The light-emitting diode (LED) has revolutionised the look of electronic products, with each new extension in the range of colours on offer providing an opportunity for designers to set new fashions. It's hard to find a consumer product that does not sport a blue LED – thanks to the rise of nitride-based devices in the past ten years – where once red and green prevailed.
The biggest revolution is coming in area illumination. LED manufacturers are still some way from the target brightness, but Shuji Nakamura, the developer of the production breakthrough that made blue LEDs possible and head of a research team at the University of California at Santa Barbara, reckons that deploying 150lumen/W LEDs in place of incandescent lightbulbs worldwide would allow 380 power stations to be decommissioned and save $100billion in energy costs. LEDs would even save energy compared to cold-cathode fluorescent lighting – the current mainstream alternative to incandescent bulbs. The saving could be up to 50% in that case.
LED backlights are already helping to extend the battery autonomy of laptops and mobile phones, partly through their greater efficiency and partly through better control over their brightness: the display driver can reduce dynamically the output of LEDs that sit behind the comparatively dark parts of an LCD.
Getting to the magic 150lumen/W figure for LEDs, however, will take some time. While experimental devices, such as those produced at UCSB, have an efficiency of around 170lumen/W, production devices are currently at around 70 to 80lumen/W and tend to lag by around 50% the 'hero' devices fabbed one at a time in a lab.
The first demonstration of LED behaviour was reported by Oleg Losev in 1927, but was not considered to be useful until after the development of the solid-state transistor. Researchers at Texas Instruments patented the first infrared LED in 1961, followed quickly by the development of a red LED by Nick Holonyak at General Electric.
The LED is exactly what its name describes: a diode that emits light. The photons are emitted as the result of conduction-band electrons arriving from the n-type half recombining with holes in the valence on the p-type side of the diode. The direct bandgap of an LED allows a photon to be emitted as a result of the loss in energy by the electron as it drops to a valence band. Conventional silicon diodes do not exhibit the same behaviour because of the material's indirect bandgap. Tuning the bandgap will provide photons of the desired colour.
Put together a red, green and blue LED and you can have a natural-looking white LED, or one suitable for making a tunable mood light for the home. Unfortunately, the materials available to engineers for the first 30 years made obtaining one of those colours – blue – very hard indeed.
The breakthrough came with a shift in chemistry, away from familiar III-V materials such as InGaAs and towards nitride-based crystals. Different combinations of aluminium, indium, gallium and nitrogen make it possible to tune the bandgap all the way from 6eV – deep in the ultraviolet range – down to around 2eV, the energy needed for red LEDs.
A further advantage of the nitride-based LEDs is that the GaN lattice is very forgiving compared with the fragile III-V materials used in most red and green LEDs. The GaN-based LEDs will still emit light reliably with a relatively large number of crystal dislocations.
There was only one problem facing engineers in the early 1990s: there was no way to make a sufficiently high-quality crystal work as a high efficiency LED, even given GaN's tolerance for dislocations. Working at Nichia in Japan, Nakamura developed the two flow metal organic chemical vapour deposition (MOCVD) process that made it possible to grow working LEDs. The number of dislocations did not reduce – it was still seven orders of magnitude higher than that of zinc selenide, the nearest competitor for blue LEDs – but hole mobility improved dramatically, allowing efficient carrier recombination and, with it, photon emission.
Normally, MOCVD uses only one gas stream: a reactive gas that carries the main materials to the target substrate. In Nakamura's system, a second inert gas blows down onto the substrate. This second flow suppresses the thermal convection that results from the high temperature – around 1000°C — needed to grow GaN crystals.
Once two-flow MOCVD had been developed, researchers quickly made variants of GaN, moving to compounds such as indium gallium nitride, allowing the production of blue, and then green, LEDs that were much brighter than zinc selenide versions and with a much longer lifetime.
