Technical Insight
Boosting brightness with hollow cavities
GaN-based LEDs are more efficient and cost-effective when they contain an array of embedded hollow cavities in the sapphire substrate
BY DUKKYU BAE AND YONGJO PARK FROM HEXA SOLUTION
The LED industry is in the doldrums. There is a worldwide oversupply of LED chips and penetration is sluggish into the latest "˜killer application', general lighting. Compounding the matter, the other major multi-billion dollar market, backlighting units for screens, is failing to grow at a healthy rate. As those within the industry increase LED performance to give their company an edge of their rivals, the number of chips required to backlight a screen falls "“ and with it, ironically, a fall in the volumes within this market.
Fortunately, there is some good news: the market for high-power LEDs, which typically produce 2-3 W, is expanding, and should continue to do so. With these devices − that are used in industrial lighting applications, such as fixtures in factories that are attached in high ceilings "“ a premium is placed on high efficiencies. That's because the higher the efficiency, the lower the heat dissipation, and ultimately, the simpler and cheaper the thermal management.
Given the state of the LED industry, it is of no surprise that many chipmakers are striving to increase the efficiency of their devices, and enjoy more success in the high-power sector that offers a higher profit margin. Even if these device makers only outperform their peers by a few percent, their superior efficiency will give them a considerable edge. This positions them to make good money, because a substantial proportion of profit margin comes from high-end LED chip production.
Another advantage of being able to produce very high efficiency devices is that if they also have a proven reliability, they can challenge for success in the automotive headlight sector. Where they will not have success, however, is in general lighting. Here, medium and low efficiency LEDs are being used, with the many makers of these devices fighting for survival.
Figure 1. The manufacturing process for forming cavity-engineered substrates involves complete crystallisation of amorphous alumina, so that it turns into sapphire. This greatly simplifies subsequent growth of GaN.
A common approach for overcoming this issue is to produce devices on patterned sapphire. A corrugated surface can scatter photons in many directions, increasing the probability for light to escape from the chip.
These substrates are formed by taking a piece of planar sapphire and etching a pattern into it with a photoresist mask and an inductively coupled plasma. With this process, the mask gradually erodes, leading to a dome-shaped surface. If cones are preferred, they can be formed by adjusting the etching conditions.
When light propagates through LEDs formed on patterned sapphire, it is reflected and diffracted, due to differences in the refractive index. The propagation of light undergoes the greatest change at interfaces with the most substantial difference in refractive index, which occurs at boundaries between GaN and air. Due to this, many research groups have tried to incorporate high-index-contrast cavity structures within their LED. This has been successful, with light extraction increasing, but up until now there have been no reports of the use of a production-level, cavity-forming process within the LED industry.
One of the research groups that have made a considerable contribution to cavity-incorporating LEDs is the team of Euijoon Yoon and co-workers from Seoul National University, Korea. After devoting several years to the fabrication of LEDs with nanometer-scale cavities, formed using hollow silica nanospheres, they have increased light extraction by incorporating a cavity within an LED. What's more, they have reduced wafer bow, thanks to relaxation of compressive stress in the GaN around the cavity. The implication is that thinner sapphire wafers can be used for LED manufacture, trimming production costs.
There is still room for improvement, however. The increase in light extraction is not as high as it could be, due to the low density of the hollow cavities and their random positions. These weaknesses must be addressed in order to create a better, more reliable cavity-forming process for LED manufacture.
From lab to fab
Addressing these weaknesses is one of the challenges that we have set ourselves at Hexa Solution Co., Ltd. of Korea. Our core technology is the cavity-engineered sapphire substrate, which we plan to sell to LED chipmakers. Several of them are currently evaluating our products, which they could either buy from us in volume, or produce themselves after licensing our technology.
Figure 2. Scanning electron microscopy reveals embedded cavities of hexagonal domes formed after GaN growth.
Our technology originates in Yoon's group, but has been refined in a partnership between that team and another research group at the same university "“ the latter is based in the Energy Semiconductor Research Center, Advanced Institutes of Convergence Technology. Together, these two groups have carried out a feasibility study, with lab-scale production revealing that LEDs incorporating cavity-engineered sapphire substrates outperformed equivalents made on patterned sapphire.
In 2014, to spur its commercialisation, the cavity engineered sapphire substrate technology was transferred to our company. By the start of the following year, we had successfully prototyped 2-inch cavity-engineered substrates for initial pilot plant production, and since then we have made further progress. Highlights include shipment of 2-inch, 4-inch and 6-inch products to LED chipmakers throughout the world. They are currently evaluating our technology.
Our process for producing cavity-engineered sapphire substrates is robust and scalable (see Figure 1). It begins with photoresist patterning, which allows the shape of a cylindrical photoresist to be easily tailored to that of a dome by a reflow process. This is followed by atomic layer deposition of an 80 nm-thick amorphous alumina layer at 120 °C over all the exposed surface. The alumina partly covers sapphire, and partly covers the photoresist.
The next step is thermal treatment in an oxidation atmosphere. This allows in-diffusion of oxygen through the porous alumina layer and out-diffusion of the oxidation by-product. The result: the creation of dome-shaped cavities.
At the same time, the amorphous alumina becomes completely crystallized. This starts at the sapphire-contacting area, and is completed at the apex of the dome over the photoresist. A very pleasing aspect of this process is that the crystalline phase of amorphous alumina is sapphire. Consequently, since alumina is crystallized to sapphire during thermal treatment, there is no need for additional processing steps to expose the sapphire seed layer and initiate GaN growth (see Figure 2 for a cavity-engineered sapphire substrate with a two-dimensional hexagonal cavity array, and a preserved cavity after GaN growth).
