SIL keynotes chart course of LED technology and supply, explore expanding applications (MAGAZINE)

MAURY WRIGHT reports on four equally compelling keynote presentations that illuminate the breadth of the Strategies in Light conference, foreshadowing the various directions that SSL industry developments will take.

Strategies in Light (SIL) 2015, along with The LED Show, took place Feb. 24-26 in Las Vegas, NV with a packed exhibit floor (Fig. 1) and a truly outstanding conference program. Highlights included market-research presentations and, of course, the keynote presentations that this year ranged from details of the state of LED technology and manufacturing, to how solid-state lighting (SSL) will change the future in smart buildings and cities, to art that expands our senses and instills civic pride. Indeed, the keynotes were sufficiently compelling that we dedicate this feature to just those talks.

Two keynote presentations took place in the Plenary Session on both Wednesday and Thursday. We will first go at length into the keynote by Jy Bhardwaj, senior vice president of research and development at Lumileds (Fig. 2), who discussed the future of innovation and what's actually happening in the LED component space. While other LED manufacturers, and perhaps other experts in the SSL arena, may disagree with some of Bhardwaj's conclusions, his methodology and framework for contemplating the future of the LED made for one of the best keynote presentations in the history of SIL.

Bhardwaj structured the talk around topics from substrates to device technology to performance to packaging. His intent was to answer key questions that he is regularly asked about the industry, with the intent of exposing or explaining common myths. Bhardwaj said, "There are a lot of myths out there, a lot of misconceptions. I aim to clear those up for everybody."

Substrates and cost

On the substrate side, Bhardwaj said he continues to be asked about gallium-nitride-on-silicon (GaN-on-Si) LEDs, or specifically silicon as a substrate. He said the common question that he fields is "when is the whole world moving to GaN-on-Si?" He said over the past five years there has been regular news on silicon-based LEDs offering a breakthrough in LED component cost.

FIG. 1. Strategies in Light drew a packed house.

Part of the silicon advantage, however, has presumably been due to substrate cost. As the table below summarizes, Bhardwaj conceded that silicon enjoyed a huge cost advantage four to five years ago relative to the most commonly-used sapphire substrates. But he said silicon wafer prices have stayed relatively level while sapphire wafer prices have dropped considerably to the point where the delta is small and still decreasing.

He warned also that you have to consider performance aspects and that at the substrate level the key parameter is the lattice constant. The mismatch for silicon is far worse and requires additional processing. "The performance of GaN-on-Si has, over the last four to five years, stayed 10% behind GaN-on-sapphire." Moreover, the fact that sapphire is opaque leads to extra processing steps in common flip-chip architectures.

Digging into details based on cost relative to wafer area, Bhardwaj said that silicon offers a baseline wafer cost advantage of about $0.002/mm². Looking for other cost factors, he addressed wafer size. He conceded that silicon is available in larger wafer sizes but said wafer size has no impact in the batch-based epitaxial-growth process where a reactor can host more smaller or fewer larger wafers. He said the advantage of the larger wafer is in the backend process when you do metallization, trenching, and other manufacturing steps. He placed the advantage in the backend at $0.008/mm², meaning that the aggregate cost advantage of silicon is around $0.01/mm².

Translating performance to cost

Bhardwaj then turned to the disadvantages of silicon, again based on cost relative to area, but with the goal of translating LED performance to that area-based metric. He said in a typical SSL application, you have a threshold efficacy measured in lumens per watt (lm/W) as a goal. He said, "As the performance of an LED increases above that threshold, you drive the devices harder back to the same number of lm/W and remove some of the LEDs." Remember that greater drive currents can increase brightness or lumen output at the expense of efficacy.

As a rule of thumb, Bhardwaj said when you get a 15% epitaxial performance improvement, the advantage translates to half the number of LEDs required in an application. Given the 10% performance lag for silicon today, he said the translated cost disadvantage for silicon is $0.02/mm². Even if the performance lag was halved, the cost delta would remain $0.01/mm².

He continued the comparison, looking at the processing complexity of silicon-based LEDs. He said the substrate removal process and bonding of a carrier substrate could add $0.005/mm² in the silicon case. He placed the best-case cost penalty for silicon at $0.015/mm², meaning that silicon-based LEDs cost more. He added, "GaN-on-Si is not going to deliver any sort of breakthrough for the industry when you take cost and performance together."

