Technion – Israel Institute of Technology

About: Technion – Israel Institute of Technology is a education organization based out in Haifa, Israel. It is known for research contribution in the topics: Population & Upper and lower bounds. The organization has 31714 authors who have published 79377 publications receiving 2603976 citations. The organization is also known as: Technion Israel Institute of Technology & Ṭekhniyon, Makhon ṭekhnologi le-Yiśraʼel.

Building devices from colloidal quantum dots.

Abstract: BACKGROUND The Information Age was founded on the semiconductor revolution, marked by the growth of high-purity semiconductor single crystals. The resultant design and fabrication of electronic devices exploits our ability to control the concentration, motion, and dynamics of charge carriers in the bulk semiconductor solid state. Our desire to introduce electronics everywhere is fueled by opportunities to create intelligent and enabling devices for the information, communication, consumer product, health, and energy sectors. This demand for ubiquitous electronics is spurring the design of materials that exhibit engineered physical properties and that can enable new fabrication methods for low-cost, large-area, and flexible devices. Semiconductors, which are at the heart of electronics and optoelectronics, come with high demands on chemical purity and structural perfection. Alternatives to silicon technology are expected to combine the electronic and optical properties of inorganic semiconductors (high charge carrier mobility, precise n- and p-type doping, and the ability to engineer the band gap energy) with the benefits of additive device manufacturing: low cost, large area, and the use of solution-based fabrication techniques. Along these lines, colloidal semiconductor quantum dots (QDs), which are nanoscale crystals of analogous bulk semiconductor crystals, offer a powerful platform for device engineers. Colloidal QDs may be tailored in size, shape, and composition and their surfaces functionalized with molecular ligands of diverse chemistry. At the nanoscale (typically 2 to 20 nm), quantum and dielectric confinement effects give rise to the prized size-, shape-, and composition-tunable electronic and optical properties of QDs. Surface ligands enable the stabilization of QDs in the form of colloids, allowing their bottom-up assembly into QD solids. The physical properties of QD solids can be designed by selecting the characteristics of the individual QD building blocks and by controlling the electronic communication between the QDs in the solid state. These QD solids can be engineered with application-specific electronic and optoelectronic properties for the large-area, solution-based assembly of devices. ADVANCES The large surface-to-volume ratio of QDs places a substantial importance on the composition and structure of the surface in defining the physical properties that govern the concentration, motion, and dynamics of excitations and charge carriers in QD solids. Recent studies have shown pathways to passivate uncoordinated atoms at the QD surface that act to trap and scatter charge carriers. Surface atoms, ligands, and ions can serve as dopants to control the electron affinity of QD solids. Surface ligands and surrounding matrices control the barriers to electronic, excitonic, and thermal transport between QDs and between QDs and matrices. New ligand chemistries and matrix materials have been reported that provide freedom to control the dynamics of excitons and charge carriers and to design device interfaces. These advances in engineering the chemical and physical properties of the QD surface have been translated into recent achievements of high-mobility transistors and circuits, high-quantum-yield photodetectors and light-emitting devices, and high-efficiency photovoltaic devices. OUTLOOK The dominant role and dynamic nature of the QD surface, and the strong motive to build novel QD devices, will drive the exploration of new surface chemistries and matrix materials, processes for their assembly and integration with other materials in devices, and measurements and simulations with which to map the relationship between surface chemistry and materials and device properties. Challenges remain to achieve full control over the carrier type, concentration, and mobility in the QD channel and the barriers and traps at device interfaces that limit the gain and speed of QD electronics. Surface chemistries that allow for both long carrier lifetime and high carrier mobility and the freedom to engineer the bandgap and band alignment of QDs and other device layers are needed to exploit physics particular to QDs and to advance device architectures that contribute to improving the performance of QD optoelectronics. The importance of thermal transport in QD solids and their devices is an essential emerging topic that promises to become of greater importance as we develop QD devices.

The relationship between luminosity and broad-line region size in active galactic nuclei

Abstract: We reinvestigate the relationship between the characteristic broad-line region size (RBLR) and the Balmer emission-line, X-ray, UV, and optical continuum luminosities. Our study makes use of the best available determinations of RBLR for a large number of active galactic nuclei (AGNs) from Peterson et al. Using their determinations of RBLR for a large sample of AGNs and two different regression methods, we investigate the robustness of our correlation results as a function of data subsample and regression technique. Although small systematic differences were found depending on the method of analysis, our results are generally consistent. Assuming a power-law relation RBLR ∝ Lα, we find that the mean best-fitting α is about 0.67 ± 0.05 for the optical continuum and the broad Hβ luminosity, about 0.56 ± 0.05 for the UV continuum luminosity, and about 0.70 ± 0.14 for the X-ray luminosity. We also find an intrinsic scatter of ~40% in these relations. The disagreement of our results with the theoretical expected slope of 0.5 indicates that the simple assumption of all AGNs having on average the same ionization parameter, BLR density, column density, and ionizing spectral energy distribution is not valid and there is likely some evolution of a few of these characteristics along the luminosity scale.

Quantized neural networks: training neural networks with low precision weights and activations

Abstract: We introduce a method to train Quantized Neural Networks (QNNs) -- neural networks with extremely low precision (e.g., 1-bit) weights and activations, at run-time. At traintime the quantized weights and activations are used for computing the parameter gradients. During the forward pass, QNNs drastically reduce memory size and accesses, and replace most arithmetic operations with bit-wise operations. As a result, power consumption is expected to be drastically reduced. We trained QNNs over the MNIST, CIFAR-10, SVHN and ImageNet datasets. The resulting QNNs achieve prediction accuracy comparable to their 32-bit counterparts. For example, our quantized version of AlexNet with 1-bit weights and 2-bit activations achieves 51% top-1 accuracy. Moreover, we quantize the parameter gradients to 6-bits as well which enables gradients computation using only bit-wise operation. Quantized recurrent neural networks were tested over the Penn Treebank dataset, and achieved comparable accuracy as their 32-bit counterparts using only 4-bits. Last but not least, we programmed a binary matrix multiplication GPU kernel with which it is possible to run our MNIST QNN 7 times faster than with an unoptimized GPU kernel, without suffering any loss in classification accuracy. The QNN code is available online.