KLA-Tencor has debuted an automated inspection system for substrates and epitaxial wafers used in high-brightness LED (HB-LED) manufacturing. The Candela 8620 provides automated defect inspection for LED materials such as gallium nitride, sapphire, and silicon carbide, enabling enhanced quality control of both opaque and transparent substrates, faster time-to-root cause, and improved metal organic chemical vapor deposition (MOCVD) reactor uptime and yield.
In an interview with SST (listen to the podcast at www.electroiq.com/index/podcasts), Frank Burkeen, VP & GM of KLA-Tencor's Candela division, cited a recent Department of Energy (DOE) study emphasizing the need for better methodology around defect inspection and root cause analysis to achieve required cost reduction targets (about 10×) that are needed to bring SSD lighting into the general lighting arena. Yield can be responsible for making up half of that cost difference.
Figure 1. Example of the impact of faster excursion detection, manual inspection vs. automated inspection. (Source: KLA-Tencor)
In LED manufacturing, just growing the epi layers (the first part of the frontend process) accounts for about half of the value of the device, before you've even started fabricating the device, said Burkeen. As HB-LED manufacturers transition production to larger wafer sizes and introduce new patterned sapphire substrate processes, the economic impact of resulting process-induced defects is estimated at millions of dollars in lost product revenue per year, and MOCVD epi process issues may result in as much as 40% of overall defect-induced yield loss. Quantum wells are constructed during the epi deposition and they determine the wavelength and luminosity of the device. "This is much different from silicon CMOS processing in which much of the value added occurs later in the process (as the structure is built with circuit lines)," he observed. Therefore, process control around an early step (in LED manufacturing), when the wafer is unpatterned, has a much greater significance than it would in CMOS manufacturing, he said (Fig. 1).
Figure 2. Excursion detection: impact of capture rate and tighter data distribution. (Source: KLA-Tencor)
Burkeen also explained how automated inspection results in higher defect capture rate and a tighter defect count distribution (Fig. 2). Being able to produce a much tighter distribution with coverage of the full wafer provides much more sensitivity to small excursions, he noted. An operator with manual inspection has virtually no sensitivity to a small excursion because of the area they are looking at, variability, etc. "A small excursion that is hard to catch is really the most dangerous type" for an LED manufacturer from a financial performance standpoint, he said (Fig. 3). LED substrate and epitaxy layers pose significant inspection challenges due to high levels of background signal and nuisance defects.
Figure 3. Economic impact of major vs. minor excursions. (Source: KLA-Tencor)
KLA-Tencor also has debuted the KLARITY LED automated analysis and defect data management system and ICOS WI-2220 wafer inspection tool designed specifically for LED defect inspection. — D.V.
Solid State Technology | Volume 54 | Issue 3 | March 2011