Publications

Although bulk silicon is not susceptible to fatigue, micron-scale silicon is. Several mechanisms have been proposed to explain this surprising behavior although the issue remains contentious. Here we describe published fatigue results for micron-scale thin siliconfilms and find that in general they display similar trends, in that lower cyclic stresses result in larger number of cycles to failure in stress-lifetime data. We further show that one of two classes of mechanisms is invariably proposed to explain the phenomenon. The first class attributes fatigue to a surface effect caused by subcritical (stable) cracking in the silicon-oxide layer, e.g., reaction-layer fatigue; the second class proposes that subcritical cracking in the silicon itself is the cause of fatigue in Si films. It is our contention that results to date from single and poly crystalline silicon fatigue studies provide no convincing experimentalevidence to support subcritical cracking in the silicon. Conversely, the reaction-layer mechanism is consistent with existing experimental results, and moreover provides a rational explanation for the marked difference in fatigue behavior of bulk and micron-scale silicon.

The synthesis and structural properties of three new hydridorhodium(III) complexes are reported. Hydrogenolysis of the cyclometalated rhodium dichloride complexes [RhCl(2)[(S,S)-benbox(Me(2))]] (2a-c) leads to formation of the new complexes [RhCl(2)(H)[(S,S)-ip-benbox(Me(2))H]] (3a-c) in 45% to 85% yield. Compounds 3a-c were found to have unusual features by NMR spectroscopy: in particular, downfield shifted aryl proton resonances (8.88-9.03 ppm) that were coupled to the rhodium hydride resonances. Using X-ray crystallographic studies, a variety of solid- and solution-state characterization techniques, and DFT calculations, these features were attributed to the presence of weak pi-type eta(1)-arene interactions in 3a-c.

Abstract The effects of solvent size, temperature, and polymer molecular weight on the swelling of poly(methyl methacrylate) (PMMA) thin films in low molecular weight alcohols were investigated using an in situ ellipsometer. Apparent activation energies were indicative of non‐Fickian diffusion, although optical data showed substantial Fickian character for swelling in methanol and moderate Fickian character in ethanol. Penetration rates were strongly dependent on the solvent molar volume for methanol, ethanol, and isopropanol, but 1‐butanol and 2‐pentanol had rates similar to isopropanol. The effective cross sections of these longer molecules are similar to isopropanol, and this apparently explains the similar penetration rates. The effect of polymer molecular weight (MW) on methanol penetration rates (21–27°C) was investigated with monodisperse PMMA ( M n = 6.4–40.0 × 10 4 g/mol). A minimum at intermediate MW was observed. Isopropanol swelling rates (45–52°C) were insensitive to MW. The swelling data were also used to determine parameters for transport models that describe the swelling of thin polymer films.

For many years, scalable routing for wireless communication systems was a compelling but elusive goal. Recently, several routing algorithms that exploit geographic information (e.g. GPSR) have been proposed to achieve this goal. These algorithms refer to nodes by their location, not address, and use those coordinates to route greedily, when possible, towards the destination. However, there are many situations where location information is not available at the nodes, and so geographic methods cannot be used. In this paper we define a scalable coordinate-based routing algorithm that does not rely on location information, and thus can be used in a wide variety of ad hoc and sensornet environments.

Policy must be informed by, but also facilitate the generation of, scientific evidence

Large language models (LLMs) have demonstrated remarkable performance, and organizations are racing to serve LLMs of varying sizes as endpoints for use-cases like chat, programming and search. However, efficiently serving multiple LLMs poses significant challenges for existing approaches due to varying popularity of LLMs. In the paper, we present MuxServe, a flexible spatial-temporal multiplexing system for efficient multiple LLM serving. The key insight behind is to colocate LLMs considering their popularity to multiplex memory resources, and leverage the characteristics of prefill and decoding phases to separate and flexibly colocate them to multiplex computation resources. MuxServe formally formulates the multiplexing problem, and proposes a novel placement algorithm and adaptive batch scheduling strategy to identify optimal colocations and maximize utilization. MuxServe designs a unified resource manager to enable flexible and efficient multiplexing. Evaluation results show that MuxServe can achieves up to $1.8\times$ higher throughput or processes $2.9\times$ more requests within $99\%$ SLO attainment. The code is available at: \url{https://github.com/hao-ai-lab/MuxServe}.

Here, copper electrodes, prepared by reduction of oxidized metallic copper, have been reported to exhibit higher activity for the electrochemical reduction of CO<sub>2</sub> and better selectivity toward C<sub>2</sub> and C<sub>3</sub> (C<sub>2+</sub>) products than metallic copper that has not been preoxidized. We report here an investigation of the effects of four different preparations of oxide-derived electrocatalysts on their activity and selectivity for CO<sub>2</sub> reduction, with particular attention given to the selectivity to C<sub>2+</sub> products. All catalysts were tested for CO<sub>2</sub> reduction in 0.1 M KHCO<sub>3</sub> and 0.1 M CsHCO<sub>3</sub> at applied voltages in the range from –0.7 to –1.0 V vs RHE. The best performing oxide-derived catalysts show up to ~70% selectivity to C<sub>2+</sub> products and only ~3% selectivity to C<sub>1</sub> products at –1.0 V vs RHE when CsHCO<sub>3</sub> is used as the electrolyte. In contrast, the selectivity to C<sub>2+</sub> products decreases to ~56% for the same catalysts tested in KHCO<sub>3</sub>. By studying all catalysts under identical conditions, the key factors affecting product selectivity could be discerned. These efforts reveal that the surface area of the oxide-derived layer is a critical parameter affecting selectivity. A high selectivity to C<sub>2+</sub> products is attained at an overpotential of –1 V vs RHE by operating at a current density sufficiently high to achieve a moderately high pH near the catalyst surface but not so high as to cause a significant reduction in the local concentration of CO<sub>2</sub>. On the basis of recent theoretical studies, a high pH suppresses the formation of C<sub>1</sub> relative to C<sub>2+</sub> products. At the same time, however, a high local CO<sub>2</sub> concentration is necessary for the formation of C<sub>2+</sub> products.

Constitutive laws for shape-memory alloys subjected to multiaxial loading, which are based on direct experimental observations, are generally not available in the literature. Accordingly, in the present work, tension–torsion tests are conducted on thin-walled tubes (thickness/radius ratio of 1:10) of the polycrystalline superelastic/shape-memory alloy Nitinol using various loading/unloading paths under isothermal conditions. The experimental results show significant variations in the mechanical response along the two loading axes. These are attributed to changes in the martensitic variants nucleated in response to the directionality of the applied loading, as well as to microstructural texture present in the parent material. Numerical simulations suggest that the characterization and modeling of the microstructure is of paramount importance in understanding the phenomenology of shape-memory alloys.