LED performance and efficiency

Next, Bhardwaj turned to a pure performance discussion focused on wall-plug efficiency (WPE) as the way to measure the performance of a basic blue LED - the efficiency inherent in injecting an electron and extracting a photon. The graph in Fig. 3 represents Bhardwaj's industry-wide view of the WPE of all power LEDs from every manufacturer in terms of WPE relative to current density.

FIG. 3. Wall-plug efficiency shows the impact of droop.

WPE is expressed as a percentage of radiometric watts out divided by electrical watts in. Peak WPE occurs at about 35 mA/mm², according to Bhardwaj. Above that level of drive current, the phenomenon called droop impacts WPE. Droop remains an obstacle to better LED performance.

Bhardwaj said he is regularly asked why the SSL industry doesn't move completely to mid-power LEDs given that droop isn't a factor at lower currents. But as Fig. 3 illustrates, even normal mid-power devices operate above the 35-mA level and are impacted by droop. So instead, the solution will rely on fully understanding and working around the cause of droop.

The industry is still not in agreement on the cause or how to deal with droop, but Bhardwaj identified the common culprits as Auger recombination and hot electron overflow. He said there are two primary ways to address droop. With the commonly used C-plane substrates, you can use multiple quantum wells (QWs) to improve carrier spreading and enable higher current density in aggregate. Or you can transition to semi- or non-polar GaN substrates, which will enable a single QW to realize high radiative recombination of holes and electrons at higher current densities and reduce Auger losses.

Having broached the subject of GaN-on-GaN LEDs, Bhardwaj dug deeper into that subject. If you refer back to the table, you can see that a GaN substrate offers a perfect match in terms of lattice constant with the epitaxial layers. But the semi- and non-polar GaN offer the added advantage of electron-hole overlap that aids in recombination and that could immediately improve the droop problem.

Referring to the table, however, Bhardwaj said, "Clearly, we get no benefit in terms of droop for silicon, silicon-carbide, sapphire, or indeed for bulk GaN." He didn't mention Soraa specifically, but clearly the comments were directed at that company's GaN-on-GaN LEDs that are today manufactured on bulk GaN wafers.

"However, when we get to GaN substrate, we have a different problem, look at the red [in the table]," said Bhardwaj. Today, the industry only has 2-in. bulk-GaN wafers and the cost of those wafers is prohibitive relative to the other options. Bhardwaj added that the semi-polar GaN is not even commercially available at this point. He said the industry would pay the premium for semi-polar GaN when the price hits a level that is five times higher than sapphire. Lumileds is working on such substrates, but Bhardwaj said commercial viability is four to five years away.

CSP packages

Next, Bhardwaj turned his attention to packaging. He said the high-power LED sector has never enjoyed standardization in packaging, and that fact has hurt the ability of the baseline technology to scale in volume and cross more applications. Conversely, packages for mid-power LEDs were standardized in the display-backlight application, and that success has allowed the manufacturers to scale production volumes and ultimately displace high-power LEDs in many lighting applications.

Mid-power LEDs are winning in two ways, according to Bhardwaj. In some applications, designers under-drive the LEDs to achieve very high efficacy. In other cases, such as retrofit lamps, designers use higher drive currents to maximize lumens per dollar, but the legacy lateral die still limits performance. While high-power LEDs still dominate in many applications such as directional lighting and high-reliability products, Bhardwaj believes that a scalable standardized package such as a chip-scale package (CSP) could lower high-power LED cost and enable broader usage.

Lumileds first announced high-power CSP LEDs at SIL 2013. Ironically, SIL 2014 featured a prominent keynote from Samsung focusing on the use of CSP technology to make mid-power LEDs even more cost effective. That theme gained momentum with several other manufacturers announcing mid-power CSP LEDs at Light+Building last year.

Bhardwaj went on to explain the advantages of the technology in terms of lower package cost. Moreover, he said that CSP can deliver performance benefits with good efficacy at high drive currents, meaning excellent control of forward voltage. Arguably, however, you would get those same benefits from mid-power CSP-based LEDs. In fact, with CSP the power-level classification may just disappear.

Still, the biggest benefit of using CSP, according to Bhardwaj, is the scalability of the LEDs to fit a broad set of applications. He said high-power CSP LEDs can be scaled down for applications such as troffers where mid-power LEDs are prevalent or up into high-output applications such as street lights.

FIG. 4. Tony Shakib, vice president of the IoT vertical solutions group at Cisco.

