Technische Universität Dresden
room CHE/0184/U
Dresden (D)

Abstract: 
Key techniques such as measurement and test-structure design for millimeter wave frequency characterization, optimization of device layout for high frequency performance, and statistical modeling technologies are reviewed. Those are essential to realize IoT applications in RF-CMOS devices. In addition, the concept of statistical modeling technology for yield management will be introduced. Author regard this technology as one of the innovation where future compact model technology is leading to. This helps to cooperate with the infrastructure of Smart-factory, which monitors and models device characteristic, and provide accurate yield information in a timely fashion as a Monte Carlo model form. To realize this methodology, it requires the efficient access to Bigdata coming from the production line and efficient way of updating typical and sigma value of the compact model. I will talk about how to establish the iteration loop by compact modeling technologies' viewpoint .

Abstract: 
Our novel approach considers how compact modeling can be used to optimize the classification accuracy of neuromorphic hardware, analogous to how traditional semiconductor devices are modeled and then used in traditional computer architectures. We focus on one of the simplest hardware implementations, a perceptron. We use a compact model for the floating gate transistor synapses and benchmark it against TCAD Sentaurus device simulations. The architecture consists of bipolar neurons and uses a classic delta learning rule. The test data uses winner take all neurons. We compare our results with an ideal linear model, which is known to provide optimal results because of the linearity of the learning rule. Using the MNIST database of hand written digits, we explore how changes in the accuracy of the test date are observed as device parameters are modified.

Abstract: 
A review, analysis, material-to-material variation and temperature influence for IV and III-V SB devices is discussed in order to offer physical insights into the device fundamentals. The physical influence of temperature down to the cryogenic regime is analyzed in a comprehensive study using Synopsys TCAD Sentaurus device simulator and benchmarked with experimental data at RT. We find reduced drive currents in SB-MOSFETS fabricated on small effective mass materials and that Schottky barrier lowering can significantly improve SB-MOSFETs, especially at low temperatures. The impact of the SBL effect influences the dominant conduction mechanism and the total amount of current. SBL can significantly improve SB-MOSFET performance and especially at low temperatures. The currents for the SB and conventional MOSFETs were found to be comparable. At room temperature higher currents for the SB-MOSFET were observed, due to the large voltage drop at the SB that leads to an increased mobility.