Aggressive New LED Lighting Roadmap Target of 8X Cost Drop by 2015 Counts on Innovation in Manufacturing Technology
By Paula Doe, SEMI Emerging and Adjacent Markets
Getting packaged LED die costs down to the new more aggressive roadmap target of $2.20/klm within five years will take some major innovations in manufacturing technology. Luckily, the manufacturing supply chain is making some major progress.
This year’s roundtable of experts gathered by the U.S. Department of Energy to review its industry roadmap decided last year’s ambitious goal of 10X cost reduction by 2020 in packaged LED die still won’t be enough. They now think an 8X reduction by 2015 will be needed to create a solid state lighting industry that can compete with fluorescents.
Can the industry achieve this aggressive goal of reducing costs from about $18/klm now down to only $2.20/klm within five years? The industry is making good progress in everything from understanding the droop in efficiency at higher currents, to controlling uniformity of epitaxial deposition, to finding and fixing defects, to simplifying packaging by using fewer components. Industry experts will discuss this progress and next key issues at the major steps along the value chain at the HB-LED manufacturing technology program at SEMICON West, July 13 in San Francisco.
Finding the Defects that Matter
The ability to detect and classify defects on the wafer or substrate surface is turning out to be fundamental in improving yields in LED manufacturing. KLA-Tencor's Candela 8620 optical inspection tool can now map the submicron defects on just the front side of the transparent sapphire wafer with dark field sensitivity, so LED and substrate makers can see the pattern of micropits, microcracks and scratches on the surface for the first time.
Submicron pits are a direct cause of defective LEDs -- they fill with metal shorting the p/n junction and making an LED that doesn’t light up -- and a direct indicator of uneven heating during epitaxial layer growth. Their pattern on the wafer map can suggest if the cause is wafer warp or a particle between the wafer and the heated pocket in the MOCVD reactor. Operators visually inspecting a few points across the wafer cannot catch these pits or these patterns.
Figure 1: Sample Yield Loss
Microcracks, meanwhile, come from the lattice mismatch between the GaN and the substrate. "It's just one atomic layer of sheared bonds, difficult to see, but becomes catastrophic reliability killers as the device heat cycles in the final application like an automotive headlight," notes Frank Burkeen, VP and GM, Candela Division, KLA-Tencor Corp. He notes that this new ability to see the microcrack defect patterns will likely be key to figuring out how to grow good nitride layers on silicon substrates.
Figure 2: The value of automated inspection for early detection of an epi reactor excursion of epi pits
While KLA was working with Philips Lumileds and others on a partially DOE-funded project to develop and apply this tool for micropits and microcracks, researchers also discovered that they could see other micro-defectivity on incoming sapphire substrates for the first time, which also turned out to directly impact deposition quality. Substrate makers have been able to make important strides in improving surface quality now that they can see the problem.
"Less than 5% of epi-ready LED substrates are now inspected with high sensitivity automated inspection," notes Burkeen. "Now that we have a more powerful toolkit, the more we use it, the more new things we see to work on." Improved yields of course translate directly into lower LED costs and open new applications for solid state lighting.
Designing Equipment Specifically for LED s to Improve Throughput and Yields
Dedicated equipment optimized for the particular needs of LED production instead of older generation semiconductor industry tools may also offer significant improvement. Ultratech, for example, figures a projection lithography tool designed specifically for LEDs could cut lithography costs in half, by things like adjusting the projection system for the warpage of sapphire wafers, changing the light source to better see alignment marks on textured wafer surfaces, modifying the shutter for the LED sector’s faster photoresist, and enclosing the tool so it works in the less stringent clean room conditions used by LED makers. CTO Andy Hawryluk estimates the lower capex and higher throughput can mean savings of around $250,000 per tool per year, while better the yields from the better alignment, handling of warped wafers and the like can mean as many as 100 million more die per tool per year as well.
Rethinking Materials Use Too
Another place with considerable margin for improved efficiency is in the delivery of precursors for the epitaxial deposition. Thiloma Perera, Dow Electronic Materials’ director of global marketing for metalorganic technologies, argues that switching from onboard cylinders to a central delivery system eliminates the space and maintenance now needed for current reactor set-ups with up to 4-5 different cylinders, each needing to be kept in different temperature baths in the tool. The central delivery system also eliminates the 4-24 hour down time typically needed when each cylinder is replaced. One central delivery system can supply multiple chambers or multiple tools with more gas flow at a more stable concentration, while allowing operators to switch cylinders without shutting down the tool. Another advantage to the central delivery system is the reduction in unusable material left at the bottom of a cylinder. Dow estimates significant savings in the cost of ownership for epitaxial growth with the use of central delivery for precursors. Perera says central delivery systems for delivering trimethylgallium are now operational at multiple LED manufacturing sites in Asia and the US. “A lot of the inefficiencies in materials usage for LED manufacturing is not so visible to people,” she notes. “We want to move the whole industry to central delivery because it will be much more cost effective.”
Users and suppliers are also making progress towards some consensus on basic requirements for tools and materials so those from different suppliers work together. Industry executives on SEMI standards committees are working towards agreement on common placement of flats and notches for marking 6-inch sapphire wafers, and common cassettes and software and hardware interfaces to enable automation.
These speakers join other experts from Philips Lumileds, Cree, University of California Santa Barbara, Veeco Instruments, Suss MicroTec, Brewer Science, Redwood Systems, Yole Développement, Fraunhofer IZM, Cascade Microtech , Intematix and NNCrystal to discuss recent progress and next issues in LED manufacturing technology at the SEMICON West event. SEMI, as a not-for-profit semiconductor industry association, offers low-cost programs that are unique in the SSL sector for their focus on the practical manufacturing technology needed to create a viable volume market for general LED lighting. More information is available at http://www.semiconwest.org.
June 7, 2011