Breaking the 200 lm/W barrier for the red LED will do far more than simply increase the efficiency of car brake lights. It will also improve the efficacy and colour quality of solid-state lighting sources as well as pico-projectors and yield a lighting product for greenhouses when combined with blue LEDs, says Martin Behringer from Osram Opto Semiconductors.

 

The term ‘solid-state lighting’ conjures up images of white emission emanating from streetlights, car headlamps and modern fixtures in subway stations. Anyone standing there and basking in such light will undoubtedly look on in awe at these diminutive illumination sources, but that does not mean that they will be thankful for their pale-looking faces produce by the collections of white-emitting LEDs.

Fortunately, we don’t live in a black and white world – we live in one of colour and warm faces. But to do justice to the world’s incredibly rich palette requires an illumination source that produces the full gamut of colours, from the deepest reds right through to blues and violets.

The traditional white LEDs that combine a blue-emitting chip with a yellow phosphor fail in this regard, because they ignore longer wavelengths. Although long-wavelength sources – directly emitting red, yellow and infrared LEDs – have been overshadowed by their blue cousins in recent years, their development stretches back far further, to the 1960s. Efforts in this direction are tremendously important, because a vast number of applications require yellow, red or hyper-red illumination: Projectors; colour-mixing systems for warm white illumination; lighting sources for green houses that marry hyper-red and blue light; and closed circuit TV, adaptive cruise control for cars and light curtains in elevators that require high-brightness infrared illumination. Despite common perceptions to the contrary, it is clear that the solid-state lighting requirements of today far transcend ‘just white light’.



LEDs are now employed for a wide variety of applications, from solid-state lighting to projectors and green house illumination

 At Osram Opto Semiconductors, which is based in Regensburg, Germany, we have recently achieved a significant technological breakthrough that will help in this regard, and one that merits a ‘red alert’: The first red LED with an efficacy exceeding 200 lm/W. This new benchmark for the red LED is a fruit of 12 years of diligent effort associated with our development of thinfilm technology.

A little history

The most widely used material for developing long-wavelength visible LEDs is AlInGaP, which can span 560 nm to 660 nm. The first devices made with this quaternary were homojunction LEDs incorporating a simple pn junction. However, over time this device has evolved from a lab curiosity to a high-performance, commercial product (see Figure 1).

 



Figure 1: The performance of red, yellow, green and blue LEDs has come on in leaps and bounds over the last 50 year. Figure adpted from Semiconductors and Semimetals 48 48 (Publisher: Elsevier/Academic Press)

The introduction of mature processes has driven up yield; and efficiency has rocketed, thanks to the introduction of more sophisticated device architectures incorporating carrier capture in quantum wells.

Growth of LED epistructures by MOCVD on high-quality substrates can routinely produce longer-wavelength emitters with internal efficiencies exceeding 90 percent. But the high refractive index of the AlInGaP LED traps most of the light: Only 4 percent leaves the chip directly and can be used for illumination; with the remaining 96 percent either re-absorbed by the material, or reflected at the interface between chip and air, before eventually being absorbed by the device (see Figure 2).

 



Figure 2: A substantial proportion of the light generated in a conventional LED,which is built on an absorbing substrate, is trapped within the chip (top). Light extraction improves dramatically with a ThinFilm LED architecture incorporating a metal mirror between the active epilayers and carrier (bottom). In this superior device, the current path, depicted by the blue line, is directed away from the bondpad to avoid shadowing. The light path (shown in red) is initially toward the substrate, but changes direction thanks to reflection by the underlying mirror. Light is redirected toward the chip surface,where it can leave the semiconductor die

 

These high levels of absorption held back the performance of red LEDs. Devices with a peak wavelength of 615 nm, for example, produce a peak efficiency of just 40 lm/W (see Figure 3).



Figure 3: The historic efficiency of red LED brightness. Blue dots indicate volume emitters on an absorbing GaAs substrate and orange triangles indicate the performance of Osram’s ThinFilm LED since 1999

To stop wasting so many photons, we have pioneered and developed ThinFilm technology (see Figure 2). With this approach, in contrast to a conventional LED, the growth substrate is removed and the active epilayers are bonded to a carrier. Absorption of light in the substrate is negated by inserting, via deposition, a highly reflecting mirror between the epilayer and carrier. And in addition, the ThinFilm LED features: A thinner active region that reduces absorption; superior light extraction, thanks to surface roughening or the introduction of microprisms; and no shadowing, because the current is directed away from the bondpad.

Our development of ThinFilm LEDs, which began in 1998, has had a tremendous impact on the efficacy of red emitters. Although the technology took some time to master, by 2010 we were able to produce 140 lm/W LEDs.

The high efficiency of our ThinFilm LEDs goes hand-in-hand with other strengths. They are also scalable. This means that, in theory, all chip sizes can be made by similar methods, possess similar characteristics, and deliver a level of performance that just depends on the dimensions of the device.

 In addition, these surface-emitting chips deliver a highly desirable Lambertian beam pattern, and can be manufactured in high volumes using cost-efficient design and manufacturing processes.

These results were obtained some time ago, and we believe that our latest generation of ThinFilm die can increase the output power of bare and packaged die by 30-50 percent and 10-30 percent, respectively.

