CRS_CityU101/24
1D van der Waals Atomic Chains for 3D-integrated Retinomorphic Optoelectronics

Hong Kong Project Coordinator: Prof Johnny Chung-yin Ho (City University of Hong Kong)
Mainland Project Coordinator: Prof Guozhen Shen (Beijing University of Technology)

One-dimensional (1D) van der Waals atomic chains have advantages such as high mobility, strong light coupling, features of atomic-level smooth surface, and low recombination rate. This project aims to utilize the advantages of 1D van der Waals atomic chains, combined with the advancements in neuromorphic devices and image sensing, to conduct research on new 1D van der Waals material systems and address key issues such as array construction, device integration, and fusion of sensing and computation. By synergistically controlling the transport characteristics of charge carriers in 1D structures through multiple-effect coupling, we will explore the three-dimensional (3D) optoelectronic integration modes of in/near computing sensors and develop intelligent sensor technology to achieve multi-dimensional light information detection, high-density single-chip integration, and low-power brain-like computation.

CRS_CityU103/24
Unbalanced Quantum Interferometers with Optical Path Difference Beyond Coherence Length and Their Applications in Sensing and Quantum Information

Hong Kong Project Coordinator: Prof Jeff Zhe-yu Ou (City University of Hong Kong)
Mainland Project Coordinator: Prof Xiaoying Li (Tianjin University)

The first quantum revolution enabled inventions such as solid-state transistors and lasers that have transformed nearly all aspects of our lives. The second quantum revolution, which promises disruptive technology that may change our daily life as well, involves designing and manipulating complex quantum systems that utilize quantum physics to bring quantum advantages in information processing, communication and sensing. The basic principle of quantum technology is quantum interference, which is the phenomenon produced when two beams of light are brought together for wave superposition and shown in the form of beautiful fringes. Quantum interferometers are the fundamental building blocks in many protocols of quantum technology.

Traditional optical interferometers are based on intensity measurement with one detector and require the balance of the paths of the two interfering beams to within the coherence length of the beams in order to have overlap in both space and time for wave superposition. This severely limits the application of traditional interferometry. Recent research revealed a new technique of intensity correlation between two detectors that can observe interference effect of some specific fields without the requirement of path-balancing, potentially extending the range of application of interferometry. Our recent work went one step further to cover more general types of fields including both continuous and pulsed waves. We also recently discovered that another measurement technique, namely, homodyne detection, which measures the amplitude of the light wave, can reveal interference even between non-overlapping beams. This is a new paradigm in interferometry and will extend further the scope for applications in quantum technology in general.

In this research program, we will apply the aforementioned two non-traditional detection methods, namely, intensity correlation and homodyne detection, to the observation of interference with the goal of revealing interference which is otherwise not present by traditional detection method due to path imbalance. We will apply the general idea of path-unbalanced interferometry to some specific systems and develop experimental platform to study this new approach in interferometry. More specifically, we will investigate theoretically and experimentally, using the two non-traditional measurement methods, various interference schemes with unbalanced paths in different path configurations. We will use different types of light sources such as classical thermal coherent laser sources as well as quantum sources of entangled photons, in both pulsed and continuous wave operation. As practical application, we apply the technique to remote sensing such as fibre dispersion measurement and synthetic aperture image systems in astronomical observation and demonstrate quantum advantages in enhanced sensitivity.

CRS_CityU104/24
Theoretically Guided Material Design, Syntheses and Device Engineering for Efficient and Stable Perovskite/Organic Tandem Solar Cells

Hong Kong Project Coordinator: Prof Xiaocheng Zeng (City University of Hong Kong)
Mainland Project Coordinator: Prof Shangfeng Yang (University of Science and Technology of China)

To address the urgent global energy demands and the imperative of achieving carbon neutrality, there is a critical need for practical solutions. The advent of organic-inorganic hybrid perovskite materials, and their subsequent integration into solar cells, holds the potential to revolutionize photovoltaic technology. These advancements could enable the production of scalable and efficient solar cells for sustainable clean energy generation. While single-junction organic solar cells and perovskite solar cells have realized impressive efficiencies of over 19% and 26% respectively, their theoretical maximum efficiency is inherently limited to around 30%. To overcome this limitation, the fabrication of multi-junction devices with multiple light absorbers of significantly varied bandgaps is a promising alternative, potentially boosting solar cell efficiency to over 40%. Perovskite/organic tandem solar cells (POTSCs) represent an exciting class of such multi-junction solar cells, offering high efficiency, ease of fabrication via solution processing, and economical raw materials. Nevertheless, the efficiency of POTSCs is currently limited to around 24%~25% due to challenges related to theoretical study, material innovation, and device optimization, highlighting the need for further development.