With a material that could emit light in the UV range, it became possible to produce white LEDs. A phosphor coating on the LED's surface converts the emitted UV into white light. However, three-colour LEDs provide the most flexibility: changing the relative intensity of the colours makes it possible to control the colour temperature to provide a more convincing daylight simulation than incandescent or fluorescent lamps and greater tunability for mood lighting. The same principle can be extended to control panels where colour-changing LEDs can provide extra feedback to a user or operator. The downside is simply one of expense: the extra LEDs and the separate drivers needed for each of the them.
There is also the quest for efficiency. It's split into two main components: quantum efficiency and light-extraction efficiency. The first quantifies how many of the energy transfers from electrons result in emitted photons. The ratio is often quite low and the situation gets worse at higher applied currents, which is bad news for high-efficiency lighting. There is also a question over the precise mechanism. Auger interactions and interband absorption are today's primary candidates for these wasteful transfers.
Without a clear culprit, increasing efficiency is problematic. However, in both cases, restricting carrier density looks to be a promising avenue. One way to do that in GaN LEDs is to change the orientation of the crystal.
Today's nitride-based LEDs are cut along the c-plane. This exposes the hexagonal faces of the GaN unit cell, rather than its rectangular sides. There are two other cuts that could be used – non-polar and semi-polar. The reason for the naming comes from the way that the hexagonal symmetry of the c-plane – which is under some stress from differences in lattice constants between the doped layers of GaN – results in a high piezoelectric field, pulling electrons and holes to opposite sides of the quantum well. The problem gets worse at longer wavelengths because the quantum well has to be made larger – which is why it has proved difficult to make yellow LEDs using nitride-based materials.
The other proposed directions reduce piezoelectric polarisation significantly. The m-plane cut, which chooses one of the rectangular faces, cuts the polarisation to zero. It is not the only option, however. The semi-polar direction cuts through the unit cell at an angle of 60°. It exhibits some polarisation, but this is also much reduced compared to the c-plane device – hence semi-polar.
The problem comes in making crystals with the required orientation. Dislocations are high in c-plane GaN and get worse with non-polar crystal growth. And the process can be incredibly slow; using conventional techniques, it took the team at UCSB six years to grow a crystal that measured 7mm across – the GaN grew at around 30µm/day.
However, one approach could yield industrial quality 2in or 3in wafers. The process is analogous to the high-pressure, high-temperature growth of quartz crystals. Instead of water, the ammonothermal process uses ammonia in a supercritical state. The process looks promising and is being used to produce devices but the substrates are still small – still less than 1in across.
The second problem with efficiency – light extraction – comes from a combination of shape and materials. Internal reflections in these semiconducting materials are high, many of the photons bounce around before being re-absorbed and never escape. The perfect LED would be a sphere a hole in it, with the electrical contacts almost touching at the base of that hole. This is not a practical shape. One alternative is to texture the surface of the LED into an array of cones. Photons transmitted towards the surface would bounce a few times along the internal surface of the cones until the angle of incidence is low enough to allow transmission.
In parallel with the quest for higher efficiency, nitride-based LEDs have pushed further into the UV end of the spectrum since the development of blue InGaN LEDs almost 20 years ago. Rather than illumination or indication, these LEDs are used for sterilisation – possibly eliminating chemical antibiotics – as well as tracking biological contaminants.
The key to extension into the cell-killing deep UV region is the presence of aluminium in the lattice. As the fraction of aluminium increases, the emitted wavelength of AlGaN LEDs decreases from around 350nm to 200nm. Unfortunately, depositing aluminium is not easy – the crystals tend to have a large number of dislocations because the lattice does not fit well with a sapphire substrate.
There is a further problem. Although simulations indicate the internal quantum efficiency can be higher than 50%, the extraction efficiency is pretty low – limited by internal reflections in the sapphire substrate, as well as absorption by the p-doped GaN regions themselves. However, researchers believe it is possible to reduce their effect with better crystal engineering.
