Revolutionising RF oscillators with GaN

GaN IMPATT diodes can enable compact and efficient solid-state oscillators that deliver higher powers in the millimetre-wave and beyond.

BY ZHENGLIANG BIAN AND SRABANTI CHOWDHURY FROM STANFORD UNIVERSITY

To advance modern RF applications, much effort is directed at realising a higher RF power within a compact form factor. There are a number of solid-state devices that can do this, but the most compelling for efficient high-RF-power generation is the impact ionisation avalanche transit time (IMPATT) diode. This form of diode, which is capable of delivering an exceptionally high output power from a few gigahertz to hundreds of gigahertz, can replace elaborate multiplier and amplifier chains.

While you may not be that familiar with the IMPATT diode, it is certainly not a new device. Its inception dates back to 1958, when W.T. Read, working at Bell Laboratories, proposed a simple p-n structure that consists of a narrow avalanche carrier multiplication region followed by a drift region (see Figure 1). Read theorised that a 180° phase shift occurs with the applied voltage, due to the combination of the time delay associated with avalanche multiplication and the transit time of the injected carriers traveling through the drift region. This phase shift gives rise to a negative resistance that contributes to RF power generation.

Figure 1. (a) The design of the Read-type IMPATT diode. (b) Voltage transient at the cathode for an IMPATT under oscillation. (c) A carrier generation waveform in the avalanche region. (d) The induced terminal current for an IMPATT under oscillation.

In 1965 the first IMPATT oscillation was observed, in a device made from silicon. This triumph laid the groundwork for significant advances in IMPATT technology over the following decades, including the deployment of this device in satellite communications and radar systems. Today, a single commercial IMPATT diode, mostly realised with silicon technology, is capable of delivering 400 mW under continuous-wave mode and 20 W under pulsed-mode operation at 94 GHz.

Embracing GaN’s benefit
Despite the success of the silicon IMPATT diode, the quest remains to develop higher-performance devices that draw on current RF technology. Throughout the evolution of the compound semiconductor industry, new materials have played a pivotal role in pushing the boundaries of device performance. It is a state-of-affairs that is no different with the IMPATT diode, where GaN in particularly is reviving interest in this device.

Due to its tendency to be deployed as an oscillator, the primary performance metric for the IMPATT diode is its power-frequency trade-off. Analysis reveals that both the output power and the oscillation frequency of an IMPATT diode are mainly determined by the critical field and the saturation velocity in the base material. The output power of the diode is proportional to both the square of the critical field and the square of the saturation velocity, and inversely proportional to the square of the frequency. Based on these relationships, GaN clearly has the upper hand over silicon (see Figure 2), thanks to a critical field that is 10 times higher, and twice the saturation velocity. Combining these attributes indicates that the GaN IMPATT diode has a 400 times higher power capability than its silicon counterpart at a given frequency.

Figure 2. Material properties for various candidate materials for IMPATT oscillators, and a power-frequency benchmark plot.

Another promise of the GaN-based IMPATT diode is that it is projected to have a much higher cut-off frequency – it is in the terahertz range – due to its shorter avalanche response time. This makes the GaN IMPATT diode an attractive solid-state terahertz source that could enable various novel applications, such as high-resolution imaging through opaque materials for medical diagnosis and security screening, and extreme wideband wireless communications.

Current status
Unfortunately, engineers have struggled to convert the tremendous promise of the GaN IMPATT diode into a high-performance device. But the good news is that this situation is starting to change. Helping to facilitate this are improvements in producing low-defect-density bulk GaN substrates, high-quality epitaxy growth, and effective edge terminations – all these advances are crucial to routinely realising avalanche in GaN p-n diodes.

Initial efforts with GaN involved embedding a p-n diode in a series resonate circuit and observing an 800 MHz oscillation. This weak, low-frequency oscillation has been attributed to limitations associated with significant device self-heating and low-quality oscillation circuitry. Improvements were wrought by placing the diode on a copper heat sink and testing it under pulsed mode in a waveguide resonate cavity. This recent work, involving a GaN IMPATT diode, enabled a peak power of 30 dBm to be extracted up to 21 GHz.

While these results are encouraging, as well as failing to unlock GaN’s potential, they don’t even exceed the performance of their silicon counterparts. Why is this? It’s because GaN is held back by immature fabrication techniques. In sharp contrast, the performance of silicon IMPATT diodes draws on refinements in device fabrication and packaging procedures that result in a robust avalanche, minimised series resistance, and adequate heat removal – all are crucial.

With GaN, the fabrication processes for excelling on these fronts are still immature. To address these weaknesses, our team at Stanford University is pursuing the following pair of goals: the development of a substrate thinning process that results in 20 µm-thick wafers that do not lead to degradation of avalanche in the fabricated diodes; and flip-chip packaging to a type IIA single crystalline diamond heat sink, to ensure enhanced heat dissipation.