Search for: All records

Accurate and computationally-viable representations of clouds and turbulence are a long-standing challenge for climate model development. Traditional parameterizations that crudely but efficiently approximate these processes are a leading source of uncertainty in long-term projected warming and precipitation patterns. Machine Learning (ML)-based parameterizations have long been hailed as a promising alternative with the potential to yield higher accuracy at a fraction of the cost of more explicit simulations. However, these ML variants are often unpredictably unstable and inaccurate in \textit{coupled} testing (i.e. in a downstream hybrid simulation task where they are dynamically interacting with the large-scale climate model). These issues are exacerbated in out-of-distribution climates. Certain design decisions such as ``climate-invariant" feature transformation for moisture inputs, input vector expansion, and temporal history incorporation have been shown to improve coupled performance, but they may be insufficient for coupled out-of-distribution generalization. If feature selection and transformations can inoculate hybrid physics-ML climate models from non-physical, out-of-distribution extrapolation in a changing climate, there is far greater potential in extrapolating from observational data. Otherwise, training on multiple simulated climates becomes an inevitable necessity. While our results show generalization benefits from these design decisions, the obtained improvment does not sufficiently preclude the necessity of using multi-climate simulated training data. 

Terrestrial experiments that use electrons in Earth as a spin-polarized source have been demonstrated to provide strong bounds on exotic long-range spin-spin and spin-velocity interactions. These bounds constrain the coupling strength of many proposed ultralight bosonic dark-matter candidates. Recently, it was pointed out that a monopole-dipole coupling between the Sun and the spin-polarized electrons of Earth would result in a modification of the precession of the perihelion of Earth. Using an estimate for the net spin polarization of Earth and experimental bounds on Earth’s perihelion precession, interesting constraints were placed on the magnitude of this monopole-dipole coupling. Here we investigate the spin associated with Earth’s electrons. We find that there are about $6\times {10}^{41}$spin-polarized electrons in the mantle and crust of Earth oriented antiparallel to their local magnetic field. However, when integrated over any spherically symmetric Earth model, we find that the vector sum of these spins is zero. In order to establish a lower bound on the magnitude of the net spin along Earth’s rotation axis we have investigated three of the largest breakdowns of Earth’s spherical symmetry: the large low shear-velocity provinces of the mantle, the crustal composition, and the oblate spheroid of Earth. From these investigations we conclude that there are at least $5\times {10}^{38}$spin-polarized electrons aligned antiparallel to Earth’s rotation axis. This analysis suggests that the bounds on the monopole-dipole coupling that were extracted from Earth’s perihelion precession need to be relaxed by a factor of about 2000. Published by the American Physical Society2025 

We report a characterization of the spatial resolution of terahertz (THz) apertureless near-field imaging of metal lines deeply buried beneath a silicon dioxide layer. We find a good resolution for edge contrast, even in the case where the capping layer is considerably thicker than the tip radius. We find that contrast and resolution depend on demodulation frequency, thickness of the capping layer, and radius of the tip. Furthermore, we observe a distinct dependence of the contrast on the direction of the incoming radiation, in both experiments and simulations. Characterization of buried features can be a valuable tool in non-contact failure analysis of semiconductor devices. 

This article addresses the quadrotors’ safety-critical landing control problem with external uncertainties and collision avoidance. A geometrically robust hierarchical control strategy is proposed for an underactuated quadrotor, which consists of a slow outer loop controlling the position and a fast inner loop regulating the attitude. First, an estimation error quantified (EEQ) observer is developed to identify and compensate for the target’s linear acceleration and the translational disturbances, whose estimation error has a nonnegative upper bound. Furthermore, an outer-loop controller is designed by embedding the EEQ observer and control barrier functions (CBFs), in which the negative effects of external uncertainties, collision avoidance, and input saturation are thoroughly considered and effectively attenuated. For the inner-loop subsystem, a geometric controller with a robust integral of the sign of the error (RISE) control structure is developed to achieve disturbances rejection and asymptotic attitude tracking. Based on Lyapunov techniques and the theory of cascade systems, it is rigorously proven that the closed-loop system is uniformly ultimately bounded. Finally, the effectiveness of the proposed control strategy is demonstrated through numerical simulations and hardware experiments. 

Abstract Previous efforts to measure atmospheric iodine have focused on marine and coastal regions. We report the first ground‐based tropospheric iodine monoxide (IO) radical observations over the central continental United States. Throughout April 2022, IO columns above Storm Peak Laboratory, Colorado (3,220 m.a.s.l.) ranged from 0.7 ± 0.5 to 3.6 ± 0.5 × 10¹²(average: 1.9 × 10¹² molec cm⁻²). IO was consistently elevated in air masses transported from over the Pacific Ocean. The observed IO columns were up to three times higher and the range was larger than predicted by a global model, which warrants further investigation into iodine sources, sinks, ozone loss, and particle formation. IO mixing ratios increased with altitude. At the observed levels, iodine may be competitive with bromine as an oxidant of elemental mercury at cold temperatures typical of the free troposphere. Iodine‐induced mercury oxidation is missing in atmospheric models, understudied, and helps explain model underestimation of oxidized mercury measurements. 

In this work, we present two embedded soft optical waveguide sensors designed for real-time onboard configuration sensing in soft actuators for robotic locomotion. Extending the contributions of our collaborators who employed external camera systems to monitor the gaits of twisted-beam structures, we strategically integrate our OptiGap sensor system into these structures to monitor their dynamic behavior. The system is validated through machine learning models that correlate sensor data with camera-based motion tracking, achieving high accuracy in predicting forward or reverse gaits and validating its capability for real-time sensing. Our second sensor, consisting of a square cross-section fiber pre-twisted to 360 degrees, is designed to detect the chirality of reconfigurable twisted beams. Experimental results confirm the sensor’s effectiveness in capturing variations in light transmittance corresponding to twist angle, serving as a reliable chirality sensor. The successful integration of these sensors not only improves the adaptability of soft robotic systems but also opens avenues for advanced control algorithms.