Can graphene 3D chips "make America great again"?

Ever since Mr. Trump set "America First" as the standard for U.S. government policy, all sectors of the U.S. industry have emerged with "Make America Great Again" programs and programs, including, of course, the electronics industry. The U.S. Defense Advanced Research Projects Agency (DARPA), as the main administration of U.S. military technology research, launched the Electronic Recovery Program in due course.
the program aims to unite U.S. industry and academia to revive the country's sleety chip industry
China
. Some media have also touted America's electronic revival plan as a way to trigger a second electronic revolution, as it claims to change the way the microelectronics industry is produced.
U.S. plan is divided into three parts:
, which is about design, including Electronic Intelligence Resources (IDEA) and Advanced Open Source Hardware (POSH), and is primarily about reducing design costs.
related to computer architecture, including software-defined hardware (SDH) and regional on-chip systems (DSSoC), focusing on the independence and compatibility between hardware and software.
last type of focus on the integration of materials, namely the integration of manufacturing chip materials, including 3D on-chip systems (3dSoC) and new computing basic requirements analysis (FRANC).
's first shortlisted project was funded by 43 teams from across the United States, with the Max Shulaker team from the Massachusetts Institute of Technology toping the list with $61 million, up from $3.1 million for the Georgia Institute of Technology team that also studied 3DSoC. The team's main focus is on using graphene materials to make carbon nanotransistors and to build 3D chips. The team's research is said to be expected to improve computing performance by 50 times at a lower cost.
(the team received the vast majority of sponsorships in the 3DSoC sub-item)
big investments, new materials and a performance boost known as order of magnitude have earned the graphene 3D chip team a lot of eyeballs. There are also a number of domestic public numbers forwarded the news, some more called it "the absolute core of the U.S. electronic recovery plan", and said that such chips will be in the field of artificial intelligence. So we can't help but ask, what is a graphene 3D chip? Is it really that powerful?
the graphene 3D chip isn't entirely made up of graphene
said Max Shulaker, a star professor at the Massachusetts Institute of Technology who led the 3D chip project. His team developed the world's first computer based on carbon nanotransistor technology and published the results in the prestigious journal Nature.
(Professor Max Shulaker)
In 2017, Professor Max again presented a 3D integrated computing and storage model on a single chip in the journal Nature, which also produced the concept of a carbon nanotube 3D chip made of graphene.
Because of Professor Max's brilliant 2013 past, almost all domestic reports have treated the 3DSoC here as a complete graphene chip, and the paper published by Max in 2017 as the development and continuation of his 2013 paper, ignoring the obvious differences between the two.
The carbon nanocrystal computer of 2013 is a pure carbon nanotechnology computer in the full sense of the meaning of the main content is to explore the use of new materials to replace silicon as a new electronic device materials, and recently published in the journal Nature graphene 3D chip is trying to use graphene materials to participate in the construction of traditional silicon chips, the two ideas are not the same.
(it's clear from the paper's drawings alone that it's not a pure graphene chip
the chip, which the professor reported in a 2017 paper in the journal Nature, holds four integrated circuit layers and has five subsysyssors. The parts responsible for the collection, transmission and processing of steam data from experimental samples are built from carbon nanotransistors, while the Resistor Random Storage Unit (RRAM) and interface circuits are constructed from silicon transistors. There is no doubt that this is a combined odor detection chip, not just a carbon nanotransistor.
Graphene chips still have a lot of problems
people want to replace existing silicon semiconductors with graphene as a chip material, in Professor Max's 2013 words: "Carbon nanotubes are smaller, more conductive, and can support fast switching, so their performance and energy consumption are much better than traditional silicon materials."
other words, graphene has better electrical, chemical, and electrical properties that silicon does not have. But are these advantages really what the electronics industry needs? In recent years, the single-core performance of the CPU as the core of the computer is no longer as great as in the past, the main reason is really because of the silicon semiconductor material's electrical, chemical and electrical performance is not good?
fact is obviously not so, today's CPU comprehensive performance does not have too complex reasons, there are main frequency difficult to continue to improve the reasons, there are also chip power barriers and bandwidth barriers. None of this is due to material problems with silicon semiconductors themselves.
the main frequency increase as an example, after the 130nm process, the chip circuit delay with the transistor shrinking trend is getting weaker and weaker. With it, the main frequency is becoming more and more difficult to improve, and the main factor restricting the main frequency has become the connection delay rather than the transistor's flip speed.
(as the process decreases, door delay decreases and connection delay increases
it can be seen that the introduction of new materials at this time can not solve the problems faced by the electronics industry, not to mention the construction of chips with graphene also faced with the old ecological incompatibility, processing difficulties. In fact, there was a discussion in the early days of semiconductor tubes about whether semiconductor substrates should be made with less power-efficient palladium. Finally, because of the cost and the accumulation of silicon circuits in the past, the industry finally abandoned this intention.
new materials introduced today, if these key issues cannot be solved, face only a smaller barrier than the niobium semiconductor materials of the year, so Max's recent research began to shift to graphene-assisted silicon.
Max declared in his recent paper: "The chip's RRAM and carbon nanotransistors are made at 200 degrees, where as traditional processes require 1,000 degrees." Low temperatures help greatly increase vertical connections between integrated circuit layers, which, according to the paper, are 1,000 times more vertical than traditional methods. This connection helps solve the problem of bandwidth barriers in large integrated circuit components.
this temperature difference is caused by the difference between graphene material and silicon semiconductor processing methods, the transistors that build the chip are not etched, but "grown" out. Graphene 3D chips are made on chemical rather than physical terms.
this approach has its advantages to some extent, on the other hand, how large-scale, uniform, the same size of carbon nanotransistors growth is also a headache.
Max Shulaker, who was born in 2013 as the world's first carbon nanotransistor computer, said: "This is the most complex electronic device that humans have produced using carbon nanotubes. "This computer has only 178 transistors and can only run operating systems that support simple functions such as counting and arranging. This is tens of millions of times different from the silicon semiconductor computers of the time.
a separate paper, Professor Max acknowledges that
carbon nanotubes (in processing) are easy to change, which reduces circuit yield, reduces the circuit's anti-jamming capability, and significantly reduces their energy and speed efficiency." In order to overcome this outstanding challenge, carbon nanotube treatment solutions and CNFET circuit design need to be explored and optimized.
says it's too early to replace traditional silicon chips
Professor Max's 2017 research is notable partly because the number of carbon nanotransistors integrated in the chips has increased dramatically to more than 2 million, and partly because the Electronic Recovery Project claims that the team's results are expected to achieve a 50-fold performance improvement at a lower cost.
the author believes that there are still many difficulties in replacing traditional silicon chips with graphene 3D chips, and the team's publicity is undoubtedly quite watery.
said this because the team did not address the yield problems that come with producing graphene chips. The so-called 2 million carbon nanotransistors consist of computing, input and output and acquisition systems, and constitute 1 million odor sensors. In other words, almost all of these transistors are used to make odor sensors, which are highly fault- 0400. Even damage to tens of thousands of sensors in a million sensors will not have a devastating effect on the chip.
the stability and reliability of carbon nanotransistor production by using chips like this is questionable.
and the team did enjoy the flamboyant style of publicity, claiming in the paper that it was several times better than the current method. In an article discussing the design of carbon nanotransistors, it even claims to be at least 100 times better than the current scheme. Therefore, the so-called 50 times performance improvement is also very questionable.