New Milestones for LED Lighting
Thin-film technology has enabled a new generation of high-brightness LEDs.
by Martin Behringer
MOST people think about solid-state lighting in terms of white light – for street lights, vehicle headlights, or subway-station lighting. Of course, we do not live in a black and white world, and our colorful world could benefit from illumination by longer wavelength light-emitting diodes (LEDs). In fact, scientists have pursued this line of research for five decades, as depicted in Fig. 1. While the first LEDs were fairly simple pn-junction devices, they were quite complicated for their time. Today’s chips are more sophisticated, as a result of several difficulties scientists had to overcome during development.
Fig. 1: Brightness development for red, yellow, green, and blue LEDs is shown for the last 50 years.
First, the conversion efficiency was fairly poor due to the low quality of the crystals themselves. Impurities and lattice defects made the conversion of electrical energy in light (photons) inefficient. Then, most of the light got absorbed again before it left the LED die. Absorption took place most of the time in the layers themselves, more severely in the substrate and also at the metal contacts. And while the first LEDs were made from InGaAlP (indium gallium aluminum phosphide) and AlGaAs (aluminum gallium arsenide) in the range between orange and infrared, today’s LEDs use additional materials and reach into the UV range of the light spectrum. Packaging concepts and designs also matured dramatically, in parallel with the LED dies. While the first LED devices were only slighty brighter than a glowing candle, LED systems now are bright enough for headlamps in cars and even trains.
Today, many applications require yellow, red, or hyper-red illumination. Such applications include projectors, color mixing for warm white illumination, hyper-red illumination coupled with blue light for greenhouses (as shown in Fig. 2), and also infrared illumination with high brightness for closed-circuit TV, adaptive cruise control for cars, or light curtains (door detectors) in elevators.
Fig. 2: LEDs are now employed for a wide variety of applications, from solid-state lighting to projectors to industrial applications. Here, multiple monochromatic LEDs are combined in a horticultural lighting system.
After more than 12 years of efforts in thin-film technology, Osram Opto Semiconductors recently achieved a significant technological breakthrough: a red LED at 200 lm/W at a dominant wavelength of about 615 nm. For red light, 200 lm/W is quite challenging, as the theoretical limit for 615 nm with 100% efficiency is about 300 lm/W. So, 200 lm/W is coming fairly close.
Currently, two companies claim to have surpassed 200 lm/W – both in a laboratory setting at specific current densities. To place these developments in context, the balance of this article briefly recounts the history of colored-LED development.
How LED Lighting Works
In solid-state lighting devices, the light is generated in the semiconductor crystal. Here, the electrical energy is converted into photons, and, in this crystal, Osram introduced improvements to make the entire system more efficient. To get the light out of a light package, the barrier of total internal reflection (TIR) at the crystal–die boundaries has to be overcome. As the light is reflected back into the die, it is incident to the surface of the die with an angle larger than a specific value. There are two ways to help emit the light. The first is to continuously change the angle of the light so if it is reflected once; the direction will be changed and it can escape in the next run. Changing the direction of the light can be done by surface roughening or by internal absorption and re-emission. The other way is to reduce the loss mechanism (equal to absorption without re-emission of light with the same wavelength) as far as possible. This gives the light more opportunities to leave the die before it is lost.
Colored light can be generated either directly (by generating the desired color through direct emission) or by conversion from one emission spectrum to another. In the first case, typically, a medium is excited electrically and emits photons in a specific wavelength. In the second scenario, a photon is emitted, then that photon is absorbed in a medium, and then another photon of a second, longer wavelength is emitted.
LEDs made from InGaAlP can directly and efficiently generate light for the wavelength range from 560 nm to about 660 nm.
The internal efficiency of a metal organic vapor-phase epitaxy (MOVPE) LED makes it the most commonly used crystal growth technology in the LED industry today. Efficiency has increased to more than 90% in the longer wavelength range. But due to the high index of refraction of InGaAlP LEDs, only 4% of the generated light can leave the chip directly to be used for illumination; the remaining 96% is either reabsorbed by the material or reflected at the chip/air interface and eventually absorbed into the chip. Encapsulation in silicon or resin reduces the disadvantage of internal reflection. But, as in InGaAlP, it has a very high index of refraction of about 3.4 and because there is no encapsulant that comes close to this value, light out-coupling remains a problem.
