Vertical Integration Streamlines Sapphire Production
It’s great for business to adopt a holistic approach to sapphire manufacturing. When a firm begins with the processing of raw materials and ends with wafer polishing, it enables a trimming of manufacturing costs, the application of proprietary processes to many steps used in sapphire substrate production, and improvements to the reliability of product supply, argues Raja Parvez from Rubicon Technology.
Through vertical integration, Rubicon Technology has been able to scale the growth of bulk sapphire crystal from 30 kg to 85 kg to 200 kg without compromising high quality or high yield
Most companies rely on other firms to provide materials, components, or process technologies. But there is another way – vertical integration. This delivers several benefits in competitive markets, including enhanced cost efficiencies, far greater control of the quality of crucial production inputs, and the ability to provide customers with assurance that their orders will be delivered on schedule.
It is not a new idea to adopt a vertically integrated approach to running a manufacturing business. Back in the 1800s, US Steel tycoon Andrew Carnegie introduced the concept by owning virtually every part of the steel-making value chain, from iron ore through steel mills to the building of railroads. By the 1920s, Ford Motor Company was also employing a vertically integrated approach – it decided to make the steel for its cars. And since then, vertical integration has been applied to almost every type of manufacturing around the globe – including sapphire.
At Rubicon of Bensenville, IL, that’s what we do. We have created the most vertically integrated business model in the sapphire industry to reliably and cost-effectively provide ultra-pure, defect-free material with diameters of 6-inches or more.
Sapphire and solid-state lighting
Our biggest market is that of substrates for the production of LEDs. These solid-state sources are needed for the backlighting of displays and the new wave of energy efficient LED lighting. The backlighting market is much more established: According to Displaybank, in 2012 total LED penetration stood at 41 percent, and it is forecast to reach 95 percent in 2014.
In comparison, the LED lighting revolution is still in its infancy.Market research firm DisplaySearch calculates that the total average LED penetration in general illumination was 1.4 percent in 2010 and forecasts that it will reach 9.3 percent in 2014. Meanwhile, IMS Research estimates that the overall LED market reached nearly $10.9 billion in 2012 with $2.9 billion coming from lighting. By 2015, this analyst projects that the LED market will reach $13.9 billion, with the lighting market nearly doubling to $5.8 billion two years from now.
This mass adoption of the LED will be accompanied by significant reduction in the price of solid-state lighting systems and components. This begins with the LED –including the sapphire wafer and chip.
One significant opportunity for LED chip manufacturers to trim their costs is to transition to larger diameter substrates. According to analysis in 2012 by the US Department of Energy, 15 percent of the cost of the LED package is in the substrate, with the package accounting for a substantial fraction of the cost of the luminaire. If LED chipmakers migrate to larger diameter substrates, they will benefit from operational savings that more than offset the cost of the larger platforms. The lower cost LEDs that result will then help to drive mass adoption of LED lighting for commercial and residential use.
Today, more than 90 percent of LEDs are built on sapphire, with the remainder on SiC, followed by other materials. To date, alternative substrates have failed to offer the performance and cost advantages of sapphire.
One platform that is attracting a lot of attention today is silicon. Proponents of this foundation are attracted to potential operational efficiencies that result from using 8-inch diameter wafers and the opportunity to use fully depreciated CMOS equipment. But these advantages must be weighed against the significant mismatches in the thermal expansion coefficients of silicon and GaN. Addressing this requires a costly, complex buffer layer to mitigate cracking and breakage. The manufacturing yields for GaN on silicon are reportedly still very low, while long-term reliability is unproven. So, until the technological hurdles are solved, silicon might be a niche solution, but is not expected to displace sapphire as the preferred substrate for LED production. Note that for cases where 8-inch diameters are important, we are capable of production volumes of 8-inch polished sapphire wafers.
Figure 1: The proportion of chips that must be discarded due to ‘edge loss’ diminishes as wafer sizes increase.
Another frequently mentioned alternative to GaN-on-sapphire is GaN-on-GaN. However, 2-inch GaN substrates cost $1,000 or more, a hundred times the cost of sapphire. These high costs stem from a very complex fabrication process, and prevent a significant volume of this material from being available.