Closing the CSP part of his discussion, Bhardwaj said CSP is real and happening now. He said Lumileds shipped more than a quarter-billion CSP LEDs in 2014 and will more than double that level in 2015.

Efficacy limits

The last topic that Bhardwaj attacked was efficacy roadmaps and how far LEDs can advance in the near term. He stated efficacy is determined by WPE multiplied by conversion efficiency (CE) or the efficiency of the phosphor. CE is characterized by the number of lumens output from the phosphor divided by the radiant watts input to the phosphor. When you multiply CE by WPE, stated as a percentage, you get lumens out per electrical watts into an LED.

Bhardwaj asserted that, for the industry and not just for Lumileds, WPE is around 78% today for mid-power LEDs and headed to around 85% in two years. High-power LEDs are around 65% today and headed to 70%. Those figures were for 450-nm blue LEDs operated at 85°C.

There may be more to gain in CE than in WPE, according to Bhardwaj. He said for a 3000K-CCT, 80-CRI phosphor the theoretical limit for a broadband phosphor is around 265 lm/W, and the theoretical limit for a mix of narrower 30-nm-wide phosphor in red and 70-nm-wide phosphor in green is 310 lm/W. Narrowband phosphor offers a potential 20% performance improvement and has led to a focus in that area for LED manufacturers.

He distilled his asserted limits in WPE and CE to conclude that mid-power efficacy will peak at 240 lm/W in 2017, and that high-power efficacy will peak at 180 lm/W in 2017 - for 4000K-CCT LEDs in both cases. Under-driving either type of LED could deliver a substantial gain with, say, a high-power LED achieving the 230-lm/W level and complementing Bhardwaj's earlier point about CSP LEDs and scale. The US Department of Energy's LED performance roadmap is largely in agreement with Bhardwaj's numbers.

Part of the motivation for the efficacy discussion, however, was clearly meant to discredit claims made by Cree late last year that it had achieved 303 lm/W in an LED in the lab. Bhardwaj considered four possible approaches to achieving such a performance level. He said an LED with white characteristics near the black-body curve would require WPE that is not feasible for either broadband or narrowband phosphor. He said an over-converted, narrowband approach could yield 300 lm/W but the LED would be green in color with a CRI value measured in negative numbers. Cree didn't specify chromaticity in its announcement, but the company does have a track record of delivering on lab milestones with commercial products lagging by about two years. Perhaps we will learn more at SIL 2017.

Internet of Things

After the lengthy LED discussion, the next keynote could not have been more different with Tony Shakib, vice president of the IoT (Internet of Things) vertical solutions group at Cisco, taking the stage (Fig. 4). "We are projecting that by the end of this decade there will be over 50 billion devices connected to the Internet," said Shakib, leading off the talk. "Today there are about 12 billion."

FIG. 5. A converged network leads to operational savings and increased functionality.

Shakib said technology disruptions create opportunity and speculated that we can converge multiple technologies to add efficiencies. "If people are changing their lighting to gain efficiency, it's a great time to change technologies and create new outcomes," said Shakib.

The goal of an expanding IoT isn't just energy efficiency but the experience for society - what he called "warp-force experiences." He said the question driving Cisco is "how do we take these things from buildings to communities to cities to governments to countries?" Cisco is currently engaged in 131 cities around the world on smart-city technology to create new citizen services.

Stepping back, Shakib said the building is the microcosm today of the ultimate goal. And he lamented, as many others have, that while there are many compelling user experiences today, building systems remain in silos in terms of security, HVAC, telephony, IT, and lighting. And the systems are ripe for convergence, some of which has already happened.

Convergence to one network

Convergence to an Internet Protocol (IP) network can lower costs in terms of the initial installation of the services required in a building and the operations of the different services over time. Cisco and the industry have been through such transitions, as shown in Fig. 5. A key has been standards such as Ethernet, as well as IP-based telephony and video. Building management is happening now using Power-over-Ethernet (PoE) technology.

"We believe that lighting is the next logical extension of how we converge those networks," said Shakib. But he added that we need to know why to do so, such as learning how we can drive down operational expenses while improving the user experience and ultimately manage everything over the cloud.

Turning to LED technology, the transition to SSL is key to Cisco's vision enabled by much-lower-power luminaires. The drop in power enables lighting to work with DC power distributed via PoE. Shakib pointed out that the new Universal PoE standard can distribute 60W of power over a network cable.

