To enhance Tsinghua University’s role as a major force in basic research and a key source of major scientific breakthroughs, foster a vibrant academic culture of innovation, advance high-quality research development, and support high-level scientific and technological self-reliance, the University launched a selection campaign for top 10 most notable research highlights of 2025.
Following organizational nominations, expert evaluations, and voting by faculty and students, ten outstanding research achievements were selected. The results were announced at the 2025 Tsinghua University Research and Innovation Exchange Conference.
Tsinghua University unveils its top 10 notable research achievements of 2025.
Brief overviews of the selected highlights are provided below, sorted alphabetically by the surnames of the lead authors.
Deng Dongling led team realizes novel quantum topological edge states on a 100-qubit quantum chip
Topologically protected edge states are typically only stable near the zero-temperature ground state. However, when subjected to finite temperature environments, they tend to destabilize due to thermal excitation, leading to the loss of quantum information. Finding ways to effectively protect topological edge states under thermal disturbances is a critical challenge in condensed matter physics and quantum information science. The team led by Deng Dongling at the Institute for Interdisciplinary Information Sciences, in collaboration with other researchers, proposed a protection scheme based on a "prethermalization mechanism." Using a 100-qubit superconducting quantum chip, they successfully realized stable novel finite-temperature topological edge states and further utilized these robust topological edge states to encode and prepare logical Bell states. This approach does not require introducing disorder but instead relies on emergent symmetries within the system to provide extra protection for the topological edge states, thus suppressing their interaction with thermal excitations.
This research establishes a viable digital simulation method, offering a new experimental tool for exploring topological materials at finite temperatures. Additionally, it demonstrates a potential pathway for achieving long-lifetime, robust boundary qubits in disorder-free systems, providing a new approach for the development of noise-resistant quantum storage and control technologies.
Chip “YuHeng” achieves sub-ångström spectral imaging on a fingertip
Developed by Prof. Lu Fang’s team at Tsinghua University and published in the journal Nature, the YuHeng chip delivers full-frame hyperspectral imaging with sub-ångström spectral resolution and ten-megapixel spatial detail, all within a compact 2 × 2 × 0.5 cm package. It operates across the visible and near-infrared spectrum at up to 88 Hz, resolving wavelength differences smaller than ten-millionths-of-a-millimeter in real time. By overcoming traditional trade-offs between resolution, throughput and integration, YuHeng opens new possibilities in astronomical surveys, biomedical diagnostics, and intelligent sensing. For spectroscopic surveys of all stars in the Milky Way, it has the potential to reduce the timeline from thousands of years to less than a decade.
Researchers modify autoactive NLRs to achieve broad-spectrum plant immunity
Professor Yule Liu’s team has pioneered and established a simple and highly efficient strategy for the rational design of plant disease-resistance genes. This approach has been successfully validated in both model plants and the important crop soybean, enabling plants to achieve complete immunity to multiple viruses, and holds great promise as a general strategy for protecting plants against viruses, bacteria, fungi, oomycetes, nematodes and pests.
Compared with existing methods, this strategy offers several key advantages: it is simple to implement, requiring modification of only a single NLR gene; it can be custom-designed to target different pathogens; it confers broad-spectrum, durable and strong resistance that is difficult for pathogens to overcome; and it is highly universal, applicable to virtually all crops. Moreover, it can be readily combined with genome-editing technologies to edit endogenous NLR genes, thereby endowing plants with broad-spectrum disease resistance. This work was published in the journal Nature.
This study represents a pioneering and transformative advance by proposing and realizing a novel disease-resistance gene design strategy based on autoactive NLRs. It demonstrates broad-spectrum, durable and complete immunity to multiple pathogens in both model plants and soybean. The findings provide a revolutionary technological route for crop disease-resistance breeding and have profound scientific significance and outstanding application potential.
Multi-source remote sensing big data: AI-enabled fusion advances global hydrological monitoring
A long-standing challenge in remote sensing hydrology is the spatiotemporal trade-off, which has constrained the ability to achieve continuous monitoring of hydrological variables at both a high spatial resolution and large spatial scales. The research team led by Di Long at the Department of Hydraulic Engineering, Tsinghua University, has overcome this fundamental bottleneck by developing a comprehensive framework that integrates massive volumes of multi-source remote sensing data with strong prior physical constraints.
The team proposed a scalable spatiotemporal fusion framework for multi-source remote sensing big data, substantially enhancing the temporal and spatial completeness of global lake monitoring. Using this framework, the spatiotemporal coverage of state-of-the-art remote sensing observations of global lakes was increased from approximately 66% to nearly 100%. This breakthrough enables, for the first time, near-continuous, high-resolution monitoring of global lake dynamics at an unprecedented scale.
