A5100369057- Semiconductor Quantum Structures and Devices
- Photonic and Optical Devices
- Advanced Memory and Neural Computing
- Ferroelectric and Negative Capacitance Devices
- Semiconductor materials and devices
- Nanowire Synthesis and Applications
- Advanced Optical Sensing Technologies
- Radio Frequency Integrated Circuit Design
- Advanced Sensor and Energy Harvesting Materials
- Phase-change materials and chalcogenides
- solar cell performance optimization
- Advanced Semiconductor Detectors and Materials
- 3D IC and TSV technologies
- Semiconductor Lasers and Optical Devices
- Organic Electronics and Photovoltaics
- Organic Light-Emitting Diodes Research
- Semiconductor materials and interfaces
- Advanced MEMS and NEMS Technologies
- Plasmonic and Surface Plasmon Research
- Quantum and electron transport phenomena
- Surface and Thin Film Phenomena
- Thin-Film Transistor Technologies
- Energy Harvesting in Wireless Networks
- Advancements in PLL and VCO Technologies
- Advanced Photonic Communication Systems
IBM Research - Thomas J. Watson Research Center
2014-2024
Analog Devices (United States)
2023
IBM (United States)
2013-2023
IBM Research - Almaden
2023
Hunan University
2023
Guangxi University for Nationalities
2022
China National Hybrid Rice R&D Central Hunan Hybrid Rice Reserch Center
2022
Alliance for Safe Kids
2022
Tokyo Institute of Technology
2010-2017
Chinese Academy of Sciences
2012-2016
Analog in-memory computing-a promising approach for energy-efficient acceleration of deep learning workloads-computes matrix-vector multiplications but only approximately, due to nonidealities that often are non-deterministic or nonlinear. This can adversely impact the achievable inference accuracy. Here, we develop an hardware-aware retraining systematically examine accuracy analog computing across multiple network topologies, and investigate sensitivity robustness a broad set...
Next generation on-chip light sources require high modulation bandwidth, compact footprint, and efficient power consumption. Plasmon-based are able to address the footprint challenge set by both diffraction limited of internal laser physics such as plasmon utilization. However, losses, large plasmonic-momentum these hinder coupling waveguides, thus, questioning their usefulness. Here we show that can be useful devices; they deliver outcoupling waveguides surpass speeds gain-compression. We...
Abstract Analogue memory-based deep neural networks provide energy-efficiency and per-area throughput gains relative to state-of-the-art digital counterparts such as graphics processing units. Recent advances focus largely on hardware-aware algorithmic training improvements circuits, architectures, memory devices. Optimal translation of software-trained weights into analogue hardware weights—given the plethora complex non-idealities—represents an equally important task. We report a...
Organic light emitting diodes (OLEDs) have attracted much attention in recent years as next generation lighting and displays, due to their many advantages, including superb performance, mechanical flexibility, ease of fabrication, chemical versatility, etc. In order fully realize the highly flexible features, reduce cost further improve performance OLED devices, replacing conventional indium tin oxide with better alternative transparent conducting electrodes (TCEs) is a crucial step. this...
We report an evanescently coupled photodiode that utilizes a short planar multimode waveguide. Very high responsivity (> 1 A/W) with polarization dependence less than 0.5 dB, 48-GHz bandwidth, and 11-mA saturation current were achieved.
Abstract Monolithic integration of III–V semiconductor lasers with Si circuits can reduce cost and enhance performance for optical interconnects dramatically. We propose investigate plasmonic nanolasers as monolithically integrated light source on chips due to many advantages. First, these sources be directly grown substrates free crystallographic defects the submicron cavity footprint (250 nm × 250 nm) being smaller than average defect region size heteroepitaxial material Si. Secondly,...
The 60GHz carrier with 9GHz bandwidth enables ultra-high-speed wireless communication in recent years [1–4]. To meet the demand from rapidly-increasing data traffic, IEEE802.11ay standard is one of most promising candidates aiming for 100Gb/s data-rate. Both higher-order digital modulation such as 128QAM and channel bonding at are considered to be used standard. However, more severe requirements LO feedthrough (LOFT) image-rejection ratio (IMRR) have satisfied, so much higher accuracy...
We report an avalanche photodiode with undepleted p-type InGaAs absorption region and a thin InAlAs multiplication layer. The motivation for utilizing layer, which is similar to that in the unitraveling carrier photodiode, reduce dark current. A current below 1 nA at gain of 10 gain–bandwidth product 160 GHz are demonstrated.
A separate absorption, charge, and multiplication In/sub 0.53/Ga/sub 0.47/As-In/sub 0.52/Al/sub 0.48/As avalanche photodiode with an impact-ionization-engineered region is reported. By implementing electric field gradient in the region, better control of impact-ionization can be achieved. Gain-bandwidth product 160 GHz excess noise factor equivalent k value 0.1 are demonstrated.
A systematic study of high-saturation-current p-i-n In/sub 0.53/Ga/sub 0.47/As photodiodes with a partially depleted absorber (PDA) has been made under front (p-side) and back (n-side) illumination. The photodiode structure consists an absorption region (450-nm p-InGaAs, 250-nm unintentionally doped InGaAs, 60-nm n-InGaAs) sandwiched between p- n-InP layers. For illumination 34-/spl mu/m-diameter at 2-V bias the saturation currents were 23 24 mA 10 1 GHz, respectively. Under similar...
