News on China's scientific and technological development.

Chinese Scientists confirm discovery of new type of niobium ore in the world’s largest rare earth deposit..

‘It could make China self-sufficient’: new strategic rare metal ore unearthed in Inner Mongolia​

• Scientists confirm discovery of new type of niobium ore in the world’s largest rare earth deposit
• Niobium is a key component in steel and superconductors, and could dramatically boost battery efficiency, experts say

China on Thursday received official confirmation that its geologists had discovered a new type of ore in the world’s largest , and that it contains niobium, a shiny, light grey metal that is prized in the steel industry for its strength, and also has superconducting properties.
The niobium ore – dubbed niobobaotite – has received an official approval number from the International Mineralogical Association’s classification committee, according to the state-owned China National Nuclear Corporation (CNNC).

The niobobaotite, which was discovered at the Bayan Obo ore deposit in the city of Baotou in Inner Mongolia, is a brown-black ore made up of niobium, barium, titanium, iron and chloride.

is a rare metal that is used mostly in the steel industry. Steel that is produced with less than 1 per cent niobium is significantly stronger with less weight, according to government organisation Geoscience Australia.

Alloys made with niobium are used to make construction materials, oil and gas pipelines, propeller blades and jet engines and more. Niobium is also known as a superconductor at low temperatures, making it useful in the production of superconducting magnets for particle accelerators and MRI machines, according to Geoscience Australia.

Geologists from a CNNC subsidiary discovered the metal ore on Tuesday, making it the 17th new mineral found in Baotou, and the 150th mineral found in the area overall.

Brazil dominates global production of niobium, accounting for nearly 89 per cent of output, followed by about 8 per cent in Canada. Brazil has more than 90 per cent of the world’s known , according to the United States Geological Survey.
Antonio H. Castro Neto, a professor of electrical and computer engineering at the National University of Singapore, told the Post that the “discovery is significant for China since most of the niobium China uses in the steel industry is imported”.

CBMM, which is partly owned by a group of Chinese companies including Citic, had several projects in China involved in the development of “niobium technologies for lithium batteries”, Ribas said.

According to Xinhua, CBMM is focused on using niobium in lithium-based batteries, and has partnered with battery manufacturers, universities and research and development centres in China to improve the technology.

Niobium is also being used in a bid to move away from lithium batteries entirely. Castro Neto, who is also the director of the National University of Singapore’s Centre for Advanced 2D Materials, said the niobium-graphene batteries they were producing “can last 30 years or more”.

The centre said niobium-based batteries could last 10 times longer than traditional lithium-ion batteries, and their improved conductivity meant the battery could perform better overall.

The batteries are also safer, because they do not contain flammable liquid electrolytes. They can also charge in just 10 minutes, compared to the few hours it takes to charge lithium batteries, according to the centre.

Recently, a research report by SDI, an internationally renowned analytical instrument think tank, shows that Guangzhou Hexin Instrument Co., Ltd. (hereinafter referred to as "Hexin"), located in the east of Yangcheng, is the only domestic mass spectrometry company that has received international attention. The market share of portable mass spectrometry is 6%. "This means that Chinese companies have a place in the field of international mass spectrometer research and development."

China targets 50% growth in computing power in race against U.S.

Tsinghua made major breakthrough in memristor-based chip-in-memory technologies. They created the world's first full-system-integrated chip that can do on-chip learning. This chip consists of multiple memristor arrays and the peripheral CMOS circuits.

Paper:

Editor’s summary​

Memristor-based computing technology has recently received considerable attention because of its potential to overcome the so-called “von Neumann bottleneck” of conventional computing architecture. In particular, memristor technology could realize time- and energy-efficient on-chip learning for various edge intelligence applications, although fully on-chip learning implementation remains challenging. To address this problem, Zhang et al. proposed a memristor-featured sign- and threshold-based learning (STELLAR) architecture. They fabricated a full-system-integrated chip consisting of multiple memristor arrays and all of the necessary complementary metal-oxide semiconductor peripheral circuits to support complete on-chip learning. The authors further demonstrated end-to-end on-chip improvement learning across various tasks, including motion control, image classification, and speech recognition, achieving software-comparable accuracy and low hardware cost. The present work is an important step in the field of computation-in-memory. —Yury Suleymanov

Abstract​

Learning is highly important for edge intelligence devices to adapt to different application scenes and owners. Current technologies for training neural networks require moving massive amounts of data between computing and memory units, which hinders the implementation of learning on edge devices. We developed a fully integrated memristor chip with the improvement learning ability and low energy cost. The schemes in the STELLAR architecture, including its learning algorithm, hardware realization, and parallel conductance tuning scheme, are general approaches that facilitate on-chip learning by using a memristor crossbar array, regardless of the type of memristor device. Tasks executed in this study included motion control, image classification, and speech recognition.