Based on this enhanced dataset, the study reveals that seasonality is the dominant driver of global lake dynamics, challenging the prevailing view that seasonal dominance is a localized phenomenon inferred mainly from long-term trend analyses. By explicitly resolving seasonal variability at the global scale, the research corrects long-standing misconceptions arising from limited temporal sampling in previous studies.
The findings were published in the journal Nature and represent a new research paradigm for AI-enabled remote sensing big data, highlighting the transition of the discipline toward greater intelligence, integration, and scalability. Beyond its scientific significance, the work holds substantial societal value, providing critical technical support for next-generation hydrological monitoring and water resource management in China.
By breaking through the spatiotemporal trade-off in remote sensing hydrology, this study establishes the most comprehensive global lake monitoring system to date, achieving the highest combined spatiotemporal resolution, the widest spatial coverage, and strongest temporal continuity. The framework is expected to significantly elevate national hydrological monitoring capabilities and offers a transferable solution for large-scale environmental monitoring globally.
Ode to Bravery: sculptural group honoring border defenders guarding the Kunlun Mountains
Created by Ma Wenjia’s Team at the Academy of Arts & Design, this sculptural group—modeled on martyr Chen Xiangrong—weaves patriotic spirit into commemorative artistic language, using a three-dimensional narrative structure that pairs the foreground figure with the Kunlun Mountains to deliver powerful emotional and artistic impact. The six-month creation process spanned the full workflow, from sketching and digital modeling to clay enlargement, bronze casting, stone carving, and installation; by leveraging digital scanning and preview technologies, the project pioneered new pathways for technology-empowered art, fully showcasing the systematic approach and professional standards of artistic creation and research in higher education.
Guided by the principle of “educating through creative and research achievements,” the project was completed under the leadership of counselors, with students who are members of the Communist Party of China and young artists working collaboratively. The process exemplifies the effective integration of ideological and political education with artistic practice. The work has been reported by Xinhua News Agency and other major media outlets, generating wide social impact. It stands as a representative example of using artistic practice to serve ideological and political education and to communicate the spirit of patriotism in contemporary China.
Centered on the theme of patriotism, the work employs a contrast between bronze and sandstone materials and incorporates digital technologies to portray the heroic image of martyr Chen Xiangrong and his spirit of bravery and fearlessness. Beyond its solemn and symbolic artistic expression, the project explores an innovative model of ideological and political education in universities by integrating artistic creation with ideological and political education, achieving the dual goals of ideological guidance and artistic innovation.
High hydrogen atom economy for olefin synthesis from syngas: A core-shell catalyst for syngas-to-olefin in-situ coupling of water-gas shift reaction
Led by Qian Weizhong’s group at the Department of Chemical Engineering, a new study targets the cutting-edge international catalytic technology route of one-step olefin synthesis from syngas, with researchers developing a novel core-shell catalyst that in-situ couples the water-gas shift (WGS) reaction with syngas-to-olefin (STO) functionality. This innovative design enables the in-situ conversion of water produced during the process into hydrogen, boosting the hydrogen atom economy (HAE) of target products to 66%–86%—far surpassing the 50% theoretical HAE and 43%–47% practical values of traditional methanol-to-olefin routes. Critically, the catalyst also achieves high CO conversion and high olefin selectivity, marking a major leap forward in efficient syngas utilization.
Compared with conventional processes, the proposed technology reduced total steam consumption, significantly decreased total wastewater generation and CO2 emissions, and lowered the complete environmental factor by 46%. It was expected to significantly promote the utilization of syngas and the green transition of advanced coal chemical industry in China, while creating conditions for the efficient utilization of renewable green hydrogen. This research was published in the journal Science.
The process of producing olefins from syngas was a revolutionary chemical technology that has been widely promoted in recent years. By ingeniously coupling WGS and STO functionalities, Qian's group achieved high conversion, high product yield, low wastewater emission, and low CO2 emission. Furthermore, the technology could be expected to utilize green hydrogen generated by renewable electricity, reduced production costs and facilitated the large-scale application.
Cheng Song’s team reveals crystal fingerprints of altermagnets, establishing a third magnetic material class
Conventional wisdom holds that the characteristics of ferromagnets and antiferromagnets are distinctly demarcated and mutually exclusive. The emergence of altermagnets challenges the prevailing viewpoint, since they combine the characteristics and advantages of both ferromagnets and antiferromagnets. It is the unique crystal symmetry that distinguishes altermagnets from conventional magnetic materials and generates the altermagnetic spin splitting. In sharp contrast to the significant attention to the spin fingerprints of altermagnets, the revelation and manipulation of the crystal fingerprints of altermagnets have remained elusive.
