LpR Article | Technologies | Jun 21, 2016

Optimization of SSL LED Devices by Osram Opto Semiconductors

There are different levers to improve LED performance. Ralph Bertram, application engineer for SSL products, and Alexander Wilm, key expert in the SSL application engineering department at Osram Opto Semiconductors, have a look into the device itself and explore the principal limitations of the different contributors to device efficacy. They show how to balance driving current and chip size to define the right LED, and explain the impact on performance and system costs of LED packages for the intended lighting task approaches for lighting strategies.

The performance of LEDs depends on different parameters such as electrical efficiency, internal quantum efficiency and package extraction efficiency. All of these aspects have limitations in principle that have to be examined to optimize LED performance. The right balance between driving current and chip size is the key to reconciling efficacy and cost targets. Other parameters such as luminance and package sizes must also be taken into account when identifying the right LED for the intended lighting task.

The most prominent and also the most important performance parameter is the efficacy of an LED. Although LEDs are small, there are lots of technological challenges inside and many small steps lead to constant improvements in LED efficiency.

When assessing the efficiency of an LED, we need to consider what is known as the efficiency chain, which is shown in figure 1. It consists of five parameters, namely electrical efficiency (ηelectr), internal quantum efficiency (ηint), extraction efficiency (ηextr), phosphor conversion efficacy and package extraction efficiency (ηpackage). Every LED manufacturer possesses strengths and weaknesses in these parameters. It is reasonable to assume all efficiencies to be at 90% (except that phosphor conversion cannot exceed around 80% due to Stokes Shift). Multiplying all these contributions yields an efficiency of 52% for a white LED, which is a value that can be achieved by R&D level LEDs already.

Screen Shot 2016-06-21 at 08.34.51.pngFigure 1: Features of an LED including the efficacy chain

Typical spectral efficacy of white LEDs is 330 lm/W, meaning that the radiated spectrum of one watt optical power contains 330 visible lumen. This results in a total efficacy of 172 lm/W for the complete LED if all efficiencies are at 90 percent. As always, extracting the last few percentage points from a parameter requires a tremendous effort.

Screen Shot 2016-06-21 at 08.34.56.pngFigure 2: The maximum achievable spectral efficacy of an LED for different CRIs

Paths of Optimization

To improve the performance of LEDs several levers can be used.

Electrical efficiency (ηelectr)
All electrical contacts and conductive parts are ohmic resistors that heat up upon current flow. For better electrical efficiency the conductivity of current-conducting layers and interfaces between different materials has to be improved. The conductivity of transparent conduction layers, in particular, is a field for constant optimization. A trade-off between transparency and conductivity is therefore needed, typically using Indium Tin Oxide (ITO) as the conducting material.

Internal quantum efficiency (ηint)
Another path of optimization is increasing the internal quantum efficiency ηint. The conversion of electrons and holes to photons (“recombination”) is the central process in an LED, taking part in the p/n junction of the diode. ηint depends largely on crystal quality. The key for best quality is the elaborated crystal growth process using MOVPE epitaxy. The thickness of the active layer is defined by the doping profile in the semiconductor. Structures to trap electrons and holes together in the crystal area (“Quantum Wells”) facilitate the recombination process and generate high quantum efficiencies. So optimizing these structures has the greatest impact on the efficacy of an LED.

Light extraction efficiency (ηextr)
Once the photons are generated they need to get out of the semiconductor. However, the refractive index of GaN (gallium nitride, used for blue and green LEDs) is about 2.5 and GaAlP’s index (gallium aluminium phosphide, used for yellow and red LEDs) is about 3.5. Together with the typical refractive index of the encapsulating material of around 1.5 it forms a interface of a refractive index difference resulting in a high amount of total internal reflection that hinders light from getting out. Strategies to improve light extraction include structuring of surfaces, introduction of mirrors or diffuse reflecting surfaces. Also, the geometry of the chip plays an important role: Volume emitting chips have larger surfaces to enable light to exit in different paths. The reduction of refractive index differences by using high-index silicones helps light to exit the chip. Only the right combination of all these improvements will significantly enhance light extraction ηextr.

