Search for: All records

We present a method whereby the finite element (FE) method is used to train physics-informed neural networks that are suitable for surrogate modeling. The method is based on a new class of mesh-based neural network operators called stenciled operators, which leverage the inverse isoparametric map of the FE method. In contrast with graph-based methods, stenciled operators can incorporate non-local information over arbitrary spatial domains with computational costs that are length-scale invariant. Using stenciled convolutional operations as an exemplar, we demonstrate the performance of the method in several training and testing scenarios with linear-elastic boundary-value problems of varying geometries. The resulting neural networks show reasonable accuracy when tested on unseen geometries that are similar to those used for training. Furthermore, when the number of training geometries is increased, the testing errors systematically decrease, demonstrating that the neural networks learn how to generalize as the training set becomes larger. Further extending the method to allow for variable boundary conditions, properties, and body forces will lead to a general-purpose surrogate modeling framework that can leverage existing FE codes for training. 

Shellfish fisheries and aquaculture within the Chesapeake Bay (hereafter 'the Bay') and its tributaries have been historically impacted by disease and climate events. Climate-driven shifts in temperature and salinity can alter host-parasite dynamics, influencing outbreaks. Here, we explore the relationship between temperature, salinity and parasite distribution and abundance in the eastern oysterCrassostrea virginica-Perkinsus marinussystem. We use long-term (30 yr) environmental data andP. marinussurveys in the Bay to identify (1) how climate affectsP. marinusprevalence and intensity, (2) seasonal and climate-driven infection patterns, and (3) regional environmental influences on disease. We found significant relationships betweenP. marinusinfection intensity, prevalence, increasing temperature and decreasing salinity. Our results indicated that there is an overall decreased abundance ofP. marinusprevalence and intensity throughout the Bay driven by decreases in salinity over time, most prominently from 2003-2020. However, these temporal trends in prevalence and intensity vary largely by region, with some regions still experiencing high disease burden. Examining monthly environmental parameters reinforced the dominant role of salinity in driving disease patterns. Salinity had significant relationships with prevalence and intensity year-round, with the largest effects in late spring/early summer. Monthly temperatures had fewer significant relationships to prevalence and intensity, but the largest significant effects were seen in late winter/early spring. Notably, this study is the first to document that winter salinity influences fall parasite prevalence, sometimes exerting a greater effect than temperature. Continued and expanded monitoring of marine disease is crucial to understand how the changing climate is impacting disease. 

The development of high-performance materials for microelectronics, energy storage, and extreme environments depends on our ability to describe and direct property-defining microstructural order. Our present understanding is typically derived from laborious manual analysis of imaging and spectroscopy data, which is difficult to scale, challenging to reproduce, and lacks the ability to reveal latent associations needed for mechanistic models. Here, we demonstrate a multi-modal machine learning (ML) approach to describe order from electron microscopy analysis of the complex oxide La1−xSrxFeO3. We construct a hybrid pipeline based on fully and semi-supervised classification, allowing us to evaluate both the characteristics of each data modality and the value each modality adds to the ensemble. We observe distinct differences in the performance of uni- and multi-modal models, from which we draw general lessons in describing crystal order using computer vision. 

Perovskite oxides such as LaFeO3 are a well-studied family of materials that possess a wide range of useful and novel properties. Successfully synthesizing perovskite oxide samples usually requires a significant number of growth attempts and a detailed film characterization on each sample to find the optimal growth window of a material. The most common real-time in situ diagnostic technique available during molecular beam epitaxy (MBE) synthesis is reflection high-energy electron diffraction (RHEED). Conventional use of RHEED allows a highly experienced operator to determine growth rate by monitoring intensity oscillations and make some qualitative observations during growth, such as recognizing the sample has become amorphous or recognizing that large islands have formed on the surface. However, due to a lack of theoretical understanding of the diffraction patterns, finer, more precise levels of observations are challenging. To address these limitations, we implement new data analytics techniques in the growth of three LaFeO3 samples on Nb-doped SrTiO3 by MBE. These techniques improve our ability to perform unsupervised machine learning using principal component analysis (PCA) and k-means clustering by using drift correction to overcome sample or stage motion during growth and intensity transformations that highlight more subtle features in the images such as Kikuchi bands. With this approach, we enable the first demonstration of PCA and k-means across multiple samples, allowing for quantitative comparison of RHEED videos for two LaFeO3 film samples. These capabilities set the stage for real-time processing of RHEED data during growth to enable machine learning-accelerated film synthesis. 

