Oommen, Vivek; Shukla, Khemraj; Goswami, Somdatta; Dingreville, Remi P.; Karniadakis, George E.
Phase-field modeling is an effective but computationally expensive method for capturing the mesoscale morphological and microstructure evolution in materials. Hence, fast and generalizable surrogate models are needed to alleviate the cost of computationally taxing processes such as in optimization and design of materials. The intrinsic discontinuous nature of the physical phenomena incurred by the presence of sharp phase boundaries makes the training of the surrogate model cumbersome. We develop a framework that integrates a convolutional autoencoder architecture with a deep neural operator (DeepONet) to learn the dynamic evolution of a two-phase mixture and accelerate time-to-solution in predicting the microstructure evolution. We utilize the convolutional autoencoder to provide a compact representation of the microstructure data in a low-dimensional latent space. After DeepONet is trained in the latent space, it can be used to replace the high-fidelity phase-field numerical solver in interpolation tasks or to accelerate the numerical solver in extrapolation tasks.
We show that piezoelectric strain actuation of acoustomechanical interactions can produce large phase velocity changes in an existing quantum phononic platform: aluminum nitride on suspended silicon. Using finite element analysis, we demonstrate a piezo-acoustomechanical phase shifter waveguide capable of producing ±π phase shifts for GHz frequency phonons in 10s of μm with 10s of volts applied. Then, using the phase shifter as a building block, we demonstrate several phononic integrated circuit elements useful for quantum information processing. In particular, we show how to construct programmable multi-mode interferometers for linear phononic processing and a dynamically reconfigurable phononic memory that can switch between an ultra-long-lifetime state and a state strongly coupled to its bus waveguide. From the master equation for the full open quantum system of the reconfigurable phononic memory, we show that it is possible to perform read and write operations with over 90% quantum state transfer fidelity for an exponentially decaying pulse.
The extreme miniaturization of a cold-atom interferometer accelerometer requires the development of novel technologies and architectures for the interferometer subsystems. Here, we describe several component technologies and a laser system architecture to enable a path to such miniaturization. We developed a custom, compact titanium vacuum package containing a microfabricated grating chip for a tetrahedral grating magneto-optical trap (GMOT) using a single cooling beam. In addition, we designed a multi-channel photonic-integrated-circuit-compatible laser system implemented with a single seed laser and single sideband modulators in a time-multiplexed manner, reducing the number of optical channels connected to the sensor head. In a compact sensor head containing the vacuum package, sub-Doppler cooling in the GMOT produces 15 μK temperatures, and the GMOT can operate at a 20 Hz data rate. We validated the atomic coherence with Ramsey interferometry using microwave spectroscopy, then demonstrated a light-pulse atom interferometer in a gravimeter configuration for a 10 Hz measurement data rate and T = 0–4.5 ms interrogation time, resulting in Δg/g = 2.0 × 10−6. This work represents a significant step towards deployable cold-atom inertial sensors under large amplitude motional dynamics.
Abstract: Due to a beneficial balance of computational cost and accuracy, real-time time-dependent density-functional theory has emerged as a promising first-principles framework to describe electron real-time dynamics. Here we discuss recent implementations around this approach, in particular in the context of complex, extended systems. Results include an analysis of the computational cost associated with numerical propagation and when using absorbing boundary conditions. We extensively explore the shortcomings for describing electron–electron scattering in real time and compare to many-body perturbation theory. Modern improvements of the description of exchange and correlation are reviewed. In this work, we specifically focus on the Qb@ll code, which we have mainly used for these types of simulations over the last years, and we conclude by pointing to further progress needed going forward. Graphical abstract: [Figure not available: see fulltext.].
