The world's first carbon nanotube universal computing chip is launched! Nature publishes three articles recommending

This article is reprinted with permission from AI new media Quantum Bit (public account ID: QbitAI). Please contact the source for reprinting.

"Hello, World! I am RV16XNano, made from CNTs".

"Hello, world! I'm RV16XNano, made of carbon nanotubes."

This sentence comes from the program executed by the 16-bit carbon nanotube chip invented by the MIT research team.

Yes, you read that right, they used the same manufacturing process as silicon to make a chip with a complete architecture based on carbon nanotubes, and then they greeted the world.

Just now, Nature published this research result and published corresponding news and comments to highlight it.

Carbon nanotubes are considered to be the first choice to replace silicon materials. They conduct electricity faster and more efficiently than silicon.

Theoretically, it would be 10 times more efficient than silicon, run three times as fast, and consume only one-third of the energy.

Moreover, it has a wider range of uses. Some scientists believe that carbon nanotubes can also be used to make microchips that can be injected into the body, or nanomachines that kill cancer cells in the human body. The prospects are immeasurable.

But carbon nanotubes also have a series of design, manufacturing and functional problems, such as "crazy growth" and difficulty in placing them in a specific location to make them play a specific role.

In the face of huge prospects and potential, these problems are being overcome.

In 2013, Stanford University built the first carbon nanotube computer, which had only 178 transistors.

Now, a research team affiliated with MIT has created the RV16X-NANO , which has 14,000 transistors.

It has increased nearly 80 times in 6 years, which is 5 times the speed of Moore's Law.

In an interview with Nature magazine, Professor Sun Yanan of Shanghai Jiao Tong University said, "This work has taken a big step forward and is closer to commercial chips."

Technology media ArsTechnica also commented that it was an impressive work.

In contrast, the comments given by netizens were not so restrained:

This is a stunt - but it's a pretty cool stunt

It's a stunt, and a very cool one at that.

Taming carbon nanotubes

Although carbon nanotubes have many advantages over silicon crystals, there are many problems in using them to make chips.

First, although carbon nanotubes are semiconductors, their manufacturing process requires metals, and the carbon nanotubes produced in this way will inevitably be mixed with metal impurities. If you want to obtain a purified semiconductor version, you need to increase the purity level to 99.999999% , which is almost impossible under current technological conditions.

Moreover, carbon nanotubes do not naturally form p-type or n-type semiconductors. In silicon, these properties are achieved by doping with small amounts of other elements. But carbon nanotubes are very small and difficult to dope.

Another problem is that making electronic components requires placing nanotubes in extremely precise locations. Scientists don't yet have a way to grow them in specific locations, so they have to be made separately and then deposited on a surface.

Unfortunately, this process typically produces a randomly oriented nanotube film consisting of a large number of carbon nanotubes and some metallic nanotubes mixed in.

Researchers at MIT and scientists at Analog Devices have found a way to address all of these issues.

The researchers have proposed a technology called DREAM , which relaxes the strict purity requirements for carbon nanotubes by about 10,000 times, which means that chips can be made with a purity of 99.99%, which is feasible with current technology.

Making a carbon nanotube chip first involves solving the problem of disordered arrangement. The researchers created a silicon surface with metallic features that was large enough to allow the nanotubes to grow between the metallic gaps.

To remove the aggregates, they deposited a layer of material on top of the nanotubes and then broke them up using ultrasound. This material carried away the aggregates but left the underlying nanotubes undisturbed.

Next, to confine the nanotubes to where they are needed, the researchers simply etch away most of the nanotubes, leaving only the desired parts.

The researchers then used a technique called atomic deposition to attach metal oxides to the nanotubes. Different metal oxides have different properties and can be used to convert the nanotubes into p-type or n-type semiconductors as needed. This process is similar to doping silicon crystals and can effectively control the behavior of individual pn junctions.

From transistors to chips

The resulting device is called a carbon nanotube field-effect transistor (CNFET), which, like the metal oxide semiconductor field-effect transistor (MOSFET), is the basic unit for building the next generation of computers.

The functions of the chip are realized by the combination of logic gates, and logic gates can be constructed by combining CNFETs.

△ Inverter made of carbon nanotubes

The researchers wanted to make certain logic operations less sensitive to metallic nanotubes, so they modified an open-source RISC design tool to account for these issues, allowing the chip design to be designed without the gates that are most sensitive to metallic carbon nanotubes.

The resulting chip is called RV16X-NANO and uses 32-bit long instructions of the RISC-V architecture. Memory addressing is limited to 16 bits, and the functional units include instruction fetch, decode, registers, execution unit, and write-back memory.

In total, the RV16X-NANO uses more than 14,000 individual transistors, with a 100% carbon nanotube yield, meaning that every one of these 14,000 transistors is functional, with no defective transistors.

The RV16X-NANO is also a 3D chip, with metal contacts below the nanotube layer used to pass signals between different transistors, while separate metal contacts above the nanotubes are used for power.

Room for improvement

The transistor channel length in the RV16X-NANO chip is about 1.5 microns, which is equivalent to the Intel 80386 in silicon chips. This processor was launched in 1985.

The 80386 runs at 16MHz, while the carbon nanotube computer maxes out at 1MHz. The difference is due to the capacitance of electronic components and the amount of current that transistors can carry.

Silicon transistors can carry about 1 milliampere of current per micron of width (1mA/μm), while carbon nanotube transistors can only carry about 6μA/μm. This is an area that needs to be improved in future versions of computers.

The first step to increase current is to reduce the length of the transistor channel. The channel length of two carbon nanotubes can be reduced to 5nm.

The second step was to increase the density of nanotubes in each channel from 10 per micron to 500 per micron. The new deposition technique boosted the current density in this network to 1.7 mA/μm.

The third step is to reduce the width of the transistor, thereby reducing the width of the source and drain, which will allow the electrodes to charge and discharge faster.

The team from 6 years ago is back

This study has two first authors, Gage Hills and Christian Lau; the corresponding author is Max M. Shulaker; both are from MIT.

Among them, Max M. Shulaker and Gage Hills were the first and second authors of the first carbon nanotube computer research results in 2013. At that time, they were still doctoral students at Stanford University.

The progress made this time is based on this research.

In July 2016, Max M. Shulaker joined MIT as an assistant professor to continue his research on carbon nanotubes. Currently, Gage Hills is a postdoctoral researcher at MIT, responsible for most chip design work.

Christian Lau, a master's student at MIT, is responsible for most of the chip manufacturing work.

In addition, two of the authors are from Analog Devices, which is reportedly one of the sponsors of this research.

According to MIT, Shulaker's next step is to bring the chip into the real world.

He said it was no longer a question of yes or no, but when.

To achieve this goal, they have applied the technology to silicon chip foundries through a project of the US DARPA to carry out research.

As for when chips made of carbon nanotubes will be commercially available, no one can give an exact time.

But Shulaker said it's probably less than five years.