In this project, our goal is to fabricate high-performance POTSCs with efficiencies exceeding 28% through a theory-guided approach to materials design and device fabrication. To fully harness the potential of POTSCs, we propose an integrated strategy that combines theoretical study with comprehensive synthesis, characterization, and optimization of materials and device preparation. Specifically, we will establish theory simulation strategies to screen perovskites, interface materials, and non-fullerene acceptors. This will guide the synthesis of materials. Subsequently, these materials will be synthesized, characterized, and selected for their superior properties in device fabrication. Additionally, we will develop a comprehensive optoelectronic model to theoretically investigate the opto-electronic properties of the tandem solar cells, providing guidelines for device fabrication. Finally, the application of these developed materials in the production of single-junction and POTSCs, guided by theoretical design, will enable the realization of highly efficient and stable POTSCs.

The successful execution of this project will significantly contribute to the development of novel research methodologies through theory-guided material design and device fabrication, advancing high-efficiency, low-cost, and scalable tandem solar cell technology. This marks a significant step towards commercialization and addresses the pressing need for practical solutions in the global pursuit of sustainable energy generation.

CRS_PolyU501/24
A Library of Polarized Van Der Waals Heterobilayers: From Prediction to Realization

Hong Kong Project Coordinator: Prof Shu-ping Lau (The Hong Kong Polytechnic University)
Mainland Project Coordinator: Prof Wei Ji (Renmin University of China)

Two-dimensional van der Waals (vdW) heterobilayers are made by stacking two different single-layer materials on top of each other. These layers are held together by weak forces called van der Waals interactions. These exciting structures have unique physical and chemical properties that can lead to new scientific discoveries and advanced technologies.

For instance, when two graphene layers are twisted at a specific angle, they can exhibit unusual superconductivity due to strong electron interactions. Similarly, certain combinations of materials can show interesting behaviors like ferroelectricity, a valuable property in electronics.

Hong Kong and Mainland China researchers recently discovered unexpected electrical properties in a specific combination of materials (MoS2 and WS2) that were not twisted using chemical vapor deposition (CVD). This finding has sparked interest in creating more heterobilayers to explore their potential in future technologies like nanoelectronics and photonics.

However, studying these materials is challenging because there are many possible combinations. To tackle this, researchers plan to use computer simulations to predict which combinations might have beneficial properties before making them in the lab. This approach will help save time and resources and could be a valuable tool for scientists working with 2D materials.

CRS_HKUST601/24
Breaking the LLM Resource Wall: A Novel Computing System based on a Heterogeneous, Disaggregated Cloud Infrastructure

Hong Kong Project Coordinator: Prof Bo Li (The Hong Kong University of Science and Technology)
Mainland Project Coordinator: Prof Hai Jin (Huazhong University of Science and Technology)

Large language models (LLMs) are making significant strides in generative AI, enabling a variety of novel applications across numerous domains. However, these models require massive amounts of computing power, data storage, and energy to train and operate, creating a “resource wall” for even a large organization. Recent restriction of advanced AI chips to Hong Kong and Mainland further exacerbates this issue. In this project, we aim to break the LLM resource wall with a novel system solution based on a heterogeneous, disaggregated cloud infrastructure. Leveraging the emerging hardware interconnect technologies, such as Compute Express Link (CXL), Huawei’s Unified Bus (UB), and Remote Direct Memory Access (RDMA), we propose to rearchitect the current cloud by breaking its monolithic servers and organizing heterogeneous hardware devices like CPU, GPU, and memory as independent, network-attached components in disaggregated resource pools. This architecture enables running LLMs on heterogeneous compute and storage devices that are previously not supported. It can significantly improve the resource utilization, elasticity, heterogeneity, and failure isolation. To fully exploit the benefits of this disaggregated architecture, we will develop a comprehensive system solution following a holistic approach. First, we will focus on efficient orchestration of heterogeneous computing devices through judicious synchronization, offloading, scheduling, load balancing, and performance and failure isolation. Second, we will develop a unified memory manager that efficiently manages disaggregated memory across heterogeneous storage devices, while minimizing the latency overhead caused by disaggregation. Third, building on the previous tasks, we will develop a disaggregated runtime system that automatically parallelizes LLM computations to achieve maximum acceleration. Finally, we will design a new cluster management framework to optimally orchestrate concurrent executions of multiple LLM tasks in a shared, disaggregated cloud.

