News on China's scientific and technological development.

SanWenYu

Captain
Registered Member
This research can have applications in controlling tubulence created by rotating machines like gas turbines, wind turbines, internal combustion engines, compressors,.

Paper:
Please, Log in or Register to view URLs content!

Deep reinforcement transfer learning of active control for bluff body flows at high Reynolds number​

Abstract​

We demonstrate how to accelerate the computationally taxing process of deep reinforcement learning (DRL) in numerical simulations for active control of bluff body flows at high Reynolds number () using transfer learning. We consider the canonical flow past a circular cylinder whose wake is controlled by two small rotating cylinders. We first pre-train the DRL agent using data from inexpensive simulations at low , and subsequently we train the agent with small data from the simulation at high (up to =1.4×105). We apply transfer learning (TL) to three different tasks, the results of which show that TL can greatly reduce the training episodes, while the control method selected by TL is more stable compared with training DRL from scratch. We analyse for the first time the wake flow at =1.4×105 in detail and discover that the hydrodynamic forces on the two rotating control cylinders are not symmetric.

News release:
Please, Log in or Register to view URLs content!

能动学院姜孝谟教授团队在湍流智能控制研究取得突破性进展​

近日,能源与动力学院姜孝谟教授团队在深度强化迁移学习对复杂流体控制方面的研究取得突破性进展,成果基于深度强化迁移学习的高雷诺数条件下钝体流动的主动控制(Deep reinforcement transfer learning of active control for bluff body flows at high Reynolds number)在流体力学国际顶刊《流体力学杂志》(Journal of Fluid Mechanics)发表。文章第一作者为王志成副教授,除了姜孝谟教授,合作者还包括西湖大学范迪夏教授,美国麻省理工学院(MIT)Michael Triantafyllou教授和布朗大学(Brown)的美国工程院院士George Karniadakis教授。

湍流是自然界和工业中最常见的流体形态,对其进行精准控制不仅可以深化对自然界湍流的科学认知,还可以在航空、航天、能源、石油等重要工业民用中避免湍流危害而产生巨大的经济效益。随着人工智能技术的发展,流体智能控制成为当前流体力学研究的热点和前沿方向。

团队采用自主研发的高置信度高精度谱元法湍流求解器,对雷诺(Reynolds)数Re=1.4x105条件下的圆柱绕流科学难题进行了深入研究,发展深度强化学习方法来控制圆柱后侧小圆柱的旋转速度和方向,控制主圆柱表面的边界层及后侧的尾迹。结果表明,无需借助任何人类的知识,针对不同的优化目标,深度强化学习可以快速学会准确控制湍流边界层。本研究还进一步提出了一种采用多个数值模拟同步产生低雷诺数训练数据,并迁移学习到高雷诺数流动控制的创新方法,这样更快速更准确地实现流动控制。

姜孝谟教授数字能源装备团队成立于2020年,由多位海归博士组成。团队依托分别于2020年和2022年创建的辽宁省工业装备数字孪生重点实验室和大工碳中和研究院等平台,面向绿色能源智能化及碳中和国家战略需求,针对大型能源动力装备及关键部件数字孪生和智慧运维中的科学问题和关键技术,与美国MIT、布朗大学等多所名校的国际著名学者合作,长期致力于数物虚融合、物理可解释AI、AI4Science、轻量化建模等理论方法及应用研究。本论文研究成果有望在面向燃气轮机、发动机、压缩机、风电等旋转机械基于数字孪生的湍流控制方面,解决高雷诺数下的小数据和高效高精度计算难题,并在湍流危害控制方面发挥重要的作用。
 

Wuhun

New Member
Registered Member
Multi-modal sensor fusion (microphone array+lidar+camera) based small drone 3D trajectory exposure system with farthest detection range and highest 3D accuracy in published literature. Semi-spherical detection range >500m with less than 1.5% error. Microphone array could achieve a maximum drone detection distance of 1300 m.

Please, Log in or Register to view URLs content!
 

sunnymaxi

Major
Registered Member
Shanghai to Nurture Synthetic Biology Startups, Become Global Biomanufacturing Hub by 2030..

Please, Log in or Register to view URLs content!

The Shanghai government will support the development of synthetic biology start-ups, which are firms that apply engineering principles to develop new biological parts, with the aim of becoming at high-end biomanufacturing industry cluster of global influence by 2030.

The mega city plans to build industrial bases, promote international tie-ups and support businesses in the field of synthetic biology, according to a document recently released by the Shanghai municipal government. By 2025, there will be more than 10 top firms with worldwide influence in the city and around three to five companies will have gone public, it said.

Shanghai will support the formation of leading enterprises, advanced research and development institutions and major industrial projects in the field of synthetic biology at home and abroad, it said. The city will encourage key firms to carry out mergers and acquisitions in the sector and cultivate high-tech companies in sub-fields, it added.

