The economic expansion of the past decade has provided funding across the entire spectrum of optoelectronic research and development. As a result, even the oldest "quantum" light sources such as ruby lasers are finding new applications and experiencing a renaissance in design. One such venerable technology experiencing a dramatic rejuvenation is the light-emitting diode (LED).
24/7 for 11 years
By conservative estimate, over 20 billion LEDs are produced annually, but this figure may be dwarfed if the potential of the latest research comes to fruition. Improvements over the past decade have resulted in LEDs of greater brightness with outputs spanning the visible spectrum, offering for the first time the prospect of obtaining white light from diodes. Compared with conventional lighting, LEDs have lower power consumption and lower voltage, longer lifetime, smaller size, and cooler operation.
The Ministry of International Trade and Industry in Japan estimates its country could save the equivalent output of six midsized power plants, and reduce the production of green-house gases, if LEDs replace half of all incandescent and fluorescent lamps currently in use. To back its claim, the Ministry is in the midst of a five-year, x5-billion ($41 million US) research-funding program called “Light for the 21st Century.”
One application in which LEDs are already replacing incandescent lamps is in traffic signals. In the United States, currently about 10% of the red lights in traffic signals use LEDs, which consume 15 W of electricity compared to 150 W typical for the older lamps. It is estimated that the new technology could save nearly 2.5 billion kilowatt-hours annually in the United States through this application alone, and the savings in maintenance resulting from the 100,000-hour projected lifetime more than justifies the higher price. The real excitement, however, and the real potential savings from LED technology derive from the production of white light (see “Solid-state lighting expands as industry awaits illumination”).
Older than they look
The emission of light by a semiconductor diode when forward-biased was reported by an assistant to Marconi in 1907. Potential devices and applications were investigated and patented by Losov in the Soviet Union, but this work was lost after his death during World War II. Modern LED development really began in the 1960s with the push to develop the diode laser.
Early LEDs were gallium arsenide (GaAs) and aluminum gallium arsenide emitting in the infrared. The first red LEDs were gallium aluminum phosphide on GaAs substrates. These devices were inefficient and had low brightness. Over the following decade brightness improved from a feeble 0.15 lumens/W to about 1 lumen/W, bright enough for use in handheld calculators and some outdoor applications. (For comparison, a 60-W incandescent bulb with a typical 15-lumens/W efficiency emits 900 lumens.)
Things started to get interesting around 1988 with the introduction of "high-brightness" LEDs in the red and orange. Breakthroughs in fabrication technology, particularly in metal-organic chemical-vapor deposition (MOCVD), allowed designs with smaller and more detailed structures, such as quantum-well gain layers. Unprecedented control in submicroscopic dimensions resulted in fewer crystalline defects at the interfaces of the alloys. Among other benefits, this allowed the use of gallium phosphide (GaP) as a substrate material, which is transparent to red.
Various gaps
The development of aluminum indium gallium phosphide (sometimes called “allengap”) by Hewlett Packard (Palo Alto, CA) and Toshiba (Tokyo, Japan) in 1990 produced greater brightness and a broader range of emission. As a quaternary material, the “allengap” alloy can be varied so that the energy band may range from 645- to 560-nm output, but still maintain the proper crystal spacing to match the substrate. Allengap LEDs under development have produced output powers as high as 100 lumens in the orange (see Fig. 1).
Research on green and blue LEDs was on a separate track of material development. Green devices based on GaP were introduced quite early, but although they had some commercial success they never achieved impressive light levels. Efficiencies remained below one lumen per Watt, even when aluminum was added to the alloy.
As with green, the first blue LEDs had a measure of commercial success simply because of the attractiveness of their wavelength. Based on silicon carbide (SiC), the brightness and efficiency of blue devices was even lower than for green diodes. The situation changed dramatically with the invention of practical blue LEDs of gallium nitride (GaN) by Shuji Nakamura, then at Nichia Chemical Industries (Anan, Japan).
When life gives you lemons...
