Search for: All records

Machine learning methods are increasingly being employed as surrogate models in place of computationally expensive and slow numerical integrators for a bevy of applications in the natural sciences. However, while the laws of physics are relationships between scalars, vectors and tensors that hold regardless of the frame of reference or chosen coordinate system, surrogate machine learning models are not coordinate-free by default. We enforce coordinate freedom by using geometric convolutions in three model architectures: a ResNet, a Dilated ResNet and a UNet. In numerical experiments emulating two-dimensional compressible Navier–Stokes, we see better accuracy and improved stability compared with baseline surrogate models in almost all cases. The ease of enforcing coordinate freedom without making major changes to the model architecture provides an exciting recipe for any convolutional neural network-based method applied to an appropriate class of problems. 

Abstract We present a proof of concept demonstration of solar thermochemical energy storage on a multiple year time scale. The storage is fungible and can take the form of process heat or hydrogen. We designed and fabricated a 4-kW solar rotary drum reactor to carry out the solar-driven charging step of solar thermochemical storage via metal oxide reduction–oxidation cycles. During the summer of 2019, the solar reactor was operated in the Valparaiso University solar furnace to effect the reduction of submillimeter cobalt oxide particles in air at approximately 1000∘C. A particle collection system cooled the reduced particles rapidly enough to maintain conversions of 84–94% for feed rates of 2.9−60.8gmin−1. The solar-to-chemical storage efficiency, defined as the enthalpy of the reduction reaction at 1000∘C divided by the solar energy input, reached 20%. Samples of the reduced cobalt oxide particles were stored in vials in air at room temperature for more than 3 years. The stored solar energy was released by reoxidizing samples in air in a benchtop reactor and by electrochemically reoxidizing samples to produce H2. Measurements of the oxygen uptake by the reduced metal oxide confirm its promise as a medium to store and dispatch solar energy over long durations. Linear sweep voltammetry and bulk electrolysis demonstrate the promise of H2 production at 0.55 V relative to the normal hydrogen electrode, 0.68 V below the 1.23 V potential required for conventional electrolysis. 

Inspired by constraints from physical law, equivariant machine learning restricts the learning to a hypothesis class where all the functions are equivariant with respect to some group action. Irreducible representations or invariant theory are typically used to parameterize the space of such functions. In this article, we introduce the topic and explain a couple of methods to explicitly parameterize equivariant functions that are being used in machine learning applications. In particular, we explicate a general procedure, attributed to Malgrange, to express all polynomial maps between linear spaces that are equivariant under the action of a group G, given a characterization of the invariant polynomials on a bigger space. The method also parametrizes smooth equivariant maps in the case that G is a compact Lie group. 

A recent computational analysis of the stabilizing intramolecular OH⋯O contact in 1,2-dialkyl-2,3-epoxycyclopentanol diastereomers has been extended to thiiriane, aziridine and phosphirane analogues. Density functional theory (DFT), second-order Møller-Plesset perturbation theory (MP2) and CCSD(T) coupled-cluster computations with simple methyl and ethyl substituents indicate that electronic energies of the c i s isomers are lowered by roughly 3 to 4 kcal mol−1 when the OH group of these cyclopentanol systems forms an intramolecular contact with the O, S, N or P atom on the adjacent carbon. The results also suggest that S and P can participate in these stabilizing intramolecular interactions as effectively as O and N in constrained molecular environments. The stabilizing intramolecular OH⋯O, OH⋯S, OH⋯N and OH⋯P contacts also increase the covalent OH bond length and significantly decrease the OH stretching vibrational frequency in every system with shifts typically on the order of −41 cm−1. 

Large-language models (LLMs) such as GPT-4 caught the interest of many scientists. Recent studies suggested that these models could be useful in chemistry and materials science. To explore these possibilities, we organized a hackathon. This article chronicles the projects built as part of this hackathon. Participants employed LLMs for various applications, including predicting properties of molecules and materials, designing novel interfaces for tools, extracting knowledge from unstructured data, and developing new educational applications. The diverse topics and the fact that working prototypes could be generated in less than two days highlight that LLMs will profoundly impact the future of our fields. The rich collection of ideas and projects also indicates that the applications of LLMs are not limited to materials science and chemistry but offer potential benefits to a wide range of scientific disciplines. 

