Parallel Processing Boosts GaN Throughput
GaN lasers perform at their best when grown on a flat native crystal that combines low dislocation densities with a high enough free carrier concentration to ensure a strong refractive index contrast between device and substrate. Making such a foundation in reasonable volumes with acceptable growth rates is tough, but we believe the most promising approach employs high nitrogen pressure and liquid gallium to form very high quality, free-standing crystals on an array of HVPE-grown seeds, says Michal Bockowski from the Institute of High Pressure Physics, Polish Academy of Sciences, in Warsaw.
If you want to make Blu-ray lasers with high yields, fabricate blue and green LEDs that set a new benchmark for brightness or produce incredibly high performance electronics, then you should build your device on a GaN substrate. Such a foundation has many benefits: So long as the defect density in this crystal is low, it should be possible to deposit nitride films with high material quality; there should be no bending and cracking of the material in the device, because differences in lattice constants and thermal coefficients of expansion between substrate and epilayer are low; and the device will benefit from the great set of intrinsic characteristics associated with the wide bandgap material that underpins it. GaN also has great thermal conductivity, an asset for all classes of device; and it is also great at dissipating heat, which helps to bolster the performance of electronic devices.
The major drawback of GaN is its high cost: A 2-inch diameter substrate retails for several thousand dollars. This high price stems from the difficulties associated with forming high-quality crystals of GaN. Melting the material requires extreme conditions – nitrogen pressures of 6 GPa and temperatures of almost 2500 K – a combination of conditions that prevents the growth of GaN from its stochiometric melt, which is the method used to make silicon and GaAs boules.
A handful of alternative methods have been developed for making GaN crystals, including HVPE deposition on foreign substrates and the growth of GaN in solution. All these approaches have their weaknesses – either in material deposition rate or crystalline quality. Minimizing these drawbacks holds the key to making affordable, high-quality GaN, and at the Institute of High Pressure Physics we believe that we have developed a novel approach that can do just that: It involves using a high nitrogen pressure solution to convert multiple GaN seed crystals grown by HVPE to free-standing GaN. Reliable blue and violet lasers have been formed on these highquality crystals.
HVPE: Pros and cons
Today HVPE is the most common approach for manufacturing GaN substrates. This involves crystallization from the vapour phase at ambient pressure, with GaN deposited on a foreign substrate through the reaction of ammonia with gallium chloride at temperatures of about 1300 K. Etching, laser lift-off and self-lift-off techniques can all be used to remove the nitride film from the foreign substrate – typically sapphire or GaAs – and yield a large-diameter, freestanding GaN substrate.
This technique has a relatively fast growth rate of up to 500 μm per hour, but suffers from a phenomenon known as parasitic nucleation. Superfluous GaN nucleation takes place within the reactor, often leading to uncontrolled changes in crystal growth conditions during the crystallization run, such as variations in the flow rate of reactants. Reducing growth time prevents degradation to the crystal, but this limits the thickness of HVPE-grown GaN to typically below 1 mm.
Another drawback is that the material quality degrades, due to either introduction of donors, such as silicon or germanium, or the addition of an acceptor, iron. This makes it difficult to produce highly n-type or semiinsulating substrates, thereby limiting the typical freecarrier concentration for free-standing GaN to 1018 cm-3.
Deposition on a foreign substrate enables the growth of large-diameter GaN crystals, but these suffer from lattice bowing. This stems from significant differences between the lattice constant and thermal expansion coefficient of the foreign substrate and the nitride film (see Figure 1). When GaN is grown by HVPE on sapphire, the bowing radii of crystallographic planes is below 10 m. This relatively low number means that there is little benefit in using HVPE-grown GaN as a seed for subsequent crystallization runs. It is possible to overcome these issues and grow crystallographically flat, free-standing HVPE-GaN by switching the growth mode during the HVPE process from a flat one to a rough one. But there is a penalty to pay: inversion domains that hamper subsequent epitaxy (see Figure 2 for details).
The best HVPE ‘free-standing’ GaN technology has been developed by Sumitomo Electric Industries. This Japanese firm can produce very good quality, freestanding GaN crystals of up to 6 inches in diameter via deposition of this wide bandgap semiconductor on GaAs wafers. The curving of GaN is not that severe, thanks in part to the similar thermal expansion coefficient of this material and GaAs. In addition, Sumitomo’s growth process aids the fabrication of crystallographically flat substrates – it is based on selective growth of the nitride, followed by re-growth on the surface containing large inverse pyramidal pits (a process described as either Dislocation Elimination by Epitaxial growth with inverse-pyramidal Pits (DEEP), or an Advanced variant known as A-DEEP). With this method the crystal is grown in the controlled rough growth mode, probably with presence of the inversion domains.
Sumitomo’s crystals feature 400 μm wide stripes that have alternating regions with defect densities of typically 104 cm-2 and 5 x 108 cm-2. High-quality lasers for Blu-Ray players are manufactured on these crystals by positioning the chips on low defect density stripes that are free from inversion domains. Due to the growth and re-growth of pits, the free-carrier concentration in Sumitomo’s substrates is relatively high – about 5 x 1018 cm-3. With HVPE-based approaches, impurities are built into GaN in an anisotropic way (see Figure 3), and the growth of pits occurs in semi-polar directions.
When sapphire is used instead of GaAs as a foundation for making GaN free-standing substrates by HVPE, the best results are obtained with a flat-growth mode and a technique known as Void Assisted Separation (VAS). With this approach that has been pioneered by Hitachi Cable, growth proceeds on a sapphire substrate coated with an ultra-thin layer of MOCVD-deposited GaN and nanometric titanium nitride. 3-inch GaN substrates with a homogeneous dislocation density of about 106 cm-2 can be formed by this technology. Free -carrier concentration is typically 1018 cm-3 and the lattice bowing radius is below 10 m.