Perhaps Cisco sees the advantages of the SSL transition opposite from the lighting industry's historic focus on light. "Once you have that LED light, that to us is a perfect sensor," said Shakib. The key is the information that you can capture for storage and analysis in the cloud, which allows the creation of new services and experiences.

Smart buildings

Shakib set out to provide some theoretical examples, using Cisco in the first one. He said the company employees 80,000 with much of that workforce in the San Francisco Bay Area spread over 50 buildings. He said it's not economically viable to have fixed offices for all of the workers, especially with many traveling much of the time.

Cisco uses flexible offices. The Wi-Fi network can recognize an employee based on their mobile phone IP address and customize an office for an employee. That functionality equates to a superior employee experience, but Shakib said the opportunity on the building management side was also significant. He said few buildings manage HVAC systems at more granular levels than a floor of a building. But a sensory network enabled by lighting could bring control to the office level. He said better granularity could bring expenses down another 50% relative to what are considered smart buildings today.

Shakib said the vision is easily extended to cities and outdoor lighting. Chicago spends 40% of its energy budget on lighting, according to Shakib. Sensors on lamp posts could change that based on controls and dimming. But Shakib said you can add features such as air-quality sensors and IP cameras triggered by activity. Many such things are being done today but in silos, thereby adding cost and limiting functionality.

He noted energy savings are very important but must be considered in the context of all building systems. He said LED lighting can deliver $0.03/ft² in a commercial building. Using PoE and realizing additional energy efficiency and optimum facility utilization is valued at $75/ft². Adding lighting as an experience could yield $100/ft².

China rises

On day 2 of the conference, the keynotes provided even more diversity in subject matter. First up was Troy Hsu, CTO of Sanan Optoelectronics (Fig. 6), presenting "The meteoric rise of the Chinese LED market." Sanan is the largest of the Chinese LED manufacturers and gained a more prominent US presence after acquiring US-based Luminus Devices in 2013. For more background on Sanan and the Chinese LED manufacturing landscape, see the feature article we published on the prospects of a Chinese manufacturer cracking the global top ten.

Hsu's keynote marked the first time that the rising Chinese LED leader has presented at a North American conference. Hsu began by characterizing the market, saying that SSL in 2014 totaled $23.9B (billion) globally and that China was responsible for $3.1B. He added that the growth rate in China is higher than growth on a global scale. Hsu projected the SSL market for 2015 at $29.9B with the China market going to $4.6B.

The value of SSL exported from China represents a far larger market for now. Hsu said that in 2014, the country exported $9.5B in SSL products - an increase of 105% from the prior year and 500% relative to 2011. The US was the target of 15% of the 2014 total.

Manufacturing capacity

Next, Hsu turned to manufacturing capacity, saying there are 2,937 MOCVD (metal-organic chemical-vapor deposition) reactors installed globally for the epitaxial growth stage of the LED manufacturing process. He said nearly 35% of that installed base is in China, with Taiwan at 25.5%, Japan at 18%, and Korea at 14%. Moreover, he said the number of reactors in China rose 22% from 2014 to 2015, and said that increase was far greater than the expansion in any other country. He said 60% of the manufacturing capacity is focused on mid-power LEDs.

Hsu said there has also been a major expansion in LED packaging capacity in China. Forest Lighting alone went from the range of 12,000 kk pieces/month in 2013 to near 20,000 kk pieces/month in 2014. Other packagers were much smaller, but capacity ramped at similar rates for at least the top six package houses.

An LED chip discussion followed. Hsu said the total 2014 market for LED chips in China was $2.3B, with Sanan responsible for about $780M (million) of that total. But Hsu said the company has a 50% share of the GaN-based chips used for lighting applications. Hsu projected that Sanan would produce 64% of the chips made in China in 2015.

Hsu also provided a brief comparison of the China chip manufacturing capacity compared to the rest of the world. He said China was responsible for 36% of the chips made globally in 2014, up from 27% in 2013. For comparison's sake, he said Taiwan produced 40% of the world's chips in 2014.

In closing, Hsu provided some insight into Sanan's operations. He said the company currently has 169 reactors producing 10M 4-in.-equivalent wafers per month. The company will add 100 more reactors in 2015 bringing capacity to 21M wafers per month. Hsu said Sanan is producing lateral, flip-chip, and CSP LED architectures and expects to improve the performance of its lateral die by 10% in 2015. In terms of intellectual property, Hsu said Sanan has 506 global patents, with 346 pending, and Luminus Devices has 100 global patents with 47 pending.

