Search

Data-driven modeling based on machine learning (ML) is showing enormous potential for weather forecasting. Rapid progress has been made with impressive results for some applications. The uptake of ML methods could be a game-changer for the incremental progress in traditional numerical weather prediction (NWP) known as the 'quiet revolution' of weather forecasting. The computational cost of running a forecast with standard NWP systems greatly hinders the improvements that can be made from increasing model resolution and ensemble sizes. An emerging new generation of ML models, developed using high-quality reanalysis datasets like ERA5 for training, allow forecasts that require much lower computational costs and that are highly-competitive in terms of accuracy. Here, we compare for the first time ML-generated forecasts with standard NWP-based forecasts in an operational-like context, initialized from the same initial conditions. Focusing on deterministic forecasts, we apply common forecast verification tools to assess to what extent a data-driven forecast produced with one of the recently developed ML models (PanguWeather) matches the quality and attributes of a forecast from one of the leading global NWP systems (the ECMWF IFS). The results are very promising, with comparable skill for both global metrics and extreme events, when verified against both the operational analysis and synoptic observations. Increasing forecast smoothness and bias drift with forecast lead time are identified as current drawbacks of ML-based forecasts. A new NWP paradigm is emerging relying on inference from ML models and state-of-the-art analysis and reanalysis datasets for forecast initialization and model training.

Power consumption is a key challenge in millimeter wave (mmWave) receiver front-ends, due to the need to support high dimensional antenna arrays at wide bandwidths. Recently, there has been considerable work in developing low-power front-ends, often based on low-resolution ADCs and low-power mixers. A critical but less studied consequence of such designs is the relatively low-dynamic range which in turn exposes the receiver to adjacent carrier interference and blockers. This paper provides a general mathematical framework for analyzing the performance of mmWave front-ends in the presence of out-of-band interference. The goal is to elucidate the fundamental trade-off of power consumption, interference tolerance and in-band performance. The analysis is combined with detailed network simulations in cellular systems with multiple carriers, as well as detailed circuit simulations of key components at 140 GHz. The analysis reveals critical bottlenecks for low-power interference robustness and suggests designs enhancements for use in practical systems., Comment: Appeared at the Asilomar Conference on Signals, Systems, and Computers 2021

Triple heterojunction (THJ) TFETs have been proposed to resolve the low ON-current challenge of TFETs. However, the design space for THJ-TFETs is limited by fabrication challenges with respect to device dimensions and material interfaces. This work shows that the original THJ-TFET design with 12 nm body thickness has poor performance, because its sub-threshold swing is 50 mV/dec and the ON-current is only 6 $\mu A/\mu m$. To improve the performance, the doping profile of THJ-TFET is engineered to boost the resonant tunneling efficiency. The proposed THJ-TFET design shows a sub-threshold swing of 40 mV/dec over four orders of drain current and an ON-current of 325 uA/um with VGS = 0.3 V. Since THJ-TFETs have multiple quantum wells and material interfaces in the tunneling junction, quantum transport simulations in such devices are complicated. State-of-the-art mode-space quantum transport simulation, including the effect of thermalization and scattering, is employed in this work to optimize THJ-TFET design.

Millimeter wave MIMO combines the benefits of compact antenna arrays with a large number of elements and massive bandwidths, so that fully digital beamforming has the potential of supporting a large number of simultaneous users with {\it per user} data rates of multiple gigabits/sec (Gbps). In this paper, we develop an analytical model for the impact of nonlinearities in such a system, and illustrate its utility in providing hardware design guidelines regarding two key challenges: the low available precision of analog-to-digital conversion at high sampling rates, and nonlinearities in ultra-high speed radio frequency (RF) and baseband circuits. We consider linear minimum mean square error (LMMSE) reception for a multiuser MIMO uplink, and provide performance guarantees based on two key concepts: (a) summarization of the impact of per-antenna nonlinearities via a quantity that we term the "intrinsic SNR", (b) using linear MMSE performance in an ideal system without nonlinearities to bound that in our non-ideal system. For our numerical results, we employ nominal parameters corresponding to outdoor picocells operating at a carrier frequency of 140 GHz, with a data rate of 10 Gbps per user.