CRS_HKUST602/24
Enabling Security Protection for Heterogeneous Blockchain Systems in Web 3.0

Hong Kong Project Coordinator: Prof Song Guo (The Hong Kong University of Science and Technology)
Mainland Project Coordinator: Prof Weizhe Zhang (Harbin Institute of Technology (Shenzhen))

The Web3.0, with the blockchain as its infrastructure, has brought about an emerging revolution in the field of the Internet and has become a strategic place influencing the economic development of the country. The research on the security protection for heterogeneous blockchain systems in Web3.0 is facing new challenges such as the lack of secure and interoperable multi-chain consensus platform, the lack of universal and effective smart contract vulnerability detection and repair methods, the difficulty of real-time protection against hidden and complex malicious transactions, and the difficulty of protecting the security of digital assets, etc. This project focuses on four major scientific issues of security protection for heterogeneous blockchain systems and proposes a comprehensive and hierarchical security architecture. We aim to: 1) develop a trustless and scalable multi-chain platform to achieve reliable and scalable interoperability across heterogeneous blockchains; 2) develop a fine-grained vulnerability detection and automatic repair scheme to ensure the security and reliability of heterogeneous cross-chain smart contracts; 3) develop a real-time intelligent monitoring and proactive defense system to protect against complex malicious attacks in Web 3.0; 4) develop a sustainable digital asset management and market governance mechanism to implement secure and fair digital asset trading. Eventually, we will build a hierarchical security protection architecture for heterogeneous blockchain systems with endogenous security, real-time reliability, intelligent protection, credible interconnection, which will significantly enhance the competitiveness of our country in the Web3.0 revolution.

CRS_HKUST605/24
Observation and Modeling of Organic Nitrogen Abundance and Molecular Composition in Aerosol and Rainwater in China

Hong Kong Project Coordinator: Prof Jianzhen Yu (The Hong Kong University of Science and Technology)
Mainland Project Coordinator: Prof Yuepeng Pan (Institute of Atmospheric Physics, Chinese Academy of Sciences)

For over a century, human activities, such as burning fossil fuels and producing food, have released substantial quantities of reactive nitrogen (Nr) into the air. This has profoundly changed regional and global nitrogen cycling patterns, significantly impacting air quality, climate, and human health. Atmospheric Nr comes back down to the earth through rain events and deposition of gases and particles, thus affecting ecosystem evolution and biodiversity. Atmospheric Nr exists in two main chemical forms: inorganic nitrogen (IN) and organic nitrogen (ON). Traditionally, discussions on the sources, sinks, and ecological impacts of atmospheric Nr have predominantly centered on the IN species, encompassing nitrogen oxides (NOx), ammonia (NH3), nitrate (NO3- ), and ammonium (NH4+). While there is extensive observation and modeling of inorganic nitrogen (IN), there has been a significant lack of attention given to ON, which has been largely overlooked in studies concerning nitrogen cycling.

Unlike IN, accurately measuring ON has always been a challenge, hindering observational and modeling studies of ON. Recently, our research team has achieved a breakthrough in ON measurement method, enabling the simultaneous measurement of aerosol IN and ON. Building on this breakthrough and access to a nationwide field observation network, this research will collect aerosols and precipitation, quantify the ON concentration levels, and characterize ON molecular composition and sources. Additionally, we will conduct experiments in controlled chamber environments to elucidate the role of NH3 in reduced ON formation. Moreover, we plan to develop an atmospheric chemistry model tailored for regional ON emissions and deposition in China, based on observational data and chemical mechanism research outcomes. This model will simulate the content, spatiotemporal distribution, phase distribution, dry/wet deposition processes of atmospheric ON and IN in China, and the response to reductions in NH3 emissions. By doing so, we aim to uncover the spatiotemporal evolution of ON on a national scale, enhance the scientific understanding of how nitrogen moves regionally. This will contribute to shaping policies for reducing air pollution in China and assist in improving the quality of the environment and progressing towards a more beautiful China.

