China's carbon-based chips are gaining predominance

According to scientists' expectations, the physical limit of silicon-based chips is 1nm. According to the process evolution of fabs, the process after 3nm is 2nm, and then the process after it is 1.4nm. Due to the limitations of silicon-based materials, the defects in the performance of silicon-based chips will become more and more obvious, and it is impossible to advance below 1nm. Like what:

• In terms of process, silicon-based chips have quantum tunneling effect, atomic doping fluctuations, power consumption walls and other phenomena.
• In terms of architecture, the memory wall phenomenon of von Neumann architecture has emerged.
• In terms of transistor principle, the Boltzmann limit of subthreshold swing and the reduction limit of the operating voltage also appeared.

Therefore, the search for materials that can replace silicon-based chips has become a hot topic, and carbon-based chips have begun to appear frequently in the public eye.

01

Three obvious advantages of carbon-based chips

First, carbon-based chips have lower limits, better performance and lower power consumption, which are more suitable for the rapid development of future technology. Carbon-based chips are made from carbon-based wafers, which are based on graphene semiconductor materials. Because graphene has excellent characteristics such as high carrier mobility and good thermal conductivity, graphene transistors run 5~10 times faster than silicon-based transistors, while the power consumption is 1/10 of silicon-based transistors. This gives carbon-based chips lower limits, better performance, and lower power consumption.

Second, carbon-based chips do not have high requirements for the process technology of the lithography machine. Carbon-based chips using the 90nm process are expected to produce silicon-based chips with performance and integration equivalent to 28nm technology nodes, and carbon-based chips using 28nm processes can realize silicon-based chips equivalent to 7nm technology nodes. That is to say, the use of 28nm lithography machine can obtain the effect of the world's most advanced EUV lithography machine.

Third, at present, there is almost no international technology in this field. Early R&D can give you a first-mover advantage.

02

The results of carbon-based chips

In the research and development of carbon-based chips, this scientific and technological chase battle is mainly between two of the world's top universities - Peking University (in conjunction with the Chinese Academy of Sciences) and the Massachusetts Institute of Technology. Many papers on carbon transistors that appeared in the journals Nature and Science were also published by these two.

Peng Lianmao, an academician of Peking University, said that compared with foreign silicon-based chips, the domestic carbon-based chips used in China to process large-scale data are fast and have low power consumption, saving at least thirty percent. And at present, the domestic graphene wafer production technology is at the international leading level in terms of the size and quality of graphene single wafers. It can be seen that China already has an advantage in the research and development of carbon-based chips.

03

The development of carbon-based chips is of great significance

As early as 2017, Peking University's work on 5nm gate carbon nanotube CMOS devices demonstrated that carbon nanotubes can overcome the short-channel effect when reaching the theoretical limit, which eliminates the need to use more complex three-dimensional transistor technologies such as FinFETs developed like silicon technology to reduce the short-channel effect. In addition, carbon nanotube technology itself is a cryogenic technology that can prepare three-dimensional chip O. Previously, Stanford University also issued a report that if carbon nanotube transistors are designed to be three-dimensional, the performance can be 1,000 times higher than that of two-dimensional silicon transistors.

At present, most of the chips use silicon-based integrated circuit technology, high-end chip technology has been monopolized by foreign manufacturers for a long time, and China's annual cost of importing chips is as high as 300 billion US dollars. Accelerating the development of the semiconductor industry represented by chips has long become a national strategy, but in the context of anti-globalization and industrial blockade, China's existing silicon-based semiconductor development is difficult, most of which are concentrated in low-profit, low-value-added links. This new path for carbon-based chips brings more possibilities for China to prepare high-end chips.

At present, the technology of the world's semiconductor manufacturers still insists on silicon wafers, and Intel, Samsung, and TSMC are investing all their resources in 5nm, 3nm manufacturing technology. That is, in the next at least 5 years, the mass production of carbon-based chips will not be born in these few. Then, the most suitable environment to become a leader in emerging factories in this field will be in China.

According to official information from the State Intellectual Property Office, in 2021, Huawei released a graphene transistor patent, and it was previously reported that Huawei would set up a special scientific research team with Peking University to work on the development of carbon-based chips. It can be seen that carbon-based chips may have become one of the future development directions of Huawei chips.

