Raising the bar for power amplification

The performance of GaN-based amplifiers operating in the X- and Ka-bands record new highs through the introduction of an AlN buffer.

BY YANG LING, HAO LU, BIN HOU, FUCHUN JIA, XIAOHUA MA AND YUE HAO FROM XIDIAN UNIVERSITY

In communication infrastructure and radar systems, there is much demand for more powerful power amplifiers (PAs). By increasing the strength of signal through greater amplification, more powerful PAs can open the door to improving the design and the capability of wireless networks, and the resolution and reach of radar.

A compelling approach to increasing the power of the PA is to enhance the buffer design. One option is to switch to compensation doping in the GaN buffer – this boosts the resistance of this layer, leading to a reduction in RF insertion loss, and in turn a hike in the operating voltage of the GaN device.

Figure 1. Improvements in RF performance are realised by introducing an iron-doped AlN buffer (a). This design is superior to the conventional iron-doped GaN heterostructure (b). Note that these illustrations are not to scale.

However, while this approach sounds easy, success is far from trivial. Our team from Xidian University initially tried to succeed in this manner, with efforts beginning by exploring compensation doping in GaN buffers. Through the integration of deep-level impurities such as either iron or carbon, we aimed to significantly increase the buffer’s resistivity, and ultimately the breakdown and power characteristics of the PA. But we came up against a number of hurdles, including the unintended introduction of carbon-related buffer traps and an iron-related doping tail effect – both threaten to degrade device performance. In addition, we found our quest for optimal material quality to be hampered by limitations in thermal management, originating from the thickness of the doped GaN buffer layer.

In response to these challenges, we switched to an AlN buffer. It’s a formidable solution: boasting a high thermal conductivity and an ultra-wide bandgap, this form of buffer has emerged as a beacon of innovation, poised to redefine the standards of PA design. We have deployed this buffer as the foundation for recording-breaking levels of amplification in the X- and Ka-bands.

Figure 2. Secondary ion mass spectrometry provides profiles of the iron, silicon and carbon elements in GaN/AlN/SiC hetero-structures.

Inserting AlN buffers
Fabrication of these devices began by loading a 75 mm semi-insulated SiC substrate into an MOCVD reactor and growing a ultrathin AlN buffer layer at high-temperature, followed by an unintentionally doped GaN channel layer, an AlN insert layer, and a Al_0.25Ga_0.75N barrier layer (see Figure 1).

To uncover the doping profiles in this design, we scrutinised our epitaxial structure with secondary ion mass spectrometry (see Figure 2). This technique revealed a sharp profile for the iron concentration at the interface between the undoped GaN and AlN, indicating the elimination of the iron-doping tail effect, thanks to incorporation of the AlN buffer.

We have investigated the interface between the SiC substrate and AlN buffer with high-angle dark-field scanning transmission electron microscopy and energy-dispersive X-ray spectrometry (see Figures 2 and 3). This pair of techniques revealed a continuous sharp boundary, as well as no element inter-diffusion between the AlN buffer and the substrate.

To produce GaN-based HEMTs with two types of buffer layer structures from our epiwafers, we began by adding an alloyed source/drain ohmic contact, prior to deposition of a 120 nm SiN_x passivation layer. Subsequent electrical isolation resulted from planar nitrogen-ion implantation, followed by the removal of SiN_x passivation under the gate stem and the formation of Ni/Au gate electrodes for the Schottky contact.

Devices with two types of epitaxial heterostructure were produced using the same process flow and device geometry. These HEMTs have a 0.5 µm gate length and a 3.1 µm gate-drain spacing. We have also fabricated transistors for millimetre-wave applications – that’s frequencies of more than 30 GHz – by scaling down the AlN buffer device using the same process platform to produce a HEMT with a 150 nm gate length.