500 GHz Transistors Based On GaN … When And How?
To take nitride transistor speeds to a completely new level, researchers must work with novel designs employing either a new pairing of materials or the unconventional nitrogen-face, argue Dong Seup Lee and Tomas Palacios from Massachusetts Institute of Technology.
GaN transistors are uniquely suited for RF power amplifiers and mixed signal applications, thanks to their unique combination of a large breakdown voltage and a high electron velocity. Researchers have discovered how to exploit these attributes over the last two decades and have managed to build devices with better and better performance figures. The fruits of this labour include class-AB amplifiers operating at 10 GHz with efficiency in excess of 70 percent, and power amplifiers operating in the Ka band (40 GHz) delivering an output power density higher than 10 W/mm.
Although the GaN transistor has come a long way, it still has the potential to operate at far, far higher frequencies, thanks to the very high electron mobility associated with GaN – it is at least three times higher than that for silicon. The great speed that electrons can zip about in this wide bandgap semiconductor indicates that is should be possible to fabricate devices operating at several hundred gigahertz or more, which would have many practical benefits.
Such devices could, for example, enable wireless communications with unprecedented speed. They could also revolutionize terahertz imaging by unlocking the door to a new generation of body scanners that reduce the threat of terrorism on aircraft, and they could also lead to advanced chemical and biological sensors to ensure the safety of our food and environment. GaN enabled anti- collision car radar systems could prevent cars from crashing, and efficient point-to-point mm- and sub-mm wave communications would help to eliminate the extensive, complex array of cables and wires that currently connect our computer’s peripherals and media centres.
One of the biggest obstacles facing engineers wanting to fabricate devices operating at unprecedented speeds, such as of 500 GHz or more, is that they cannot work with conventional nitride transistors that employ the pairing of GaN and AlGaN. Conventional transistor structures require AlGaN barriers with a thickness in excess of 20 nm to induce a high enough carrier density in the channel. This thickness has a major downside for high speeds: It leads to a relatively large distance between the gate metal and the channel, and ultimately reduces the ability of the gate electrode to efficiently modulate the channel electrons. Gate recesses have been proposed to mitigate this problem, but the the damage introduced by the gate recess introduces new challenges.
The good news is that there are three alternative architectures that promise to yield transistors operating at ultra-high frequencies: InAlN/GaN heterostructures, which are being investigated by several research teams, including our group at Massachusetts Institute of Technology; nitrogen-face GaN/AlGaN devices; and AlN/GaN structures. In all three cases, researchers are reaching higher and higher speeds through aggressive scaling of the dimensions of the transistor in both lateral and vertical directions.
Such efforts are aided by DARPA, which is funding the Nitride Electronic NeXT Generation Technology (NEXT) programme. US agencies have a good track record in helping to advance GaN technology, and through The Office of Naval Research it previously funded the Millimeter-wave Initiative in Nitride Electronics, Multi- University Research Initiative (MINE MURI). Of these three nascent technologies, arguably the most established is the InAlN/GaN heterostructure, which was first proposed by Jan Kuzmik from the Slovak Academy of Science in 2001.
Transistors made with InAlN and GaN can realize extremely high charge densities, due to the large polarization discontinuity between the InAlN barrier and the GaN channel. However, the benefits of polarization discontinuity are not limited to a reduction in the access resistance of these devices – this architecture also suppresses short-channel effects in deep-submicron transistors. It is also worth noting that this suppression does not require the use of a gate recess process that can introduce plasma damage during the dry etching step.
By taking advantage of these strengths, several groups have been able to attain outstanding results for shortchannel InAlN/GaN transistors. Last year a Swiss partnership between ETH-Zurich and EPFL reported a 100-nm gate length InAlN/GaN transistor with a cut-off frequency (fT) of 144 GHz, and a maximum oscillation frequency (fMAX) of 137 GHz. It did not take long for them to trump that effort and hit an fT / fMAX of 205 / 191 GHz with a 55-nm gate-length device. More recently, we have set a new benchmark for InAlN/GaN transistors, working in close collaboration with the University of Notre Dame and the companies TriQuint and IQE. Our 30 nm gate length device produces an fT of 300 GHz, the highest cut-off frequency ever reported for GaN transistors (see Figure 1).
