Carbon nanotube transistors are a step closer to the commercial reality, because researchers from Massachusetts Institute of Technology have proved that these devices can be rapidly manufactured in commercial facilities, and their devices are the same as those for manufacturing silicon-based transistors, silicon-based transistors are the backbone of today's computing industry.
Carbon nanotube field effect transistor (CNFET) is more energy-saving than silicon field effect transistor and can be used to construct new three-dimensional microprocessor. But until now, most of them exist in academic laboratories, and only a small amount of production is used for academic experiments.
However, in a study published in Nature Electronics on June 1, scientists showed how to mass manufacture CNFET on a 200mm wafer of the computer chip design industry standard. CNFET is manufactured in a commercial silicon manufacturing plant and a semiconductor foundry in the United States.
Max Schulak, assistant professor of electrical engineering and computer science at Massachusetts Institute of Technology, and his colleagues made some changes after analyzing the deposition technology used to make CNFET, the manufacturing process is more than 1100 times faster than the traditional method, and the production cost is also reduced. This technology deposits the edge of carbon nanotubes on the edge of the wafer, 14,400 CFNET from 14,400 arrays are distributed on multiple wafers.
Schulak (Shulaker) has been designing CNFET since his doctor's time. He said new research shows that "this is a big step towards production-grade facilities."
Bridging the gap between laboratories and industry is something researchers "rarely have opportunities to do," Schulak added. "But it is an important touchstone for emerging technologies.
Other researchers at the Massachusetts Institute of Technology include the first author Mindy D Bishop. Students of Harvard-Massachusetts Institute of Technology health science and technology program, Gage Hills (Gage Hills), Tathagata Tani (Tathagata Srimani) and Christian Liu (Christian Lau).
Solve the problem of spaghetti
For decades, improvements in silicon transistor manufacturing have reduced prices and improved the energy efficiency of computing. However, this trend may come to an end because more and more transistors are encapsulated in integrated circuits and there is no room to improve energy efficiency.
Schulak said CNFET is an attractive alternative technology because they are "an order of magnitude more energy efficient" than silicon-based transistors ".
Unlike silicon-based transistors, they are manufactured at a temperature of about 450 to 500 degrees Celsius, and CNFET can also be manufactured at a near-chamber temperature. "This means that you can actually directly build a circuit layer above the previously manufactured circuit layer to create a three-dimensional chip." Schulak explained, "Silicon-based technology cannot do this because it will melt the circuit layer below."
He said that a three-dimensional computer chip, which can combine logic and memory functions, is expected to "beat the most advanced two-dimensional chip by the order of magnitude produced by silicon".
One of the most effective methods to construct CFNET in the laboratory is the method of depositing nanotubes, called incubation. In this method, the wafer is immersed in the nano-tube Bath until the nano-tube sticks to the surface of the wafer.
Bishop said that the performance of CNFET depends to a large extent on the deposition process, which affects both the number of carbon nanotubes on the wafer surface and its direction. She said they "either paste them on the wafer in random direction like cooked spaghetti, or arrange them in the same direction like uncooked spaghetti".
Perfect alignment of nanotubes in CNFET can bring ideal performance, but it is difficult to obtain alignment. "It is really difficult to place billions of tiny nanometers in diameter in a perfect direction on a 200mm wafer. "Bishop explained," to put these tiny things perfectly in a wafer is just like trying to cover the whole New Hampshire with fully directional dry pasta."
Although the incubation method is suitable for industry, it does not align the nanotubes at all. Bishop said that they were eventually more like cooked pasta slices, and researchers initially believed that this pasta could not provide high enough CNFET performance. However, after their experiments, she and her colleagues concluded that a simple incubation process would help to produce CNFET better than silicon-based transistors.
CNFET other than beaker
Careful observation of the incubation process showed researchers how to change the process to make it more suitable for industrial production. For example, they found that dry circulation (a method of intermittently drying immersed wafers) can significantly reduce incubation time from 48 hours to 150 seconds.
Another new method called ACE (artificial concentration through evaporation) deposits a small amount of nanotube solution on the wafer instead of immersing the wafer in the groove. Slow evaporation of the solution increases the concentration of carbon nanotubes and the total density of the nanotubes deposited on the wafer.
Bishop said that these changes must be made before the process can be tried on an industrial scale: "In our laboratory, we can put the wafer in a beaker for a week, the effect is very good, but for the company, there is no such luxurious time."
She said, "very simple tests" have helped them understand and improve the incubation method, "for solutions that scholars may not have when considering establishing new processes, but there must be industry concerns, facts have proved that this is very important. "
Researchers cooperated with Analog Devices of commercial silicon manufacturing plants and SkyWater Technology semiconductor foundry to use improved methods to manufacture CNFET. They can use the same equipment in two factories that manufacture silicon-based chips, and at the same time ensure that the nano-tube solution meets the strict requirements of the factory for chemistry and pollutants.
Bishop said: "We are very lucky to work closely with our industry partners to understand their requirements and iterate our development through their investment." Bishop pointed out that this partnership helped them develop automated, mass, and low-cost processes.
Schulak added that the two facilities showed "a serious commitment to research and development and exploration of cutting-edge technologies".
He said the next step will be to use CNFET to build different types of integrated circuits in an industrial environment and explore some new functions that 3-D chips can provide. "The next goal is to transform academically interesting things into things that people will use. I think this is a very important step in this direction."
