producing a strain at the interface between the substrate and overlayer that can result in a curvature or “bowing” of the wafer. The curvature is accentuated by the mismatch in lattice constant between the sapphire and the LED overlayers (see section below). This curvature in turn aggravates non-uniformity in temperature and the flux of materials seen by each wafer. The most important issue for the MOCVD growth of nitride LEDs is control of the growth temperature. Wafer temperature is important for determining the growth rate as well as the composition of the indium-containing layers in the structure. Even small changes in temperature are enough to change the optical emission (color) of the LED.
The technology that has been developed to accomplish monitoring of MOCVD is based largely on optical reflectivity. Early technology development was supported by DOE through Sandia National Laboratories (Sandia National Laboratories, 2004) and has resulted in a U.S.-based start-up company, k-space Associates, that manufactures MOCVD monitoring products. There is a similar company in Europe, LayTec, which manufactures similar technology. Both k-Space1 and LayTec2 have developed specific systems, which can in “real time” monitor either the wafer temperature or the growth rate or the curvature (hence the strain) of the growing layer. However, a complete picture of the state of the reactor, and therefore of the material being grown, requires information about all of these parameters. It is possible to incorporate several of these monitors into a single reactor, but this requires the careful design of the reactor to best accommodate and integrate the monitors. In addition, careful cross-calibration of the monitor outputs with the actual grown materials is required in order to extract meaningful data about the monitored growth conditions and the actual material characteristics. There are currently no systems, which can directly monitor the composition of the growing film. Initial efforts have been made to implement many of systems with feedback reactor control (Haberland, 2008). However, for commercial reactors only wafer temperature has been implemented as part of the control loop.
Current sapphire substrates are 4 inches in diameter. As the production reactors scale up to 8-inch-diameter sapphire substrates, the interrelated problems of substrate temperature, wafer bow, and change in materials composition can only get worse. In order to achieve the desired wafer yield, optical monitoring needs to continue to develop. Goals for future optical monitoring system include the following: (1) integration of growth rate and wafer bow measurements with reactor control in commercial systems, (2) development of tools to make real time composition measurements on the growing layers, and (3) development of tools that will provide full wafer maps of the important parameters (temperature, composition, strain, and growth rate).
FINDING: Production-scale MOCVD growth of LEDs is a complex process. The uniformity and yield of the structures grown (and hence of the optical performance of the LEDs) is strongly and negatively affected by small variations in the MOCVD growth process. The thermal and lattice mismatch between substrate and overlayer exacerbates the sensitivity of the growth process. Further difficulties of growth control are anticipated with use of substrates with increased diameter.
RECOMMENDATION 3-3: The Department of Energy should fund research to develop instrumentation for in situ monitoring and dynamic control of the metal organic chemical vapor deposition growth process.
Epitaxial growth processes work best when the substrate (e.g., sapphire) that serves as the template for the material growth has a structure (lattice constant) that matches that of the finally formed material. Without the one-to-one registry of the overgrown material to the template, there will be a strain in the overlayer that may eventually give rise to dislocations and defects in the material (107 to 108 cm-2 dislocations in the best case for GaN on sapphire). Such defects will compromise the performance and reliability of the devices formed from the material, and this in turn leads to “uncertainty in the long-term performance of the luminaire system” and “makes it difficult to estimate and warrant the lifetime of LED-based luminaires” (DOE, 2012a, p. 30).
Thus, a key issue in the growth of III-nitrides is the lack of a native or lattice-matched substrate. Currently, the substrates used for the III-nitrides are sapphire, SiC, or Si; at the moment, there are no GaN substrates of suitable size and quality available. Were they available, they might provide a better match to the overlayer III-nitride material, within the limitations imposed by different lattice constants for InGaN, GaN, and so forth. Besides improvement in device efficiency and reliability, a better thermal match between the substrate and overlayer would reduce wafer bowing, increasing the yield of fabricated devices.
From the discussion above it is easy to see that there is a great impetus for the development of GaN native substrates— among which is the growth of GaN bulk substrates. DOE is considering several competing technologies for substrates. There are a variety of approaches to the growth of GaN bulk substrates, which have been recently reviewed in a special issue of the Proceedings of the IEEE and in other journals (Avrutin et al., 2010; Ehrentraut and Fukuda, 2010; Paskova et al., 2010). Further progress in the formation of GaN bulk substrates has considerable technological challenges; furthermore, GaN substrates will have to compete with the lower-cost availability of (lattice-mismatched) substrates such as 8-inch sapphire.
1 See, for example, Data Sheets KSA 400, KSA MOS, KSA Rate Rat Pro, and KSA BandiT (k-Space Associates, 2012).
2 See, for example, Application Notes 49, 53, 45, 50, and 34 at http://www.laytec.de/compounds-applicationnotes.html.