Given this backdrop, most LED chipmakers are looking to large diameter sapphire wafers to cut costs. Using this particular platform will enable more throughput in each run of the MOCVD reactor, making better use of the reactor ‘real estate’ and ultimately diminish the cost per unit of area processed. In addition, large wafers reduce edge loss, and also provide post-MOCVD efficiencies.
Depending on the type of MOCVD reactor used, LED chip manufacturers using 6-inch wafer platforms can achieve up to 48 percent greater usable area, per reactor run, compared with 2-inch wafers. The overall surface area of a 6-inch wafer is nine times that of a 2-inch wafer, and its outer curvature is less, enabling greater use of the surface area, culminating in a reduction in edge loss (see Figure 1). What’s more, when placed in an MOCVD reactor, there is greater coverage area in the reactor, resulting in further gains that are due to less waste of processing materials. Put all this together and all these efficiency gains associated with production on larger substrates become very compelling when LED chip production ramps up in large volumes to support a high growth market like general lighting.
Leading MOCVD tool makers, such as Aixtron, offer multi-wafer reactors with a variety of configurations. Loading the reactors with 6-inch wafers, rather than 2-inch wafers, leads to more efficient use of gases and pre-cursors.
Our substrates are also being consumed in a second significant market for sapphire – silicon-on-sapphire (SoS) RFICs. Sales of SoS RFIC chips are ramping up, because they combine high RF performance with low power consumption, a small form factor, and significantly reduced crosstalk in antenna applications that are pervasive in smart phones and other consumer devices. Sapphire is highly insulating, which helps the fabrication of devices that are fast, frugal, and offer high levels of isolation. In the last few years, SoS RFICs have gained a significant share of the smart phone antenna chip market, especially in the rapidly growing LTE networks, and they are now being marketed for other applications within these devices.
Vertical integration holds the key to our cost structure and the reliable supply of high-quality products. This integrated approach influences every step in the growth of sapphire crystals and their processing into wafers. Our end-to-end manufacturing capability, with strong intellectual property at each step of the manufacturing process, produces an advantageous cost structure and provides better control of product quality and delivery schedules. Vertical integration is also central to our ability to grow larger and larger sapphire and be the first firm to market with large-diameter sapphire wafers. To date, we have shipped more than 400,000 6-inch wafers.
One of the great strengths of our vertically integrated approach is that we have the materials on hand to meet our customers’ expectations. Production begins with the processing of powdered Al2O3 to yield purified, ‘densified’ material, which is then fed into the furnace to produce a large sapphire crystal or boule. Recently, we started a transition to on-premise processing of Al2O3.
Processing our material in-house gives us greater control of our raw material supply, reduces our costs and enables us to ensure the quality of our starting materials for making sapphire crystals. If we were not able to do this, we would have to rely on commercially purchased crackle, which can be highly expensive and subject to an unreliable, fluctuating supply. Even high purity crackle can be plagued with impurities from transition metals such as silicon, chromium and titanium, resulting in lower quality sapphire boules. For example, the addition of small amounts of titanium can lead to a pink boule - titanium and chromium give the red sapphires found in nature their ‘ruby’ red colour.
The processed Al2O3 produced in-house is fed into custom-built, proprietary furnaces, named ES2-XLG3.0. We make these furnaces for less than half the cost of merchant furnaces. These growth tools, which have all been recently upgraded, are installed in our facilities in Batavia, Franklin Park and Bensenville, Illinois, where we maintain tight control over this valuable intellectual property.
Ourcustomized furnaces are equipped with automation for monitoring all the vital functions and crystal growth rates. This is the key to greater yield consistency. Our proprietary ES2 crystal growth methodology is automated, requiring operator intervention only at pre-set points during the growth process. Thanks to this, it requires less operator intervention than competing methods – with our approach, the operator must be present for less than 10 percent of total cycle time. The primary role of the technician is to initiate crystal growth.
