By Jiang Qingling, Trainee Reporter of China Science Daily
In the clean room of the National Key Laboratory of Integrated Chips and Systems at Fudan University, researchers in lab coats are demonstrating the chips in their hands to reporters from China Science Daily.
Twenty-some golden chips are neatly arranged on a transparent tray, looking unremarkable at first glance, but they hide great secrets. Each chip integrates 5,900 transistors based on two-dimensional semiconductor materials, which is currently the largest-scale verification record for two-dimensional logic functions internationally. This surpasses the previous highest record of 115 transistors by 51 times.
The research team led by Professor Peng Zhou and Researcher Zhongwen Bao from Fudan University has broken through the integration bottleneck of two-dimensional semiconductor electronics, completing the full-chain independent R&D from materials to architecture to tape-out, successfully developing the world's first 32-bit RISC-V architecture microprocessor based on two-dimensional semiconductor materials. They named it "WUJI" (meaning boundlessness), symbolizing creation from nothingness and endless possibilities.
Research artwork.
Under the control of 32-bit input instructions, "WUJI" can perform up to 4.2 billion addition and subtraction operations between data, support GB-level data storage and access, and allow the writing of programs with up to 1 billion reduced instruction set computer (RISC) instructions. The relevant findings were published in Nature on April 2nd.
Bringing the 'Ultimate Form' of Transistors Out of the Lab
"To talk about this research achievement, we have to trace back 10 years," recalled Zhongwen Bao. In July 2015, Zhongwen Bao joined the Microelectronics College at Fudan University and dived into the study of controllable growth of wafer-scale two-dimensional semiconductors and their practical applications.
This year marks 11 years since the discovery of graphene, the first two-dimensional material, and other two-dimensional semiconductor materials such as molybdenum disulfide (MoS2) began to catch the attention of scientists.
So-called two-dimensional (2D) refers to materials that only extend in the plane. With a thickness of just one to several atomic layers, this "ultra-thin" characteristic makes these materials exhibit very unique electronic and optical properties, holding potential application value in fields like nanoelectronics, optoelectronics, flexible devices, and sensors.
MoS2 is one of the most popular two-dimensional semiconductor materials, known for its natural bandgap and adjustable bandgap type, giving it unique advantages in transistor and optoelectronic device manufacturing.
"At the nanoscale, silicon materials are not the best channel materials," pointed out Zhongwen Bao. "Faced with the global challenge of Moore's Law approaching physical limits, two-dimensional semiconductors are currently recognized worldwide as the key to breaking through this limitation and can be considered the 'ultimate form' of transistors."
However, there is still a vast distance between individual transistors and functional integrated circuits. It's akin to a violinist excelling in solo performances; when dozens of equally talented musicians come together to form an orchestra, it takes considerable time to rehearse before they can play harmoniously. If the number increases to tens of thousands or even billions, the difficulty grows exponentially.
Therefore, although the individual capabilities of single transistors made from MoS2 have been widely acknowledged, due to bottlenecks in process accuracy and scale uniformity as well as overall yield control, the highest integration level previously reached was only in the hundreds of transistors, never crossing the technical threshold for functional microprocessors.
"Can two-dimensional semiconductor materials actually be made into chips, and what kind of performance do they have?" Peng Zhou told China Science Daily.
Paving a New Path for Chip Development
In 2021, the research task was handed over to PhD students Mingrui Ao and Xiucheng Zhou from Fudan University. Previously, the team had accumulated extensive experience in process production and achieved rapid growth of uniform and monolayer MoS2 on industrial mainstream 12-inch wafers using chemical vapor deposition methods.
As the saying goes, if you want to do something well, you must first sharpen your tools. Industrial 7-nanometer and 2-nanometer integrated circuits rely on advanced lithography machines and other equipment. The team’s goal is an integrated circuit system for industrial production, but most of the equipment used is at the research level. With insufficient hardware, they could only actively adapt, continuously accumulating experience through trial and error.