CRS_HKU701/24
Study of Thermodynamic and Dynamical Properties of Novel One-dimensional Quantum Systems in Cold Atoms

Hong Kong Project Coordinator: Prof Shizhong Zhang (The University of Hong Kong)
Mainland Project Coordinator: Prof Yunbo Zhang (Zhejiang Sci-Tech University)

In this proposal, we explore a set of interesting physics questions related to the internal degrees of freedom in one-dimensional quantum systems. These one-dimensional quantum systems can occur in solid state systems as well as the artificial quantum systems composed of ultracold atomic gases. This last system is unique in that much of its parameters can be tuned broadly, giving rises to an expanded parameter space. Specially, we will consider two types of internal degrees of freedom, one is internal spin that can be expanded to SU(N) and the other is the transverse orbital states that can lead to axial interaction that is both even and odd parity. Using exact solution, field theory and numerical simulation, we will try to investigate the correlation and transport properties of these systems.

CRS_HKU702/24
Space-ground Integrated Intelligent Computing and Networking for 6G Large-scale Constellations

Hong Kong Project Coordinator: Prof Kaibin Huang (The University of Hong Kong)
Mainland Project Coordinator: Prof Min Sheng (Xidian University)

The International Telecommunication Union (ITU) identifies space-ground integrated networks (SGI-Nets) and integrated AI and communications as key features of 6G. SGI-Nets aim to provide global coverage, while edge AI involve ubiquitous deployment of AI algorithms at the network edge to automate various IoT services. At the same time, countries and companies are racing to launch mega satellite constellations to build next-generation information infrastructures. The fusion of SGI-Nets and edge AI, facilitated by mega constellations, presents a unique opportunity to expand the 2D edge-AI architecture into space, thereby providing anywhere-and-anytime intelligent services ranging from environmental monitoring to autonomous navigation and crime prevention.

This project proposes a novel “Fluid AI” framework to address challenges in satellite networks, including fast-changing network topologies and communication bottlenecks caused by massive access and long distances. Using the Mainland-developed satellite “Xidian No. 1” and a mega-constellation simulation platform, the project will: (1) design a practical system with realistic models and predictions; (2) develop Fluid AI technologies to support learning, inference, and AI model downloading with fluid-like migration capabilities in SGI-Nets; (3) create goal-oriented networking techniques customized for Fluid AI to achieve high end-to-end performance and overcome communication bottlenecks; and (4) establish a public simulation platform for implementing and refining these technologies.

Upon successful completion, this research will position Hong Kong at the forefront of 6G research and create a socio-economic impact through knowledge transfer to telecommunication conglomerates headquartered in the region.

CRS_HKU703/24
Next Generation Visual Perception in Open-world by Learning from Multi-Source Data

Hong Kong Project Coordinator: Prof Yizhou Yu (The University of Hong Kong)
Mainland Project Coordinator: Prof Xilin Chen (Institute of Computing Technology / Chinese Academy of Sciences)

As one of the major driving forces of economic growth, Artificial Intelligence (AI) is revolutionizing our lives. Particularly, deep learning based visual perception technologies, which aim to mimic the human brain's capacity to automatically perceive and understand different types of visual data including images and videos from cameras as well as point clouds from LiDAR sensors, have substantially contributed to the rapid development of a wide variety of robots (e.g., service robots). The next-generation visual perception technologies to be developed from this project will assist service robots to take accurate and timely actions for better navigation and interaction with humans and eventually enable robots to provide better functionalities and services to improve the quality of our lives.

Led by world-renowned researchers from both Hong Kong and Mainland in computer vision, computer graphics and deep learning, this collaborative research project will develop next-generation visual perception technologies in the open world by learning from a mix of synthetic and real-world data to support service robots in better perceiving and understanding people, objects and scenes in previously unseen and evolving environments as well as their interactions. Particularly, to address the data collection challenge in conventional visual perception systems (e.g., expensive human annotation costs), in Task I, we will study innovative diffusion model based techniques to achieve high quality and controllable visual data generation with freely available semantic labels at different annotation levels. To substantially improve the generalization capability of learnt visual perception models for both recognition and localization tasks in new and evolving environments, in Task II and Task III, we will study advanced domain adaptation and continuous learning frameworks to address the challenges of conventional visual perception technologies (e.g., lack of annotated visual data in new environments for both previously seen and unseen objects, and for data distribution discrepancy between different environments). In Task IV, we will investigate highly innovative technologies for understanding human emotion and intention by exploiting multiple clues (e.g., facial expression, gaze direction, and gestures) automatically predicted from visual data. In Task V, we will integrate the newly proposed technologies from both teams into one service robot platform to improve its perception capability for better navigation and interaction with humans under different levels of social privacy.