At mm-wave frequency, the layout of CMOS transistors has a larger effect on device performance than ever before in low frequency. In this work, distance between gate and drain contact (D <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">gd</sub> ) been enlarged to obtain better maximum available gain (MAG). A 0.6 dB MAG improvement is realized when D changes from 60 nm 200 nm. By using asymmetric-layout transistor, four-stage common-source noise...
A helium-3 ion bombardment technique is proposed to realize highQ inductors by creating locally semi-insulating substrate areas. dose of 1.0 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">13</sup> cm xmlns:xlink="http://www.w3.org/1999/xlink">-2</sup> increases a Si resistivity from 4 Ω · above 1 kQ cm, which improves the quality factor 2-nH inductor with 140-μm diameter 38% (Q = 16.3). An aluminum mask used for covering active areas, and...
Abstract Phase change memory (PCM) is one of the most promising candidates for non‐von Neumann based analog in‐memory computing–particularly inference previously‐trained deep neural networks (DNN). It shown that PCM electrical properties can be tuned systematically using a projection liner, which designed resistance drift mitigation, in manufacturable mushroom PCM. A systematic study properties‐including values, window, drift, read noise, and their impact on accuracy large various types with...
Flexible and stretchable electronics are becoming increasingly important in many emerging applications. Due to the outstanding electrical properties of single crystal semiconductors, there is great interest releasing thin films fabricating flexible with these conventionally rigid materials. In this study authors report a universal layer release process, called "3D spalling," extending beyond prior art. contrast conventional way removing blanket layers from their substrates, new process...
Franckeite, a natural van der Waals heterostructure with excellent physicochemical characteristics, has shown great potential for high-performance optoelectronic device. Here, the few-layer franckeite nanosheets have been prepared by liquid phase-exfoliation method successfully, and nonlinear optical absorption carrier dynamics behaviors of studied experimentally. The show wavelength- intensity-dependent exhibit ultrafast dynamic process. nonlinearity in further validated via an Er3+-doped...
We employ the organic semiconductor 3,4,9,10-perylene-tetracarboxylic-dianhydride (PTCDA) as a nanometer thick window layer for p-InP/indium tin oxide (ITO) Schottky barrier diode solar cells. The power conversion efficiency is enhanced compared to ITO/InP cells lacking PTCDA layer, primarily due neutralizing InP surface state charges via hole injection from PTCDA. This leads an increased ITO/p-InP height, and hence open circuit voltage. of increases 13.2±0.5% cell 15.4±0.4% ITO/4 nm...
Thin, lightweight and flexible electronics are being regarded as an important evolutionary step in the development of novel technological products. Interestingly, this trend has emerged a wide range industries; from microelectronics to photovoltaics even solid state lighting. Historically, most attempts enable flexibility have focused on introduction new material systems that, so far, severely compromise performance compared state-of-the-art The few approaches that do attempt render...
Abstract Advancement in microelectronics technology enables autonomous edge computing platforms the size of a dust mote (<1 mm), bringing efficient and low‐cost artificial intelligence close to end user Internet‐of‐Things (IoT) applications. The key challenge for these compact high‐performance computers is integration power source that satisfies high‐power‐density requirement does not increase complexity cost packaging. Here, it shown dust‐sized III–V photovoltaic (PV) cells grown on Si...
Analog memory presents a promising solution in the face of growing demand for energy-efficient artificial intelligence (AI) at edge. In this study, we demonstrate efficient deep neural network transfer learning utilizing hardware and algorithm co-optimization an analog resistive random-access (ReRAM) array. For first time, illustrate that open-loop (DNN) image classification tasks, convergence rates can be accelerated by approximately 3.5 times through utilization co-optimized ReRAM...
In-memory computing using analog non-volatile memory (NVM) devices can improve the speed and reduce latency of deep neural network (DNN) inference. It has been recently shown that neuromorphic crossbar arrays, where each weight is implemented conductance values phase-change devices, achieve competitive accuracy high power efficiency. However, due to large amount NVMs needed challenge for making NVM these chips typically include some failed from fabrication or developed over time. We study...
The potential benefit from the combination of optoelectronic and electronic functionality III–V semiconductors with silicon technology is one most desired outcomes to date. Here we have systematically investigated optical properties InAs quantum structure embedded in GaAs grown on patterned sub-micron nanosize holes Si(001). material tends accumulate a depletion region observed around when GaAs/InAs/GaAs deposited directly By use 60 nm SiO2 layer patterning through oxide substrate,...
Although historically studied as a failure mode, substrate spalling can be transformed into versatile layer transfer method by carefully controlling the thickness and intrinsic stress of surface layer, mechanically guiding fracture path. This Controlled Spalling process requires no specialized equipment applied to essentially any brittle substrate. We have successfully demonstrated i) device fabrication in Si, Ge, III-V based materials, ii) removal fully-processed CMOS circuits, iii)...