During the last 11 years, we have refined our ES2 growth process that is based on the Kyropoulos method (see “Building on the Kyropoulos method" for details of this growth technology). Our now-perfected methodology involves a top-seeded approach that allows a clear, contaminant-free sapphire crystal to grow unconstrained. Thanks to this, growth is stress-free and defect densities are incredibly low. We employ a master seed crystal, made from one of our own high-quality boules, over and over again to produce consistently high-quality material.
Additional advantages stemming from our furnace design and growth process include: A low thermal gradient; in-situ annealing; minimal bubble formation, due to no crystal rotation; and continual monitoring of the entire growth process, which is possible with a weight sensor. Thanks to this impressive set of attributes, it is possible to routinely yield material with very little stress and a low dislocation density, which is of the order of 10-100 defects/cm2. This is far lower than that found in sapphire formed by either the Czochralski method or the heat exchanger method, which both yield defect densities of 1000-10,000/cm2. High crystal quality with a high yield is even possible when scaling the growth of bulk crystal. The mass of our boules has steadily increased from 30 kg to 85 kg and then to 200 kg.
Processing our crystals into substrates involves high-precision core drilling, wafer slicing, surface lapping, large diameter polishing and wafer cleaning. We have strong intellectual property in many aspects of the finishing processes, including a recently awarded patent for in-situ orientation technology. This allows us to fulfil the differing needs of our customers in the LED, SoS/RFIC and optical markets. These clients all have specific, distinct requirements for the crystal planar orientation of the sapphire products used in their applications. Two of the key benefits of our new orientation process are enhanced precision associated with sapphire planar orientation, and the elimination of time-consuming steps, because orientation is performed at the fabrication tool.
Recently, we have also been awarded a patent for our lapping technology. This is associated with platens that are continuously self-conditioned and self-optimized to maintain high performance. Efficiencies resulting from this technology will translate into savings for our customers.
In early 2011 we opened a facility at Penang, Malaysia, that is used to carry out most of our finishing processes. Operating from this location helps to reduce costs, and it is near to LED chip manufacturers in Asia. We believe that this facility will enable us to reduce the costs associated with making a 6-inch wafer by 20 percent – so far, we have got halfway towards that goal, and we expect to hit this target by summer 2013.
Patterns for the future
Engineers in our research and development team are now focusing on the fabrication of patterned sapphire substrates. Today, many LED chipmakers etch a pattern into the sapphire substrate, prior to MOCVD growth, to increase light extraction efficiency from this device.
This year we will give our customers the opportunity to eliminate this step by doing it for them. If they take up this opportunity, they can devote more time to their core technology. Patterns etched into the sapphire vary from LED makers to LED maker, so we will offer them their own bespoke design, a step that will bring us closer to them. Patterned substrates are already available from several suppliers, but wafer sizes tend to be small. In contrast, we will focus on the production of large diameter substrates, which are more challenging to process. This means that our provision of them adds greater value.
Our move to providing patterned sapphire is in keeping with our general philosophy: As sapphire and LED chip manufacturing evolve, we will continue to invest in capabilities that provide ever-increasing value to this industry. We already employ industry-leading expertise at every stage of sapphire production, which begins with raw materials and goes right through to wafer finishing. This vertically integrated approach, working in tandem with a deep understanding of our customers’ manufacturing processes, delivers two major benefits to our business: It contributes to excellence in quality and reliability; and it draws us closer to our key customers.
Building on the Kyropoulos process
Rubicon employs a sapphire growth process based ofthe Kyropoulos method. This well-established technique involves placing pure alumina raw material in a crucible and heating it so that it melts. The sapphire crystal is formed on the seed, from the top down. The crystal solidifies as it is very slowly pulled up, with the thermal gradient tightly controlled, to form a free and natural shape as the crystal grows unconstrained.
The Kyropoulos technique is ideal for materials with low thermal conductivity and a high degree of thermal expansion, the combination of which can make crystal material vulnerable to various imperfections unless grown and cooled in a low-stress environment. With this highly controlled thermal-gradient, the Kyropoulos method yields large-diameter boules of very high optical quality due to high purity. The resulting boules can be cut to any crystallographic orientation or plane.