Hydrogen continues to show promise as a viable contributor to achieving energy storage goals such as energy security and decarbonization in the United States. However, many new and expanded hydrogen use applications will require identifying methods of larger-scale storage than the solutions that currently exist for smaller storage applications. One possibility is to store large quantities of gaseous hydrogen below ground level. Underground storage of other fuels such as natural gas is already currently utilized, so much of the infrastructure and basic technologies can be used as a basis for underground hydrogen storage (UHS). A few commercial UHS facilities currently exist in the United States, including salt caverns owned and operated by Air Liquide, Linde, and Conoco Philips, but UHS is still a relatively new concept that has not been widely deployed. It is necessary to understand the safety risks and hazards associated with UHS before its use can be expanded and accepted more broadly. Many of these risks are addressed through regulations, codes, and standards (RCS) issued by governing bodies and organizations with expertise in certain hazards. This report is a review of RCS documents relevant to UHS, with a particular lens on potential technical gaps in existing guidance. These gaps may be specific to the physical properties of hydrogen or due to the different technologies relevant for hydrogen vs. natural gas storage. This is meant to be a high-level review to identify relevant documents and potential gaps. Formally addressing the individual gaps identified here within the codes and standards themselves would involve a more intensive analysis and differ based on the code or standard revision processes of the various publishing organizations. Therefore, presenting specific recommendations for revising the verbiage of the documents for UHS applications is left for future work and other publications.
In Baraldi (Math Program 20:1–40, 2022), we introduced an inexact trust-region algorithm for minimizing the sum of a smooth nonconvex function and a nonsmooth convex function in Hilbert space—a class of problems that is ubiquitous in data science, learning, optimal control, and inverse problems. This algorithm has demonstrated excellent performance and scalability with problem size. In this paper, we enrich the convergence analysis for this algorithm, proving strong convergence of the iterates with guaranteed rates. In particular, we demonstrate that the trust-region algorithm recovers superlinear, even quadratic, convergence rates when using a second-order Taylor approximation of the smooth objective function term.
Due to its tunable bandgap, anisotropic behavior, and superior thermoelectric properties, device applications using layered tellurene (Te) are becoming more attractive. Here, we report a thinning technique for exfoliated tellurene nanosheets using thermal annealing in an oxygen environment. We characterize different thinning parameters, including temperature and annealing time. Based on our measurements, we show that controlled layer thinning occurs in the narrow temperature range of 325-350 °C. We also show a reliable method to form β-tellurene oxide (β-TeO2), which is an emerging wide bandgap semiconductor with promising electronic and optoelectronic properties. This wide bandgap semiconductor exhibits a broad photoluminescence (PL) spectrum with multiple peaks covering the range of 1.76-2.08 eV. This PL emission, coupled with Raman spectra, is strong evidence of the formation of 2D β-TeO2. We discuss the results obtained and the mechanisms of Te thinning and β-TeO2 formation at different temperature regimes. We also discuss the optical bandgap of β-TeO2 and show the existence of pronounced excitonic effects evident by the large exciton binding energy in this 2D β-TeO2 system that reach 1.54-1.62 eV for bulk and monolayer, respectively. Our work can be utilized to have better control over the Te nanosheet thickness. It also sheds light on the formation of well-controlled β-TeO2 layered semiconductors for electronic and optoelectronic applications.
We present a comprehensive study of transport coefficients including DC electrical conductivity and related optical properties, electrical contribution to the thermal conductivity, and the shear viscosity via ab initio molecular dynamics and density functional theory calculations on the “priority 1” cases from the “Second Charged-Particle Transport Coefficient Workshop” [Stanek et al., Phys. Plasmas (to be published 2024)]. The purpose of this work is to carefully document the entire workflow used to generate our reported transport coefficients, up to and including our definitions of finite size and statistical convergence, extrapolation techniques, and choice of thermodynamic ensembles. In pursuit of accurate optical properties, we also present a novel, simple, and highly accurate algorithm for evaluating the Kramers-Kronig relations. These heuristics are often not discussed in the literature, and it is hoped that this work will facilitate the reproducibility of our data.
Understanding temperature-dependent material decomposition and structural deformation induced by combined thermal-mechanical environments is critical for safety qualification of hardware under accident scenarios. Seeing in with X-rays elucidated the physics necessary to develop X-ray strain and thermometry diagnostics for use in optically opaque environments. Two parallel thermometry schemes were explored: X-ray fluorescence and X-ray diffraction of inorganic doped ceramics– colloquially known as thermographic phosphors. Two parallel surface strain techniques–Path-Integrated Digital Image Correlation and Frequency Multiplexed Digital Image Correlation–were demonstrated. Finally, preliminary demonstration of time-resolved digital volume correlation was performed by taking advantage of limited view reconstruction techniques. Additionally, research into blended ceramic-metal coatings was critical to generating intrinsic thermographic patterns for the future combination of X-ray strain and thermometry measurements.
ORIGEN is one of the main transmutation software packages used in nuclear engineering Modeling Tungsten Boride Neutronics in ORIGEN for Z-Facilityproblems. For the case of this study, tungsten borides are studied using a coupled framework between MCNP and the ORIGEN package of scale. The input used four compositions of tungsten boride: WB with natural boron- 10 abundance, WB with 80wt% B-10 per isotope of boron, WB4 with natural boron-10 abundance, and WB4 with 80wt% B-10 per isotope of boron. Isotopic inventories were produced for WB which show the time dependent change up to 2 years after a 6-Month irradiation. This will allow for further studies of the materials to assess things material composition changes, dose contribution, and waste management requirements.
We characterize the performance of two pixelated neutron detectors: a PMT-based array that utilizes Anger logic for pixel identification and a SiPM-based array that employs individual pixel readout. The SiPM-based array offers improved performance over the previously developed PMT-based detector both in terms of uniformity and neutron detection efficiency. Each detector array uses PSD-capable plastic scintillator as a detection medium. We describe the calibration and neutron efficiency measurement of both detectors using a 137Cs source for energy calibration and a 252Cf source for calibration of the neutron response. We find that the intrinsic neutron detection efficiency of the SiPM-based array is (30.2 ± 0.9)%, which is almost twice that of the PMT-based array, which we measure to be (16.9 ± 0.1)%.
In this work, we introduce a family of novel activation functions for deep neural networks that approximate n-ary, or n-argument, probabilistic logic. Logic has long been used to encode complex relationships between claims that are either true or false. Thus, these activation functions provide a step towards models that can efficiently encode information. Unfortunately, typical feedforward networks with elementwise activation functions cannot capture certain relationships succinctly, such as the exclusive disjunction (p xor q) and conditioned disjunction (if c then p else q). Our n-ary activation functions address this challenge by approximating belief functions (probabilistic Boolean logic) with logit representations of probability and experiments demonstrate the ability to learn arbitrary logical ground truths in a single layer. Further, by representing belief tables using a basis that associates the number of nonzero parameters with the effective arity of each belief function, we forge a concrete relationship between logical complexity and sparsity, thus opening new optimization approaches to suppress logical complexity during training. We provide a computationally efficient PyTorch implementation and test our activation functions against other logic-approximating activation functions on both traditional machine learning tasks as well as reproducing known logical relationships.