Solutions for GaN?
It is also possible to form GaN crystals from solution in supercritical ammonia. This approach, which is known as the ammonothermal method, is analogous to hydrothermal crystallization of quartz or oxide crystals such as ZnO. However, ammonia is used in the place of water. One of the biggest drawbacks of the ammonothermal approach is that it is very slow: Growth is at best 0.1 mm per day. It is also necessary to secure high-quality GaN seeds that can initiate growth and find a suitable ‘mineralizer’ to aid dissolution of GaN.
The world-leader in this technology is our neighbour, the well-known Polish company Ammono. This firm manufactures 1 cm2 and 1 inch crystals and is on the road to start the production of 2 inch variants. These crystals have many great attributes: They are extremely flat, with bowing radii of the crystallographic planes reaching up to 100 m; defect density is of the order of 104 cm-2; and free carrier concentration does not exceed 2 x 1019 cm-3. This material has already provided a good foundation for making high-power lasers by various groups, including the Polish company TopGaN.
Another way to grow GaN is from solution, using gallium-sodium mixtures held at temperatures from 700- 900 °C and a nitrogen pressure of up to 5 MPa. Osaka University has trail-blazed this approach, and produces bulk single crystals that are a few millimetres thick, have a diameter of up to 3-inch, and exhibit a defect density of the order of 105 cm-2.
Piling up the pressure
But if low defect density is the primary goal, then by far the best approach is the High Nitrogen Pressure Solution (HNPS) method. Dislocations of just 102 cm-2 can be realized via a direct synthesis reaction between a liquid of gallium and nitrogen held at up to 1800 °C, and nitrogen at pressures of up to 2 GPa. A spontaneous reaction yields hexagonal platelets typically 1 cm2 in size, which make a good foundation for laser diodes. We are the pioneers of this growth technology, and our experience of making lasers on these hexagonal platelets in conjunction with our spin-off company, TopGaN, has enabled us to determine the most important characteristics for GaN substrates that are used for making laser diodes. Our findings are that GaN crystals must have: High structural quality, including a bowing radius exceeding 20 m and a dislocation density below 106 cm-2; and high electric conductivity with freecarrier concentration of more than 5 x 1019 cm-3, because this enables the preparation of a lowresistance ohmic bottom laser diode contact and also makes the GaN substrate ‘plasmonic’, aiding optical confinement (more about this later).
The HNPS approach yields GaN of great quality, but crystal size is small and material throughput is low. So to address these issues, we have recently developed a method for growing GaN that we call multi-feed-seed (MFS). This involves the conversion of free-standing HVPE-grown GaN crystals to free-standing HNPS GaN, which has a much higher quality than the seed material. The great strength of this approach is that it yields several GaN crystals from one run – what’s more, all of these crystals satisfy our criteria for laser manufacture.
Production of this material begins by positioning several (0001)-oriented HVPE-GaN seeds in a vertical stack, separated by liquid gallium (see Figure 4). The lowest seed is placed above the bottom of the crucible and the distance between individual seeds can be varied. Under nitrogen pressures of typically 1 GPa an axial temperature gradient is applied along the crucible, leading to overgrowth of seeds on their (0001) surface. On each seed, liquid gallium is dissolved from its lower (0001) surface, and atomic nitrogen is supplied into the solution, where it is transported to the underlying crystal. In other words, each seed is overgrown and dissolved at the same time, but at a slightly different temperatures varying from 1420 °C to 1450 °C.
The upshot of all this activity is the formation of stable, macroscopically flat crystals (see Figure 5 for a typical example). Pinholes are visible on the bottom (nitrogen surface) of the seed, but absent on the surface of the pressure-grown material. The temperature gradient that’s applied during the process determines the growth mode and surface morphology.
If the temperature gradient is large, growth rates of 5 μm/hour are possible, leading to macrosteps on the surface, plus voids and gallium inclusions within the crystals. Reduce this gradient and the density of voids and inclusions decreases from 103 cm-2 to 100 cm-2, growth rate falls to 1 μm/hour or less, and hillocks are formed on the surface. GaN crystals made by our MFS process are far better than those made by HVPE. However, this can only be appreciated after residual HVPE-GaN is taken away from the nitrogen surface of the crystal. Benefits include an increase in the bowing radius from 2 m to 30 m, which results from the removal of the HVPE substrate, a step that helps to release stress in the HNPS-grown GaN.
We have fabricated 500 μm-thick GaN using our MFS technology. Lasers made on this platform, which was formed after 500 hours of crystallization emit at 390 nm to 440 nm, have a typical threshold voltage of 4.5 V at 2.5 kA/cm2 and have a lifetime of up to of 5000 hours. Defect-selective etching reveals that the dislocation density in the laser diode is below 5x106 cm-2. It is about one-fifth of that value in the substrate.
One way to improve laser performance is to increase the refractive index contrast between device and substrate, because this suppresses optical mode leakage into the substrate and optimises the optical mode in the active region. It is possible to do this with our ‘plasmonic’ GaN substrate, which has a free carrier concentration up to 7x1019 cm-3, and is produced using growth temperatures of 1440 °C (see Figure 7). This approach is superior to using a conventional GaN substrate and adding a thick AlGaN layer or an AlGaN super lattice beneath the active region to prevent light propagating from here into the substrate. Go down this more common route and strain increases in the epistructure, leading to macroscopic bowing, cracking and the creation of misfit dislocations.
The only major weakness of our approach is its slow growth rate. However, we believe that this can be increased to 50 μm/hour by increasing nitrogen solubility in gallium through the addition of sodium or lithium impurities into the solution. We plan to try this in the coming months.