This paper investigates the effect of oscillator phase noise on a multiuser millimeter wave (mmWave) massive MIMO uplink as we scale up the number of base station antennas, fixing the load factor, defined as the ratio of the number of simultaneous users to the number of base station antennas. We consider a modular approach in which the base station employs an array of subarrays, or "tiles." Each tile supports a fixed number of antennas, and can therefore be implemented using a separate radio frequency integrated circuit (RFIC), with synchronization across tiles accomplished by employing a phased locked loop in each tile to synthesize an on-chip oscillator at the carrier frequency by locking on to a common lower frequency reference clock. Assuming linear minimum mean squared error (LMMSE) multiuser detection, we provide an analytical framework that can be used to specify the required power spectral density (PSD) mask for phase noise for a target system performance. Our analysis for the phase noise at the output of the LMMSE receiver indicates two distinct effects: self-noise for each user which is inversely proportional to the number of tiles, and cross-talk between users which is insensitive to the number of tiles, and is proportional to the load factor. These analytical predictions, verified by simulations for a 140 GHz system targeting a per-user data rate of 10 Gbps, show that tiling is a robust approach for scaling. Numerical results for our proposed design approach yield relatively relaxed specifications for phase noise PSD masks.

We report on the selective-area chemical beam epitaxial growth of InAs in-plane, one-dimensional (1-D) channels using patterned SiO$_{2}$-coated InP(001), InP(111)B, and InP(110) substrates to establish a scalable platform for topological superconductor networks. Top-view scanning electron micrographs show excellent surface selectivity and dependence of major facet planes on the substrate orientations and ridge directions, and the ratios of the surface energies of the major facet planes were estimated. Detailed structural properties and defects in the InAs nanowires (NWs) were characterized by transmission electron microscopic analysis of cross-sections perpendicular to the NW ridge direction and along the NW ridge direction. Electrical transport properties of the InAs NWs were investigated using Hall bars, a field effect mobility device, a quantum dot, and an Aharonov-Bohm loop device, which reflect the strong spin-orbit interaction and phase-coherent transport characteristic in the selectively grown InAs systems. This study demonstrates that selective-area chemical beam epitaxy is a scalable approach to realize semiconductor 1-D channel networks with the excellent surface selectivity and this material system is suitable for quantum transport studies.

A high performance triple-heterojunction (3HJ) design has been previously proposed for tunneling FETs (TFETs). Compared with single heterojunction (HJ) TFETs, the 3HJ TFETs have both shorter tunneling distance and two transmission resonances that significantly improve the ON-state current ($I_{\rm{ON}}$). Coherent quantum transport simulation predicts, that $I_{\rm{ON}}=460\rm{\mu A/\mu m}$ can be achieved at gate length $Lg=15\rm{nm}$, supply voltage $V_{\rm{DD}}=0.3\rm{V}$, and OFF-state current $I_{\rm{OFF}}=1\rm{nA/\mu m}$. However, strong electron-phonon and electron-electron scattering in the heavily doped leads implies, that the 3HJ devices operate far from the ideal coherent limit. In this study, such scattering effects are assessed by a newly developed multiscale transport model, which combines the ballistic non-equilibrium Green's function method for the channel and the drift-diffusion scattering method for the leads. Simulation results show that the thermalizing scattering in the leads both degrades the 3HJ TFET's subthreshold swing through scattering induced leakage and reduces the turn-on current through the access resistance. Assuming bulk scattering rates and carrier mobilities, the $I_{\rm{ON}}$ is dropped from $460\rm{\mu A/\mu m}$ down to $254\rm{\mu A/\mu m}$, which is still much larger than the single HJ TFET case.