Layers of GaN are difficult to grow because of the lack of a substrate material that matches the lattice spacing of the nitride crystal. The most commonly used substrate, sapphire, has a 16% mismatch, whereas advanced epitaxial techniques can accommodate at most a mismatch of around 1%. The 1000°C temperature required by MOCVD had defeated efforts to finesse this problem, which was solved by inserting a 20-nm-thick aluminum-nitride buffer layer grown at lower temperatures to isolate defects at the substrate interface.
Another difficulty lay in the tendency of GaN when doped p-type to form defects in the crystal to “correct” for the introduction of the positive charge. It was discovered that bombarding the doped GaN with a relatively low-energy electron beam causes the p-type dopant to bond instead with hydrogen (which is present as a byproduct of MOCVD). Development was greatly aided by the material strength of GaN, which allows successful LED operation at defect levels that would quickly destroy devices made from other alloys.
Gallium nitride has an energy gap corresponding to ultraviolet wavelengths. By adding indium (In) to narrow the gap, Nichia was able to introduce the first commercial blue GaN LEDs in 1993. These devices had efficiencies more than 100 times greater than previous blue diodes. The addition of more indium produced LEDs in the green that greatly outshone the old GaP devices.
Interestingly, it appears that a characteristic of indium in GaN that has been a hindrance to the development of green diode lasers actually proves beneficial for green LEDs. When the fraction of indium in the InGaN alloy rises above a certain threshold, the indium tends to cluster randomly. These clusters protect the electron-hole recombination that produces photons from the harmful nonradiative defect centers, of which there are many in GaN.
Understanding the behavior of indium clusters is a major goal of nitride research. Research is focusing also on increasing the concentration of p-dopant, and identifying new dopants with lower ionization energy to improve efficiency and allow nitride diodes to handle greater input power. In another development, Cree (Durham, NC) produces its blue LEDs on SiC substrates, which simplifies device design.
Balancing act
With the addition of high-brightness blue and green LEDs to the spectrum, several avenues are now open to produce a white light device. A straightforward concept is to combine three primary-color LED sources. Maintaining color balance is a critical consideration in this approach, especially considering that the emission of the individual devices changes with temperature, current, and age.
Color balance is also a prime concern in a more elegant design aimed at obtaining white light from a single device. This concept uses the shorter-wavelength blue emission from a nitride layer to double as a pump source for separate layers of allengap material, which produce the other primary colors. Again, how the emission of the individual layers changes with time and operating parameters remains a question.
Concerns about color balance and device simplicity have led Nichia to favor obtaining white light by exciting a phosphor with an ultraviolet LED. Nichia produces UV LEDs by omitting the indium from the GaN alloy. The tradeoff in this design is lower efficiency in the final white-light output. For this reason a main concern of the Japanese program is the development of new phosphors.
But is it enough?
Although its efficiency may be less than 10%, a common 100-W incandescent bulb emits 1500 lumens and may cost only 25 cents in quantity. By contrast, the ultraviolet LED used by Nichia to excite its white phosphor costs about $30. A GaN diode may produce 30 lumens per Watt, but is limited in the input power it can accept.
Obtaining more light from an LED is possible primarily through two methods—increasing the amount of input power the device can handle, and increasing the output efficiency. Although the quantum efficiency of an LED is 90% or better, the high dielectric coefficients of the alloys result in a large fraction of the light remaining trapped inside the crystal by total internal reflection.
One approach to obtaining more output therefore has been to shape the LED so as to let the light escape. Combined with other improvements, LEDs with 200-lumen output have been predicted within a few years.
You can't save your way to prosperity
It is attractive to think that, along with all of the energy to be saved, white-light LEDs may mean that someday forgetting to turn off the headlights on your car won't kill the battery, or that no longer will you suffer burned fingers from changing a still-hot light bulb. But the new capabilities of a new technology almost demand brand-new applications. Entire walls that light up, dishes that change color, microscopic UV-light sources for medical applications—there will be many uses for a miniature, brilliant light source cool enough to touch that might be color variable.
Color Kinetics (Boston, MA), which makes commercial lighting fixtures using LED technology, has produced a fanciful fish as part of a celebration of Boston artwork. Titled "White Cod Nebulae," the fish radiates an infinite variety of colors from three primary-color LEDs inside the cod (see Fig. 2). It might be a symbol of a new era for an old technology.
Next month the series will review developments in low-power CO2 lasers.