Technical Insight
Research Review: Swiss speed AlInN HEMTs
A Swiss partnership between Colombo Bolognesi’s group at ETH-Zürich and Nicolas Grandjean’s group at EPFL have broken the speed record for AlInN/GaN HEMTs. When deposited on SiC, their devices can deliver a current gain cut-off frequency (fT) of 144 GHz, an extrinsic transconductance of 480 mS/mm and a maximum current density of 1.84 A/mm. Silicon offers a cheaper, but inferior foundation that leads to a fT of 113 GHz.
AlInN/GaN HEMTs promise faster speeds than their AlGaN/GaN cousins, but until now they have lagged a long way behind. “Our work has closed the gap for 100 nm gate lengths, presumably thanks to progress in the crystal quality of AlInN/GaN materials,” explains Bolognesi.
He claims that the AlInN/GaN HEMT has the potential to outperform its AlGaN/GaN equivalent because its heterostructure can realize a higher channel electron density with a thin top barrier. “This allows one to place the gate electrode closer to the channel and thereby maintain a better electrostatic control of the transistor, so as to minimize the so-called ‘short-channel effects’ that curtail performance in shorter gate devices.” Thanks to the combination of small gate-to-channel distance and a high electron density in the channel, these transistors can realize very high transconductances, and ultimately higher cut-off frequencies.
The AlInN/GaN HEMT fabricated by ETH Zürich and EPFL features Ni/Au T-shaped 100 nm gates. Credit: ETH Zurich
One of the weaknesses of the team’s AlInN/GaN HEMT is its residual gate leakage, which prevents complete channel pinch-off at source-drain voltages above 3V. The source of this leakage might be caused by tunneling through the barrier, or leakage through either dislocation cores or the surface of the structure.
“If the gate leakage problem can be fixed, it will be interesting to explore how these devices behave in power applications,” says Bolognesi. Another target for the partnership is the optimization of the epitaxial and device fabrication processes. “And of course, we are interested in finding out just how fast we can make these devices go.”
H. Sun et al. IEEE Electron Device Lett. 31 292 (2010)