News release:

集成电路学院高滨课题组在支持片上学习的忆阻器存算一体芯片领域取得重大突破

当前，生成式人工智能已引爆新一轮智能革命的发展浪潮，大算力支撑下的人工智能技术极大改变着人类的生产生活方式。可随之而来的海量参数令算力需求持续攀升，如何解决庞大的算力缺口，实现能效比的大幅提升，正在变得日益迫切。高算力、高能效芯片作为算力的具体载体，已成为驱动本轮智能革命发展的核心底座，更是推动人类社会不断发展的动力源泉。

面向传统存算分离架构制约算力提升的重大挑战，清华大学集成电路学院吴华强教授、高滨副教授聚焦忆阻器存算一体技术研究，探索实现计算机系统新范式。忆阻器存算一体技术从底层器件、电路架构和计算理论全面颠覆了冯·诺依曼传统计算架构，可实现算力和能效的跨越式提升，同时，该技术还可利用底层器件的学习特性，支持实时片上学习，赋能基于本地学习的边缘训练新场景。当前国际上的相关研究主要集中在忆阻器阵列层面的学习功能演示，然而实现全系统集成的、支持高效片上学习的忆阻器芯片仍面临较大挑战，至今还未实现，主要在于传统的反向传播训练算法所要求的高精度权重更新方式与忆阻器实际特性的适配性较差。

为解决上述难题，课题组基于存算一体计算范式，创造性提出适配忆阻器存算一体实现高效片上学习的新型通用算法和架构（STELLAR），有效实现大规模模拟型忆阻器阵列与CMOS的单片三维集成，通过算法、架构、集成方式的全流程协同创新，研制出全球首颗全系统集成的、支持高效片上学习的忆阻器存算一体芯片。该芯片包含支持完整片上学习所必需的全部电路模块，成功完成图像分类、语音识别和控制任务等多种片上增量学习功能验证，展示出高适应性、高能效、高通用性、高准确率等特点，有效强化了智能设备在实际应用场景下的学习适应能力。相同任务下，该芯片实现片上学习的能耗仅为先进工艺下专用集成电路（ASIC）系统的3%，展现出卓越的能效优势，极具满足人工智能时代高算力需求的应用潜力，为突破冯·诺依曼传统计算架构下的能效瓶颈提供了一种创新发展路径。

Franklin said:

The biggest mistake that the West has made in regards to China is the mixing of science and technology with politics and ideology. The idea that you can only progress in science and technology with liberal democracy and freedom of speech is just false. There is no historical evidence for this and they didn’t say that about the Soviet Union. But they made up this rule just for China. And now after more than two decades of this the idea is stuck in their heads and it can’t get out.

Physics, math, biology and chemistry works the same way in China just as in America, Europe and elsewhere in the world. So China will figure things out sooner or later. And at this point its going to be sooner rather than later as China has a lot of money, people and tools to help them do so.

Click to expand...

The beauty is there is no strategic loss.

Essentially, US gifted China atleast 10 years worth of accelerated indigenization and only gained a 4 year delay in tech, which makes no discernable difference in strategic balance. SMIC and Huawei are still alive. They lost.

We can thank the non-semiconductor experts making policies.

Franklin said:

The biggest mistake that the West has made in regards to China is the mixing of science and technology with politics and ideology. The idea that you can only progress in science and technology with liberal democracy and freedom of speech is just false. There is no historical evidence for this and they didn’t say that about the Soviet Union. But they made up this rule just for China. And now after more than two decades of this the idea is stuck in their heads and it can’t get out.