Light conversion efficiency
In terms of phosphors, there is also room for improvement. Part of the blue light from the LED chip is converted to yellow, green and red wavelengths to generate white light. Here, efficient phosphor materials must be developed and embedded in suitable materials to enhance the conversion efficiency. The converted light is emitted in all directions, including back to the chip and to all package surfaces. Engineering the whole system to recycle as much of this light as possible before it hits absorbing surfaces is essential to improve the conversion process. Although the photon conversion process itself can be very efficient (near unity), around 20 to 30 percent of the light energy is irrecoverable due to the inevitable loss of photon energy in the conversion process. The exact amount depends on the color target. Warm white and high-color rendering LEDs need plenty of red light, having higher losses.

Up to now, only physical parameters have been addressed. In order to match the light spectrum to the sensitivity of the human eye (“spectrum engineering”), the right mixture of phosphors needs to be found. Concentrating the light in the green part of the spectrum where the eye is most efficient, and reducing the red and blue content results in higher spectral efficacies. However, a color target on the Planckian locus (“white light”) is desired as well as good color rendering. A trade-off between efficacy and light quality is therefore needed. In figure 2, the maximum achievable spectral efficacy is shown for different CRIs, calculated for a theoretical optimum light spectrum. A realistic spectral efficacy for the widely used CRI 80 is 300-330 lm/W.

In addition to all the issues mentioned above, parameters such as internal quantum efficiency, phosphor efficiency and, to a lesser extent, resistive losses, deteriorate with increasing temperature. Proper thermal management of the LED chips as well as phosphors therefore help to maintain high efficacy during operation. To facilitate thermal design, LED manufacturers are working on chips and phosphors that can be operated at higher temperatures as well as on packages with enhanced stability. LEDs with an efficacy maintenance of more than 95 percent at 100°C compared to 100 percent at 25°C will soon hit the market and enable tight packing with minimum efficacy impact.

The Importance of Current Density

The biggest lever in LED efficacy, however, is the current density. The reduction in the number of electrons and holes in the device results in stronger confinement of charge carriers in the active region. Because of this, there are few ways for them to escape without generating a photon. LED efficacy consequently increases at lower currents. Below a certain threshold there are not enough electrons and holes to recombine, so efficacy drops again at extremely low current densities. The loss of efficacy with increasing current density is known as “droop” (see figure 3).

Screen Shot 2016-06-21 at 08.35.01.pngFigure 3: LED efficiency as a function of current density (droop)

The latest research results demonstrated the physical effect responsible for the droop. Although the physical mechanism of the droop effect has been well understood, it is far from being overcome. However, the scope of future research projects can now be narrowed, with the focus particularly on measures to eliminate the Auger effect. In LEDs based on the indium gallium nitride (InGaN) material system, the “bipolar Auger effect” is limiting the efficiency that converts charge carriers into light.

By using a bigger chip surface for a required lumen package, the LEDs can be driven in the low current density regime (“underdrive”) to greatly improve efficacy with respect to standard currents. This is most obvious in the following diagram showing the requirements in cost (lm/$) versus LED efficacy (lm/W) of an LED retrofit lamp. The different points that are located around the black line symbolize one LED driven at different currents. Located top left, the LED is overdriven to get more light at the expense of efficacy. If the LED is located bottom right, it is underdriven. This means that a greater efficacy is possible, but also that more LEDs are required to achieve the same light output.

Knowing this, an LED chip manufacturer has two ways to improve the efficacy of LEDs: Either by improving the LED efficacy itself at a given current density or by reducing production costs to get more chip surface for the same price, allowing underdriving to achieve acceptable efficacy.

Evidently, both ways need to be used in parallel, but the focus may lie more on one or the other side for different manufacturers and applications. Most renowned LED manufacturers offer several variants of LED packages containing different chip sizes to enable the customer to choose their optimum operating conditions and light output.

Screen Shot 2016-06-21 at 08.35.05.pngFigure 4: Impact on cost and efficacy by underdriving/ overdriving typical LEDs

LEDs with High Luminance

There are applications where underdriving is not reasonable, simply because a lot of light has to come from a very small source. Examples are LEDs used for tight spotlights, and LEDs embedded in automotive headlamps or in street lighting applications, where luminance and controllability of the light distribution are even more important than luminous flux. But how can the luminance be improved? There are several technologies available to provide white LED light. In the case of phosphor-converted LEDs, two different concepts prevail in the market:

• The first concept uses many volume-emitting blue chips mounted on a mirror, embedded in a phosphor-filled silicone matrix. Since light from volume-emission is also directed to the sides, the chips cannot be packed closely but need space between them to allow the light to escape and be converted by the phosphor. This fundamentally limits the luminance achievable with this type of LED.