To achieve these goals, the municipal government will strengthen funding guarantees, it said. It will ensure continued investment in special funds, improve diversified fund guarantee mechanisms and set up market-oriented industrial guidance funds with the local government as the guarantee institution. More innovative regulatory policies should also be explored, it added.
 

SanWenYu

Captain
Registered Member
Nanjing University created moiré synaptic transistor from 2D materials and demonstrated a full-moiré physical neural network with classification accuracy of 90.8% for the MNIST handwritten digits database.

Paper:
Please, Log in or Register to view URLs content!


Moiré Synaptic Transistor for Homogeneous-Architecture Reservoir Computing​

Abstract​

Reservoir computing has been considered as a promising intelligent computing paradigm for effectively processing complex temporal information. Exploiting tunable and reproducible dynamics in the single electronic device have been desired to implement the "reservoir" and the "readout" layer of reservoir computing system. Two-dimensional moiré materials, with an artificial lattice constant many times larger than the atomic length scale, are one type of most studied artificial quantum materials in community of material science and condensed-matter physics over the past years. These materials are featured with gate-tunable periodic potential and electronic correlation, thus varying the electric field allows the electrons in the moiré potential per unit cell to exhibit distinct and reproducible dynamics, showing great promise in robust reservoir computing. Here, we report that a moiré synaptic transistor can be used to implement the reservoir computing system with a homogeneous reservoir-readout architecture. The synaptic transistor is fabricated based on an h-BN/bilayer graphene/h-BN moiré heterostructure, exhibiting ferroelectricity-like hysteretic gate voltage dependence of resistance. Varying the magnitude of the gate voltage enables the moiré transistor to switch between long-term memory and short-term memory with nonlinear dynamics. By employing the short- and long-term memories as the reservoir nodes and weights of the readout layer, respectively, we construct a full-moiré physical neural network and demonstrate that the classification accuracy of 90.8% can be achieved for the MNIST (Modified National Institute of Standards and Technology) handwritten digits database. Our work would pave the way towards the development of neuromorphic computing based on moiré materials.

News release:
Please, Log in or Register to view URLs content!


物理学院梁世军、缪峰团队首次实现莫尔突触晶体管​

近日,南京大学物理学院梁世军、缪峰教授团队联合南京理工大学程斌教授团队,以“原子乐高”的方式搭建了六方氮化硼与双层石墨烯对齐的莫尔超晶格异质结,首次实现了可模拟生物突触短时可塑性与长时可塑性的莫尔突触晶体管(Moiré synaptic transistor)。进一步,合作团队基于莫尔突触晶体管的动力学高度可调谐的特性,提出了能够执行同质架构储备池计算的全莫尔物理神经网络(Full-moiré physical neural network, MPNN )。该工作为莫尔电子学的未来发展提供了重要思路。

莫尔材料,是通过堆叠二维原子晶体所形成的一类具有新奇强关联和拓扑物性的新型低维量子材料体系。与传统量子材料不同的是,莫尔材料拥有丰富的量子态(例如,关联绝缘体、轨道磁性、界面铁电性等),可被电场、光场、应力场等外场进行调控。这使得莫尔材料不仅成为了物性探索的新型理想平台,而且在莫尔电子器件应用方面展现出巨大的潜力。目前的研究主要集中在莫尔材料中新型量子态的探索与调控方面,如何利用莫尔材料的独特量子态与调控规律设计莫尔电子器件是一个广泛关注的议题。

面对上述机遇与挑战,合作团队首先利用二维材料异质结转移技术人工搭建了基于六方氮化硼封装的双层石墨烯莫尔突触器件,其中由六方氮化硼与石墨烯晶格对齐产生的电场可调的莫尔势场,与生物神经系统中的信号刺激引发的突触可塑性具有一定的相似性(图1a-d)。合作团队发现,由于可调外加电场与莫尔势场的叠加,双层石墨烯的转移特性曲线展现出栅压可控的记忆窗口(图1e-f),并且当栅压超过某个阈值电压后,记忆窗口开始出现(图1g),这标志着器件从易失向非易失存储行为的转变。器件的这一可调记忆特性能够被用来模拟生物突触功能。
 

SanWenYu

Captain
Registered Member
Another team from Nanjing University pushed the operating frequency of 2D semiconductor ICs to 2.65GHz, 200 times of the previous record.

Paper:
Please, Log in or Register to view URLs content!

Two-dimensional semiconductor integrated circuits operating at gigahertz frequencies​

Abstract​

Two-dimensional transition metal dichalcogenides could potentially be used to create transistors that are scaled beyond the capabilities of silicon devices. However, despite progress on the single-transistor level, the development of high-frequency integrated circuits remains a challenge and the operating frequency of integrated circuits based on transition metal dichalcogenides has so far been limited to the megahertz regime; this is well below the silicon complementary metal–oxide–semiconductor technology, as well as emerging technologies such as carbon nanotubes. Here we report two-dimensional semiconductor integrated circuits—five-stage ring oscillators—that operate in the gigahertz regime (up to 2.65 GHz) and are developed using a design-technology co-optimization process. The circuits are based on monolayer molybdenum disulfide field-effect transistors that have an air-gap structure, which leads to doping-free ohmic contacts and low parasitic capacitance. Technology computer-aided design simulations also suggest that our air-gap structure can potentially be scaled to the 1 nm technology node and could reach the targets set out in the IEEE International Roadmap for Devices and Systems for 2031.