The team led by Cheng Song grew CrSb films, a new type of room-temperature altermagnet. The magnetic space group can be switched by the crystal distortion, thereby enabling control of the electrical read-out and write-in. They formulated a theoretical criterion for the all-electrical switching of altermagnetic order. Replacing the magnetic field with the unique crystal symmetry, they achieved the long-sought all-electrical manipulation of altermagnetism. Their work demonstrates that “altermagnetic order = Néel vector × crystal symmetry”, and provides the critical experimental evidence for establishing the altermagnets as a third class of magnetic materials, separated from ferromagnets and antiferromagnets. The relevant work was published in the journal Nature.
The magnetization compensation and spin splitting are mutually exclusive in conventional magnetic materials, with ferromagnets and antiferromagnets as the representative examples. Altermagnets combine the characteristics and advantages of both ferromagnets and antiferromagnets, holding great promise for developing the next-generation magnetic memory technology. This work concentrated on the physical underpinnings of altermagnetism, as well as the revelation and manipulation of the crystal fingerprints of altermagnets. It not only lays foundations for promoting altermagnets to be the third class of magnetic materials, but also opens new avenues for developing altermagnet-based memories.
Spin configurations trigger breakthrough in antiferromagnetic quantum anomalous Hall effect
A research team led by Professor Yayu Wang from the Department of Physics has unveiled how manipulating spin configurations can control a unique quantum phenomenon, the quantum anomalous Hall effect (QAHE), in an antiferromagnetic material. This work overcomes a major experimental hurdle for QAHE device fabrication and opens new pathways for controlling topological quantum phase transitions.
The material at the heart of this study, MnBi2Te4, is a landmark system in condensed matter physics, uniquely combining a two-dimensional structure, intrinsic antiferromagnetic order, and topological electronic properties. Despite its promise, progress has been stalled for years due to difficulties in creating high-quality experimental devices.
After a five-year effort, Professor Wang's group, through refined single-crystal growth and advanced device fabrication techniques—notably the application of a protective AlOx capping layer—has dramatically improved device performance and reproducibility.
The team's high-quality devices, based on 7-layer MnBi2Te4, yielded several groundbreaking results. Most significantly, they observed a quantized Hall resistance plateau at zero magnetic field, a hallmark of the QAHE. Their research systematically demonstrates how the specific spin configurations (spin flips and flops) in a 2D antiferromagnet govern topological electron transport. Furthermore, they discovered that applying an in-plane magnetic field can surprisingly enhance both the stability and precision of QAHE quantization.
Published in the journal Nature, this research marks a landmark achievement in magnetic topological insulators. By solving persistent technical challenges and expanding the experimental toolkit, the work provides crucial insights into topological quantum phases. It also establishes a vital foundation for future applications in low-energy-consumption antiferromagnetic spintronics.
Pioneering "Storage-for-Compute": beaking barriers in high-performance LLM inference
Prof. Wu Yongwei's team at Tsinghua University's Department of Computer Science and Technology has achieved a breakthrough in AI infrastructure with their novel "Storage-for-Compute" and "Full-System Synergy" design paradigms. Addressing the critical "Memory Wall" challenge in Large Language Model (LLM) inference, the team has successfully developed a robust solution for high-performance computing.
In collaboration with leading industry partners, the team co-developed advanced inference systems such as Mooncake (https://github.com/kvcache-ai/Mooncake) and KTransformers (https://github.com/kvcache-ai/ktransformers). These innovations significantly enhance inference throughput and lower the deployment barrier for sparse models. The work has been recognized with the Best Paper Award at FAST 2025, a top-tier system conference, and the prestigious global "Olympus Award" in the storage domain.
Currently open-sourced with over 20,000 stars on GitHub, the core technology has been adopted by major Internet and AI enterprises. It is now deployed at scale on clusters exceeding tens of thousands of GPUs, powering the serving of trillion-parameter models.
Tsinghua researchers prototype wafer-scale chips for AI computing
A team of researchers led by Professor Shouyi Yin at Tsinghua University’s School of Integrated Circuits has unveiled new research exploring wafer-scale approaches to large-area AI computing and system integration, with the team designing and evaluating a wafer-scale AI chip prototype and developing an architectural framework that addresses computing organization, on-wafer interconnect, and integration at an expanded physical scale. As computing systems continue to grow in complexity and scale, the work examines how architectural and system-level design choices can extend performance and efficiency beyond conventional single-die boundaries. The study establishes a coherent technical framework spanning architecture, interconnect, and advanced integration, and validates the feasibility of achieving high computing capability through coordinated design across these layers using mature semiconductor technologies.
The research further demonstrates how such a framework can support reusable design methodologies and structured design rules, enabling systematic exploration of wafer-scale implementations. Through academic–industry engagement, the work contributes practical insights into the alignment of architectural innovation with manufacturable design flows.
Overall, this study provides a reference framework for future investigations into large-scale AI computing architectures and offers a scalable perspective on system integration beyond traditional chip-level design.
Editors: Li Han, John Paul Grima
