturbations in different aspects of the dataset are explored, including input, label, and sampling distribution. Single-factor and multi-factor experiments are conducted to assess individual perturbations and their combinations. The findings provide insights into constructing high-quality training datasets for optimizing robustness and recommend the appropriate degree of training set perturbations that balance robustness and correctness, and contribute to understanding model robustness in deep learning and offer practical guidance for enhancing model performance through perturbed retraining, promoting the development of more reliable and trustworthy deep learning systems for safety-critical applications.
RESUMEN
Polyvinyl chloride (PVC) is ubiquitous in everyday life; however, it is not recycled because it degrades uncontrollably into toxic products above 250°C. Therefore, it is of interest to controllably dechlorinate PVC at mild temperatures to generate narrowly distributed carbon materials. We present a catalytic route to dechlorinate PVC (~90% reduction of Cl content) at mild temperature (200°C) to produce gas H2 (with negligible coproduction of corrosive gas HCl) and carbon materials using Ga as a liquid metal (LM) catalyst. A LM was used to promote intimate contact between PVC and the catalytic sites. During dechlorination of PVC, Cl is sequestrated in the carbonaceous solid product. Later, chlorine is easily removed with an acetone wash at room temperature. The Ga LM catalyst is reusable, outperforms a traditional supported metal catalyst, and successfully converts (untreated) discarded PVC pipe.
RESUMEN
A convenient method for the synthesis of perdeuterated alkyl amides/amines is disclosed. Perdeuterated acetyl amides can be achieved by a hydrogen-deuterium (H/D) exchange protocol with Pt/C as a catalyst and D2O as a deuterium source under mild conditions. After removal or reduction of the acetyl group, this protocol can provide perdeuterated primary, secondary, and tertiary amines, which are difficult to achieve via other methods.
RESUMEN
Tuning the anionic site of catalyst supports can impact reaction pathways by creating active sites on the support or influencing metal-support interactions when using supported metal nanoparticles. This study focuses on CO2 hydrogenation over supported Cu nanoparticles, revealing a 3-fold increase in methanol yield when replacing oxygen anions with hydrides in the perovskite support (Cu/BaTiO2.8 H0.2 yields ~146â mg/h/gCu vs. Cu/BaTiO3 yields ~50â mg/h/gCu). The contrast suggests that significant roles are played by the support hydrides in the reaction. Temperature programmed reaction and isotopic labelling studies indicate that BaTiO2.8 H0.2 surface hydride species follow a Mars van Krevelen mechanism in CO2 hydrogenation, promoting methanol production. High-pressure steady-state isotopic transient kinetic analysis (SSITKA) studies suggest that Cu/BaTiO2.8 H0.2 possesses both a higher density and more active and selective sites for methanol production compared to Cu/BaTiO3 . An operando high-pressure diffuse reflectance infrared spectroscopy (DRIFTS)-SSITKA study shows that formate species are the major surface intermediates over both catalysts, and the subsequent hydrogenation steps of formate are likely rate-limiting. However, the catalytic reactivity of Cu/BaTiO2.8 H0.2 towards the formate species is much higher than Cu/BaTiO3 , likely due to the altered electronic structure of interface Cu sites by the hydrides in the support as validated by density functional theory (DFT) calculations.
RESUMEN
Polar surfaces undergo polarity compensation, which can lead to significantly different surface chemistry from their nonpolar counterparts. This process in turn can substantially alter the binding of adsorbates on the surface. Here, we find that hydrogen binds much more strongly to the polar (110) surface than the nonpolar (100) surface for a wide range of ABO3 perovskites, forming a hydroxyl layer on the O24- termination and a hydride layer on the ABO4+ termination of the (110) surface. The stronger adsorption on the polar surfaces can be explained by polarity compensation: hydrogen atoms can act as electron donors or acceptors to compensate for the polarity of perovskite surfaces. The relative stability of the surface terminations is further compared under different gas environments and several perovskites have been found to form stable surface hydride layers under oxygen-poor conditions. These results demonstrate the feasibility of creating stable surface hydrides on perovskites by polarity compensation which might lead to new hydrogenation catalysts based on ABO3 perovskites.
RESUMEN
"Single - atom" catalysts (SACs) have been the focus of intense research, due to debates about their reactivity and challenges toward determining and designing "single - atom" (SA) sites. To address the challenge, in this work, we designed Pt SACs supported on Gd-doped ceria (Pt/CGO), which showed improved activity for CO oxidation compared to its counterpart, Pt/ceria. The enhanced activity of Pt/CGO was associated with a new Pt SA site which appeared only in the Pt/CGO catalyst under CO pretreatment at elevated temperatures. Combined X-ray and optical spectroscopies revealed that, at this site, Pt was found to be d-electron rich and bridged with Gd-induced defects via an oxygen vacancy. As explained by density functional theory calculations, this site opened a new path via a dicarbonyl intermediate for CO oxidation with a greatly reduced energy barrier. These results provide guidance for rationally improving the catalytic properties of SA sites for oxidation reactions.
