New Transistor Design Combines Best of High and Low Performance Versions

Researchers were motivated by trying to sidestep the Boltzmann limit, which implies that MOSFET transistors can't operate below a certain voltage.

The phone in your hand right now packs the punch a computer three decades ago would have. This is because transistors, the core components of any electronic device, have been shrinking in size. However, the power they consume to perform their functions has not reduced proportionally because of a constraint imposed by thermodynamics.

Researchers at the Indian Institute of Science (IISc), Bengaluru, led by Navakanta Bhat have reported a way around this problem. They’ve designed a new kind of transistor that has high performance and consumed a lot of power when required and then switches down a low-performance mode that consumes less power. This way, the total power used by the device is lower than building and including a high-performance transistor that uses more power all the time but is required by the device only from time to time.

True to Richard Feynman’s prediction in 1959 that “there’s plenty of room at the bottom”, and subsequently to Gordon Moore’s famous law, the size of a transistor has been decreasing continuously since it was first made in 1947. In 1971, the Intel 4004 microprocessor packed 2,300 transistors. Today, almost 100 million transistors can fit into a millimetre-sized chip. This trend allowed electronic devices to become more powerful as they also became faster, smaller, thinner and lighter.

However, they don’t tend to consume that much less power. This is because of the Boltzmann limit.

A transistor is a switch that controls the flow of current. When a suitable voltage is applied to a part of the transistor called the gate, the flow of current flow can be turned on or off.

In a metal-oxide semiconductor field-effect transistor (MOSFET) – a commonly used kind of transistor – this switching on and off doesn’t happen instantaneously: it doesn’t go from ‘on’ to ‘off’, or vice versa, in a single moment. This is because the Boltzmann limit caps the number of electrons that can be emitted thermally at a particular temperature. For a MOSFET transistor, this means that no matter how small the currents are, one has to apply a potential of at least 60 millivolts (mV) to change the current 10 times. So, the minimum voltage required for operating these transistors is about 1 V regardless of their size.

“Power consumption in modern electronic devices is a major technological and environmental concern,” Shubhadeep Bhattacharjee, of the Centre for Nano Science and Engineering at IISc and the lead author of the study, told The Wire. “This is because although we are able to reduce the dimensions of the transistor, we are unable to reduce the voltage required for its operation.”

To get around this problem, Bhattacharjee and his colleagues designed a transistor that acts like a MOSFET at times and like a tunnel field-effect transistor (TFET) at others. The TFET transistor doesn’t work on the thermionic principle, so it sidesteps the Boltzmann limit.

TFETs are based on a quantum mechanical phenomenon called tunnelling; it comes into play only at the scale of a few nanometers. There’s a simple analogy for it: a ball can’t roll up a hill if it doesn’t have the kinetic energy necessary. But if the ball were really small and the hill were a few nanometers thick, it will have the option of tunneling through the hill to the other side. So the TFET can operate at much lower voltages than those determined by the Boltzmann limit and hence consume lower power. But on the flip side, it has a lower performance.

To combine the benefits of both types of FETs in the same transistor, the researchers used a switching element called a Schottky junction. This junction – an interface between two different surfaces – allows both thermionic and tunnelling emission of electrons simply by modifying how tall the barrier to the electrons’ flow is. Instead of using one gate – the part of the transistor that controls whether or not current flows – the researchers used two: one to turn the transistor on or off and another to decide whether the transistor would be thermionic or tunnelling.

“This is a novel idea where the authors decouple the tunnelling and thermionic emission mechanisms to get the best of both worlds,” Swaroop Ganguly, a professor of electrical engineering at IIT Bombay, told The Wire.

There are several transistor technologies that have tried to get around the Boltzmann limit, but they have many disadvantages, according to Bhattacharjee. For example, some can only break the limit either going from ‘off’ to ‘on’ or the other way but not both ways. Others require large operational voltages.

According to the study’s authors, their new design works both ways and can also operate at voltages many hundreds of times smaller. To achieve this, they placed a thin flake of molybdenum disulphide, a few nanometers thick, below the metal contacts that allow current to enter or leave the transistor. This creates the Schottky junction. They enhanced its performance using a special sulphur-based treatment. Based on the voltage applied to the back gate, the barrier to electron flow becomes higher or lower. If the barrier is high, the transistor operates by tunnelling; if it’s low, the transistor switches to the thermionic mode.

The new design isn’t without its shortcomings, however. According to Ganguly, the currents are still very low for practical use in devices, although this can be fixed by choosing appropriate materials. Second: manufacturing this design on a large scale could be a challenge – a general truism for all designs that have an additional back gate. The back gate is common in all transistors that sit on a wafer. Making individual back gates that sit under the molybdenum disulfide flake, which by itself would also be difficult to manufacture, could be challenging.

But Bhattacharjee is optimistic about the manufacturability. Transistors based on molybdenum disulphide are quite new and engineers are still at work trying to make them better. “It will have a developmental cycle. We have not figured out all the problems,” Bhattacharjee said. It seems to be something that can be solved in the near future.

He also said that their technology is compatible with current planar semiconductor manufacturing technologies. For most applications, transistors won’t need to be controlled individually but rather millions at a time. And making a common back gate for each such batch wouldn’t require new manufacturing techniques.

Moreover, the voltages required at the moment to operate the back gate need to be reduced as well. The design uses about 40 V, considered to be very high. This is because the back gate is thick, Bhattacharjee said, and it can be brought down considerably by thinning the gate. In the future, they hope to have a design where the transistor can work both as a MOSFET and a TFET simultaneously instead of having to switch between the two states.

Some applications of such transistors include low-cost devices that need to work with higher currents, such as large-area electronics or flexible display screens. They could also be used in smartphones and computers that need to provide a good gaming experience for a few hours in a day and lower performance at others.

The study was published in the journal Applied Physics Letters in October 2017.

Lakshmi Supriya is a freelance science writer based in Bengaluru.