万博manbetx网页版在线登录 _ 2022welcome

2018-12-24

“未来计算系列论坛”(一)成功举办

    2018年12月14日9:00~17:30, “未来计算系列论坛”(一)在北京大学微纳电子大厦103报告厅成功举办。本次论坛邀请了来自瑞士苏黎世大学的Giacomo Indiveri教授、来自国立台湾师范大学的Min-Hung Lee教授、来自韩国国立首尔大学的Cheol Seong Hwang教授,以及来自韩国科学技术院的Kyung Min Kim教授,并依次做了题为 “Neuromorphic electronic circuits for building autonomous cognitive systems”、题为“Ferroelectric Hf-based Oxide: Steep Slope Transistors and FeRAM Applications”、题为“Understanding the negative capacitance in nanoscale by two-dimensional phase field simulations”和题为“Stateful In-Memory Computing Technique in Emerging Crossbar Memories for Future Computing”的学术报告。

微纳电子学研究院院长黄如院士、黎明研究员、蔡一茂研究员、杨玉超研究员、叶乐副教授、崔小欣副教授、刘军华副教授、许晓燕副教授,以及相关研究方向的学生参加了报告会,并与报告人进行了深入、热烈的讨论。 

                                                                

Lecture #1

Neuromorphic electronic circuits for building autonomous cognitive systems

Prof. Giacomo Indiveri

the University of Zurich, Switzerland

Neural networks and deep learning algorithms are currently achieving impressive state-of-the-art results on computing tasks that operate on stored data sets. However, artificial computing systems are still vastly outperformed by biological neural processing ones for tasks that involve processing of sensory data acquired in real-time in complex and uncertain settings, and closed-loop interactions with the environment. This difference is remarkable especially when size and energy consumption are factored in.

One of the reasons for this gap is that, as opposed to conventional computing architectures, in biological neural systems computation is tightly linked to the to the physics of their computing elements and to their temporal dynamics. In this talk, I will present hybrid analog/digital microelectronic circuits that use their physics to directly emulate the biophysics of the neural processes and memory elements they model. I will demonstrate examples of brain-inspired architectures that integrate massively parallel arrays of such circuits to implement on-chip on-line spike-based learning and computation, and will describe the advantages and disadvantages of these types of computing architectures compared to conventional computing systems. I will argue that the circuits proposed represent a promising approach for building intelligent and energy-efficient autonomous cognitive agents that need to process input data as it arrives, in real-time, without having to use eternal memory storage.

 

Lecture #2

Ferroelectric Hf-based Oxide: Steep Slope Transistors and FeRAM Applications

Prof. Min-Hung Lee

National Taiwan Normal University, Taiwan

The prospect of ferroelectric Hf-based oxide by ALD (Atomic Layer Deposition), discovered about one decade, has been wide and intensive studied due to lots of applications. For the IoT (Internet of Things) era, the requirement of scaling down supply voltage VDD and power consumption for low power devices is the pursued goals for CMOS and memory applications. The ferroelectric gate stack is integrated into FETs with negative capacitance effect for subthreshold swing (SS) improvement and FeRAM for memory applications. The feasible concept of coupling the polarization Hf-based oxide is practicable to following current CMOS architectures.

 

Lecture #3

Understanding the negative capacitance in nanoscale by two-dimensional phase field simulations

Prof. Cheol Seong Hwang

Seoul National University, Korea

The negative capacitance (NC) effects in ferroelectric materials have emerged as the possible solution to low-power transistor devices and high-charge-density capacitors. This is fundamentally based on the total energy argument of dielectric/ferroelectric (DE/FE) stacked film using the Landau-Ginzburg-Devonshire (LGD) theory. Although the subthreshold swing

The most critical problem for implementing the NC effect in the actual device is the involvement of the poly-domain structure, which fundamentally hampers the desired functionality of the FE materials. When the poly-domain structure presents in an FE film, the FE switching is usually mediated by the growth of domains having the aligned polarization with the external field through the domain wall motion. In this case, the almost entire volume of the FE film maintains a fully polarized state, which can hardly involve the desired NC effect. Nonetheless, as the film thickness becomes closer to the critical thickness where the fully poled state becomes unstable due to the strong depolarization effect, there could be different physical circumstances where the NC effect can stably emerge. However, a precise understanding of such a complicated physical situation is challenging because the system size is too large to be examined by the first principles study. Macroscopic simulation using a one-dimensional model could not capture the critical assets of this intriguing phenomenon mainly because it can hardly describe the lateral interaction between the neighboring domains. In nano-scale, the electrostatic interaction between the neighboring domains plays a critical role in determining the total energy and stability.

In this work, therefore, a two-dimensional phase field simulation (PFS) technique was adopted, where the evolution of the domain configuration as a function of time was captured by using the time-dependent Ginzburg Landau (TDGL) formalism. This TDGL-PFS framework provided a highly feasible simulation platform to precisely understand the evolution of the domain patterns with time and system size. It was intriguingly found that there could be a feasible nano-scale material and device configurations which can show the stable NC effect even with the poly-domain structure. The presentation will discuss the critical features of such an intriguing phenomenon.

 

Lecture #4

Stateful In-Memory Computing Technique in Emerging Crossbar Memories for Future Computing

Prof. Kyung Min Kim

Department of Materials Science and Engineering, KAIST, Korea

Emerging memories such as MRAM, PRAM, and RRAM have been extensively studied due to its various advantages over the conventional memories. Because their performances are yet better than the conventional memories as DRAM and NAND Flash, researchers are primarily trying to find their applications at embedded memory or storages class memory applications. As such, when the emerging memories are used for memory or data storage, its application can be very limited to one of the computing elements in the conventional computing hierarchy. If an entirely new function—a computing function—can be implemented in the emerging memories, it could destroy the traditional computing hierarchy and reconstruct the computing paradigm. The stateful in-memory computing technology provides such capability to the emerging memories. The first concept of stateful logic was proposed in 2010 by a group of HP using the crossbar RRAM. Afterward, there have been many advancements for putting the technology into practical use. In this talk, the details of the stateful logic technology and its technical challenges and solutions are presented. The stateful in-memory computing technology can apply to any emerging memories based on the crossbar architecture. Therefore, it would be an additional beneficial option for the emerging memories strengthening its functionality more than memory or storage.


返回