The Development of Thin-film Technology
Thin-film technology was invented by the lighting industry to increase efficiency and reduce light loss. (Many companies worked on this technology, and many now offer thin-film devices, but Osram was one of the first to commercialize it, winning a German Future Prize for its technological efforts.) In thin-film LEDs, the light is also generated in a quantum-well structure. In contrast to conventional LEDs, the growth substrate is removed and the active epilayers are bonded onto a carrier, which can be germanium, silicon, or another material that matches the requirements for this carrier. A highly reflecting mirror can be deposited in between the epilayer and the carrier, which prevents the light from being absorbed in the substrate. In addition, the active epilayers can be so thin that the light is absorbed differently from before, when thick layers were required to extract light efficiently. Normally, the surface is roughened to enhance light extraction or microprisms are included to increase efficiency (see Fig. 3).
In Fig. 3(a) and Fig. 3(b), the blue line shows that the current is directed away from the bond pad to avoid shadowing. The red arrows depict the light path. The light, which is initially emitted toward the substrate, is reflected by the underlying mirror and redirected toward the chip surface where it can then leave the semiconductor die.
Fig. 3: A schematic sketch of a conventional LED is shown on an absorbing substrate (a) next to a thin-film LED with a metal mirror between the active epilayer and carrier (b).
Figure 4 highlights the historic efficiency of red LEDs (dominant wavelength, 615 nm). As the triangles and diamonds demonstrate, conventional LEDs reached a peak efficiency of approximately 30 lm/W.
Beginning in 1998, Siemens/Osram began research to develop thin-film LEDs for high-brightness emission. Initially, a number of difficulties arose. Perhaps the biggest problem was transferring the thin semiconductor layer (~ 5 µm, which is about 1/10 of the thickness of a human hair) to a carrier in a large area (5 µm / 100 mm = an aspect ratio of 20,000). Because this layer is very fragile and requires a high yield for an economically meaningful concept, about 10 years transpired between the first publication of the idea in 1993 and Osram’s first commercial ThinFilm LED in 2003. Regardless of this and other thin film problems, it quickly became obvious that by using thin-film technology, higher brightness can be obtained (see the triangles in Fig. 4). By 2010, levels of approximately 140 lm/W were reached.
Fig. 4: The historic efficiency of red LED brightness is shown by black diamonds indicating volume emitters on an absorbing GaAs (gallium arsenide) substrate. Red points indicate thin-film LEDs since 1999.
Some advantages of thin-film LEDs include:
• Scalability: In theory, all chip sizes can be made similarly, and, in turn, possess similar characteristics and performance, according to their size.
• Surface emitters have Lambertian beam patterns, and thus make it possible to combine many chips side by side without changing the beam pattern. Also, a Lambertian beam pattern means that the LED seems to have the same brightness from all viewing angles, and a chip with this profile is suitable for both reflector and focusing optics.
• High efficiency.
• Cost-efficient design and manufacturing processes (based on lm/$).
During the last few years, further investigations into loss mechanisms within the devices have not only improved efficiency, but also reduced cost and increased reliability. An analysis of different materials and semiconductor compositions used within the diode for electrical conductivity and optical transmission helped in the replacement of absorbing layers with non-absorbing layers of similar or even better functionality.
Figure 5 depicts the current curve and efficiency for a 1-mm2 die mounted into a laboratory package, optimized for efficient out-coupling of its generated light.
In Fig. 5(a), the light output for currents up to 350 mA is shown. A linear curve indicates efficient light generation over a wide current range. In Fig. 5(b), the efficiencies are given. The orange curve depicts the efficacy in lm/W, while the black curve provides wall-plug efficiency.
Fig. 5: (a) The light-current curve and (b) the efficiency/efficacy over the current of a 1-mm² LED in a laboratory package.
Both curves show a distinct maximum at about 50 mA, with approximately 201 lm/W and 61%, respectively. At 350 mA, the values reduce to 168 lm/W and 52%. The explanation for this decline can be found in the red curve, which shows the external quantum efficiency. It shows a very broad maximum, and the value is almost unchanged between 50 and 350 mA and between 58% and 59%. The decline in efficacy and wall-plug efficiency must therefore be attributed to an increase in operating voltage due to still-existing ohmic resistances within the die. The high lm/W values were reached with a red die, which emits at a dominant wavelength of about 609 nm at room temperature. For longer wavelengths, eye sensitivity lessens and, therefore, similarly high lm/W values are harder to realize. On the other hand, high wall-plug efficiency values are easier to realize with longer wavelengths. Therefore, the high value of 61% is quite remarkable.
These improvements were made possible in the following ways:
• By reducing the absorption of the layers as much as feasible, it is possible to extract more light, even if it travels 10 times through the die before escaping the semiconductor.
