Semiconductor alloys exhibit a strong dependence of effective thermal conductivity on measurement frequency. So far this quasiballistic behavior has only been interpreted phenomenologically, providing limited insight into the underlying thermal transport dynamics. Here, we show that quasiballistic heat conduction in semiconductor alloys is governed by Lévy superdiffusion. By solving the Boltzmann transport equation (BTE) with ab initio phonon dispersions and scattering rates, we reveal a transport regime with fractal space dimension and superlinear time evolution of mean-square energy displacement . The characteristic exponents are directly interconnected with the order of the dominant phonon scattering mechanism and cumulative conductivity spectra resolved for relaxation times or mean free paths through the simple relations . The quasiballistic transport inside alloys is no longer governed by Brownian motion, but instead is dominated by Lévy dynamics. This has important implications for the interpretation of thermoreflectance (TR) measurements with modified Fourier theory. Experimental values for InGaAs and SiGe, determined through TR analysis with a novel Lévy heat formalism, match ab initio BTE predictions within a few percent. Our findings lead to a deeper and more accurate quantitative understanding of the physics of nanoscale heat-flow experiments.
Understanding nanoscale thermal transport is of substantial importance for designing contemporary semiconductor technologies. Heat removal from small sources is well established to be severely impeded compared to diffusive predictions due to the ballistic nature of the dominant heat carriers. Experimental observations are commonly interpreted through a reduction of effective thermal conductivity, even though most measurements only probe a single aggregate thermal metric. Here, we employ thermoreflectance thermal imaging to directly visualise the 2D temperature field produced by localised heat sources on InGaAs with characteristic widths down to 100 nm. Besides displaying effective thermal performance reductions up to 50% at the active junctions in agreement with prior studies, our steady-state thermal images reveal that, remarkably, 1–3 μm adjacent to submicron devices the crosstalk is actually reduced by up to fourfold. Submicrosecond transient imaging additionally shows responses to be faster than conventionally predicted. A possible explanation based on hydrodynamic heat transport, and some open questions, are discussed.
The recent developments in nanoscale solid-state cooling are reviewed. This includes both theoretical and experimental studies of different physical concepts, as well as nanostructured material design and device configurations. We primarily focus on thermoelectric, thermionic and thermo-magnetic coolers. Particular emphasis is given to the concepts based on metal–semiconductor superlattices, graded materials, non-equilibrium thermoelectric devices, Thomson coolers, and photon assisted Peltier coolers as promising methods for efficient solid-state cooling. Thermomagnetic effects such as magneto–Peltier and Nernst–Ettingshausen cooling are briefly described and recent advances and future trends in these areas are reviewed. The ongoing progress in solid-state cooling concepts such as spin-calorimetrics, electrocalorics, non-equilibrium/nonlinear Peltier devices, superconducting junctions and two-dimensional materials are also elucidated and practical achievements are reviewed. We explain the thermoreflectance thermal imaging microscopy and the transient Harman method as two unique techniques developed for characterization of thermoelectric microrefrigerators. The future prospects for solid-state cooling are briefly summarized.
Nearly all experimental observations of quasiballistic heat flow are interpreted using Fourier theory with modified thermal conductivity. Detailed Boltzmann transport equation (BTE) analysis, however, reveals that the quasi-ballistic motion of thermal energy in semiconductor alloys is no longer Brownian but instead exhibits Lévy dynamics with fractal dimension . Here, we present a framework that enables full three-dimensional experimental analysis by retaining all essential physics of the quasiballistic BTE dynamics phenomenologically. A stochastic process with just two fitting parameters describes the transition from pure Lévy superdiffusion as short length and time scales to regular Fourier diffusion. The model provides accurate fits to time domain thermoreflectance raw experimental data over the full modulation frequency range without requiring any “effective” thermal parameters and without any a priori knowledge of microscopic phonon scattering mechanisms. Identified values for InGaAs and SiGe match ab initio BTE predictions within a few percent. Our results provide experimental evidence of fractal Lévy heat conduction in semiconductor alloys. The formalism additionally indicates that the transient temperature inside the material differs significantly from Fourier theory and can lead to improved thermal characterization of nanoscale devices and material interfaces.
Smart manufacturing systems are considered the next generation of manufacturing applications. One important goal of the smart manufacturing system is to rapidly detect and anticipate failures to reduce maintenance cost and minimize machine downtime. This often boils down to detecting anomalies within the sensor data acquired from the system which has different characteristics with respect to the operating point of the environment or machines, such as, the RPM of the motor. In this paper, we analyze four datasets from sensors deployed in manufacturing testbeds. We detect the level of defect for each sensor data leveraging deep learning techniques. We also evaluate the performance of several traditional and ML-based forecasting models for predicting the time series of sensor data. We show that careful selection of training data by aggregating multiple predictive RPM values is beneficial. Then, considering the sparse data from one kind of sensor, we perform transfer learning from a high data rate sensor to perform defect type classification. We release our manufacturing database corpus (4 datasets) and codes for anomaly detection and defect type classification for the community to build on it. Taken together, we show that predictive failure classification can be achieved, paving the way for predictive maintenance.