Intense infrared LEDs

Powerful illumination for optimum image quality. Applications such as video surveillance of public places call for infrared LEDs with a wavelength of 850 nm and extremely high optical output.

BY MARTIN BEHRINGER, MARKUS BRÖLL, JÖRG HEERLEIN, CLAUS Jà„GER AND THOMAS KIPPES FROM OSRAM OPTO SEMICONDUCTORS

There are parallels between what is happening in the mainstream LED industry and that associated with the 850 nm chip, which emits at wavelengths far enough into the infrared to be virtually invisible to the human eye. Both the visible and the infrared classes of chip can serve myriad applications, and makers of these emitters are trying to set a new benchmark for performance to win sales in new markets.

For the 850 nm LED, one of the biggest markets is providing illumination for infrared cameras, which can be used for the likes of various forms of surveillance, license plate recognition and industrial image processing. In addition, the infrared LED can be deployed in 3D systems and biometric identification "“ all these applications are outlined in more detail in the panel "Opportunities for infrared LEDs".

To address all these applications, our team of engineers at Osram Opto Semiconductors launched infrared LEDs and subsequently expanded this portfolio. We have many sources operating at 850 nm, an ideal wavelength for infrared cameras and gesture-recognition, and we have also brought to market a cousin emitting at 810 nm, designed to deliver more reliable results for iris scanning.

In 2008, we brought our first 850 nm high-power LEDs, housed in the Dragon package, to market. Since then we have increased output at 1A from 440 mW to 800 mW (see Figure 1). Even high powers are possible with our nanostack technology, which we have developed in-house. By turning to two emitters per chip, single-chip LEDs can exceed an output of 1W by a significant margin.

Figure 1. Since the launch of Osram's first 850 nm high-power LED in 2008, optical output has almost doubled, and even trebled, when stack chip technology is considered.

Of all the gains in output over the last six years, the greatest hike came in 2014, with the introduction of the Oslon SFH 4715A (see Figure 2). This boosted the emitted power by 30 percent, with new developments in chip technology accounting for two-thirds of the increase and the remainder resulting from refinements to package design. Another attribute of this particular chip is a 48 percent electrical efficiency, defined in terms of electrical input to optical output. This level of efficiency makes the SFH 4715A the most efficient infrared power LED on the market at a 1A drive current.

Figure 2. The optical output of the 850 nm emitter has been increased by 30 percent, and efficiency by 20 percent. Improvements in chip technology account for two-thirds of this increase, with the remainder resulting from an improved package.

This record-breaking LED and its predecessors are based on an AlGaInAs thin-film technology (see Figure 3). This architecture is superior to that of a conventional chip, and delivers two major benefits: far less light is absorbed in the chip, enhancing optical output; and emission occurs from almost all of the top surface of the LED, eliminating the need for reflectors within the package. With this design, the area of this surface is proportional to the output from the chip − for our high-power devices we use a surface of 1 mm².

Figure 3. Thin-film chip technology: A highly reflective mirror between the substrate and the epitaxial layer prevents light being absorbed by the carrier; a roughened surface extracts the light efficiently (a). Thanks to improvements in the mirror, at the p-contacts and in the epitaxial layers, the optical output of the chip (measured in the package) has been increased by 20 percent (b).

One defining feature of our thin-film technology is the highly reflective dielectric mirror between substrate and light-emitting region. To create such a structure, the light-absorbing substrate is removed after epitaxial growth, and is replaced by a new carrier and mirror (see Figure 3). Turning to this geometry lowers light absorption, but radiation can still be trapped in the chip, bouncing back and fourth between the mirror and the surface. So, to enhance extraction, the surface of the chip is roughened to increase light output.

The spectral width of our 850 nm LEDs is relatively narrow. This is intentional: With the half-value width of the emission peak at just 30 nm, the lower wavelength tail cannot deliver visible illumination (note that only 4 percent of peak output occurs at 800 nm, implying that the latest chips do not produce any red glimmer detectable to the human eye).

Over the last few years we have devoted much effort into increasing the current-handling capability of our high-power LEDs. One goal has been to develop components that could be driven by either 5 A pulses or a DC current of 1 A. After realizing this by trimming substrate thickness and consequently thermal resistance, the target for our latest generation of thin-film chips has been to retain this current-handling capability while delivering high efficiency, especially when the LED is driven in DC mode.

These efforts have not involved reducing the light path, as this is already as short as possible, thanks to the roughened surface. Instead, there has been a cutting of the internal absorption of the epitaxial layer and an increasing of the reflectivity of the mirror between the substrate and the epitaxial layers. On top of this, by reducing the area of the p-type contacts, we have minimized absorption losses at the metal-semiconductor boundary.