Physics, math, biology and chemistry works the same way in China just as in America, Europe and elsewhere in the world. So China will figure things out sooner or later. And at this point its going to be sooner rather than later as China has a lot of money, people and tools to help them do so.

Click to expand...

And to make things even worse, Chinese companies have an strong incentive to survive, they will innovate their way out to survive if they are force to do so and in top of that they have strong support in country.

DJI coming out new version of LiDAR for drones.

全新禅思L2进一步采用一体化设计，集成了激光雷达、高精度惯导及4/3 CMOS可见光测绘相机，并配备三轴云台，搭载大疆M300 RTK及M350 RTK行业无人机，即可开启精准、高效、可靠的三维数据空中采集作业任务。

与上一代相比，新一代激光雷达量程提升30%，典型作业航高可达150米，进一步提升作业安全性及作业效率。

与此同时，光斑更小，能量更集中，每秒可发射24万个激光点，可识别更细小的目标物、获取更精细的模型；最大支持5次回波，可穿透更加茂密的植被，采集更多地面点云信息。

可见光测绘相机具备2000万有效像素，点云着色细节更丰富，支持使用机械快门，寿命多达20万次，作业成本进一步降低。无需采集点云时，可见光相机还可单独采集图像，用于可见光建图。

其融合GNSS、高精度自研惯导，在150米典型作业高度下，可实现高程精度4厘米，平面精度5厘米，可满足1:500高精度地形测量要求。

单架次作业面积高达2.5平方公里，配合大疆智图一键处理，可轻松复刻点云世界。

Click to expand...

The capabilities of this aerial Lidar along with light mapping camera & high precision navigation system is pretty good. It can emit 240k lasers per second.

t a typical operating altitude of 150 meters, it can achieve an elevation accuracy of 4 cm and a plane accuracy of 5 cm, meeting the 1:500 high-precision terrain measurement requirements.

Single sortie covers 2.5 km^2 area.

tphuang said:

Please, Log in or Register to view URLs content!

DJI coming out new version of LiDAR for drones.

The capabilities of this aerial Lidar along with light mapping camera & high precision navigation system is pretty good. It can emit 240k lasers per second.

t a typical operating altitude of 150 meters, it can achieve an elevation accuracy of 4 cm and a plane accuracy of 5 cm, meeting the 1:500 high-precision terrain measurement requirements.

Single sortie covers 2.5 km^2 area.

Click to expand...

Battle tested and proven /s

caudaceus said:

Battle tested and proven /s

Click to expand...

No wonder there is an export ban — actual national security concerns.

Prof. Pan of USTC and his team released the third version of their Jiuzhang optical quantuum computer. In this latest iteration, the computer now consists of 255 optical qubits, a new height in advancing optical quantuum technologies. Jiuzhang 3.0 also sets a new record in quantuum supermacy. It completes Gaussian boson sampling at a speed that is 1 million times faster than its predecessor, Jiuzhang 2.0.

This curve shows how the number of optical qubits of quantuum computers increases over time. The red dots are the systems from USTC.



Paper:

Gaussian Boson Sampling with Pseudo-Photon-Number-Resolving Detectors and Quantum Computational Advantage​

Abstract​

We report new Gaussian boson sampling experiments with pseudo-photon-number-resolving detection, which register up to 255 photon-click events. We consider partial photon distinguishability and develop a more complete model for the characterization of the noisy Gaussian boson sampling. In the quantum computational advantage regime, we use Bayesian tests and correlation function analysis to validate the samples against all current classical spoofing mockups. Estimating with the best classical algorithms to date, generating a single ideal sample from the same distribution on the supercomputer Frontier would take ∼600  yr using exact methods, whereas our quantum computer, Jiǔzhāng 3.0, takes only 1.27  μs to produce a sample. Generating the hardest sample from the experiment using an exact algorithm would take Frontier∼3.1×1010  yr.