• The second concept is based on surface-emitting blue chips with a phosphor-containing layer on top of the chip. Almost 100 percent of the light is emitted from the top surface only. Here, the single chip has the highest luminance achievable and, for example, embedded in automotive headlamps they are driven at extremely high current densities to generate even higher luminance. However, in general lighting applications many of these chips have to be closely packed and driven at normal current densities in order to maintain sufficient efficacy and high light output.

Both volume and surface emitters are available in SMD LED packages; high-power LEDs in both concepts use Chip-on-Board technology (CoB).

As an exemplary calculation, for a 4,000 lm light source 42 packaged LEDs with a common chip size are necessary, covering a light emitting surface (LES) of 28 mm in diameter. To achieve a 24° beam angle, a reflector of 120 mm in diameter would be necessary. Moving to larger chip sizes or, alternatively, to volume-emitter CoBs, a reduction of the LES to 19 mm can be achieved, enabling 24° optics to be smaller than 100 mm. As discussed above, only direct mounting of surface-emitters on a metal core board enables tighter spacing of the chips. This results in the same flux at 13 mm in diameter, enabling the optics to be 60 mm or tighter beam angles to be achieved. Due to the space required for the placement and wire-bonds, they are still not packed with maximum density. Further improvements in chip assembly and connection technologies have been developed at a research level and 4,400 lm from an LES of 9 mm has been demonstrated recently.

The Right LED Package for Each Application

Screen Shot 2016-06-21 at 08.35.10.pngFigure 5: Light emitting surfaces (LES), when using different LED types, indicated by dashed lines – from LES 28 to LES 19 down, to LES 13, and to the LES 9 lab demonstrator (from left to right).

With this overview of the factors that influence LED efficacy right now, there is still a variety of packages to choose from. In order to decide on the perfect solution, the following questions have to be answered:

• Is the focus more on efficacy or on cost?

• Are there special requirements with regard to optics?

• Are high operating temperatures and/or is long lifetime important?

• What is the required level of robustness? Are extreme conditions expected (humidity, pollution, high temperature differences)?

• Are there restrictions to the assembly process? Is SMD soldering an option?

Starting from the easiest conditions, if there are no special requirements Sapphire-based volume emitters currently perform best in terms of the cost/efficacy ratio. Built into high-volume backlighting packages there is almost no match in lm/W versus lm/$. Is it really? Currently, multi-chip SMD packages can save even more on both package costs and assembly costs. Delivering up to 500 lm from one package, they are more cost-effective than most CoBs. In addition, the package material is more stable and can withstand higher temperatures, and they achieve longer lifetimes.

If higher luminous flux levels are required or SMD assembly is not an option, CoB LEDs are the right choice. They also offer good optical properties and many accessories such as holders and lenses that are available as standard components.

However, if it comes to very long lifetimes and rough conditions, there is almost no alternative to ceramics based high power LEDs. These exhibit almost perfect stability against outdoor conditions, and even in tunnel lighting applications, where corrosive gases are present, they show no signs of corrosion. Unlike volume-emitter chips, they do not need a highly reflective mirror below the chip, which is usually made out of corrosion-sensitive silver. Instead, the complete LED is made from inorganic material such as ceramics and silicone, and the electrical contacts are gold-plated.

The Best Ways to Improve LED Performance

As already stated, there are many levers that can be used to improve LED performance. On the one hand by improving LED efficacy using physical parameters such as electrical efficiency ηelectr and phosphor conversion efficacy ηpackage that influence LED efficacy. These parameters get worse with increasing temperatures so manufacturers are working on LED chips and phosphors that can be used at higher operating temperatures. The most important aspect for improving LED efficacy is the current density: LED efficacy increases at lower currents but may not fall under a certain threshold because of the droop effect. Better LED efficacy can therefore also be achieved by reducing production costs to get a greater chip surface for the same price. On the other hand, the best package for an LED is needed to get the best performance. For up to 500 lm one can choose a multi-chip package that is more cost-effective than most CoB LEDs. For higher luminous fluxes or to avoid SMD assembly, CoB LEDs are the right choice because of their excellent optical properties and standardized accessories. For maximum lifetime and harsh conditions there is no alternative to ceramics-based packages.

As can be seen, there is no one way to optimize LED performance, but many, depending on the applications and the conditions. One can therefore focus on one or two parameters or exploit as many as possible to get the best performance from the LEDs.