RESUMEN
Spiking neural network (SNN) is a brain-inspired model with more spatio-temporal information processing capacity and computational energy efficiency. However, with the increasing depth of SNNs, the memory problem caused by the weights of SNNs has gradually attracted attention. In this study, we propose an ultra-low latency adaptive local binary spiking neural network (ALBSNN) with accuracy loss estimators, which dynamically selects the network layers to be binarized to ensure a balance between quantization degree and classification accuracy by evaluating the error caused by the binarized weights during the network learning process. At the same time, to accelerate the training speed of the network, the global average pooling (GAP) layer is introduced to replace the fully connected layers by combining convolution and pooling. Finally, to further reduce the error caused by the binary weight, we propose binary weight optimization (BWO), which updates the overall weight by directly adjusting the binary weight. This method further reduces the loss of the network that reaches the training bottleneck. The combination of the above methods balances the network's quantization and recognition ability, enabling the network to maintain the recognition capability equivalent to the full precision network and reduce the storage space by more than 20%. So, SNNs can use a small number of time steps to obtain better recognition accuracy. In the extreme case of using only a one-time step, we still can achieve 93.39, 92.12, and 69.55% testing accuracy on three traditional static datasets, Fashion- MNIST, CIFAR-10, and CIFAR-100, respectively. At the same time, we evaluate our method on neuromorphic N-MNIST, CIFAR10-DVS, and IBM DVS128 Gesture datasets and achieve advanced accuracy in SNN with binary weights. Our network has greater advantages in terms of storage resources and training time.
RESUMEN
Understanding the structural dynamics/evolution of catalysts and the related surface chemistry is essential for establishing structure-catalysis relationships, where spectroscopic and scattering tools play a crucial role. Among many such tools, neutron scattering, though less-known, has a unique power for investigating catalytic phenomena. Since neutrons interact with the nuclei of matter, the neutron-nucleon interaction provides unique information on light elements (mainly hydrogen), neighboring elements, and isotopes, which are complementary to X-ray and photon-based techniques. Neutron vibrational spectroscopy has been the most utilized neutron scattering approach for heterogeneous catalysis research by providing chemical information on surface/bulk species (mostly H-containing) and reaction chemistry. Neutron diffraction and quasielastic neutron scattering can also supply important information on catalyst structures and dynamics of surface species. Other neutron approaches, such as small angle neutron scattering and neutron imaging, have been much less used but still give distinctive catalytic information. This review provides a comprehensive overview of recent advances in neutron scattering investigations of heterogeneous catalysis, focusing on surface adsorbates, reaction mechanisms, and catalyst structural changes revealed by neutron spectroscopy, diffraction, quasielastic neutron scattering, and other neutron techniques. Perspectives are also provided on the challenges and future opportunities in neutron scattering studies of heterogeneous catalysis.
RESUMEN
Boron-enriched scaffolds have demonstrated unique features and promising performance in the field of catalysis towards the activation of small gas molecules. However, there is still a lack of facile approaches capable of achieving high B doping and abundant porous channels in the targeted catalysts. Herein, construction of boron- and nitrogen-enriched nanoporous π-conjugated networks (BN-NCNs) was achieved via a facile ionothermal polymerization procedure with hexaazatriphenylenehexacarbonitrile [HAT(CN)6 ] sodium borohydride as the starting materials. The as-produced BN-NCN scaffolds were featured by high heteroatoms doping (B up to 23â wt. % and N: up to 17â wt. %) and permanent porosity (surface area up to 759â m2 g-1 mainly contributed by micropores). With the unsaturated bonded B species acting as the active Lewis acid sites and defected N species acting as the active Lewis base sites, those BN-NCNs delivered attractive catalytic performance towards H2 activation/dissociation in both gaseous and liquid phase, acting as efficient metal-free heterogeneous frustrated Lewis pairs (FLPs) catalysts in hydrogenation procedures.
RESUMEN
A photocatalyst TiO2 /Ti-BPDC-Pt is developed with a self-grown TiO2 /Ti-metal-organic framework (MOF) heterojunction, i.e., TiO2 /Ti-BPDC, and selectively anchored high-density Pt single-atomic cocatalysts on Ti-BPDC for photocatalytic hydrogen evolution. This intimate heterojunction, growing from the surface pyrolytic reconstruction of Ti-BPDC, works in a direct Z-scheme, efficiently separating electrons and holes. Pt is selectively anchored on Ti-BPDC by ligands and is found in the form of single atoms with loading up to 1.8â wt %. The selective location of Pt is the electron-enriched domain of the heterojunction, which further enhances the utilization of the separated electrons. This tailored TiO2 /Ti-BPDC-Pt shows a significantly enhanced activity of 12.4â mmol g-1 h-1 compared to other TiO2 - or MOF-based catalysts. The structure-activity relationship further proves the balance of two simultaneously exposed domains of heterojunctions is critical to fulfilling this kind of catalyst.
