Gallium nitride alternative to silicon for more powerful electronics

Materials World magazine
1 Jul 2008

Scientists at the Rensselaer Polytechnic Institute in Troy, USA, claim to have developed the world’s first gallium nitride metal-oxide-semiconductor field-effect transistor (GaN MOSFET).

They hope to exploit the higher breakdown voltages and more efficient electric energy conversion capabilities of GaN to replace silicon MOSFETs for more powerful and efficient electronic devices that can operate in extreme conditions.

Layer upon layer

The manufacturing process adopted resembles that of silicon MOSFET with an n+ doped polysilicon (poly-Si) gate, silicon oxide (SiO2) gate insulator, implanted source and drain regions. In this case an un-doped layer of GaN (about three micrometres) is deposited onto a sapphire substrate, followed by the poly-Si gate and SiO2 gate. The latter is annealed at between 900-1,100ºC. The device then follows processes of poly-Si etching, source implantation, interlayer dielectric deposition, implantation activation, and Ohmic and Schottky metal deposition.

Researcher Weixiao Huang says, ‘Because it is so resilient, the device could open up the field of electronic engineering’. The team claims to have demonstrated GaN MOSFETs with lower power consumption, smaller chip size and higher power density than their silicon counterparts.

‘Experimentally, our GaN MOSFETS already have a lower specific on-resistance,’ says Dr T Paul Chow, who supervised the project and is Professor of Electrical, Computer and Systems Engineering at Rensselaer.

He explains, ‘Most researchers have focused on high electron mobility transistors (HEMTs) which use aluminium GaN or GaN heterojunction as the conducting channel. The main drawbacks are that [the HEMT] is normally on and has a high off-state leakage current.’

‘This approach has its root in [work on] gallium arsenide (GaAs), where HEMTs have been routinely employed because GaAs MOSFETs are difficult to produce. Instead, he says, ‘our approach uses a metal-oxide-semiconductor (MOS) interface for the conducting channel (like Si MOSFET). [It] is normally off and has lower off-state leakage current. Several groups have tried GaN MOS before [but they have used] silicon nitride. We use deposited SiO2’.

The research has received interest from American and Japanese automotive companies for headlamps and engine control.

Reality check

However, Professor Christopher Snowden, Vice Chancellor of the University of Surrey, UK, and a specialist in semiconductor materials and devices, argues that although the research ‘is undoubtedly technically very impressive, it does not prove that this technology can be easily commercialised’.

While scientists worldwide have probed the superior electrical properties of non-silicon semiconductors such as GaN, GaAs and germanium, silicon substrates are still the most cost effective for mass manufacture.

‘It will be immensely difficult for GaN MOS technology to seriously compete in all but niche applications such as high frequency communications, microwave radar and optoelectronics,’ says Snowden.

‘Many people saw silicon-germanium displacing silicon MOS [a while ago], it has not, despite huge investment and a major market position by IBM and others’.
For the team at Rensselaer, areas of further development include improved gate oxide reliability, complementary MOS logic, performance enhancement of the transistor and a better understanding of the SiO2/GaN interface.

‘While intrinsically GaN can withstand high temperatures (>400ºC), the present limitation is the gate oxide reliability which is limited to 250ºC or less,’ says Chow. ‘The most appealing aspect [however] is that GaN can be, though not easily, grown on silicon substrates.’