Abstract Ice shelf fracture is responsible for roughly half of Antarctic ice mass loss in the form of calving and can weaken buttressing of upstream ice flow. Large uncertainties associated with the ice sheet response to climate variations are due to a poor understanding of these fracture processes and how to model them. Here, we address these problems by implementing an anisotropic, nonlocal integral formulation of creep damage within a large‐scale shallow‐shelf ice flow model. This model can be used to study the full evolution of fracture from initiation of crevassing to rifting that eventually causes tabular calving. While previous ice shelf fracture models have largely relied on simple expressions to estimate crevasse depths, our model parameterizes fracture as a progressive damage evolution process in three‐dimensions (3‐D). We also implement an efficient numerical framework based on the material point method, which avoids advection errors. Using an idealized marine ice sheet, we test the creep damage model and a crevasse‐depth based damage model, including a modified version of the latter that accounts for damage evolution due to necking and mass balance. We demonstrate that the creep damage model is best suited for capturing weakening and rifting over shorter (monthly/yearly) timescales, and that anisotropic damage reproduces typically observed fracture patterns better than isotropic damage. Because necking and mass balance can significantly influence damage on longer (decadal) timescales, we discuss the potential for a combined approach between models to best represent mechanical weakening and tabular calving within long‐term simulations. 

Abstract We develop a generalized interpolation material point method (GIMPM) for the shallow shelf approximation (SSA) of ice flow. The GIMPM, which can be viewed as a particle version of the finite element method, is used here to solve the shallow shelf approximations of the momentum balance and ice thickness evolution equations. We introduce novel numerical schemes for particle splitting and integration at domain boundaries to accurately simulate the spreading of an ice shelf. The advantages of the proposed GIMPM‐SSA framework include efficient advection of history or internal state variables without diffusion errors, automated tracking of the ice front and grounding line at sub‐element scales, and a weak formulation based on well‐established conventions of the finite element method with minimal additional computational cost. We demonstrate the numerical accuracy and stability of the GIMPM using 1‐D and 2‐D benchmark examples. We also compare the accuracy of the GIMPM with the standard material point method (sMPM) and a reweighted form of the sMPM. We find that the grid‐crossing error is very severe for SSA simulations with the sMPM, whereas the GIMPM successfully mitigates this error. While the grid‐crossing error can be reasonably reduced in the sMPM by implementing a simple material point reweighting scheme, this approach it not as accurate as the GIMPM. Thus, we illustrate that the GIMPM‐SSA framework is viable for the simulation of ice sheet‐shelf evolution and enables boundary tracking and error‐free advection of history or state variables, such as ice thickness or damage. 

Abstract The springtime transition to regional‐scale onset of photosynthesis and net ecosystem carbon uptake in boreal and tundra ecosystems are linked to the soil freeze–thaw state. We present evidence from diagnostic and inversion models constrained by satellite fluorescence and airborneCO₂from 2012 to 2014 indicating the timing and magnitude of spring carbon uptake in Alaska correlates with landscape thaw and ecoregion. Landscape thaw in boreal forests typically occurs in late April (DOY111 ± 7) with a 29 ± 6 day lag until photosynthetic onset. North Slope tundra thaws 3 weeks later (DOY133 ± 5) but experiences only a 20 ± 5 day lag until photosynthetic onset. These time lag differences reflect efficient cold season adaptation in tundra shrub and the longer dehardening period for boreal evergreens. Despite the short transition from thaw to photosynthetic onset in tundra, synchrony of tundra respiration with snow melt and landscape thaw delays the transition from net carbon loss (at photosynthetic onset) to net uptake by 13 ± 7 days, thus reducing the tundra net carbon uptake period. Two globalCO₂inversions using aCASA‐GFEDmodel prior estimate earlier northern high latitude net carbon uptake compared to our regional inversion, which we attribute to (i) early photosynthetic‐onset model prior bias, (ii) inverse method (scaling factor + optimization window), and (iii) sparsity of available AlaskanCO₂observations. Another global inversion with zero prior estimates the same timing for net carbon uptake as the regional model but smaller seasonal amplitude. The analysis of Alaskan eddy covariance observations confirms regional scale findings for tundra, but indicates that photosynthesis and net carbon uptake occur up to 1 month earlier in evergreens than captured by models orCO₂inversions, with better correlation to above‐freezing air temperature than date of primary thaw. Further collection and analysis of boreal evergreen species over multiple years and at additional subarctic flux towers are critically needed.