Search for: All records

Environments associated with severe hailstorms, compared to those of tornadoes, are often less apparent to forecasters. Understanding has evolved considerably in recent years; namely, that weak low-level shear and sufficient convective available potential energy (CAPE) above the freezing level is most favorable for large hail. However, this understanding comes only from examining the mean characteristics of large hail environments. How much variety exists within the kinematic and thermodynamic environments of large hail? Is there a balance between shear and CAPE analogous to that noted with tornadoes? We address these questions to move toward a more complete conceptual model. In this study, we investigate the environments of 92 323 hail reports (both severe and nonsevere) using ERA5 modeled proximity soundings. By employing a self-organizing map algorithm and subsetting these environments by a multitude of characteristics, we find that the conditions leading to large hail are highly variable, but three primary patterns emerge. First, hail growth depends on a favorable balance of CAPE, wind shear, and relative humidity, such that accounting for entrainment is important in parameter-based hail prediction. Second, hail growth is thwarted by strong low-level storm-relative winds, unless CAPE below the hail growth zone is weak. Finally, the maximum hail size possible in a given environment may be predictable by the depth of buoyancy, rather than CAPE itself.

 

It is critical that machine learning (ML) model predictions be trustworthy for high-throughput catalyst discovery approaches. Uncertainty quantification (UQ) methods allow estimation of the trustworthiness of an ML model, but these methods have not been well explored in the field of heterogeneous catalysis. Herein, we investigate different UQ methods applied to a crystal graph convolutional neural network to predict adsorption energies of molecules on alloys from the Open Catalyst 2020 dataset, the largest existing heterogeneous catalyst dataset. We apply three UQ methods to the adsorption energy predictions, namelyk-fold ensembling, Monte Carlo dropout, and evidential regression. The effectiveness of each UQ method is assessed based on accuracy, sharpness, dispersion, calibration, and tightness. Evidential regression is demonstrated to be a powerful approach for rapidly obtaining tunable, competitively trustworthy UQ estimates for heterogeneous catalysis applications when using neural networks. Recalibration of model uncertainties is shown to be essential in practical screening applications of catalysts using uncertainties.

 

Breathing behaviour involves the generation of normal breaths (eupnoea) on a timescale of seconds and sigh breaths on the order of minutes. Both rhythms emerge in tandem from a single brainstem site, but whether and how a single cell population can generate two disparate rhythms remains unclear. We posit that recurrent synaptic excitation in concert with synaptic depression and cellular refractoriness gives rise to the eupnoea rhythm, whereas an intracellular calcium oscillation that is slower by orders of magnitude gives rise to the sigh rhythm. A mathematical model capturing these dynamics simultaneously generates eupnoea and sigh rhythms with disparate frequencies, which can be separately regulated by physiological parameters. We experimentally validated key model predictions regarding intracellular calcium signalling. All vertebrate brains feature a network oscillator that drives the breathing pump for regular respiration. However, in air‐breathing mammals with compliant lungs susceptible to collapse, the breathing rhythmogenic network may have refashioned ubiquitous intracellular signalling systems to produce a second slower rhythm (for sighs) that prevents atelectasis without impeding eupnoea.image

A simplified activity‐based model of the preBötC generates inspiratory and sigh rhythms from a single neuron population.

Inspiration is attributable to a canonical excitatory network oscillator mechanism.

Sigh emerges from intracellular calcium signalling.

The model predicts that perturbations of calcium uptake and release across the endoplasmic reticulum counterintuitively accelerate and decelerate sigh rhythmicity, respectively, which was experimentally validated.

Vertebrate evolution may have adapted existing intracellular signalling mechanisms to produce slow oscillations needed to optimize pulmonary function in mammals.

 

2D materials have attracted broad attention from researchers for their unique electronic proper-ties, which may be been further enhanced by combining 2D layers into vertically stacked van der Waals heterostructures. Among the superlative properties of 2D systems, thermoelectric energy (TE) conversion promises to enable targeted energy conversion, localized thermal management, and thermal sensing. However, TE conversion efficiency remains limited by the inherent tradeoff between conductivity and thermopower. In this paper, we use first-principles calculation to study graphene-based van der Waals heterostructures (vdWHs) composed of graphene layers and hexagonal boron nitride (h-BN). We compute the electronic band structures of heterostructured systems using Quantum Espresso and their thermoelectric (TE) properties using BoltzTrap2. Our results have shown that stacking layers of these 2D materials opens a bandgap, increasing it with the number of h-BN interlayers, which significantly improves the power factor (PF). We predict a PF of ~1.0x10 11 W/K 2 .m.s for the vdWHs, nearly double compared to 5x10 10 W/K 2 .m.s that we obtained for single-layer graphene. This study gives important information on the effect of stacking layers of 2D materials and points toward new avenues to optimize the TE properties of vdWHs. 

Copper-based catalyst is uniquely positioned to catalyze the hydrocarbon formations through electrochemical CO₂reduction. The catalyst design freedom is limited for alloying copper with H-affinitive elements represented by platinum group metals because the latter would easily drive the hydrogen evolution reaction to override CO₂reduction. We report an adept design of anchoring atomically dispersed platinum group metal species on both polycrystalline and shape-controlled Cu catalysts, which now promote targeted CO₂reduction reaction while frustrating the undesired hydrogen evolution reaction. Notably, alloys with similar metal formulations but comprising small platinum or palladium clusters would fail this objective. With an appreciable amount of CO-Pd₁moieties on copper surfaces, facile CO^*hydrogenation to CHO^*or CO-CHO^*coupling is now viable as one of the main pathways on Cu(111) or Cu(100) to selectively produce CH₄or C₂H₄through Pd-Cu dual-site pathways. The work broadens copper alloying choices for CO₂reduction in aqueous phases.

 

A simultaneously accurate and computationally efficient parametrization of the potential energy surface of molecules and materials is a long-standing goal in the natural sciences. While atom-centered message passing neural networks (MPNNs) have shown remarkable accuracy, their information propagation has limited the accessible length-scales. Local methods, conversely, scale to large simulations but have suffered from inferior accuracy. This work introduces Allegro, a strictly local equivariant deep neural network interatomic potential architecture that simultaneously exhibits excellent accuracy and scalability. Allegro represents a many-body potential using iterated tensor products of learned equivariant representations without atom-centered message passing. Allegro obtains improvements over state-of-the-art methods on QM9 and revMD17. A single tensor product layer outperforms existing deep MPNNs and transformers on QM9. Furthermore, Allegro displays remarkable generalization to out-of-distribution data. Molecular simulations using Allegro recover structural and kinetic properties of an amorphous electrolyte in excellent agreement with ab-initio simulations. Finally, we demonstrate parallelization with a simulation of 100 million atoms.