While chipmakers struggle with growing crystals of GaN, some researchers have shifted their attention to two different groups in the periodic table and a much more common material: zinc oxide. ZnO – a II-VI material, rather than a III-V – has a similar bandgap to GaN. It is easier to etch and, perhaps most importantly, it is a lot easier to make large-area substrates. Toxicity, however, might prove to be an issue. Although ZnO itself is not much of a danger, two of the important dopants for today's experimental ZnO are cadmium and beryllium, although other dopants under investigation include more benign metals, such as magnesium.
Most devices made so far have been UV LEDs, with some reaching into the blue and green parts of the visible spectrum – so, if they can be made in volume, they are likely to be paired with phosphors for use in white LEDs or for the same medical applications as the latest generation of nitride-based devices. One potentially big advantage of ZnO is that it is transparent, so may find applications where see-through electronics are desirable or where the LEDs are formed along with other active transparent devices.
The downside to ZnO is the lack of a high-quality p-type material, so earlier work on making LEDs concentrated on heterojunction devices, depositing n-type ZnO on layers of GaN, silicon or other conductive oxides. More recent work has used careful growth of p-type ZnO using layer-by-layer techniques such as molecular beam epitaxy.
In contrast to GaN LEDs, substrate size is far from the biggest concern for the organic LED (OLED). By printing or spin coating liquids onto a plastic substrate, it is possible, at least in principle, to make single LEDs the size of wall posters. Displays based on OLEDs promise to be simpler and, because they use light emissions rather than absorption, provide much brighter, stronger images and more energy-efficiently than backlit LCDs.
The structure is conceptually very simple: just place a layer of light-emitting material between two electrodes – at least one of them being a transparent material such as indium tin oxide or a conductive polymer.
OLED displays have had a rocky start. Sony launched into the market with a small, expensive but attractive OLED-based TV that it pulled from stores just months later. Samsung, however, has continued investment in OLED displays for mobile phones and, if yields hold up, could be a major challenger to small LCDs in the near term. Ultimately, light wall-sized TVs should be the result from ongoing work into OLED technology.
The OLED could revolutionise the way architects think about lighting. Instead of discrete light bulbs, you have sheets of glowing plastic. Industrial designers such as Ingo Maurer in Germany have created novel designs, including a light in the shape of a soccer ball.
One consequence of a shift to OLED-based lighting would be a change in how the products are made and supplied. Most of the raw materials would have to be shipped and used by the maker of the end equipment – on web or sheet-fed printers they have bought or leased – rather than pre-assembled in dedicated factories, as is the case with conventional inorganic LEDs.
It will take a while before OLED lighting becomes commonplace. Moisture in the air is proving to be problematic for flat-panel and flexible lighting made from OLEDs. Manufacturers are struggling to find effective barrier materials that can prevent water from destroying the electrodes in OLEDs and causing them to fail prematurely.
Many of the known barrier films allow too much water through. Without any protection, a typical OLED would fail in less than a day in typical northern European humidity. The light-emitting films are not the problem; rather, it is the charge-injecting layers. Typically, the cathode degrades on contact with water, with pinholes forming that steadily grow and cause the surrounding area to go dark.
Recent work in European research projects has increased the lifetime of barrier films up by four orders of magnitude, but it is still a struggle to get to the lifetime expected for lighting panels that can be become part of a building's infrastructure.
Although most printed displays and lights are based on organic materials, future products may combine them with inorganic compounds, such as ZnO. There are a number of inorganic materials, oxides in particular, that do not need high-temperature processing and which can be spin coated or printed onto a flexible substrate in the same way as organic small molecules. Some researchers have combined OLED polymers with ZnO, claiming they can achieve improvements in luminescence from the combination as well as better resistance to moisture.
The future of the LED could well be with these hybrid chemistries as manufacturers strive to boost luminance and slash production costs – and so move away from the relatively expensive and energy-intensive steps of crystal ingot growth and clean-room processing.