• This was accomplished, in part, by increasing the bandgap and, in part, by adjusting the doping. To avoid high- ohmic resistances at these high operating voltages, increasing thickness improved the conductivity of the layers. The contact design was optimized to keep the operating voltage low. Fig. 6 shows a 1-mm2 die. The dense arrangement of these metals’ current paths enables strong electrical performance despite adjusted doping levels.
Fig. 6: Shown is the illumination pattern of a 1-mm² die from the latest thin-film generation.
Finally, these improvements resulted in a very-broad process window, which allows for high yield and, therefore, lower overall cost.
This new technology was applied to chip sizes of varying lengths: 250 µm, 300 µm, 500 µm, 750 µm, and 1 mm. A 150 µm and a 2-mm² die followed in 2012. The dominant wavelength range is from 590 to 645 nm; 560 nm and infrared range followed in 2012 and 2013, respectively.
As mentioned above, maximum efficacy is reached at about 610 nm, while wall-plug efficiency is highest at a wavelength of 645 nm.
Figure 7 depicts the characteristics for a 1-mm2; die at a dominant wavelength of 645 nm (λ = 660 nm peak) in the same laboratory package as used for the 615-nm emitting chip. The output power over current and the wall-plug efficiency are shown.
Fig. 7: Light output and wall-plug efficiency are demonstrated over the operating current for a 1-mm² die emitting at 660-nm peak.
The output power rises linearly with current and reaches about 437 mW at 350 mA (at about 2.1 V). The wall-plug efficiency reaches about 59% at 350 mA and is above 70% between 5 and 60 mA. Due to reduced eye sensitivity (photopic curve), the maximum output at 350 mA is only 21 lm. Lifetime depends on operating conditions such as current and temperature. The entire LED community is continuously working to improve the reliability of these devices.
With the new generation of thin-film dies, however, output power could be increased by 30–50% for a bare die compared to the previous generation and about 10–30% for a packaged die in an LED. This is quite significant. With this kind of increase, new and more numerous applications can be addressed. An already widespread application is the combination of highly efficient – but aesthetically unpleasant – cold white with amber or red, which changes the color temperature to a more visually pleasant and high-quality white. As a result, both higher overall efficiency and good color rendering can be achieved simultaneously.
Meanwhile, projection and signaling applications absolutely require high-efficiency red light and will doubtlessly profit from these breakthroughs. As far as mobile applications are concerned, either more power or longer battery life will provide tremendous customer benefits. For industrial applications, lower energy consumption will reduce operating costs and brighter devices will minimize initial installation expenses.
In this context, it is clear that improvement in 660-nm LEDs will also have a positive impact on LED-illuminated greenhouses. With such high efficiency, energy costs can be reduced by almost half compared with conventional lighting; furthermore, LED cooling is made easier. For example, when comparing a light with 40% efficiency to the above-mentioned record values, it is realistically possible to generate 75% more light with the same amount of electrical energy and up to 250% more light at the same amount of waste heat.
Due to the improved performance of the chips themselves, package and luminaire designers are emboldened to design bigger and higher-powered packages and lamps that are driven at higher operating currents. While the first LEDs were for the mA range, today often several amperes are run through the die. Even though efficiency has been greatly improved, these higher operating conditions still require good thermal management, due to higher total load. Thus, modern LEDs and luminaires must provide low-ohmic resistance in terms of electrical supply; they must efficiently remove the heat with low-temperature increase, and they must also guarantee efficient light out-coupling and beam shaping to fulfil needs for modern light sources.
What Comes Next?
Considering 60% efficiency, it is clear that incremental brightness improvements will be harder to realize as we approach the ideal of 100%. With that said, other parameters continue to require improvement. In particular, companies will want to see better production and process values.
Also important will be improvements in high-current and high-temperature operation. Certain devices are commonly characterized at room temperature, while approximately 80–100°C is a more realistic operating condition. Therefore, work is being done to optimize output power at these temperatures, as well as reducing the shift in operating para-meters when changing the temperature.
Last but not least, light out-coupling remains a constantly evolving challenge. On the package level as well as on the bare die, there is ongoing research dedicated to obtaining every photon that is generated out of the die. Certainly, out-coupling in bare die configurations offers a wide range of possible improvements.
To summarize, since 1998, thin film has become a preferred solution for generating high-brightness LEDs. It was recently determined how to reduce optical and electrical loss mechanisms to realize a maximum efficacy output of 200 lm/W. This enables far-broader LED use in many applications, which will increase lighting quality considerably.
References
1Source: Elsevier/Academic Press. This image was published in M. George Craford, “High Brightness Light Emitting Diodes, Semiconductors and Semimetals,” Overview of Device Issues in High Brightness Light Emitting Diodes (Elsevier/Academic Press, 1997), p. 48. •