LONDON – Researchers at the University of Pittsburgh have come up with a device structure that allows a switch back to vacuum, in contrast to the solid-state, as the medium for electron transport in transistors.
The team is proposing a MOS vertical structure with a triple layer of metal/silicon dioxide/silicon exposed on the side by a deep trench. The metal and silicon layers form the anode and cathode of the device, separated by the insulating silicon dioxide, and the electron transport occurs in the vertical direction through the vacuum.
The work is discussed in a research paper entitled Metal-oxide-semiconductor field effect transistor with a vacuum channel, published in Nature Nanotechnology July 1.
The work represents a return to the roots of electronics. The solid-state transistor was invented in 1947 as a replacement for the bulky, unreliable vacuum tube. Vacuum tube style electronics in miniature made using solid-state semiconductor manufacturing techniques have been tried before, but the concept has struggled to overcome requirements for high voltage and issue of compatibility with the incumbent solid-state CMOS technology.
A team under Hong Koo Kim, principal investigator on the project and a Professor in the University of Pittsburgh's Swanson School of Engineering, has redesigned the structure of the vacuum electronic device. With the assistance of PhD candidate Siwapon Srisonphan and postdoctoral fellow Yun Suk Jung Kim and his team discovered that electrons trapped inside a semiconductor at the interface with an oxide or metal layer can be easily extracted out into the air. The electrons at the material interface form a sheet of charges, a two-dimensional electron gas and Kim found that the Coulombic repulsion of the electrons for each other enables the easy emission of electrons out of the silicon.
This allows the creation of a low-voltage device in which the electrons travel ballistically in air in a nanometer-scale channel without any collisions or scattering.
The channel length is of the order of 20-nm and the team measured a transconductance of 20-nS per micron and an on/off ratio of 500 and turn-on gate voltage of 0.5-V under ambient conditions, according to the paper's abstract.
"The emission of this electron system into vacuum channels could enable a new class of low-power, high-speed transistors, and it's also compatible with current silicon electronics, complementing those electronics by adding new functions that are faster and more energy efficient due to the low voltage," said Professor Kim, in a statement.
This is not a new Idea, in fact a whole conference series was devoted to integrated vacuum microelectronics. I believe the last one was in 2003. It was the IEEE 16th International Conference on Vacuum Microelectronics.
One other comment.
This device should probably be benchmarked against multi-gate FinFet transistors.
Intel is the only company signed up to go to FinFet, but that would be the incumbent technology to beat.
All devices will work poorly when scaled below 20nm, where direct tunneling of electrons will make it hard to turn any switches off.
Moores law is near the end.
I am encouraged by the significant progress of quantum computing devices using qubits, but that transition is going to take a huge investment in resources and most of the EE community would need significant education just to understand how they work and how to make designs with them.
avtometals appears to be based on a very different mechanism. This does not constitute prior art for patents in the U.S.
I am frankly not very impressed by avtometals technology. It will be incredibly sensitive to small variations in geometry, and the electron momentum. I see a lot of problems with reducing this to practice.
I believe one of the key advantages of this kind of device is that the electrical properties of the channel of the device could be well controlled.
This device may mitigate transistor mismatch, which is a very significant challenge at 20nm.
I believe transistor mismatch largely prevents supply voltage reduction, which used to be effective for reducing power (not lately).
All the other technologies mentioned below suggest low switching voltages, but when you make a billion of them on a chip, it is hard to get them all to switch at the same voltage.
It would be interesting to see if this device addresses or mitigates device mismatch.
That was my question too. From what the article implies, although it doesn't provide any support for this, it sounds like the turn on voltage is lower and the device is faster?
The gate turn-on voltage is 0.5 V for this new device. In a bipolar silicon transistor, the turn-on voltage at the base is 0.7 V, is it not? Germanium transistors had a turn-on voltage of something lower, like 0.2 V. Don't remember off the top what this is for a regular FET.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.