LED Droop: The Role Of The Phosphor
Phosphors contribute to droop, but their energy-sapping impact can be minimised through LED design, or by casting them in a ceramic form.
BY OLEG SHCHEKIN FROM LUMILEDS
During the last decade, there has been a widespread, on-going debate over the origin of droop.
Attempts to uncover the cause of droop have focused on the blue-emitting chip, which overlooks one of the key processes in the solid-state light bulb. As white light is produced by mixing the blue emission from the chip with that generated by an optically pumped phosphor, it is crucial to consider whether the phosphor is also prone to droop.
Figure 1. Efficiency droop is more severe in white LEDs than it is in blue-emitting chips.
One way to illustrate that droop in white LEDs can be appreciably higher than that produced by the blue pump LED alone is to compare the normalized external quantum efficiency of a warm-white, 1mm2 thin-film-flip-chip LED and its constituent blue pump LED at a range of drive currents (see Figure 1). Pulsed driving conditions distinguish between the impact of drive current density and thermal effects, and reveal that droop is significantly more severe in a white LED than a blue-emitting chip.
These measurements have led our team at Lumileds to scrutinise Europium-doped red nitride phosphors, which are commonly used in state-of-the-art LEDs to provide emission at longer wavelengths. This investigation involved measurements of quantum conversion efficiency for the LED phosphor at different temperatures and different intensities of blue light irradiance. To distinguish between the influence of temperature and irradiance, we employed pulsed excitation. Laser pulses were long enough to account for phosphor luminescence rise and fall times, and heating was avoided with a 1 kHz duty cycle.
Our study has not been limited to the europium-doped nitride red phosphors, and includes the commonly used cerium-doped aluminium garnets. Both phosphors suffer from droop, but this is more pronounced in the europium-doped nitride reds "“ here phosphor photo-quenching is severe enough to contribute between 20 percent and 25 percent of the droop in warm-white LEDs. Compounding matters, the phosphor-related droop produces a shift in spectral output towards higher colour-correlated temperatures.
As cerium-doped aluminium garnet is less prone to droop than europium-doped nitride red phosphors, at first glance it would appear that the solution is to use this in white-emitting LEDs. But such an approach is not practical, because the spectral output of this phosphor is not suitable for realising a high colour-rendering index. A better way forward is to understand the limitations of europium-doped nitride red phosphors and devise approaches to minimising the droop associated with them.
Figure 2: The droop in efficiency is accompanied by the shift in white LED spectra, which gives insights into the process of photo-quenching in phosphors.
To gain greater insight into the underlying physics of europium-doped nitride reds, and the impact of photo quenching on LED performance, we prepared (Ba,Sr)2Si5N8: Eu2+ powders with varying europium concentration. Our quantum efficiency measurements revealed that the rate of quenching in (Ba,Sr)2Si5N8:Eu2+ increases with temperature and europium concentration (see Figure 4). These findings can guide phosphor material and LED architecture design, and lead to a better understanding of the physics of the quenching mechanism.
We have fitted the experimental rate of phosphor quenching in our samples with a model for radiative and non-radiative recombination processes. This gives a close fit when using a term that is close to the quadratic of the concentration of excited Europium activators. This model replicates results for different concentrations and thicknesses, so long as the non-linear, non-radiative coefficient is kept constant.
Figure 3: The drop in quantum efficiency with excitation varies with the phosphors used in state of the art LEDs.
These insights suggest that of the two most probable candidates for the non-radiative processes âˆ' Foerster/Dexter cross relaxation and excited state absorption "“ it is the latter that is the likely culprit in phosphor quenching. As excited state absorption is excitation dependent, it is similar to the very familiar non-linear process of droop in InGaN quantum wells. This intrinsic loss mechanism for the phosphor cannot be eliminated, but it is possible to minimise its impact via engineering at the material and device level.
Figure 4: Photo-quenching in Eu2+ red nitrides shows strong dependence on temperature and activator concentration.
