2019, Milo, V., Zambelli, C., Olivo, P.

Training and recognition with neural networks generally require high throughput, high energy efficiency, and scalable circuits to enable artificial intelligence tasks to be operated at the edge, i.e., in battery-powered portable devices and other limited-energy environments. In this scenario, scalable resistive memories have been proposed as artificial synapses thanks to their scalability, reconfigurability, and high-energy efficiency, and thanks to the ability to perform analog computation by physical laws in hardware. In this work, we study the material, device, and architecture aspects of resistive switching memory (RRAM) devices for implementing a 2-layer neural network for pattern recognition. First, various RRAM processes are screened in view of the device window, analog storage, and reliability. Then, synaptic weights are stored with 5-level precision in a 4 kbit array of RRAM devices to classify the Modified National Institute of Standards and Technology (MNIST) dataset. Finally, classification performance of a 2-layer neural network is tested before and after an annealing experiment by using experimental values of conductance stored into the array, and a simulation-based analysis of inference accuracy for arrays of increasing size is presented. Our work supports material-based development of RRAM synapses for novel neural networks with high accuracy and low-power consumption. © 2019 Author(s).

2019, Kissinger, G., Kot, D., Costina, I., Lisker, M.

PECVD nitride layers with different layer stress ranging from about 315 MPa to −1735 MPa were deposited on silicon wafers with similar concentration of interstitial oxygen. After a thermal treatment consisting of nucleation at 650°C for 4 h or 8 h followed annealing 780°C 3 h + 1000°C 16 h in nitrogen, the profiles of the oxide precipitate density were investigated. The binding states of hydrogen in the layers was investigated by FTIR. There is a clear effect of the layer stress on oxide precipitate nucleation. The higher the compressive layer stress is the higher is a BMD peak below the front surface. If the nitride layer is removed after the nucleation anneal the BMD peak below the front surface becomes lower. It is possible to model the BMD peak below the surface by vacancy in-diffusion from the silicon/nitride interface. With increasing duration of the nucleation anneal the vacancy injection from the silicon/nitride interface decreases and with increasing compressive layer stress it increases. © The Author(s) 2019.

2019, Steglich, Patrick, Villringer, Claus, Dietzel, Birgit, Mai, Christian, Schrader, Sigurd, Casalboni, Mauro, Mai, Andreas

A novel method to determine the dispersion of the quadratic electro-optic effect in nonlinear optical materials by using a silicon-on-insulator microring resonator is presented. The microring consists of a silicon slot waveguide enabling large dc electric field strength at low applied voltages. The dispersion of third-order hyperpolarizability of a linear conjugated dye is approximated by using a two-level model for the off-resonant spectral region. As an example, the dispersion of the resonance wavelength of the resonator filled with a dye doped polymer was measured in dependence of the applied dc voltage. The polymer was poly (methylmethacrylate) doped with 5 wt% disperse red 1 (DR1), and the measurements have been carried out at the telecommunication wavelength band around 1550 nm (optical C-band). The described measurements represent a new technique to determine the dispersion of the third-order susceptibility and molecular hyperpolarizability of the material filled into the slot of the ring-resonator. © 2019 IEEE.

2019, Sammak, Amir, Sabbagh, Diego, Hendrickx, Nico W., Lodari, Mario, Wuetz, Brian Paquelet, Tosato, Alberto, Yeoh, LaReine, Bollani, Monica, Virgilio, Michele, Schubert, Markus Andreas, Zaumseil, Peter, Capellini, Giovanni, Veldhorst, Menno, Scappucci, Giordano

Buried-channel semiconductor heterostructures are an archetype material platform for the fabrication of gated semiconductor quantum devices. Sharp confinement potential is obtained by positioning the channel near the surface; however, nearby surface states degrade the electrical properties of the starting material. Here, a 2D hole gas of high mobility (5 × 10 5 cm 2 V −1 s −1 ) is demonstrated in a very shallow strained germanium (Ge) channel, which is located only 22 nm below the surface. The top-gate of a dopant-less field effect transistor controls the channel carrier density confined in an undoped Ge/SiGe heterostructure with reduced background contamination, sharp interfaces, and high uniformity. The high mobility leads to mean free paths ≈ 6 µm, setting new benchmarks for holes in shallow field effect transistors. The high mobility, along with a percolation density of 1.2 × 10 11 cm −2 , light effective mass (0.09m e ), and high effective g-factor (up to 9.2) highlight the potential of undoped Ge/SiGe as a low-disorder material platform for hybrid quantum technologies. © 2019 The Authors. Published by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

2017, Bramburger, Stefan, Zinke, Benny, Killat, Dirk

This paper presents a modified interpolation algorithm for signals with variable data rate from asynchronous ADCs. The Adaptive weights Conjugate gradient Toeplitz matrix (ACT) algorithm is extended to operate with a continuous data stream. An additional preprocessing of data with constant and linear sections and a weighted overlap of step-by-step into spectral domain transformed signals improve the reconstruction of the asycnhronous ADC signal. The interpolation method can be used if asynchronous ADC data is fed into synchronous digital signal processing.