 Towards new highs

During the last few years we have made further strides in terms of efficiency, cost and reliability. This has been accomplished by carrying out a detailed investigation of loss mechanisms and introducing new designs to combat these losses.

One aspect of the LED that has been improved is a reduction in its internal absorption. This pays dividends even if light has to travel ten times through the die before escaping the semiconductor. Absorption has been trimmed by increasing the bandgap and adjusting the doping in some layers.

These steps could also drive up operating voltage and ohmic resistance, so to address this we have improved the conductivity in certain layers.

Another of our improvements is optimisation of contact design, which helps to lower the operating voltage. With more current paths spaced closer together, the electrical current can be evenly distributed across the entire chip. Images of 1mm2 die reveal the dense arrangement of these metals’ current paths, and strong electrical performance despite adjusted doping levels.

 



1mm2 die of latest Osram ThinFilm generation. Surface image (left) and illumination pattern (right)

Thanks to all these enhancements in device design, the latest generation of ThinFilm LEDs set a new performance benchmark. Our 1mm2 die, which was mounted into a laboratory package and optimised for efficient out-coupling, produces high levels of efficacy and wall plug efficiency (WPE) over a wide current range (see Figure 4).

 



Figure 4: Light current curve (top) and efficiency /efficacy over current (bottom) of 1mm2 LED in laboratory package

Both these characteristics peak at a drive current of 50 mA, hitting 201 lm/W and 61 percent and falling to 168 lm/W and 52 percent at 350 mA. The reason behind these declines in efficacy and WPE is revealed by the orange curve, the plot of external quantum efficiency. This has a very broad maximum, with values of 58 percent at 50 mA and 59 percent at 350 mA.

Since the external quantum efficiency is essentially constant, declines in efficacy and WPE must be attributed to increases in operating voltage, due to ohmic resistances within the die.

 Record-breaking efficacies were produced with a red die with a dominant wavelength of 609 nm at room temperature. Realising similarly high efficacies at longer  wavelengths is even more challenging, due to declines in eye sensitivity. But, counter-balancing this, high-wall-plug- efficiency values are easier to realize with longer wavelengths.

We have also produced 1mm2 die housed in identical laboratory package that have a 645nm dominant wavelength and a peak emission at 660nm (see Figure 5). WPE exceeds 70 percent between 5 mA and 60 mA, and falls to 59 percent at 350 mA. Output power rises linearly with current to reach about 437 mW at 350 mA, using a forward voltage of 2.1 V. Due to reduced eye sensitivity, the lumen output at this drive current is just 21 lm.

 



Figure 5: Light output and wall plug efficiency over operating current for a Osram’s 1mm2 die emitting at a peak wavelength of 660 nm

The improvements resulting from out latest ThinFilm technology can be implemented with a very broad process window. This allows for high yield, which in turn lowers overall cost. We have already applied this new technology to chip sizes of varying lengths: 250 μm, 300 μm, 500 μm, 750 μm and 1 mm.

Later this year, 150μm and a 2mm2 die will be added to our portfolio.

The wavelength range, in terms of the dominant wavelength, is currently 590 nm to 645 nm, and 560 nm and 570 nm LEDs will follow in the latter part of 2012.

Commercial opportunities

An already-widespread application is the combination of a highly efficient – but aesthetically unpleasant – cold-white LED with amber or red one, because this lowers the colour temperature to a more visually pleasant and high-quality white. This modification is something of a win-win, because it increases the overall efficiency and the colour-rendering index.

Our breakthrough in red LED efficiency will also aid projection and signalling applications, which value a highly efficient red source. Mobile applications are also set to benefit, with customers enjoying brighter screens, or benefitting from longer battery life. While in industry, lower energy consumption will trim operating expense, and brighter devices will cut the cost of initial installation.

One area where highly efficient LEDs seem assured to make a positive impact is in LED-illuminated green houses. Thanks to incredibly high efficiencies, energy costs can plummet by almost a factor of two compared with conventional lighting; and on top of this, LED-cooling is made easier. For example, our state-of-the-art LEDs can generate 75 percent more light at a given electrical power than a source with 40 percent peak efficiency. And if waste heat is the chief concern, our LEDs can generate two-and-a-half times as much light compared with the present lighting fixtures.

What next?

Wringing out further improvements in efficiency is increasingly challenging as the devices creep towards the ideal goal of 100 percent. Efforts will undoubtedly continue in this direction, nudging efficiency from 60 to 70 percent, but this will be carried out in conjunction with programmes to improve production and processes.

There is also a need to improve LED performance at high currents and temperatures. Characterizing devices at room temperature is common practice, even though many devices actually operate in far hotter environments, such as 100°C. To address this, we are working to optimise output power at this temperature, as well as reducing the shift in operating parameters with temperature.

 Another challenging area, where issues continue to evolve, relates to optimisation of the coupling of light out of the LED. Research is ongoing to extract every generated photon, at both the package and the die level. New die configurations promise to deliver significant progress in this direction.

As we continue to develop and improve our long-wavelength LEDs using the approaches just outlined, our devices will be able to serve an increasing broad portfolio of applications. Further improvements are imminent, so don’t be surprised if its soon time for us to issue another ‘red alert’! A brighter future surely lies ahead for all of us.

© 2012 Angel Business Communications. Permission required.