Non-polar GaN reaches tipping point

However, despite their tremendous success, these devices have a fundamental flaw – an internal electric field that restricts laser and LED performance. The devices are produced by growing GaN along the c-axis of the wurtzite crystal, leading to spontaneous and piezoelectric polarization within the heterostructure, which causes the electric field. This field pulls apart electrons and holes, so that they are spatially separated within the quantum well. The result is a reduction in recombination rate and light emission, possibly combined with a decrease in internal quantum efficiency. The internal field also means that the emission wavelength shifts with drive current. This can limit the use of GaN devices in certain applications, such as those that involve the precise mixing of red, green and blue light to produce white illumination.

To overcome these issues, the scientific community has been developing structures grown on non-polar planes, such as the m-plane and the a-plane, which are free from polarization-related internal electric fields (see figure 1). This began with hetero-epitaxial growth of GaN on foreign substrates. Unfortunately, these films were plagued with a high density of stacking faults (SFs) and threading dislocations (TDs), which impaired device performance. TDs, for example, act as non-radiative recombination centers, which quench luminescence and limit output power.

Defects are what hampered the performance of the first m-plane LED, which was fabricated by our group in 2004 at the Solid State Lighting and Display Center at the University of California, Santa Barbara (UCSB). Many of these defects would have originated in the HVPE-grown free-standing m-plane GaN substrate, which had a TD density of 4 × 109 cm–2 and a SF density of 1 × 105 cm–1. This 450 nm LED produced just 0.24 mW at an external quantum efficiency (EQE) of 0.43%, when driven at 20 mA.

Higher outputs from non-polar LEDs were reported last year by researchers at Rohm Corporation, Japan, which is a partner of ours. This team fabricated an LED on a dislocation-free m-plane GaN free-standing substrate that operated at the slightly shorter wavelength of 435 nm. It delivered 1.79 mW and an EQE of 3.1% at 20 mA. But even with this improvement, the efficiency of the GaN non-polar LED still trailed that of its conventional c-plane cousin by an order of magnitude.

Recently, however, there has been a tremendous hike in non-polar LED performance. Our devices, fabricated at the end of 2006, have an EQE in excess of 40% and our very latest emitters can deliver an efficiency of up to 46% (see figure 2). The significant jump in efficiency has been driven by improvements in the quality of bulk m-plane GaN substrates. The latest LEDs were fabricated on Mitsubishi Chemical Company s free-standing bulk m-plane GaN substrate, which was produced using chemical and mechanical surface treatments on the exposed surface. Transmission electron microscopy images indicated that these substrates were free of dislocations, and placed an upper limit on the dislocation density of 106 cm–2.

With these superior substrates our LED s output at 20 mA has rocketed to 28 mW, with an EQE of 45%, according to room-temperature DC measurements taken using an integrating sphere (see figure 3 for device details). The EQE peaks at a drive current 10 mA (see figure 4), and slightly decreases at higher currents, which is likely to be due to device heating. The fairly flat response of the EQE over this current range is probably related to the absence of polarization-related electric fields in these structures, but it might also be a consequence of the relatively short wavelength (407 nm) of these emitters.

Laser diodes produced on m-plane GaN have the potential to share many of the attributes enjoyed by their LEDs equivalents. They could also benefit from higher gain, due to the structure s unbalanced biaxial stress and the absence of parasitic waveguides that disturb the beam profile. The lack of strong internal fields within the device also allows thicker InGaN quantum wells to be included in the design, which can provide effective waveguiding of the laser s transverse optical mode.

In this year s February 27 issue of the Japanese Journal of Applied Physics, our group and Rohm independently announced the fabrication of the first non-polar GaN laser. Our initial device was a broad-area gain-guided laser that was driven in pulsed mode and had a threshold-current density of 7.5 kA cm–2. In comparison, Rohm s device had an index-guided ridge geometry and operated in the preferred, continuous-wave mode. The maximum output power for this device was 10 mW, and current densities were as low as 4.0 kA cm–2. Both of the devices were fabricated by MOCVD on Mitsubishi s high-quality low-defect-density GaN substrates, which were undoubtedly critical to these breakthroughs.

More recently, we have started to investigate the performance of non-polar m-plane InGaN/GaN laser diodes that do not contain AlGaN cladding layers, which have the potential to operate at lower threshold-current densities. The removal of the AlGaN layer – needed for optical-mode confinement in conventional structures – frees the laser from the severe constraints associated with c-plane device growth due to tensile strain in the AlGaN layers. The device can then be grown and fabricated in a similar way to the InGaN/GaN LEDs, which ultimately offers a more straightforward method for the manufacture of GaN laser diodes.

To compare the performance of this design with that featuring cladding layers, we have produced three different broad-area lasers with growth conditions similar to those used for c-plane LEDs. Each of these lasers, which share the same basic structure (see figure 5), were fabricated by using lithographic patterning to define a thin metal stripe that provides a current injection area of 15 × 1000 μm2. Reactive ion etching defined the mesa and laser facet; a current blocking layer was added by patterning a SiO2 film; and n-metal contacts and p-metal contact pads were formed by evaporation.

The first device that was built for comparison has a structure similar to a traditional c-plane GaN-based laser diode and has AlGaN/GaN superlattice waveguide cladding layers on either side of the active region. However, this active region contains thicker quantum wells and features five 8 nm InGaN wells separated by 8 nm GaN barriers. This design produced a laser threshold voltage of 11.7 V and a threshold-current density of 7.2 kA cm–2, which is a slight improvement on our earlier design (see figure 6).

The second device differs in only one aspect from the first – the lack of an AlGaN waveguide cladding layer. This change reduces the operating voltage to 7.6 V and cuts the threshold current to 5.6 kA cm–2.

Our final design also omits the AlGaN cladding layer. It differs from the second design by featuring a reduced magnesium concentration in the p-type GaN and a new active region, which consists of three 13 nm thick InGaN quantum wells separated by 8 nm GaN barriers. The result cut the threshold voltage to 6.7 V and the current density to 3.7 kA cm–2. The 411.3 nm lasing peak from this device has a full-width half-maximum of less than 0.5 nm, making it suitable for optical recording.

The lasing characteristics of the unclad devices – which we believe are the first GaN-based laser diodes that do not contain AlGaN cladding layers – are particularly encouraging because they feature uncoated etched facets, rather than cleaved facets. We used ion-beam etching to create smooth, vertical facets and further improve the laser performance. This has reduced the threshold density of our broad-area lasers to 2.3 kA cm–2 using processing steps that deliver a very high yield.

We believe that the threshold current can be further reduced by introducing a ridge-waveguide structure and by optimizing the magnesium-doping levels. When we built a laser that combined the active region of the third device with the magnesium-doping levels used in the first two designs, we produced an emitter with a performance similar to that of the second device. This implied that the laser s performance is predominantly governed by the magnesium-doping level, which impacts optical absorption losses and is not strongly influenced by the quantum-well design. Since this magnesium doping has not been optimized, further improvements in diode performance are highly likely.

Our improvements to the performance of non-polar LEDs, which are mainly thanks to significant gains in m-plane material quality, means that these non-polar devices now match their conventional cousins. We anticipate further refinements in substrate quality, alongside improvements in device optimization, which will propel these emitters and their laser diode equivalents to higher performance levels, and ultimately increase deployment of GaN devices.