Ideal, completely coherent quantum transport calculations had predicted that superlattice MOSFETs may offer steep subthreshold swing performance below 60mV/dec to around 39mV/dec. However, the high carrier density in the superlattice source suggest that scattering may significantly degrade the ideal device performance. Such effects of electron scattering and decoherence in the contacts of superlattice MOSFETs are examined through a multiscale quantum transport model developed in NEMO5. This model couples NEGF-based quantum ballistic transport in the channel to a quantum mechanical density of states dominated reservoir, which is thermalized through strong scattering with local quasi-Fermi levels determined by drift-diffusion transport. The simulations show that scattering increases the electron transmission in the nominally forbidden minigap therefore degrading the subthreshold swing (S.S.) and the ON/OFF DC current ratio. This degradation varies with both the scattering rate and the length of the scattering dominated regions. Different superlattice MOSFET designs are explored to mitigate the effects of such deleterious scattering. Specifically, shortening the spacer region between the superlattice and the channel from 3.5 nm to 0 nm improves the simulated S.S. from 51mV/dec. to 40mV/dec. I. INTRODUCTION, Comment: 16 pages, 8 figures

GaSb/InAs heterojunction tunnel field-effect transistors are strong candidates in building future low-power integrated circuits, as they could provide both steep subthreshold swing and large ON-state current ($I_{\rm{ON}}$). However, at short channel lengths they suffer from large tunneling leakage originating from the small band gap and small effective masses of the InAs channel. As proposed in this article, this problem can be significantly mitigated by reducing the channel thickness meanwhile retaining a thick source-channel tunnel junction, thus forming a design with a non-uniform body thickness. Because of the quantum confinement, the thin InAs channel offers a large band gap and large effective masses, reducing the ambipolar and source-to-drain tunneling leakage at OFF state. The thick GaSb/InAs tunnel junction, instead, offers a low tunnel barrier and small effective masses, allowing a large tunnel probability at ON state. In addition, the confinement induced band discontinuity enhances the tunnel electric field and creates a resonant state, further improving $I_{\rm{ON}}$. Atomistic quantum transport simulations show that ballistic $I_{\rm{ON}}=284$A/m is obtained at 15nm channel length, $I_{\rm{OFF}}=1\times10^{-3}$A/m, and $V_{\rm{DD}}=0.3$V. While with uniform body thickness, the largest achievable $I_{\rm{ON}}$ is only 25A/m. Simulations also indicate that this design is scalable to sub-10nm channel length., Comment: 4 pages, 8 figures

A triple-heterojunction (3HJ) design is employed to improve p-type InAs/GaSb heterojunction (HJ) tunnel FETs. The added two HJs (AlInAsSb/InAs in the source and GaSb/AlSb in the channel) significantly shorten the tunnel distance and create two resonant states, greatly improving the ON state tunneling probability. Moreover, the source Fermi degeneracy is reduced by the increased source (AlInAsSb) density of states and the OFF state leakage is reduced by the heavier channel (AlSb) hole effective masses. Quantum ballistic transport simulations show, that with V_{DD} = 0.3V and I_{OFF} = 10^{-3}A/m, I_{ON} of 582A=m (488A=m) is obtained at 30nm (15nm) channel length, which is comparable to n-type 3HJ counterpart and significantly exceeding p-type silicon MOSFET. Simultaneously, the nonlinear turn on and delayed saturation in the output characteristics are also greatly improved.

Control of faceting during epitaxy is critical for nanoscale devices. This work identifies the origins of gaps and different facets during regrowth of InGaAs adjacent to patterned features. Molecular beam epitaxy (MBE) near SiO2 or SiNx led to gaps, roughness, or polycrystalline growth, but metal modulated epitaxy (MME) produced smooth and gap-free "rising tide" (001) growth filling up to the mask. The resulting self-aligned FETs were dominated by FET channel resistance rather than source-drain access resistance. Higher As fluxes led first to conformal growth, then pronounced {111} facets sloping up away from the mask., Comment: 18 pages, 7 figures