Asunto(s)
Estructuras Metalorgánicas , Titanio , Ingeniería , HidrógenoRESUMEN
Since the discovery of strong metal-support interactions (SMSIs) over supported metal catalysts in the 1970s, researchers have studied ways to harness this type of catalyst reconstruction to achieve enhanced stability of metal particles against sintering and to create catalytic sites with novel electronic and bonding properties. The motivation to elucidate performance-structure relationships in catalytic transformations has led researchers to take a closer look into catalytic surfaces under reaction conditions rather than a postreaction analysis. These investigations of operating catalysts have made it clear that SMSIs are more common than initially thought. Recent reports show how various adsorbed species, rather than traditional H2/O2 treatment, can promote SMSI in various catalytic systems, a phenomenon named adsorbate-induced SMSI (A-SMSI). Researching the occurrence of A-SMSI has allowed fundamental understanding of catalyst stability, catalytic rates, and product selectivity. The present Perspective discusses the state-of-the-art regarding A-SMSI, the current challenges, and the opportunities ahead in heterogeneous catalysis.
RESUMEN
The development of facile methodologies to afford robust supported metal nanocatalysts under mild conditions is highly desirable yet challenging, particularly via strong metal-support interactions (SMSI) construction. State-of-the-art approaches capable of generating SMSI encapsulation mainly focus on high temperature annealing in reductive/oxidative atmosphere. Herein, ultra-stable metal nanocatalysts based on SMSI construction were produced by leveraging the instantaneous high-energy input from ultrasonication under ambient conditions in H2 O, which could rapidly afford abundant active intermediates, Ti3+ ions, and oxygen vacancies within the scaffolds to induce the SMSI overlayer formation. The encapsulation degree could be tuned and controlled via the reducibility of the solvents and the ultrasonication parameters. This facile and efficient approach could be further extended to diverse metal oxide supports and noble metal NPs leading to enhanced performance in hydrogenation reactions and CO2 conversion.
RESUMEN
Supported Pd single atom catalysts (SACs) have triggered great research interest in methane combustion yet with contradicting views on their activity and stability. Here, we show that the Pd SAs can take different electronic structure and atomic geometry on ceria support, resulting in different catalytic properties. By a simple thermal pretreatment to ceria prior to Pd deposition, a unique anchoring site is created. The Pd SA, taking this site, can be activated to Pdδ+ (0<δ<2) that has greatly enhanced activity for methane oxidation: T50 lowered by up to 130 °C and almost 10 times higher turnover frequency compared to the untreated catalyst. The enhanced activity of Pdδ+ site is related to its oxygen-deficient local structure and elongated interacting distance with ceria, leading to enhanced capability in delivering reactive oxygen species and decomposing reaction intermediates. This work provides insights into designing highly efficient Pd SACs for oxidation reactions.
RESUMEN
Introducing transition-metal components to ceria (CeO2) is important to tailor the surface redox properties for a broad scope of applications. The emergence of high-entropy oxides (HEOs) has brought transformative opportunities for oxygen defect engineering in ceria yet has been hindered by the difficulty in controllably introducing transition metals to the bulk lattice of ceria. Here, we report the fabrication of ceria-based nanocrystals with surface-confined atomic HEO layers for enhanced catalysis. The increased covalency of the transition-metal-oxygen bonds at the HEO-CeO2 interface promotes the formation of surface oxygen vacancies, enabling efficient oxygen activation and replenishment for enhanced CO oxidation capabilities. Understanding the structural heterogeneity involving bulk and surface oxygen defects in nanostructured HEOs provides useful insights into rational design of atomically precise metal oxides, whose increased compositional and structural complexities give rise to expanded functionalities.
RESUMEN
Correction for 'The synergic effect between Mo species and acid sites in Mo/HMCM-22 catalysts for methane aromatization' by Ding Ma et al., Phys. Chem. Chem. Phys., 2005, 7, 3102-3109, https://doi.org/10.1039/B502794B.
RESUMEN
Featuring high olefin selectivity, hexagonal boron nitride (h-BN) has emerged recently as an attractive catalyst for oxidative dehydrogenation of propane (ODHP). Herein, we report that dispersion of vanadium oxide onto BN facilitates the oxyfunctionalization of BN to generate more BO x active sites to catalyze ODHP via the Eley-Rideal mechanism and concurrently produce nitric oxide to initiate additional gas-phase radical chemistry and to introduce redox VO x sites to catalyze ODHP via the Mars-van Krevelen mechanism, all of which promote the catalytic performance of BN for ODHP. As a result, loading 0.5 wt % V onto BN has doubled the yield of light alkene (C2-C3) at 540-580 °C, and adding an appropriate concentration of NO in the reactants further enhances the catalytic performance. These results provide a potential strategy for developing efficient h-BN-based catalysts through coupling gas-phase and surface reactions for the ODHP process.