Heavy metals released from kerogen to produced water during oil/gas extraction have caused major enviromental concerns. To curtail water usage and production in an operation and to use the same process for carbon sequestration, supercritical CO2 (scCO2) has been suggested as a fracking fluid or an oil/gas recovery agent. It has been shown previously that injection of scCO2 into a reservoir may cause several chemical and physical changes to the reservoir properties including pore surface wettability, gas sorption capacity, and transport properties. Using molecular dynamics simulations, we here demonstrate that injection of scCO2 might lead to desorption of physically adsorbed metals from kerogen structures. This process on one hand may impact the quality of produced water. On the other hand, it may enhance metal recovery if this process is used for in-situ extraction of critical metals from shale or other organic carbon-rich formations such as coal.
Commercial vendors, trying to tap into the physical protection of critical infrastructure, are offering nuclear facilities the opportunity to borrow detection counter-unmanned aircraft systems (CUAS) equipment to survey the airspace over and around the facility. However, using one vendor or method of detection (e.g., radio frequency [RF], radar, acoustic, visual) will not necessarily provide a complete airspace profile since no single method can detect all UAS threats. Using several detection technologies, the unmanned aircraft systems (UAS) Team, who supports the U.S. National Nuclear Security Administration (NNSA) Office of International Nuclear Security (INS), would like to offer partners a comprehensive airspace profile of the types and frequency of UAS that fly within and around critical infrastructure. Improved UAS awareness will aid in the risk assessment process.
This report represents completion of milestone deliverable M2SF-23SN010309082 Annual Status Update for OWL due on November 30, 2022. It provides the status of fiscal year 2022 (FY2022) updates for the Online Waste Library (OWL).
Unmanned aircraft systems (UAS/drones) are rapidly evolving and are considered an emerging threat by nuclear facilities throughout the world. Due to the wide range of UAS capabilities, members of the workforce and security/response force personnel need to be prepared for a variety of drone incursion situations. Tabletop exercises are helpful, but actual live exercises are often needed to evaluate the quick chain of events that might ensue during a real drone fly-in and the essential kinds of information that will help identify the type of drone and pilot. Even with drone detection equipment, the type of UAS used for incursion drills can have a major impact on detection altitude and finding the UAS in the sky. Using a variety of UAS, the U.S. National Nuclear Security Administration (NNSA) Office of International Nuclear Security (INS) would like to offer partners the capability of adding actual UAS into workforce and response exercises to improve overall UAS awareness as well as the procedures that capture critical steps in dealing with intruding drones.
We present Q Framework: a verification framework used at Sandia National Laboratories. Q is a collection of tools used to verify safety and correctness properties of high-consequence embedded systems and captures the structure and compositionality of system specifications written with state machines in order to prove system-level properties about their implementations. Q consists of two main workflows: 1) compilation of temporal properties and state machine models (such as those made with Stateflow) into SMV models and 2) generation of ACSL specifications for the C code implementation of the state machine models. These together prove a refinement relation between the state machine model and its C code implementation, with proofs of properties checked by NuSMV (for SMV models) and Frama-C (for ACSL specifications).
Platinum@hexaniobate nanopeapods (Pt@HNB NPPs) are a nanocomposite photocatalyst that was selectively engineered to increase the efficiency of hydrogen production from visible light photolysis. Pt@HNB NPPs consist of linear arrays of high surface area Pt nanocubes encapsulated within scrolled sheets of the semiconductor HxK4–xNb6O17 and were synthesized in high yield via a facile one-pot microwave heating method that is fast, reproducible, and more easily scalable than multi-step approaches required by many other state-of-the-art catalysts. The Pt@HNB NPPs’ unique 3D architecture enables physical separation of the Pt catalysts from competing surface reactions, promoting electron efficient delivery to the isolated reduction environment along directed charge transport pathways that kinetically prohibit recombination reactions. Pt@HNB NPPs’ catalytic activity was assessed in direct comparison to representative state-of-the-art Pt/semiconductor nanocomposites (extPt-HNB NScs) and unsupported Pt nanocubes. Photolysis under similar conditions exhibited superior H2 production by the Pt@HNB NPPs, which exceeded other catalyst H2 yields (μmol) by a factor of 10. Turnover number and apparent quantum yield values showed similar dramatic increases over the other catalysts. Overall, the results clearly demonstrate that Pt@HNB NPPs represent a unique, intricate nanoarchitecture among state-of-the-art heterogeneous catalysts, offering obvious benefits as a new architectural pathway toward efficient, versatile, and scalable hydrogen energy production. Potential factors behind the Pt@HNB NPPs’ superior performance are discussed below, as are the impacts of systematic variation of photolysis parameters and the use of a non-aqueous reductive quenching photosystem.