Our experiments point to an approach for reducing phosphor droop âˆ' cutting the activator concentration. The downside is that to realise a desired colour point, there must be a longer optical path for pump light through the phosphor. In practice, this means thicker or heavier-loaded converting layers, which lead to excessive scattering and heating and ultimately lower LED efficiency.
One way to mitigate these unwanted consequences is to turn to ceramic phosphor materials. Their attributes include excellent thermal properties and greatly reduced optical scattering, which allows for thicker converting layers without efficiency penalties. This is a solution that we have adopted in a number of products that feature our Lumiramic phosphor technology (see Figure 5).
Figure 5: Lumiramic phosphor technology is well suited for high power high temperature applications. Here are some examples of Lumileds automotive LED products which leverage the Lumiramic technology.
LED devices can also be designed to minimise the quenching and its increase with temperature (the effect revealed in our measurements (see Figure 4)). The key is to reduce blue light irradiance and to draw the heat out of the phosphor layer as effectively as possible, while maintaining a low overall temperature. An illustration of this mitigation of photo-thermal quenching through LED design is provided by a comparison of two die-on-ceramic LED packages "“ one is a thin-film flip-chip, and the other a flip-chip architecture. With the thin-film flip-chip design âˆ' and any thin-film LED architecture for that matter âˆ' blue light is extracted through one side of the die. So, due to this, phosphor particles are positioned near the emitting surface of the die, in a thin layer of silicone.
In contrast, with a flip-chip architecture, the phosphor-filled converting layer can be placed on all five sides of the transparent sapphire substrate. This means that the blue light emitted by the active region of the pump LED can be distributed over a larger area than it would be with a thin-film design. What's more, with a flip-chip, the heat of down-conversion is conducted through a larger contact area.
The superiority of the flip-chip architecture over the thin-film flip-chip design is proven in our measurements of normalized optical power output (see Figure 6). As expected, the reduction in the intensity of the pump light on the phosphor, as well as better phosphor heat-sinking in the flip-chip LED, enables this type of design to be less afflicted by droop than its thin-film flip-chip cousin.
For both designs, droop is more prevalent when the device is driven continuously, rather than in pulsed operation. That's because the device driven continuously runs at a higher temperature, and photo-thermal quenching of the europium-doped nitride phosphors is more prevalent.
It is worth noting that at 1.5 A, which is a typical maximum operating current for a 1 mm2 active-area high-power LED, although the junction temperatures of both classes of devices are within 5Â°C of the estimated 125 Â°C, the difference in phosphor temperatures is significantly higher. This contributes to the difference in light output and efficiency between the two architectures, which is more than 15 percent.
The merits of a lower phosphor temperature and a lower irradiance of phosphor are very welcome, because they enable the flip-chip design to maintain high efficiency into even higher current densities.
Figure 6: Thin-film architectures excite and dissipate phosphor heat through "˜one side', roughly equal to the pump-die area. Multiple sides of a flip-chip allow lower average irradiance and a larger area to conduct out the heat of conversion for the same die area as thin-film. Thanks to these benefits, in DC operation, the five-sided flip-chip emitter is over 15 percent more efficient at 1.5 A.
The flip-chip design is not the only architecture for reducing photo-quenching. There are also the remote-phosphor configurations that are often seen in lighting modules, and mid-power LED or chip-on-board-type emitters. In these architectures, blue light is distributed over a large phosphor volume, which helps to maximize efficiency. The downside is a compromise in source brightness, limiting emitter usability.
For high brightness applications âˆ' such as automotive LEDs, laser-based sources, and forms of architectural lighting that require narrow, collimated beams âˆ' photo-thermal quenching of phosphors is a basic limitation. In these applications there is a need for sources with the lowest etendue possible, which means minimizing the emitting area and the angle of the source.
In short, as photo-thermal quenching is a basic property of phosphor materials, it will impact LED efficiency. While LED architecture design can be used to mitigate this form of droop, ultimately, the best solution is at the material level. We have shown success in this area through the use of ceramic phosphors.