Making these electrical and optical refinements has driven up device performance. Brightness, measured with a chip mounted in a silicone-encapsulated  package without a lens, is up 25 percent. The smaller surface of the p-contacts impairs forward voltage, which rises by 7 percent at 1 A, but the key figure of merit is an 18 percent increase in electro-optical efficiency. Testing in a lab produces impressive results, including a record efficiency of 72 percent at 70 mA for a chip with a mounted lens. Optical output can hit 930 mW at 1A, and when mounted in our Oslon package, the latest 850 nm thin-film flip chip provides a 10 percent increase in efficiency and a 20 percent hike in output compared to its most recent predecessor. Another of our improvements is to optimize the n-contact of our new chip for continuous current and high efficiency. This involved switching from a centrally located bond pad to a contact in the corner, with current distribution via a thin metal grid (see Figure 3). The new geometry eliminates the bond wire from the optical path, allowing more uniform illumination to be delivered by the chip.

Perfecting the package

An additional 10 percent increase in optical output has been driven by improvements to the package. This was accomplished by changing the material composition of the silicone lens in the Oslon Black series. Results are an increase in output efficiency and a trimming of scatter losses. This has led to the release of improved Oslon Black sources with two different options for the lens: beam angles of ±45° or ±75° (see Figure 6).

Figure 4. Angular distribution of the optical output of the Oslon Black series for a beam angle of ±45° and ±75° compared to a Lambertian emitter (± 60°) in the Dragon package. The Oslon Black series is one of the most compact designs for infrared LEDs with an optical output of more than 500 mW. It has been used by Osram for high-power LEDs of all colours and chip technologies, and by employing a silicon lens specially adapted to the package and the chip, it can emit around 15 percent more optical output than a comparable low-profile LED, such as the Dragon. Other strengths of this package are good heat dissipation, which ensures that the chips are highly resistant to aging, and a low thermal resistance of typically 6.5 K/W that is optimised for continuous operation with high currents.

Figure 5. One of the most powerful infrared LEDs with a wavelength of 850 nm: the Oslon Black series SFH 4715AS features a 1mm² thin-film chip in stack technology and emits an optical output of around 1370 mW at a current of 1 A.

Figure 6. The output of the 850 nm Oslon Black remains virtually linear for moderate pulse lengths of 100 µs up to a current of 5 A.

The narrow beam angle is suited to coupling to external lenses, allowing the resultant beam angle to be shaped to suit the particular application. Meanwhile, the ±75° version enables far broader illumination in the near field over distances of only a few meters, making it well suited to gesture recognition. This option is also ideal for forming compact designs that realise very narrow-angle emission characteristics, thanks to reflector-based optics that fill the reflector with light, due to the wide emission of the +/-75° design. Such a source can produce a narrow light beam for sending light over long distances.

Our latest version of the infrared Oslon Black features the latest-generation chip, which is distinguishable by the contact in the corner (see Figure 5). This particular design is ideal for optical imaging applications, because the bond wire and contact are no longer in the centre of the beam, thereby eliminating interference effects from the detected light signal. Turning to the combination of the Oslon package and the latest thin-film flip chip increases current electro-optical efficiency by 20 percent at 1 A and optical output by 30 percent.

Setting up a stack

To squeeze even more light out of our chips, we have turned to our nanostack technology, which we have previously used to improve laser diode performance. Brighter chips result from two p-n junctions in the epitaxial layer, as this doubles the number of regions generating light, leading to a higher total output.

Several years ago we transferred our stack technology from laser diodes to infrared LEDs, and this increased output from the component by 70 percent (see Figure 7). In 2009, we started to offer 850 nm LEDs with our chip stack technology on a range of LEDs "“ those with edge lengths of 1 mm, 0.75 mm, 0.3 mm and 0.2 mm. Since 2013, the portfolio has grown to include large-format 940 nm stack chips, and will soon include the SFH 4715AS, a latest-generation thin-film flip chip with stack technology that delivers 1370 mW at 1 A.

Figure 7. Nanostack chip technology features two p-n junctions in one chip. A component with a stack chip has an output that is 70 percent higher for the same forward current.

Another breakthrough associated with this product is that it is the first infrared thin-film chip on a silicon substrate. We have used this foundation for some time in our blue and green InGaN chips, but our infrared emitters have been traditionally mounted on a germanium substrate. Switching to silicon slashes the typical thermal resistance of 5.5 K/W by around 1K/W, and thus reduces the rise in LED temperature, allowing these devices to be driven at higher currents before they hit the maximum allowed operating temperature. Alternatively, the component can be run under the same operating conditions as before, as this will make it more efficient and longer lasting than its predecessor.

Our plans for the future include increasing the range of wavelengths of devices exploiting our new generation of thin-film technology, so that we increase the portfolio of higher power, more efficient products.  These high-power infrared chips will migrate to a silicon substrate, with packages further optimized for operation at high currents. We will also listen to our customers, and if there is demand for new versions of infrared LEDs with different emission angles, we will bring such devices to market. The opportunities for this class of LED appear to be endless, and we will continue to innovate and set new benchmarks for these infrared emitters.