Figure 1. Three different device structures designed for high speeds and fabricated under the sponsorship of the DARPA NEXT program: a state-of-the-art InAlN/GaN transistor fabricated by Tomas Palacios and coworkers at MIT (a); an N-face transistor fabricated at UCSB (b) an AlN/GaN high-speed transistor from HRL (c)
Fabrication of this record-breaking device stemmed from a combination of gate-length scaling and the introduction of novel technologies that were able to squeeze a few more gigahertz from this transistor. These technologies included vertical scaling of the device, which holds the key to reduced short channel effects, and also led to an increase in the modulation efficiency of the gate electrode and higher frequency performance.
Another feature of our transistor is its barrier thickness of only 9 nm, which leads to a significant improvement in the modulation efficiency of the electrons in the GaN channel by the gate electrode. Our device also contains a 3.3 nm InGaN back-barrier structure underneath the GaN channel that helps to mitigate short-channel effects top barrier scaling. And last but by no means least, we have introduced an oxygen plasma treatment step prior to gate metal deposition that increases the frequency of the transistor by at least 30 percent, due to elimination of transconductance dispersion at high frequencies.
Another device that is showing great promise for submm wave applications is the GaN/AlGaN transistor that has a nitrogen-face GaN structure. This device has several advantages over its conventional equivalent with a gallium-face. One of the biggest benefits results from the formation of the two-dimensional electron gas on top of the AlGaN layer, which occurs because the polarization in this transistor is inverted compared to the standard Ga-face AlGaN/GaN/AlGaN one.
Thanks to this switch in the direction of polarization, a top nitride barrier is not needed – an omission that paves the way to obtaining a very low contact resistance. What’s more, the bottom AlGaN layer that induces the two-dimensional electron gas also leads to excellent channel confinement, a characteristic that enables a high output resistance even in deepsubmicron devices. Unlocking the benefits of these nitrogen-face devices has traditionally been very tough, because it is tricky to grow high-quality GaN with a nitrogen-face. But continued efforts in this direction have recently led to substantial improvements in material quality, and researchers at the University of California, Santa Barbara, have reported some excellent results.
That team has produced N-face devices with a 0.7 μm gate with a power-added efficiency of 74 percent at 4 GHz and fT/ fMAX of 15 / 42 GHz. Devices with a 100 nm gate length have shown a maximum fT of 163 GHz, while the optimization of the gate structure allows an fMAX as high as 310 GHz in transistors with a gate length of 70 nm. These transistors also feature an unalloyed ohmic contact with a contact resistance of just 0.027 ohm-mm, which was formed via InGaN re-growth. This technology has also been used to produce selfaligned devices.
The third class of novel device – a heterostructure transistor formed from the pairing of AlN and GaN – has also been piquing the interest of the wide bandgap microelectronics community, due to its potential for forming highly scaled transistors. Thanks to a very large polarization discontinuity between the two nitrides, such structures can yield a charge density well in excess of 2 x 1013 cm-2, which in turn leads to a room-temperature sheet resistance of 150-180 ohm/-. In addition to this strength, the ultra-thin AlN barriers that are typically used in these devices can significantly mitigate degradation caused by short-channel effects. Devices made from this material system are delivering promising results. Transistors made by a team at the National Institute of Information and Communication Technology in Japan with gate lengths of 250 nm and 60 nm have delivered fT and fMAX combinations of 52 GHz and 60 GHz, and 107 GHz and 133 GHz, respectively. However, even more impressive results are possible by applying a combination of a re-growth contact and back barrier structure. Engineers at HRL have pioneered this approach, and produced a 45-nm gate length device with an fT of 260 GHz and a fMAX of 394 GHz. The latter value is a record for GaN transistors.
The results outlined above showcase the tremendous improvements in all three classes of novel nitride transistor over the last few years. And the race is certainly on to break the 500 GHz barrier, a target that would have seemed far-fetched in the not-to-distant-past. Once that record has been claimed, many will rejoice at the fabrication of transistors that combine speed with great efficiencies and output power levels at mm- and sub-mm wave frequencies that are orders of magnitude higher than what exists today. Hopefully the wait will be a short one.
Figure 2: Small signal, high frequency performance of the InAlN/GaN HEMT produced by MIT. The gate length of this transistor is just 30 nm
© 2011 Angel Business Communications. Permission required.