A great strength of our cavity-engineered substrates is their unique optical properties. They produce strong interference, leading to a range of colours that can be seen by the naked eye. Compared to patterned sapphire, more vivid interference colours are produced under both incandescent and fluorescent illumination, thanks to strong diffraction by the cavity (see Figure 3).
Figure 3. Stronger diffraction in cavity-engineered sapphire, compared with patterned sapphire, leads to brighter colours under illumination.
Experimental and theoretical investigations into the strong diffraction produced by our substrates have been undertaken by Sun-Kyung Kim's research group at Kyung Hee University, Korea. Their work included careful transmission experiments, which revealed that cavity-engineered sapphire produces higher transmission than patterned sapphire over a wide range of wavelengths.
A key insight provided by the finite-difference time-domain simulations of the team of Sun-Kyung Kim is that the high-index-contrast cavity interplays very strongly with the in-coming plane waves. The upshot is that the cavity is effective at changing the propagation direction of the light (See Figure 4).
Figure 4. Finite-difference time-domain simulations show that there is stronger scattering in cavity-engineered sapphire than patterned sapphire. Plane waves with an incident angle of 0° (top) and 45° (bottom) are used in the simulations.
Another consequence of the air cavity is that it acts as an optical disturbance against the plane waves, imparting significant phase distortion inside and outside the structure. The array of embedded air cavities is able to extract more light out of the LED, because each individual air cavity generates a strong secondary wave, according to Huygens' principle. Light generated within the LED interacts strongly with the high-index-contrast cavity array, so, rather than being trapped by total internal reflection "“ as it would be in a conventional device "“ it leaves the chip, due to interaction with the periodic array of cavities. Or to put it another way, the light is diffracted (see Figure 5 for finite-difference time-domain simulations that clearly visualize that the cavity-engineered sapphire substrate is better than patterned sapphire at extracting light out of the chip, particularly under total internal reflection conditions).
Figure 5. Finite-difference time-domain simulations show that LEDs incorporating cavity-engineered sapphire are more efficient at extracting light than equivalents with a patterned sapphire foundation.
Proven performance
To demonstrate the superb performance of lateral and flip-chip LEDs on cavity-engineered sapphire, compared with those formed on patterned sapphire, we have partnered with Semicon Light Co., Ltd. Korea. Together we have simultaneously produced device epistructures on the two types of substrate in an MOCVD reactor. Following this, both wafers were processed concurrently to produce large-area, lateral-type blue-emitting InGaN/GaN LEDs with dimensions of 1075 μm by 750 μm.
The features in both types of sapphire were markedly different. The cavity-engineered sapphire contained hemispheres 1.5 μm high and 2.4 μm in diameter, while the cones on the patterned sapphire were 1.7 μm high and 2.7 μm in diameter. LED chips were diced from these wafers (see Figure 6), with individual chips mounted on the same package.
Figure 6. Lateral LEDs, with dimensions of 1075 μm by 750 μm, incorporating cavity-engineered sapphire technology.
Measurements of optical power, using a standard integrating sphere, showed that the power produced by LEDs on cavity-engineered sapphire were typically significantly higher than those made on patterned sapphire (see Figure 7). Collecting spectra of individual LED chips, driven at 240 mA, revealed that those made on cavity-engineered sapphire deliver peak emission at 468 nm, and produce optical power almost 40 percent higher than variants formed on patterned sapphire. Devices grown on cavity-engineered sapphire that emit dominant wavelengths of 456 nm and 462 nm also outperformed cousins on pattern sapphire, but by a smaller margin (see Figure 8).
Figure 7. The optical power distribution of batches of LEDs produced with cavity-engineered sapphire and patterned sapphire technologies.
Figure 8. LEDs incorporating cavity-engineered sapphire produce more power than those formed with patterned sapphire at the three dominant wavelengths of 456 nm, 462 nm and 468 nm.
A great virtue of our cavity-engineered technology is that the surface of our substrates is made entirely of sapphire. This means that the chemical reactions associated with MOCVD growth are identical to those for planar and patterned sapphire. Although a slight adjustment in growth temperature might be needed to optimise device performance, no significant alternation to the manufacturing process is required to adopt our technology, and to produce more competitively priced LEDs.
Lower prices, compared to LEDs made on patterned sapphire, stem from a lower-cost process. Patterning sapphire by plasma etching is relatively expensive, due to the incredible hardness of sapphire, which leads lengthy etching times under harsh conditions. Making matters worse, yields for patterned sapphire are unsatisfactory. In comparison, a batch of cavity-engineered sapphire substrates can be mass-produced using a combination of a more stable, less expensive atomic layer deposition process and thermal treatment in a furnace.
We are confident that in the next few years our cavity-engineered sapphire substrates will form the foundation for a new generation of high-performance LEDs. We have proven that they can deliver a significant increase in light extraction in lateral LEDs, and we anticipate that the benefits will be even greater in more powerful flip-chip designs. Here the distance between the bottom mirror and the active layer is comparable to the wavelength of light, so it is possible to exploit resonant cavity effects. We are exploring this opportunity with Semicon Light Co., Ltd. through a joint development of silver-free, flip-chip, cavity-engineered sapphire LEDs.
Other exciting opportunities of our technology include the fabrication of ultrathin sapphire membranes. These could be used for less defective GaN growth and the mechanical lift-off and bonding of individual processed, dicing-less LED chips on to arbitrary substrates, including those that are flexible. Our technology is clearly in its infancy, and has a great deal to offer in helping to rejuvenate the LED industry.
