The fast moving HB LED sector is starting discussions to consider the kinds of consistent materials characterization, testing protocols, tool interoperability, or other common practices that typically enable a mature high volume industry, reports Paula Doe from SEMI.

Driven by rapid progress in technology, fast growth in display backlighting markets, and a potentially huge market for general lighting about to come, the high brightness LED sector is hurtling towards becoming a more mature volume manufacturing industry within the next few years. Though companies will of course maintain their unique core processes, the mature high volume manufacturing business will also require some changes, towards efficient supply chain management, towards more automation and towards more emphasis on tuning a controllable manufacturing process for consistent high yields.

“The industry is maturing very fast, much faster than the IC industry did, as it can  learn from that experience,” says Iain Black, Philips Lumileds VP of Worldwide Manufacturing Engineering & Innovation. “This is becoming a serious business. In the future we expect that with a smaller number of key players and a consolidating supply base, some of the custom variation will have to come out of industry. That will require standards, and they will need to be defined early enough to avoid delaying the development of the market, perhaps not in 2011, but certainly sooner rather than later.”

Some industry leaders are gathering to start to discuss these issues at the first meeting of the newly formed SEMI North American HB LED Standards Committee November 11 in San Jose, California. The group is led by Black; Bill Quinn, chief technologist for MOCVD at Veeco Instruments; and Chris Moore, CEO of Semilab USA. SEMI invites participation from all interested parties. Parallel activities are expected to follow with committees in Europe, Taiwan and Japan.

“I see standards as somewhat inevitable,” adds Black, noting that it won’t be realistic to keep using all bespoke materials. “So it’s important to be involved in the process to have some input. It’s time to get the conversation started around what might be possible.”

Finding common ground

One key area of common ground is the characterization of incoming materials. This could begin with consistent measurement of purity for chemicals and benchmarks for LED grade materials like indium, gallium, metal hydrides, and packaging materials, which all differ a bit from supplier to supplier. This might not be too difficult because there are a limited number of suppliers.

Wafer standards are more challenging, with multiple different diameters and substrate materials in production employing radically different processes, so some argue that even agreeing on a common thickness for 150 mm sapphire wafers is unlikely. But others counter that one thinner ~1 mm standard for those who thin the wafer down and prefer a thinner substrate, and another thicker ~ 1.3 mm standard for those who remove the epi layer and prefer a thicker substrate, could probably handle most production needs while doing away with much of the individual customization.

Even a few more standard products would allow efficient supply chain management, letting both wafer users and suppliers stock products and buy and sell off the shelf, and to deal with multiple sources in times of shortage. It would also allow tool suppliers to improve process measurement and control. The timing could be good for the fast transition about to come in substrate diameter.

Tool hardware specifications may be another area of common ground, as susceptors, graphite and SiC from different vendors aren’t the same, and measurements of conditions in the tool aren’t measured the same way across suppliers, so results can’t be compared.

Better binning

Metrology and test is another key area for potential gain from consensus on best practice. As the HB LED sector matures and moves from the realm of the development engineer to the manufacturing engineer it will begin to move away from its current focus on volume. The focus will shift to control of the established manufacturing process, and to greater attention to the operational benefits of yield, identifying and tracking yield issues as early in the process as possible. But the process is currently hindered by everyone measuring different things in different ways.

To measure something as basic as wavelength uniformity on the wafer, for example, some LED makers use electroluminescence tests, others use photoluminescence tests, some measure peak wavelength, and others measure dominant wavelength — and all of those give different results.

“If everyone could agree on one of those measures as best practice, it would get everyone talking about the same issues,” argues Veeco’s Quinn. “If everyone measured things the same way, it would be easier for the epi guys and the rest of the fab suppliers to improve the process for everyone. The potential to improve yields with standard testing and feedback is huge. It’s definitely not too early to start talking about standards.”

“Every company’s fixturing is different, and will give different results,” concurs Dan Morrow, president of Op- Test, citing the difference between probe tests and integrating spheres, and multiple different set ups for each. Morrow suggests that common testing protocols — like standards for thermal junction temperature management during production test — would give makers more useful feedback for process control, and make spec sheet data more useful and meaningful for both LED makers and their customers.