Metal-assisted chemical etching (MACE) is a flexible technique for texturing the surface of semiconductors. In this work, we study the spatial variation of the etch profile, the effect of angular orientation relative to the crystallographic planes, and the effect of doping type. We employ gold in direct contact with germanium as the metal catalyst, and dilute hydrogen peroxide solution as the chemical etchant. With this catalyst-etchant combination, we observe inverse-MACE, where the area directly under gold is not etched, but the neighboring, exposed germanium experiences enhanced etching. This enhancement in etching decays exponentially with the lateral distance from the gold structure. An empirical formula for the gold-enhanced etching depth as a function of lateral distance from the edge of the gold film is extracted from the experimentally measured etch profiles. The lateral range of enhanced etching is approximately 10–20 µm and is independent of etchant concentration. At length scales beyond a few microns, the etching enhancement is independent of the orientation with respect to the germanium crystallographic planes. The etch rate as a function of etchant concentration follows a power law with exponent smaller than 1. The observed etch rates and profiles are independent of whether the germanium substrate is n-type, p-type, or nearly intrinsic.
As machine learning (ML) models are deployed into an ever-diversifying set of application spaces, ranging from self-driving cars to cybersecurity to climate modeling, the need to carefully evaluate model credibility becomes increasingly important. Uncertainty quantification (UQ) provides important information about the ability of a learned model to make sound predictions, often with respect to individual test cases. However, most UQ methods for ML are themselves data-driven and therefore susceptible to the same knowledge gaps as the models themselves. Specifically, UQ helps to identify points near decision boundaries where the models fit the data poorly, yet predictions can score as certain for points that are under-represented by the training data and thus out-of-distribution (OOD). One method for evaluating the quality of both ML models and their associated uncertainty estimates is out-of-distribution detection (OODD). We combine OODD with UQ to provide insights into the reliability of the individual predictions made by an ML model.
Neural operators [1–5] have recently become popular tools for designing solution maps between function spaces in the form of neural networks. Differently from classical scientific machine learning approaches that learn parameters of a known partial differential equation (PDE) for a single instance of the input parameters at a fixed resolution, neural operators approximate the solution map of a family of PDEs [6,7]. Despite their success, the uses of neural operators are so far restricted to relatively shallow neural networks and confined to learning hidden governing laws. In this work, we propose a novel nonlocal neural operator, which we refer to as nonlocal kernel network (NKN), that is resolution independent, characterized by deep neural networks, and capable of handling a variety of tasks such as learning governing equations and classifying images. Our NKN stems from the interpretation of the neural network as a discrete nonlocal diffusion reaction equation that, in the limit of infinite layers, is equivalent to a parabolic nonlocal equation, whose stability is analyzed via nonlocal vector calculus. The resemblance with integral forms of neural operators allows NKNs to capture long-range dependencies in the feature space, while the continuous treatment of node-to-node interactions makes NKNs resolution independent. The resemblance with neural ODEs, reinterpreted in a nonlocal sense, and the stable network dynamics between layers allow for generalization of NKN's optimal parameters from shallow to deep networks. This fact enables the use of shallow-to-deep initialization techniques [8]. Our tests show that NKNs outperform baseline methods in both learning governing equations and image classification tasks and generalize well to different resolutions and depths.