However, no company in the world has yet been able to build a commercial carbon-based chip. Why didn't "carbon-based chips" enter the market as expected?

04

Carbon-based chips cannot replace silicon-based chips for the time being

Carbon nanotubes are difficult to purify

For chip manufacturing, carbon nanotubes have excellent performance, but they are not easy to use. It is reported that the purity of carbon-based chips is at least 6 to 8 9's (99.999999%), so that the performance of carbon-based chips can be comparable to traditional silicon-based chips. This is also the reason why the MIT team's CPU obviously contains 14,000 carbon nanotube crystals, but its performance is not as good as Peking University's CPU composed of about 2,500 transistors.

For carbon nanotubes, it is difficult for a single purification method to completely remove impurities in carbon nanotubes, especially single-walled carbon nanotubes. After the preparation of carbon nanotubes, they often contain a large number of impurities, and their impurities mainly include carbon impurities and metal impurities. The presence of impurities in carbon nanotubes limits their application, so the purification of carbon nanotubes has become an urgent technical problem.

Mass production is difficult

In addition to process limitations, the mass production of carbon-based chips is still much more difficult than silicon-based chips. On the one hand, the research on graphene in countries around the world has not yet reached the point where carbon-based wafers can be mass-produced to meet the needs of major enterprises; On the other hand, there are also differences in the related technologies and process equipment of carbon-based semiconductors and silicon-based semiconductors, although 90% of the existing silicon-based semiconductor processing equipment can be directly applied to the carbon-based, but some processes or equipment need to be debugged to adapt to the production of carbon-based semiconductor devices.

Mass production is a very important prerequisite for determining commercialization, without large-scale production, it cannot be truly commercialized; If the production is insufficient, the price cannot be brought down.

As for the mass production time of existing products, Chinese carbon-based chip researchers said that since 2018, the entire line of the first 4-inch carbon-based wafer was put into experiments from preparation, circuit design, lithography, and packaging, 4-inch 5-micron gate-length carbon-based wafers have been produced on the product side. These achievements can be directly linked to front-end RF device manufacturers. The relatively low difficulty of IoT carbon tube chips is expected to be commercially available in the next 3 to 5 years. Carbon-based chips applied to mobile phones and servers take longer.

The integration of production, education and research is not easy

According to the current situation of China's industry, education and research, domestic companies are more willing to cooperate with universities when the research results are mature. Because of the uncertain cooperation, it is difficult to bring profits to the company in the short term, and there are certain risks in the long run, and the industry cannot get enough motivation to invest funds. Carbon-based semiconductor technology from the laboratory to the industry also needs to carry out engineering research, and the challenges include continuous funding assurance, concept change, and compatibility with existing industries. If industry, academia and research cannot be better integrated, these problems will be difficult to solve by relying only on the research team.

At present, although most domestic companies prefer to cooperate with universities when their research solutions are mature, Huawei, which is in a leading position in the industry, will pay more attention to basic research and development, and has already contacted and docked with Peng Lianmao's team.

summary

With the wide application of mobile smart devices, cloud storage and big data processing, the rapidly developing information industry has put forward unprecedented requirements for future semiconductor chips and information processing technologies. In order to continue Moore's Law and cope with the post-Moore's Law era, and continue to improve chip performance, it is necessary to develop new semiconductor chips with faster speed and lower energy consumption.

China has gradually gained an advantage on the road of carbon-based chips, but the integration of carbon-based chips currently developed is still far from the silicon-based chips commonly used in the world. At present, the school laboratory can use carbon nanotube materials to prepare some medium-scale or even large-scale integrated circuits, but it is not possible to use it to make ultra-large-scale integrated circuits. Peng Lianmao said that under the condition that the state attaches importance to it and scientific research funds are sufficient, it is expected that carbon-based technologies can be applied on a small scale in some special fields in 3-5 years; It is expected that after 10 years, carbon-based chips are expected to gradually become mainstream chip technology with product changes.

It is worth noting that the domestic frenzy for new concepts is often too fanatical. Carbon-based chips do have great potential in many fields, but from discovering potential to industrialization, what is needed is down-to-earth research, and the public's tolerance of new technology research and development failures. Not only "new" materials are needed, but also a "new" industry-university-research system and a more open and inclusive investment environment.