RESUMEN
Precise control of charge transfer between catalyst nanoparticles and supports presents a unique opportunity to enhance the stability, activity, and selectivity of heterogeneous catalysts. While charge transfer is tunable using the atomic structure and chemistry of the catalyst-support interface, direct experimental evidence is missing for three-dimensional catalyst nanoparticles, primarily due to the lack of a high-resolution method that can probe and correlate both the charge distribution and atomic structure of catalyst/support interfaces in these structures. We demonstrate a robust scanning transmission electron microscopy (STEM) method that simultaneously visualizes the atomic-scale structure and sub-nanometer-scale charge distribution in heterogeneous catalysts using a model Au-catalyst/SrTiO3-support system. Using this method, we further reveal the atomic-scale mechanisms responsible for the highly active perimeter sites and demonstrate that the charge transfer behavior can be readily controlled using post-synthesis treatments. This methodology provides a blueprint for better understanding the role of charge transfer in catalyst stability and performance and facilitates the future development of highly active advanced catalysts.
RESUMEN
The construction of heterogeneous frustrated Lewis pairs (FLPs) with performance comparable to or surpassing the homogeneous counterparts in H2 activation is a long-standing challenge. Herein, sterically hindered Lewis acid ("B" center) and Lewis base ("N" center) sites were anchored within the rigid lattice of highly crystalline hexagonal boron nitride (h-BN) scaffolds. The active sites were created via precision defect regulation during the molten-salt-involved (NaNH2 and NaBH4) h-BN construction procedure. The as-afforded h-BN scaffolds achieved highly efficient H2/D2 activation and dissociation under ambient pressure via FLP-like behavior, and attractive catalytic efficiency in hydrogenation reactions surpassing the current heterogeneous analogues.
RESUMEN
Due to tunable redox properties and cost-effectiveness, copper-ceria (Cu-CeO2 ) materials have been investigated for a wide scope of catalytic reactions. However, accurately identifying and rationally tuning the local structures in Cu-CeO2 have remained challenging, especially for nanomaterials with inherent structural complexities involving surfaces, interfaces, and defects. Here, a nanocrystal-based atom-trapping strategy to access atomically precise Cu-CeO2 nanostructures for enhanced catalysis is reported. Driven by the interfacial interactions between the presynthesized Cu and CeO2 nanocrystals, Cu atoms migrate and redisperse onto the CeO2 surface via a solid-solid route. This interfacial restructuring behavior facilitates tuning of the copper dispersion and the associated creation of surface oxygen defects on CeO2 , which gives rise to enhanced activities and stabilities catalyzing water-gas shift reaction. Combining soft and solid-state chemistry of colloidal nanocrystals provide a well-defined platform to understand, elucidate, and harness metal-support interactions. The dynamic behavior of the supported metal species can be further exploited to realize exquisite control and rational design of multicomponent nanocatalysts.
RESUMEN
Perovskite oxyhydrides such as BaTiO2.5H0.5 have been found to be able to catalyze NH3 synthesis, but the mechanism and the role of the catalyst's lattice hydrides in the catalytic reaction remain unknown. Here we employ first principles density functional theory to investigate the mechanism of ammonia synthesis and the role of lattice hydrides on a prototypical perovskite oxyhydride, BaTiO2.5H0.5 (BTOH). Two mechanistic hypotheses, the distal and alternating pathways, have been tested on the Ti2O2 termination of the BTOH (210) surface, previously determined to be the most stable surface termination under the reaction conditions considered. In the distal pathway, H atoms hydrogenate N2 to form the *N-NHx key intermediates, followed by N-N bond breaking. In the alternating pathway, H atoms hydrogenate N2 in an alternating fashion to form the *NHx-NHy intermediates before N-N bond breaking and formation of co-adsorbed *NHx/*NHy on the surface. We find that the subsurface hydride vacancy formed after reaction of *N2 with the lattice hydride is key to the distal pathway, leading to surface nitride formation after breaking the *N-NH3 bond, while the neighboring surface Ti sites are key to bridging and stabilizing the *NNH intermediate in the alternating pathway. In both pathways, desorption of NH3 is the most uphill in energy. Our results provide important insights into the role of hydrides and surface vacancies in hydrogenation reactions over BTOH, which will be useful to guide future spectroscopic experiments such as operando IR and inelastic neutron scattering to verify the key intermediates.