Lasers and LEDs that are grown on semi-polar planes deliver very impressive performance at green wavelengths, but commercial success of these devices is hampered by a lack of affordable, high-quality substrates with appropriate orientations. To address this, a German team is developing various methods to make semi-polar material, and studying its properties in detail. Richard Stevenson reports.

Semi-polar planes are promising orientations for making green-emitting structures. Scanning electron microscopy reveals the surface of the wafer containing inverse GaN pyramids with semi-polar facets.  There is no question that GaN LEDs and lasers are a great success. They have backlit billions of screens, they lie at the heart of countless Blu-ray players, and they are driving a revolution in energy-efficient lighting. However, that is not to say that these devices are without fault. In fact, they have several downsides, including internal electric fields that pull apart the electrons and holes in the quantum wells, hampering light emission (see Figure 1). Figure 1. Conventional GaInN quantum wells are embedded in GaN barriers and grown in the c-direction (left). The GaInN quantum well gets compressively strained due to the different lattice constants of the two materials. This leads to an internal piezoelectric field and a tilting of the conduction (CB) and valence band (VB) edges. One consequence of this is a spatial separation of the electrons in the CB (which move to the right) and the holes in the VB (moving to the left). Moreover, the effective band gap is slightly reduced. This effect is called the quantum confined Stark effect. Removing the internal electric field allows the electron and hole wave functions to overlap perfectly, improving light emission from this structure (right). Separation of the carriers by internal fields, which is referred to as the quantum confined Stark effect, has impeded the development of conventional GaN-based green lasers that could be deployed in red-green-blue laser displays. Producing nitride devices that emit at this wavelength requires indium-rich InGaN quantum wells, but the greater the indium content, the stronger the internal electric fields that pull the carriers apart. Internal fields are also bad news for LEDs. One of the weaknesses of this device is droop, a reduction in light-emitting efficiency at higher drive currents. The origin of this mysterious malady is a hotly debated, but its two most popular explanations –Auger recombination and a spilling over of electrons from the quantum well – suggest that internal electric fields are detrimental. That’s because these fields increase the likelihood that electrons will spill over into the hole-emitting region of an LED; and they also prevent efficient operation of devices with wide wells, which enable a reduction in carrier density and lower Auger recombination rates. To overcome the problems associated with these electric fields, some researchers have switched from conventional substrates to those that are described as semi-polar or non-polar. Growing devices on alternative platforms either reduces substantially or eliminates the internal electric fields in the device (see Figure 2). Milestones in the advancement of such devices include: The first reports of non-polar lasers in 2007, independently developed by Rohm and the University of California, Santa Barbara; semi-polar green lasers fabricated by a partnership between Sony and Sumitomo Electric that emit a continuous output above 100 mW at wavelengths longer than 530 nm; and non-polar LEDs announced by Panasonic at the International Electron Devices Meeting that deliver a light output efficiency of almost 40 percent at a current density of 1 kA cm-2. Figure 2. Polarization (left axis) and wave function overlap (right axis) for a GaInN quantum well between GaN barriers. 0° refers to c-plane quantum wells, 90° to non-polar planes. Two low-index, semi-polar planes are marked at about 60° These results highlight the benefits that result from a move from growing devices on semi-polar and non-polar substrates. However, switching to these novel planes pays a heavy price: A hike in the cost of the substrate, which stems from the difficulties associated with making it. GaN substrates that provide a platform for the growth of c-plane devices, mainly lasers, have been available for several years, and prices are falling, with 2-inch material now costing around $1000. These substrates tend to be made by a HVPE process, leading to the deposition of a relatively thick layer of GaN on a foreign substrate, such as sapphire or GaAs. The wide bandgap crystal is subsequently removed and sliced into wafers. Cutting perpendicular to the growth direction yields c-plane substrates, while slicing in other directions produces semi-polar or non-polar material. The downside of this approach is that because it is not easy to grow a very thick GaN crystal, the sizes of the semi-polar and non-polar substrates that are sliced from it are limited. They are typically just 10 mm by 20 mm in size, and sometimes just 10 mm by 10 mm, and they retail for around $1000. In addition to the high cost of the real estate on these planes of