Morrow suggests the traditional transformed CIE XY photometric measurements of what the eye sees, from the traditional lighting industry, may not give precise enough data of what energies are at what wavelengths to drive yield improvements and guarantee that things that look the same really are the same. Spectral power distribution may be the more useful data to improve the production process. “Few are using process control feedback yet, but big players are coming into the market and I expect that will bring big changes,” he adds. Wafer level test systems are still largely a custom business, points out Mark Cejer, Keithley Instruments director of marketing. He notes that the different degrees of precision in color uniformity needed even for edge-lit versus back-array television backlights applications, to say nothing of automobile lights versus cell phone displays, all requiring different tradeoffs of complexity and costs. So LED makers tend to put together their own systems of source-measure units, probers, spectrometers and custom software, or have a local systems integrator do so. To make it even more complex, companies are also going to larger devices, or to multiple-device die cut from the original wafer, which require higher voltages or currents and shorter pulses for testing.

“We have to remain flexible to all these different ways of doing things because there are no established methods for test yet,” says Cejer. “But the relentless price pressures will drive the industry towards efficiency, and if we put our heads together maybe we can do it efficiently.”

Going forward, one solution might be finding enough correlation to characterize optical quality from only electrical tests, at least in some applications where the trade offs were acceptable. Metrology and test are also easier for LED makers to talk about with suppliers. There is no advantage in not being able to measure things, so no one loses if a company works with a supplier to develop technology to test things that can’t be tested now.

Materials, substrates, automation and metrology are usually the first areas to standardize as an industry matures, says Semilab’s Chris Moore, noting the accelerating speed with which some other sectors have moved up to consistent volume manufacture. The solar industry, which also differentiates on process IP, was skeptical of manufacturing standards only a few years ago. However, it has now embraced standards with amazing rapidity, compared to the slow and painful process of the IC and LCD sectors before. Now two years in to starting standards discussions, there are some 400 photovoltaic industry experts working to facilitate efficient volume manufacture. This team is starting by coming to a consensus on the best measurement methodology for purity and setting benchmarks to define PV grade materials, and agreeing on carrier and equipment interfaces to facilitate automation. First agreement on defining consistent characteristics of solar grade silicon was reached in only three meetings over one year. They’ve reached consensus on 12 standards so far, with another 6 expected to be published by the end of the year, and more than a dozen more under development.

Automation

Though small wafers, cheap labor and long batch processes have limited the need for automation so far, the LED industry is transitioning to more and more automation as it moves to larger wafers and higher volumes. Robotic wafer handling, automated glove boxes, interbay automation, mini-environments and standardized carriers such as SMIFS are all technologies that will become pervasive in LED fabs of the future, argues Clint Haris, Senior Vice President of the Systems Solution group at Brooks Automation. He points out that LED makers are starting to look towards automation to improve yields and traceability. “The industry is rapidly evolving from manual operation to fully automated factories,” he says, noting that the LED industry has seen change in the last five years that took 40 years in the semiconductor industry. “Things are moving so quickly, standards need to focus on the leading edge, 6-inch wafers, 6 inch cassettes, and some sort of wafer or carrier-level identification for traceability as the basics to enable automation.”

“Automation always increases yield,” points out Quinn, noting that Veeco has found that when its experienced technicians instead of its interns load the tools, yields are improved by as much as 50 percent. Automation also may enable a faster ramp to volume than finding or developing all the qualified operators and engineers to run all the new epi reactors now out in the field. Though HB LED manufacturing is unlikely to move to the expensive full automated materials handling required for the heavy cassettes of 300 mm semiconductor wafers, it is likely to move towards the semiconductor automation of the 200 mm generation of the 1990s.

Automating data collection, analysis and even correction will also be key to getting more die into the bins that bring high margins, notes Applied Materials’ Phil Walker, global product manager for automation products. “We need to get the data out of the tool and metrology, and linked to the right wafer to be able improve bin yields,” he says, pointing to the potential for tracking the parameters of the production process, mining the data for the root cause of defects, and making needed adjustments to the tool as soon as possible.

“We also need the software to compare tools and tune them to match those that provide the best performance,” he added. “The first step to unlocking hidden performance is collecting the data and understanding it.”

 



HB LEDs production is fast transitioning to larger wafer diameters, moving to mass adoption of 4-inch (actually 100 mm) wafers next year, and to 6-inch (150 mm) within a few years, as the better use of reactor carrier real estate and reduced edge losses can increase throughput by as much as 30 percent, and allow use of the other semiconductor equipment available at 6-inch. Source: Yole Développement Sapphire Market 2010 Q4 Update