Novel chip geometries, such as triangular and hexagonal devices, can deliver massive increases in light extraction by cutting optical confinement in both the vertical and horizontal directions, says Hoi Wai Choi from the Semiconductor Lighting and Display Laboratory at The University of Hong Kong.

 

LEDs have evolved from candela-class monochromatic point sources that are only good for indicator lights to 120 lm/W high-power, white-light lamps widely used as LCD backlighting. The tremendous progress can be attributed to great progress in material epitaxy skills, device design and processing techniques, together with advanced packaging strategies. With dozens of advanced features loaded into today’s LED chip, isn’t it strange that chip geometries have hardly changed over the last decade?

Rectangular and square geometries are probably the norm because it is easy to process chips in these shapes. Since sawing involves only continuous linear cuts, rectangular chips seem the logical way to go. However, geometries with high degrees of symmetries generally do not favour light extraction.

Maximising light extraction is one of the keys to enabling the LED to reach is full potential. Inside this device, a significant proportion of the light can be trapped, because any light that impinges on the semiconductor-air interface beyond the critical angle will be completely reflected. The value of critical angle is governed by the difference in the refractive index of the two materials, which is quite large for the pairing of GaN and air. Today light extraction efficiencies rarely exceed 20 percent, but if this figure can double, LEDs will then be in a great position to dominate the general lighting market.

Excessive reflection of light at interfaces results in light confinement and subsequent re-absorption, resulting in heat generation. So improvements in light extraction have the added bonus of directly combating heat dissipation issues.

One of the most common approaches to minimise optical confinement in the vertical direction is surface texturing. This process may involve either random roughening, or the introduction of regular and periodic arrays of microstructures and nanostructures, which are also known as photonic crystals.

However, little attention has been paid to lateral optical confinement, which depends on the dimensions and geometry of the LED chip. In the highly symmetrical rectangular or square structure, a light ray that is reflected due to total internal reflection at one interface will most likely be reflected at subsequent interfaces. This can be understood from the fact that all interfaces, or facets, of such chips are either parallel or orthogonal to each other; the incident angle of a light ray remains unchanged after successive reflections unless scattered, so that such confined photons are eventually reabsorbed.

However, it is possible to stop wasting photons in this way. What’s needed is a switch to a chip geometry involving facets that are not parallel to each other, which is an approach that we have pioneered at the Semiconductor Light and Display Laboratory at The University of Hong Kong.

Any shape you want

The enabling technology for realising LEDs of unconventional geometries is laser micromachining. Although lasers have now been widely adopted for chip dicing, they have continued to be used for making linear cuts even though laser beams can easily be steered to carve virtually any shape. Our chip-shaping experiments are conducted with our custom designed and assembled optical setup. The workstation comprise of a thirdharmonic ultraviolet, diode-pumped solid-state laser source that features a collimated beam that is focused into a tiny spot onto an x-y-θ motorised stage. By translation or rotation of the stages controlled by our inhouse developed software, we can form chips of any arbitrary shapes. These are cut from LED dies fabricated on InGaN/GaN wafers grown on c-plane sapphire substrates.

Our belief in the inherent advantages of these novel chip geometries is supported by optical ray-trace simulations. By examining the proportion of light rays that can be extracted from LEDs of different geometries, their efficiencies can be predicted. The results of our set of simulations on LED chips of circular, triangular, squarish, pentagonal, hexagonal, heptagonal, octagonal and circular (the circle is in fact a polygon with an infinite number of sides) geometries are summarised in Table 1.

What stands out from the set of figures is the fact that square LEDs have distinctively lower extraction efficiencies than any other shapes. Considering the fact that LED chips in the market are invariably diced into squares or rectangles, device designers should give serious thought to redesigning the chip.

Our findings from these simulations are backed up by measurements on packaged but un-encapsulated LED chips laser micro-machined into various shapes (see Figure 1). These results show that the square LED is on average 16 percent less efficient than other polygons. Of course, one could argue that the chip packing density, and thus chip count across a wafer, might be compromised. Therefore, we propose the use of triangles and hexagons; such shapes can be closed-packed into any array without sacrificing chip space and thus make economic sense.