GaN, their small sizes are incompatible with wafer processing lines. Together, this pair of weaknesses forms a major barrier to the commercial progress of non-polar and semi-polar lasers and LEDs. Slashing prices Since 2006, a group of German Universities have been collaborating to try and develop low cost foundations for the growth of semi-polar and non-polar optoelectronic devices. Their programme, which is named PolarCoN and is co-ordinated by Ferdinand Scholz from Ulm University, won funding from the German Research Foundation in 2008. Now in its second phase, this €4.5 million project involves, in addition to the University of Ulm, seven other universities: Stuttgart University, Otto-von-Guericke University Magdeburg, TU Braunschweig, TU Berlin, Regensburg University, Freiburg University and Kassel University. “Originally, our main target was the green laser,” admits Scholz, who explains that researchers outside of the project, such as those at Sumitomo, have now succeeded in that endeavour. “We would still like to get such a device – a green laser is always a good demonstrator that you are successful – but we now focus on more basic relations of non-polar and semi-polar material.”  Efforts have been directed at producing high-quality material – principally on semi-polar planes, but also on non-polar planes – and understanding the bandstructure of this material and how its influences electrical transport. During the project the team has looked at the growth of bulk GaN and developed techniques to form semi-polar GaN films and novel lasers on sapphire substrates. In addition, the group has provided new insights into growth conditions on different planes and explanations for the differences in photoluminescence of quantum wells grown on different types of substrate. The most attractive approach to forming semi-polar GaN is to deposit this on r-plane sapphire. “[This] is more expensive than c-plane, but it is still in the range of $70 for 2-inch,” explains Scholz. Producing high-quality GaN on this plane of sapphire is very tricky, however, because GaN tends to be plagued with stacking faults that propagate throughout the material. These stacking faults occur due to local differences in crystalline structure. GaN is a hexagonal material, and its atomic units follow the sequence ABAB… The other common crystalline structure for compound semiconductor materials is cubic: GaAs crystals are an example of this, and their units align in the order ABCABC. Stacking faults arise in GaN when the units have a sequence ABC, which involves a shift in the position of the planes. “For c-plane devices, [stacking faults] don’t matter so much,” says Scholz. “You have a lot of stacking faults close to the foreign substrate, such as sapphire, but they run parallel to the interface and you don’t find them later if the material grows nicely.” In stark contrast, in non-polar material, the planes of stacking faults are aligned perpendicularly to the plane of the wafer, and they will always reach the surface. Similarly, in semi-polar material the planes of stacking faults are inclined at an angle to the substrate, so they also reach the surface. In both cases, no growth technology is currently capable of eliminating these faults. Growth on triangular pyramids Scholz and his co-workers at the University of Ulm have developed a technique that slashes the density of stacking faults through modifications to the substrate. Back in 2004, they were improving the quality of GaN with a well-known technique called epitaxial lateral overgrowth. This begins by taking a sapphire wafer and depositing onto it a GaN film and a SiO2 mask. The wafer is then patterned, with SiO2 selectively removed to create open stripes in the mask, before the structure is put back in the MOCVD chamber. Using optimized growth conditions, GaN stripes with triangular cross-section selectively grow out of these windows. Hence, Scholz’s team have realized pyramid-shaped GaN, which has semi-polar facets but is formed by c-plane growth – so the problems of stacking faults are avoided. This team has gone on to form a range of LEDs on these facets. Early results included 425 nm LEDs producing more than 3 mW when driven at 110 mA. “We later focused on a longer wavelength,” says Scholz. “You get problems with indium incorporation, and the power drops drastically. We are currently trying to improve it.” Another problem with this type of LED is that it is very difficult to optimise its p-type doping. “You can’t easily measure the p-type doping in such structures, because SIMS (Secondary Ion Mass Spectroscopy) and Hall measurements do not work,” explains Scholz. One technique that can be used to identify problems is transmission electron microscopy (TEM), but it can take two months to get results with this approach. However, although TEM doesn’t provide fast feedback, it has played a very important role in characterizing these semi-polar LEDs. It is able to determine the thickness of the quantum wells grown on the facets, and revealed that this is thicker near the apex of the stripe. Compositional fluctuations are also present in the InGaN quantum wells, according to locally resolved high-resolution X-ray