Obtaining insight into the constituents of dark matter and their interactions with normal matter has inspired a wide range of experimental efforts. Several approaches, particularly those involving searches for ultralight bosonic dark matter (UBDM) fields, involve the use of quantum systems or measurements performed at the limits imposed by quantum mechanics. While a classical treatment of UBDM and its detectors is satisfactory, a fully quantum description would assist in developing future detection strategies. Here, we present an open quantum systems approach that accomplishes this while providing intuition into the quantum nature of the detection process itself. Furthermore, we apply the quantum theory of optical coherence to characterize the statistical properties of the UBDM field. Using representative examples, we show that this theoretical treatment has implications in uncovering signatures of the cosmological production mechanism of the UBDM field and its galactic merger history. By adapting tools from quantum optics, this work will help facilitate the creation of novel methods to extract astrophysically relevant information from correlation measurements. Published by the American Physical Society2025 

Abstract A series of Co^2+/3+and Fe^2+/3+complexes is prepared using three variants of a hexadentate tris(imidazole)triazacyclononane ligand bearing different 4‐alkyl substituents on the imidazole rings. The steric bulk of the alkyl substituent (R=H,ⁱPr, or^tBu) alters the preferred size of the ligand binding cavity by inhibiting close approach of the imidazole donors with bulky substituents. The resulting changes in geometry, redox potentials, spin states, and optical properties are catalogued across the series, demonstrating redox potential tuning over at least 670 mV as well as spin state switching based on the choice of substituent. The ligand field splitting of the complexes decreases with increasing bulk of the substituents. Tuning of the steric bulk of the substituents in these positions therefore allows for the electronic properties of the complexes to be fine‐tuned in a manner orthogonal to the donor properties of the substituents. 

Abstract Oceanic plateaus are common in modern oceanic basins and will ultimately collide with continental subduction zones. Despite the frequency of these events, complete sedimentary records of oceanic plateau collision and accretion have remained limited to only a few Cenozoic examples with excellent exposure and tectonic context. Our study focuses on building a stratigraphic record of plateau collision using the sedimentary strata deposited on the Siletzia oceanic plateau, which accreted to the Pacific Northwest at ca. 50 Ma. By combining previously published provenance and stratigraphic data with new lithofacies and geologic mapping, measured stratigraphic sections, conglomerate clast counts, and U-Pb zircon geochronology, we were able to divide the strata of the northern Olympic Peninsula in Washington, USA, into precollisional, syn-collisional, and postcollisional stages. Precollisional strata include early Eocene deep-marine hemipelagic to pelagic mudstones of the Aldwell Formation that were deposited directly on Siletzia basalts. These strata were deformed during collision and are separated from the overlying syn-collisional middle Eocene sandstone and conglomerate of the marine (?) Lyre Formation by an angular unconformity. Postcollisional strata were deposited by submarine fans and include interbedded sandstone and siltstone of the Hoko River and Makah formations. These units initially record the filling of isolated trench-slope basins by late Eocene time before eventual integration into an Oligocene regional forearc basin as the accreted Siletzia plateau began to subside. Our chronostratigraphy permits the correlation of basin strata across tectonic domains and provides more general insight into how forearc sedimentary systems evolve following the accretion of a young, buoyant oceanic plateau. 

Thin film deposition is a fundamental technology for the discovery, optimization, and manufacturing of functional materials. Deposition by molecular beam epitaxy (MBE) typically employs reflection high-energy electron diffraction (RHEED) as a real-time in situ probe of the growing film. However, the state-of-the-art for RHEED analysis during deposition requires human observation. Here, we present an approach using machine learning (ML) methods to monitor, analyze, and interpret RHEED images on-the-fly during thin film deposition. In the analysis workflow, RHEED pattern images are collected at one frame per second and featurized using a pretrained deep convolutional neural network. The feature vectors are then statistically analyzed to identify changepoints; these changepoints can be related to changes in the deposition mode from initial film nucleation to a transition regime, smooth film deposition, and in some cases, an additional transition to a rough, islanded deposition regime. The feature vectors are additionally analyzed via graph analysis and community classification. The graph is quantified as a stabilization plot, and we show that inflection points in the stabilization plot correspond to changes in the growth regime. The full RHEED analysis workflow is termed RHAAPsody and includes data transfer and output to a visual dashboard. We demonstrate the functionality of RHAAPsody by analyzing the precaptured RHEED images from epitaxial depositions of anatase TiO2 on SrTiO3(001) and show that the analysis workflow can be executed in less than 1 s. Our approach shows promise as one component of ML-enabled real-time feedback control of the MBE deposition process.