Illuminate the Arts

The final keynote came courtesy of Ben Davis (Fig. 7), the founder and chief visionary officer of the Illuminate the Arts organization that was behind the dynamic LED art project, The Bay Lights, installed on the Bay Bridge between San Francisco and Oakland, CA. Davis had the idea for the project while sitting under the Western span of the bridge that opened six months before the more-famous Golden Gate Bridge but which Davis said "fell into a Cinderella hard-working role."

"This elegant expression has completely reoriented the City of San Francisco and the Bay Area's sense of self - a new North Star," said Davis. "It's activated the waterfront. It's visible to tens of thousands of residents from miles and miles away, as far away as Sausalito. And it's deeply beloved." Moreover, Davis has said independent analysis has documented a greater than $100M boost to the annual regional economy that is directly attributable to the project.

By the time you read this article, however, The Bay Lights will have gone dark, ending a planned two-year run. But Davis used SIL to announce Version 2 of the project. Some of the lights have to be removed for cable maintenance. But Davis said even more-capable fixtures will be reinstalled in time for a relaunch coincident with Super Bowl 50 that will be played in Santa Clara early next year.

Davis also used the keynote to introduce and champion a new LED-based public art project that is in the planning stages in San Francisco. He is leading the effort to install a near-2-mile linear stretch of color-tunable lighting along each side of Market Street from the Ferry Building on the bay to Van Ness Street. Davis called Market Street "the civic spine of San Francisco and the most important street on the West Coast." But he said that in recent years the street had lost some of its luster with vagrants in the area that can dissuade visitors.

Subway-responsive LED art

Called LightRail, the new project will feature dynamic movement of light sequences that are synchronized with the subway trains that run in each direction below the street (Fig. 8). Pedestrians will feel the rumble of the trains from below and experience the synchronized and colorful show overhead.

"It's not only celebrating mass transit but really using light, energy, and movement to seamlessly permeate some of those economic and emotional divides that have segregated and separated the street," said Davis. "It's an amazing vision, a very powerful vision, and also a deeply challenging vision to manifest."

About the challenges of LightRail, Davis said, "It would not be possible if The Bay Lights were not shining as a beacon of possibility and maybe even a bureaucratic taunt." City government was perhaps pressured by the success that CalTrans (the California Department of Transportation) has had with The Bay Lights to pursue an iconic project on Market Street.

In December 2014, the San Francisco Board of Supervisors unanimously approved a major encroachment permit to Illuminate the Arts for the installation of LightRail. Davis said approval required cooperation by the San Francisco Department of Public Works, the Public Utilities Commission, the Metropolitan Transportation Authority, the planning department, the arts commission, the Historic Preservation Commission, and the Fire Department. The diverse group has joined to pursue the project. And Parsons Brinckerhoff, the design firm behind The Bay Lights, has developed a design for LightRail.

Obstacles

The challenges to the project range from historical preservation to public safety to technical obstacles. Davis said the light stanchions and globe-style street lights have to be treated with historical reverence, and the Historic Preservation Commission had to be convinced that the project would not negatively impact the ambience. The team turned to The Maker community for help with 3D modeling so the commission could see and even touch the way LightRail will interact with the existing lighting.

The Fire Department raised the issue of the new lighting being an obstacle to the use of hook-and-ladder trucks on Market Street. But after evaluating the situation, the LightRail team learned that only buildings less than 75-ft high were an issue, because the ladders aren't usable with buildings taller than that. The design team has developed a breakaway solution that allows the Fire Department to immediately drop the lighting out of the way for each area in which there is a shorter building to achieve the safety goals and still have the installation bring the benefits of artistic architectural lighting to the area.

Ultimately, the project will encompass the street lights as well. High-pressure sodium (HPS) lighting installed on Market Street when BART was built in the 1970s has put the street into a "sense of illumination doldrums for nearly half a century," said Davis. "It becomes a place of hurried passages rather than lingering destination." The LightRail team needed to install new lighting anyway to allow the power system to handle the load, and Davis emphasized that the planned upgrade will deliver 80% energy savings.

Still, the visitor experience is the primary goal of the lighting project. "Market Street will come back," said Davis. "It will be a place you want to go to. And the illumination of that street will be key to your sense of experience."

What should be clear from the keynotes is that Strategies in Light and the LED lighting industry are changing from a focus of promise to one of reality. Moreover, the SSL transition is very much a global revolution in terms of participation, application, and opportunity encompassing multiple paths, some diverging and some developing in parallel along the way.