measurements, which show increasing indium richness near the apex. Scholz and his co-workers have attributed these variations in the composition and thickness of the wells to gas diffusion effects. It is thought that as the precursor molecules diffuse down to the bottom of the stripes, the effective diffusion length of the indium molecules is shorter than that for the gallium-containing species, and this reduces the growth rate for the well and its indium content. One upshot of these variations within the well is a broadening of the LED output. It’s not clear if this is beneficial for commercial applications, and when Scholz has discussed this with colleagues working in industry, they have been concerned that differences in drive current lead to changes in the colour emitted by this device. Novel lasers What may raise a few eyebrows is that the team from Ulm are also trying to develop semi-polar lasers on these triangular pyramids (see Figure 3). “It seems to be very difficult to think about a laser,” admits Scholz, “but it depends on which kind of laser you hope to realise, and how much stripe material you need. You can consider whether you can realise a laser that runs along the stripe.” Figure 3. Masking and regrowth on the c-plane leads to the formation of triangular pyramids, with semi-polar facets. Researchers at the University of Ulm are trying to develop lasers on these structures. To optimise the design, the dimensions of the stripe should be tailored for waveguiding, and the feature sizes should be reduced to minimize variations in the thickness and indium composition of the quantum well. Judicious selection of the spacing of the structures and their dimensions may also enable them to form gratings for distributed feedback lasers (see Figure 4). Figure 4. Lasers formed on semi-polar facets can have feature sizes that could enable distributed feedback. The team from Ulm are working on that, and now grappling with the problem form good waveguides via high-quality overgrowth of AlGaN on the stripes. “Our goal is to get optically pumped lasers this year,” says Scholz. “We have some optical gain measurements, but they are not that great yet.” In addition to forming triangular pyramids, Scholz and his co-workers have fabricated arrays of hexagonal pyramids by creating hexagonal apertures in the dielectric mask, and then growing the structures out of these holes. This work allowed the team to investigate how the angle of the facet impacts variations in the thickness and composition of the wells. It also led to the fabrication of a more uniform surface of material, which had a greater area of semi-polar planes. One application of such structures is luminescence conversion. A partner in that work Osram Opto Semiconductors of Regensburg, Germany. “[Osram] are interested in getting green light by optically pumping such material with highly efficient blue LEDs. They are currently comparing their polar quantum well structures grown conventionally in c-direction with our semi-polar material,” explains Scholz. From hills to plains From 2008 onwards, the researchers at Ulm have also been trying to form engineered substrates with a flat, semi-polar surface using c-plane growth. “Triangular shaped devices are not liked in industry, because you have to produce contacts on a very fancy surface,” explains Scholz. His team’s efforts have followed in the footsteps of researchers in Japan: Nobuhiko Sawaki’s group from Aichi Institute of Technology, Japan, which have developed flat {1011} GaN surfaces by etching trenches with {111} sidewalls in silicon; and more recently,  Kazuyuki Tadatomo’s team at Yamaguchi University, that produced pure {1122} GaN by patterning r-plane sapphire with 3 µm wide, 1 µm deep stripes running along the in-plane m-direction and separated by 3 µm-wide terraces.  If the growth on these structured substrates were perfect, material would just grow on one type of sidewall, known as the +c-wing (see Figure 5). In this case GaN would grow out of the trench, grow laterally over the ridges separating the trenches and eventually coalesce, creating a flat surface with a semi-polar nature. In practice, however, GaN also grows laterally in the opposite direction after having filled the trench in the wafer thus forming a -c-wing, which is riddled with defects (see Figures 6, 7 and 8). In comparison, the quality of material outside this area is far higher. According to Scholz, material produced by his team has: “in average stacking faults below 104, per  centimetre, and dislocation density below 109 cm-2”. Figure 5. One way to form semi-polar GaN wafers is to initiate GaN growth on c-plane side-walls of trenches etched into sapphire wafers with specific non-c-plane orientations. Figure 6. Transmission electron microscopy highlights the reduction in the crystal quality of GaN in the –c-wing. Figure 7. Cathodoluminescence measurements by a team at Otto-von-Guericke-University Magdeburg show that basal plane stacking faults (BSFs) are only present in the –c wing. Near band edge (NBE) emission is strong outside this region, due to far higher crystal quality. Figure 8. Inserting SiN interlayers improves the quality of GaN. Comparisons with companies.