2019, Perez, E., Zambelli, C., Mahadevaiah, M.K., Olivo, P., Wenger, C.

Achieving a reliable multi-level operation in resistive random access memory (RRAM) arrays is currently a challenging task due to several threats like the post-algorithm instability occurring after the levels placement, the limited endurance, and the poor data retention capabilities at high temperature. In this paper, we introduced a multi-level variation of the state-of-the-art incremental step pulse with verify algorithm (M-ISPVA) to improve the stability of the low resistive state levels. This algorithm introduces for the first time the proper combination of current compliance control and program/verify paradigms. The validation of the algorithm for forming and set operations has been performed on 4-kbit RRAM arrays. In addition, we assessed the endurance and the high temperature multi-level retention capabilities after the algorithm application proving a 1 k switching cycles stability and a ten years retention target with temperatures below 100 °C.

2019, Ng, Herman Jalli, Hasan, Raqibul, Kissinger, Dietmar

A scalable four-channel multiple-input multiple-output (MIMO) radar that features a modular system architecture and a novel frequency-division multiplexing approach is presented in this article. It includes a single 30-GHz voltage-controlled oscillator (VCO) for the local oscillator signal generation, four cascaded 120-GHz transceivers with a frequency quadrupler, and on-board differential series-fed patch antennas. The utilized uniform antenna configuration results in 16 virtual array elements and enables an angular resolution of 6.2°. The vector modulators in the transmit (TX) paths allow the application of complex bit streams of second-order delta-sigma modulators easily generated on a field-programmable gate array (FPGA) to implement single-sideband (SSB) modulation on the TX signals resulting in orthogonal waveforms for the MIMO operation. Only one phase-locked loop and no digital-To-Analog converter is required. The waveform diversity also allows the simultaneous transmission of the TX signals to reduce the measurement time. The application of the SSB modulation on the frequency-modulated continuous-wave MIMO radar requires only half of the intermediate frequency bandwidth compared with the double-sideband modulation. The issue of the phase and amplitude mismatches at the virtual array elements due to the scalable radar architecture is addressed and a calibration solution is introduced in this article. Radar measurements using different numbers of virtual array elements were compared and the digital-beamforming method was applied to the results to create 2-D images. © 1963-2012 IEEE.

2019, Wang, Qiang, Niu, Gang, Roy, Sourav, Wang, Yankun, Zhang, Yijun, Wu, Heping, Zhai, Shijie, Bai, Wei, Shi, Peng, Song, Sannian, Song, Zhitang, Xie, Ya-Hong, Ye, Zuo-Guang, Wenger, Christian, Meng, Xiangjian, Ren, Wei

There was an error in the author list of this published article. The corresponding authors for this paper are Gang Niu (gangniu@xjtu.edu.cn) and Wei Ren (wren@mail.xjtu.edu.cn). The footnote indicating that Qiang Wang and Gang Niu contributed equally to the work was not intended. The corrected author list and notations are shown here. The Royal Society of Chemistry apologises for these errors and any consequent inconvenience to authors and readers. © The Royal Society of Chemistry 2019.

2018, Zaumseil, P., Hou, Y., Schubert, M.A., von den Driesch, N., Stange, D., Rainko, D., Virgilio, M., Buca, D., Capellini, G.

We report on the direct observation of lattice relaxation and Sn segregation of GeSn/Ge/Si heterostructures under annealing. We investigated strained and partially relaxed epi-layers with Sn content in the 5 at. %-12 at. % range. In relaxed samples, we observe a further strain relaxation followed by a sudden Sn segregation, resulting in the separation of a β-Sn phase. In pseudomorphic samples, a slower segregation process progressively leads to the accumulation of Sn at the surface only. The different behaviors are explained by the role of dislocations in the Sn diffusion process. The positive impact of annealing on optical emission is also discussed.

2017, Kreiser, Dan, Dyka, Zoya, Kornemann, Stephan, Wittke, Christian, Kabin, Ievgen, Stecklina, Oliver, Langendoerfer, Peter

The advent of industry 4.0 with its idea of individualized mass production will significantly increase the demand for more flexibility on the production floor. Wireless communication provides this type of flexibility but puts the automation system at risk as potential attackers now can eavesdrop or even manipulate the messages exchanged even without getting access to the premises of the victim. Cryptographic means can prevent such attacks if applied properly. One of their core components is the distribution of keys. The generation of keys from channel parameters seems to be a promising approach in comparison to classical approaches based on public key cryptography as it avoids computing intense operations for exchanging keys. In this paper we investigated key generation approaches using channel parameters recorded in a real industrial environment. Our key results are that the key generation may take unpredictable long and that the resulting keys are of low quality with respect to the test for randomness we applied.