Carbon Nanotube Transistors
Successful Production in Realworld Fabs
Now carbon nanotube transistors have made the leap from the laboratory to production: a decisive step on the road to commercialization.
Carbon nanotube transistors can be produced in existing fabs on the same machines as are used to produce conventional ICs based on silicon transistors.
This would bring them a significant step closer to commercial use, which would be desirable. This is because carbon nanotube field effect transistors (CNFETs) are an order of magnitude more energy-efficient than their silicon brothers. Three-dimensional architectures could be realized on the basis of CNFETs. Until now, however, they could only be produced in the laboratory in very small quantities.
Researchers at MIT have now produced CNFETs on 200 mm wafers in high volumes in two existing fabs in the USA - without having to change the equipment. Compared to the conventional production method, they have been able to speed up the process by no less than a factor of 1,100 and reduce costs considerably. "This is a huge step forward towards commercialization, we're making the leap into normal fabs," says Max Shulaker, assistant professor of electrical engineering and computer science at MIT, who has been involved in CNFET design since completing his doctorate. "Not many researchers are lucky enough to make the leap from the lab to real-world manufacturing environments." But any new technology that has a chance of succeeding has to pass this test.
The researchers tested their method in collaboration with Analog Devices and Foundry SkyWater Technology. They made no changes to the equipment of the two IC manufacturers and ensured that the nanotube solution complied with Fabs' regulations, for example, to rule out contamination.
"We are now pleased to be able to build the infrastructure necessary to bring the CNFETs into commercial processes. This will also help bring to the U.S. the latest process technologies for the components needed to build the most advanced devices," said Thomas Sondermann, president of SkyWater.
In the next steps, various types of ICs based on CNFETs will now be manufactured in industrial environments and the new 3D chips whose construction has now become possible will be investigated. "In this way, we are bringing carbon nanotubes out of the academic environment and into the real world, people are learning how to use them, a decisive step," says Thomas Sondermann.
This is why CFNETs are so promising
Because CNFETs are significantly more energy-efficient than silicon transistors, they could solve a problem that is becoming more and more urgent with ever smaller structure sizes: the power consumption of silicon transistors no longer decreases with falling structure sizes to the extent known from the past, which is why quite a few experts now really see the end of Moore's Law coming.
CNFETs have another advantage: Unlike silicon transistors, which are manufactured at temperatures of 450 to 500 °C, CNFETs can be formed at room temperatures. "That's why it's very easy to form layers of switching circuits on top of each other to create three-dimensional chips," explains Max Shulaker. "This would not be possible with circuits made of silicon transistors, because the lower layers would inevitably melt during the production of the upper layer. Such 3D ICs, on which logic and memory functions could be combined, would beat the performance of 2D chips by several orders of magnitude.
Until now, CNFETs have been manufactured by dipping a wafer into a solution containing carbon nanotubes. They adhere to the surface of the wafer. This process determines the number of nanotubes on the wafer, their orientation and thus the performance of the CNFETs based on them. It is important to get the orientation of the nanotubes as accurately as possible, especially when it comes to billions of nanotubes with a diameter of 1 nm, whose orientation must be precisely aligned over the 200 mm wafer. It would be tantamount to covering the state of New Hampshire with perfectly aligned - uncooked, so to speak - spaghetti. But by the immersion method, the nanotubes look more like cooked spaghetti, with only a few forming with an orientation that would be orderly enough to form CNFETs with performance superior to silicon transistors. To the surprise of the researchers, however, this method is quite sufficient to form CNFETs with the required performance.
The researchers have also succeeded in adapting the process to the requirements of real production. For example, they have discovered that the process can be greatly accelerated by briefly drying the wafers after wetting and then re-immersion: from 148 hours to 150 seconds.
Another modification of the method is to apply small amounts of the nanotube solution to the wafer instead of completely immersing it (Artificial Concentration through Evaporation). This increased the concentration of the nanotubes and their density over the wafer.