In 2010, the progress of the German team would have been compared to that of two companies outside of Europe that announced tremendous progress towards the manufacture of semi-polar GaN substrates. In the summer of that year, Ostendo Technologies Inc. and Technologies and Devices International Inc. (at that time part of the Oxford Instruments Group, but now owned by Ostendo) announced their joint development of semi-polar (1122) GaN layers on sapphire substrates. And in November, Sumitomo Electric unveiled its large-scale production of the world's first 2-inch semi-polar/nonpolar GaN substrates for green lasers. This platform had a dislocation density of the order of just 105 cm-2.

Impressive announcements from Sumitomo and TDI could have jeopardized further funding of PolarCoN, but they didn’t.  “In Germany, we are in this good situation where the funders do not kill a project after such a message,” says Scholz, who believes that continued backing of the project was aided by its broad aims: Not only make semi-polar GaN, but to also understand the nature of this material. According to him, there are still opportunities to improve the technologies for making semi-polar and non-polar GaN, and the findings that stem from PolarCoN could benefit companies from Germany or other parts of Europe that may make these materials in future. Although it is now more than two years since these announcements from both Sumitomo and the collaboration between TDI and Ostendo, little is known about the material produced by these companies. “I would think that TDI is just growing on r-plane sapphire,” says Scholz. “Maybe they have found a method to get just one phase of semi-polar material. But there is no scientific publication about that, so it’s very hard to discuss this.” Sumitomo is just as secretive. “From what I see, there is not even a publication from groups who may have used that substrate,” claims Scholz. He and his co-workers are continuing to develop their flat semi-polar substrates, and deposit heterostructures on them. They began with quantum wells, and the strong intensity of the photoluminescence emanating from them encouraged the development of LEDs on this semi-polar platform, which are compared to devices on sapphire. “The intensity of the electroluminescence [from these semi-polar devices] is less than that from c-plane counterparts, which is to some extent due to non-optimised p-doping,” says Scholz, who plans to continue to develop these devices. Growth mechanics One of the biggest questions surrounding the growth of non-polar and semi-polar material is this: How do the deposition conditions compare to those for the growth of conventional GaN? Some groups claim that they are considerably different, but in general terms, Scholz believes that they are broadly similar. “It was a very controversial discussion in our group, and to some extent the results were different in different groups,” explains Scholz, who admits that the details of the growth conditions play a major role in drawing conclusions. He believes that the growth conditions for most facets are very similar, but studies conducted by TU Berlin have confirmed earlier studies done in Ulm that incorporation of indium is markedly enhanced on the {1011} plane, compared to other semi-polar orientations. Another important contribution from a PolarCoN team partner –TU Braunschweig - is an explanation for the substantially shorter emission wavelength for InGaN quantum wells grown on m-plane SiC, compared to c-plane sapphire and c-plane and m-plane SiC. X-ray diffraction measurements reveal very similar levels of indium incorporation for all the wells, and Scholz and his co-workers argue that the shift in emission wavelength stems from growth on SiC. This leads to high levels of stacking faults, which create quantum-wire-like structures. “That means you produce a local different bandgap,” explains Scholz, with emission from this structure depending on the alignment of this bandgap to that of the host material. In this case, a type-II bandstructure results. Funding for this study and other efforts within the PolarCoN project runs until 2015, which gives the team some more time optimise its substrates and devices. Scholz has set his team in Ulm two targets for the remainder of the programme: “One is to produce a DFB laser, making use of triangular stripes, operating at least by optical pumping, if not electrical pumping. And the other is to produce a nice wafer, which means optimising HVPE growth of semi-polar material.” Success on both fronts would be a fitting way to finalise this six-year effort.