News story:
Please, Log in or Register to view URLs content!


南大团队将二维半导体集成电路推向千兆赫兹​

南京大学电子科学与工程学院王欣然教授、施毅教授带领的团队在二维半导体集成电路领域取得突破性进展。通过设计-工艺协同优化(DTCO),开发出空气隔墙晶体管结构,大幅降低寄生电容,在国际上首次实现了GHz频率的二维半导体环形振荡器电路,比原有记录提升200倍,并预测了二维半导体应用于1nm节点集成电路的潜力与技术路径。

由于短沟道效应,硅基互补金属氧化物半导体(CMOS)器件的微缩化越来越具有挑战性。以MoS2为代表的二维过渡金属二硫属化物 (TMD) 具有原子级超薄厚度、高载流子迁移率和免疫短沟道效应等优点,是亚1nm节点集成电路的重要候选材料。相对于硅沟道材料,单层TMD可以维持晶体管尺寸进一步缩小,满足国际器件和系统路线图 (IRDS)设定的目标。过去10余年,尽管TMD材料生长和场效应晶体管器件取得了系列重大进展,但是高频集成电路的开发仍然是一个挑战,基于TMD的集成电路工作频率迄今为止仅限于MHz,远远低于硅基CMOS以及碳纳米管等技术,成为限制二维材料走向集成电路应用的关键瓶颈之一。

面对上述挑战,王欣然、施毅教授领导的国际合作团队将DTCO应用于二维器件领域,进行多项突破创新。团队在MoS2场效应晶体管中创新性引入空气隔墙(Air-gap)结构,不仅避免了对接触部分进行掺杂的额外工艺步骤,更重要的是大幅度降低器件的寄生电容。根据TCAD模型计算,引入空气隔墙的器件结构与没有隔墙的结构相比,寄生电容降低了34%。同时结合团队之前报道的半金属Sb(011?2)接触技术,在降低寄生电容的基础上保持了高性能:本次报道的空气隔墙晶体管具有同等尺寸器件中的最高电流密度。为了获得低延迟高频率的电路,团队对器件结构进行了TCAD建模仿真,获得了接触重叠长度、掺杂水平等重要参数的最优设计区间。基于器件工艺和TCAD模型的DTCO,团队成功在大面积单层MoS2上实现了GHz频率的五级环形振荡电路阵列,平均工作频率达2.1GHz,最高工作频率达2.65GHz,对应单级反向器延迟降低至37ps。

该工作不仅首次实现了GHz二维半导体集成电路,而且展示了DTCO在减少非理想寄生效应、在众多权衡中找到性能/功耗/面积最优解的关键作用,为高性能二维集成电路发展指明了方向。相关工作以“Two-dimensional semiconductor integrated circuits operating at gigahertz frequencies”为题发表在《自然?电子学》期刊。
 

tacoburger

Junior Member
Registered Member
Nanjing University created moiré synaptic transistor from 2D materials and demonstrated a full-moiré physical neural network with classification accuracy of 90.8% for the MNIST handwritten digits database.

Paper:
Please, Log in or Register to view URLs content!

Another team from Nanjing University pushed the operating frequency of 2D semiconductor ICs to 2.65GHz, 200 times of the previous record.

Paper:
Please, Log in or Register to view URLs content!
This is big deal. This 2D TMD semiconductors are the true next generation semiconductor that will be likely to replace silicon. TSMC has said that they are looking to start making 2D TMD semiconductors once they have maxed out existing CFET architecture on silicon around the 2030s.
 

SanWenYu

Captain
Registered Member
This is big deal. This 2D TMD semiconductors are the true next generation semiconductor that will be likely to replace silicon. TSMC has said that they are looking to start making 2D TMD semiconductors once they have maxed out existing CFET architecture on silicon around the 2030s.
Yep China has not been sitting idle to wait for the breakthrough in lithgraphy machines. In addition to these two recent news on 2D semiconductors above, we have seen these progress on the next gen semiconductor materials by CAS and various universities in China since last year. And these are just the ones I came across and by no means complete.

Approaching the quantum limit in two-dimensional semiconductor contacts by NJU

One-dimensional semimetal contacts to two-dimensional semiconductors by THU

Ballistic two-dimensional InSe transistors by PKU

2D fin field-effect transistors integrated with epitaxial high-k gate oxide by PKU

Low power flexible monolayer MoS2 integrated circuits by CAS

Modularized Batch Production of 12-inch Transition Metal Dichalcogenides by Local Element Supply by CAS, PKU and Songshan Lake Materials Lab

Vertical organic electrochemical transistors for complementary circuits by UESTC
 
Top