 

 



Figure 1. Laser-micromachined LED chips

 

An additional feature introduced into our optical setup enables machining of three dimensional freeform structures. A mirror inserted into the optical path between the focusing optics and the machining plane allows the focused beam to strikes the wafer at an oblique angle. LEDs of truncated pyramidal (TP) structure, as depicted in Figure 2(a), are formed simply by applying four successive oblique cuts along the four edges of devices. On average light enhancement is increased by 89 percent over similar chips with vertical facets, consistent with our theoretical prediction of a 85 percent gain obtained by ray-tracing.

 

 



Figure 2. LEDs of (a) inverted pyramidal and (b) inverted conical geometries

 

The inclined sidewalls serve as reflectors to redirect laterally-propagating photons into the escape cone. Devices of TP-geometry also emit with a wider divergence angle. Combining oblique angle machining and rotary machining, LEDs of truncated conical structures can also be formed (see Figure 2(b)). Subsequently, a reflective metallic layer is deposited onto the bottom and inclined sidewall surfaces of the chip to form an integrated reflector cup.

This structure is particularly beneficial for building white LEDs with homogeneous emission. In a typical white LED, an epoxy mixture containing phosphor particles is coated prior to encapsulation. The phosphor mixture is allowed to reflow in order to cover both the chip surface and sidewall facets, resulting in a dome-shape-like nonuniformity, the effects of which are particularly prominent at the edges. Such thickness non-uniformities give rise to non-homogeneity of colour emission at different viewing angles.

With the integrated reflector cup design, light emission via sidewalls is suppressed effectively. The majority of the light rays are emitted from the top planar surface, so that the divergence angle of the device can be reduced by 16°. Consequently, only the top planar surface needs to be phosphor coated; planar coating produces significantly improved uniformity that paves the way for superior colour homogeneity.

Coupling LED emission to fibres

It is possible to incorporate a hemispherical lens by inverting and flip-chip bonding the truncated conical chip to a package. The lens can be attached via liquid capillary bonding to produce a nanometre scale air-gap with exceptionally high transmission across the interface. At close proximity the lens is also able to capture most of the emitted light, focusing this to a tight spot with high power density. Such an assembly is ideal for optical coupling to a fibre (see Figure 3(a)).

 

 



Figure 3. (a) A fibre coupled hemispherical LED assembly and (b) RGB LED stack

 

To test its efficiency, the aperture of a 2mm-diameter plastic optical fibre was aligned to the focal spot of the LED assembly. Measurements determined a coupling efficiency of 53.8 percent, the highest value reported to the best of our knowledge. fibre-coupled LED sources find uses in a diverse range of applications from optical microscopy to short-range optical communications. Apart from the integration of optics, the monolithic integration of LED chips has also been demonstrated as a viable solution for building colour-tuneable LED assemblies.

Three LEDs chips of the primary colours, cut into truncated pyramidal geometries, were vertically stacked on top of each other in the order of decreasing wavelength from the bottom to the top, as illustrated in Figure 3(b). The inclined sidewalls are mirror-coated to suppress lateral leakage.

With this stacking arrangement, light emitted from chips at the bottom of the stack emits through the upper chips; the top blue LED chip serves as the output window. As the optical paths of the three chips overlap with each other, their colours are naturally mixed along the optical path without requiring additional optics. By individually controlling the intensities of the three colours via bias current, a wide range of colours can be produced with excellent homogeneity and performance (see Figure 4 for examples).

 



Figure 4. Wide range of colors emitted by a stacked LED

 

 



Table 1. Light extraction efficiencies of LED chips with different geometries

Such stacked devices are obviously useful for building high-resolution LED display panels. However, they are also excellent candidates as plain white light LEDs. Being free of phosphor, the stacked LEDs emit white light with high efficiency, excellent homogeneity, high colourrendering index and long-term stability; all of this is achieved without colour-conversion losses.

So let’s get moving and get in shape! With smart tweaks in design, extended functionalities and enhanced efficiencies can be achieved with amazing results, without the need for additional components.

© 2011 Angel Business Communications.

Further reading

X.H. Wang et al. Journal of Vacuum Science and Technology B 27 2048 (2009)

X.H. Wang et al. Journal of Applied Physics 10 023110 (2010)

W.Y. Fu et al. IEEE Photonics Technology Letters 21 1078 (2009)

L. Zhu et al. IEEE Photonics Technology Letters 22 513 (2010)

K.N. Hui et al. OSA Optics Express 17 9873 (2009)