News release:

中国科大实现“九章三号”光量子计算原型机

中国科学技术大学中国科学院量子信息与量子科技创新研究院潘建伟、陆朝阳、刘乃乐等组成的研究团队与中国科学院上海微系统所、国家并行计算机工程技术研究中心合作，成功构建了255个光子的量子计算原型机“九章三号”，再度刷新了光量子信息的技术水平和量子计算优越性的世界纪录。科研人员设计了时空解复用的光子探测新方法，构建了高保真度的准光子数可分辨探测器，提升了光子操纵水平和量子计算复杂度。根据公开正式发表的最优经典精确采样算法，“九章三号”处理高斯玻色取样的速度比上一代“九章二号”提升一百万倍。“九章三号”在百万分之一秒时间内所处理的最高复杂度的样本，需要当前最强的超级计算机“前沿”（Frontier）花费超过二百亿年的时间。这一成果进一步巩固了我国在光量子计算领域的国际领先地位。

量子计算是后摩尔时代的一种新的计算范式，它在原理上具有超快的并行计算能力，可望通过特定量子算法在一些具有重大社会和经济价值的问题方面相比经典计算机实现指数级别的加速。因而，研制量子计算机是当前世界科技前沿的最大挑战之一。

为此，国际学术界制定了三步走的发展路线。其中，第一步是实现“量子计算优越性”，即通过对近百个量子比特的高精度量子调控，对特定问题的求解展现超级计算机无法比拟的算力，这标志着40年前 Feynman等人的梦想成为现实。“量子计算优越性”实验还可用于检验计算科学的“扩展的丘奇—图灵论题”。同时，在此过程中，发展出可扩展的量子调控技术，为具备容错能力的通用量子计算机的研制提供技术基础。

2019年，美国谷歌和加州大学发布了53比特“悬铃木”超导量子计算处理器，宣称用200秒求解的随机线路采样问题需要超级计算机一万年时间求解。然而，这一宣称随后受到了中国科学家的挑战，改进后的经典算法使得超算上的计算时间从一万年缩短到数十秒，快于“悬铃木”量子处理器。

2020年，中国科大团队成功构建76光子的“九章”光量子计算原型机[Science 370, 1460 (2020)]，首次在国际上实现光学体系的“量子计算优越性”，并克服了谷歌实验中量子优越性依赖于样本数量的漏洞。2021年，中国科大团队进一步成功研制了113光子的可相位编程的“九章二号”[PRL 127, 180502 (2021)]和56比特的“祖冲之二号”量子计算原型机[PRL 127, 180501 (2021)]，使我国成为唯一在光学和超导两种技术路线都达到了“量子计算优越性”的国家。

在这个“量子计算优越性”战略高地，国际竞争呈现出白热化。位于加拿大多伦多的Xanadu公司与美国国家标准与技术研究院合作，采用与“九章”光量子计算原型机相同的高斯玻色取样路线，在2022年发布了216光子的“北极光”量子处理器，在国际上第二个实现了光学体系“量子计算优越性”。

中国科大团队在理论上首次发展了包含光子全同性的新理论模型，实现了更精确的理论与实验的吻合；同时，发展了完备的贝叶斯验证和关联函数验证，全面排除了所有已知的经典仿冒算法，为量子计算优越性提供了进一步数据支撑。在技术上，研制了基于光纤时间延迟环的超导纳米线探测器，把多光子态分束到不同空间模式并通过延时把空间转化为时间，实现了准光子数可分辨的探测系统。这一系列创新使得研究团队首次实现了对255个光子的操纵能力，极大地提升了光量子计算的复杂度，处理高斯玻色取样的速度比“九章二号”提升了一百万倍。在激烈的国际竞争角逐中，“九章三号”的实现进一步巩固了我国在光量子计算领域的国际领先地位。

进一步，在构建“九章”系列光量子计算原型机的基础上，中国科大研究团队揭示了高斯玻色取样和图论之间的数学联系，完成对稠密子图和Max-Haf两类具有实用价值的图论问题的求解，相比经典计算机精确模拟的速度快1.8亿倍[PRL 130, 190601 (2023)]。此外，又在国际上首次演示了无条件的多光子量子精密测量优势[PRL 130, 070801(2023)]。

量子计算优越性的研究是一个复杂而富有挑战性的工作，量子计算硬件与经典算法之间存在着长期竞争。研究人员期待这项工作一方面能够激发更多关于经典算法模拟的研究工作，另一方面有助于逐步解决量子计算研究中的各种科学和工程挑战。

上述项目受到了安徽省、上海市、科技部、中国科学院和基金委的支持。