Meanwhile, the extremely thin nature of two-dimensional semiconductor materials poses significant challenges in processing. Zhongwen Bao used sculpture as an analogy—while silicon materials are like marble blocks that can be sculpted into figures with axes and chisels, two-dimensional semiconductors are more like tofu, easily damaged with light touches, requiring special equipment for processing.
"The production process of two-dimensional semiconductors is interdependent," said Zhongwen Bao. "Not only does one step affect the next, but subsequent steps also impact earlier ones."
In recent years, Mingrui Ao and Xiucheng Zhou have conducted experiments day after day, meticulously refining details, adjusting parameters, controlling laboratory environments, and optimizing process steps.
"We paused at a certain step for quite some time, only to find out later that it was caused by a problem in an earlier process. Therefore, after completing each process step, we conduct corresponding tests. If issues arise, we adjust them promptly," added Xiucheng Zhou.
It is worth mentioning that combining the team's extensive accumulation of material and process data along with Fudan University's AI layout, they developed an AI-driven integrated process optimization technology, achieving precise control from material growth to integration processes through dual engines of "atomic-level interface precision control +全流程AI algorithm optimization." "With AI algorithm-recommended combinations, we can more efficiently produce chips in the lab," said Zhongwen Bao.
Thus, under the joint efforts of the members and with the support of new technologies, the team ultimately solved the precise coupling regulation problems of two-dimensional materials-contact-gate dielectric-post-processing and verified scalable digital circuits using atomically precise processing and characterization techniques.
Research diagram. Provided by Fudan University.
"This is not a disruptive overall change. It's like building an office building where the foundation and main structure remain the same, except that a few floors are designed as malls instead of offices," explained Zhongwen Bao. "About 70% of the integration process of 'WUJI' can directly adopt mature technologies from existing silicon-based production lines, while the remaining core two-dimensional specific processes combine the team's independently designed specialized process equipment and systems, including over 20 process invention patents."
Each test result indicates that the integration process optimization and scalability of 'WUJI' have reached the international optimal level at the same period. Zhongwen Bao confidently stated, "We utilized domestically produced semiconductor equipment and open-source RISC-V architecture, without relying on advanced EUV lithography machines, but instead integrated our independently developed complete set of two-dimensional semiconductor integration processes, laying the foundation for opening a new path of independent R&D for China's chips."
Two-Dimensional Semiconductors Will Not Replace Silicon
"Just as the advent of subways still values buses, two-dimensional semiconductor chips and silicon-based chips are complementary relationships," said Peng Zhou. "'WUJI' uses micrometer-level processes, with power consumption comparable to nanometer-level chips. If better lithography machines are used, power consumption will further decrease, making it more advantageous in devices requiring higher power efficiency in the future."
Peng Zhou emphasized, "'WUJI' is merely a concept validation prototype, and its overall performance still has a gap compared to commercial chips currently available. It does not yet possess market competitiveness at present."
Currently, the team is striving for the conversion and implementation of 'WUJI.' On one hand, they aim to further enhance the performance and integration of two-dimensional electronic devices, breaking through the current transistor integration bottleneck, making it more competitive in a wider range of applications. On the other hand, in the industrialization process, the team strengthens the integration with existing silicon-based production lines, promoting the industrial application of core two-dimensional specific processes, and collaborates with related enterprises and institutions to enable it to play a role in actual products as soon as possible.
Zhongwen Bao stated that over the past few decades, the development of integrated circuits has accumulated rich experience for the industrialization of two-dimensional semiconductor chips. "There is reason to believe that the performance of two-dimensional semiconductor chips can catch up with silicon-based chips in a relatively short time, eventually forming a long-term coexistence and application complementarity with silicon-based chips."
Relevant paper information:
http://doi.org/10.1038/s41586-025-08759-9
Original article: https://www.toutiao.com/article/7490489712134636095/
Disclaimer: This article represents the author's personal views. You are welcome to express your attitude by clicking the 'like/dislike' buttons below.