User:Frankushima/sandbox/Carbon Nanotube Computer

A Carbon Nanotube Computer is a computer whose processor consists of transistors made with carbon nanotubes (CNTs) instead of silicon.

CNT computers are not commercially available as they are still in the early stages of development. Carbon nanotubes (CNTs) are atom-thick tubes made of graphene—that is, a sheet made of carbon atoms arranged in a 2-D honeycomb-like formation. Due to their small size, they conduct electricity faster than silicon which could potentially lead to faster processors. As research into silicon-based processors suffer from diminishing returns, CNTs offer a possible solution. However, current CNT computers have several issues with the manufacturing process that make them difficult to implement on a larger scale.

Development

Although attempts at creating a CNT-based modern processor have been successful, they are still rudimentary when compared with current computer processors. The most recent carbon nanotube-based processor contains around 14,000 transistors, while the modern Intel i7 contains around three billion transistors. In addition, creating a CNT-based CMOS( Complementary metal–oxide–semiconductor) that would allow for storage of system settings between reboots was extremely difficult. This means that carbon nanotube computers were not programmable at first and large changes to the hardware had to be performed in order for the computer to execute different sets of computational tasks.

Issues with CNTs

• Carbon nanotubes (CNTs) are prone to material defects. During creation, the thickness and diameter tends to vary which results in the creation of "Metallic CNTs" that are too thick to function as proper semiconductors. This may cause a leakage of current and lead to incorrect outputs— that is, the processor could take in more electricity than necessary, leading to incorrect logic.
• Carbon nanotubes are prone to manufacturing defects during wafer fabrication. In other words, when the carbon nanotubes are fused together to make parts of the computer processor, they tend to bundle together. This makes it extremely difficult to form circuits with them as they tend to overlap which could cause circuit failure.
• Carbon nanotubes have variable properties when created. Properties such as the threshold voltages and polarities of semiconductors must be precisely controlled to create a CMOS digital system. However, current manufacturing processes of carbon nanotubes lack this precision.

Cedric

In 2013, researchers at Stanford University created the first carbon nanotube computer and named it Cedric, a rough acronym for "carbon nanotube digital integrated circuit". Researchers were able to create simple computer circuits by baking a semiconducting wafer in a high-temperature chamber. However, they found that carbon nanotubes tended to intertwine with each other instead of growing parallel to each other. In order to work around this, researchers decided to make their circuits that would work properly, even with the existence of imperfections. Although functional, this heavily limited the scale of their computer, as it only contained 178 transistors, each consisting of 10-200 carbon nanotubes. In addition, the computer could only operate on a single instruction on a single bit of data, limiting the speed at which it could run a program. In comparison, most modern processors run on a 32-bit or 64-bit basis.

RV16X-NANO

In 2019, researchers at the Massachusetts Institute of Technology (MIT) created a larger scale carbon nanotube computer and a process to limit many of the platform's issues. Although far from modern computer processors, the computer was able to display the message: "Hello, world! I am RV16XNano, made from CNTs" and contained around 14,000 transistors. In addition, the computer was able to run 32-bit instructions using 16-bit data and addresses.

In order to accomplish this, researchers at MIT had to develop a robust manufacturing system that would allow the computer to bypass the many issues with using carbon nanotubes as semiconductors. Known as the MMC(Manufacturing Methodology for CNTs), it consists of a three step method that could be done by existing facilities that manufacture computer processors .

1. RINSE (Removal of Incubated Nanotubes through Selective Exfoliation). By using a CNT adhesion promoter and adhesive on the wafer, the non-metallic carbon nanotubes are able to firmly stick onto the wafer before being placed in a solvent and sonicated—that is, they scraped off the metallic CNTs through ultrasonic vibration.
2. MIXED (Metal Interface Engineering Crossed with Electrostatic Doping). Researchers were able to use dielectric oxide to add or subtract electrons from the nanotube, allowing for the creation of the positive and negative polarity CNTs needed to create a CMOS digital system.
3. DREAM (Designing Resiliency Against Metallic CNTs). As a last measure, circuits had to be designed with the fact that several carbon nanotubes would not function to ensure correct functionality of these circuits in case of a metallic CNT or some other defect during the manufacturing process.

Comparison to Silicon-Based Computers

Energy Efficiency

Modern silicon-based computers spend around 80-90% of their electricity usage on fetching data from storage and memory instead of preforming actual computations. As emerging data-driven applications, such as artificial intelligence and machine learning, become more commonplace, this bottleneck has become more apparent. Due to their better conductivity and their small size, CNTs have shown that they can be three times as fast and use three times less energy for the same memory access when compared to their silicon-based counterparts, possibility leading to increasingly energy efficient data-intensive applications in the future.

3-D Circuits

One of the easiest ways to make a processor more powerful is to cram more transistors into it, as stated by Moore's Law. However, recent advances in silicon-based processors no longer offer the same performance gains for each new generation of chips—only around 10–15%, as manufacturers reach the theoretical limit for the amount of transistors they can fit on a processor. One of the possible ways to get around this is to scale circuits vertically, creating 3-D circuits. However, this is not possible with silicon-based processors as the high temperatures used to process the chips would melt the wiring used to connect each layer. On the other hand, CNT-based processors are able to operate as much lower temperatures that would allow for such integration. In addition, 3-D circuits would offer shorter pathways for data to travel, further increasing computing speed.