Continuing the series, here is a look at light-emitting diodes and how to make them.
LEDs fundamentally are like photovoltaic cells operated in reverse: a current is applied and light is produced instead of the other way around. This is the first device in the series that requires photoresist and etching.
The takeaway here is that LED manufacturing is only modestly more challenging than single-junction monosilicon PV cells. Equipment for spin coating, making masks, a stepper and high-intensity light source are all required. Also, LEDs require potting (encasing in plastic) to function properly, so a supply of suitable plastic must also be available. Otherwise all the equipment for CVD, making ingots, saws, polishers, acid etch, etc. are used in much the same way and can be used in the same vacuum chamber.
Details after the jump.
(part 1) (part 2) (part 4)
In the same way that bandgap is an important value for PV cells, bandgap for an LED device determines what wavelength (and therefore color) of light is produced. LED devices work best with a direct bandgap, so most are made with gallium compounds of arsenic, aluminum, indium, phosphorus and nitrogen.
LEDs produce a very narrow range of wavelengths (a very sharply defined color), so white or multicolor devices use either several chips of different colors in one package or a short wavelength (blue or UV) chip with a phosphor layer to spread out the range.
Another concern is that LEDs emit light in all directions, more or less. They are not very functional unless they are packaged properly, with optical materials to help extract photons from the semiconductor material and lenses to focus or disperse the light.
There are many different LED technologies in development today, but at present there are no viable thin-film production techniques. All practical LEDs are wafer-based, even if they are later modified into thin devices.
The process starts with a 0.3 to 0.5 mm wafer of single-crystal sapphire (aluminum oxide crystal) produced in the same way as single-crystal silicon wafers. (Due to the high melting point of alumina, crucibles for this process are typically made of molybdenum or iridium.) A thick layer of n-type gallium nitride is epitaxially grown via moCVD (meaning it forms in the same crystal structure and orientation as the substrate). Next a thin layer of p-type gallium nitride is deposited by the same method.
Since light is emitted in all directions, little 'islands' of the active junction are best for getting the most light out of the device. To do this, the wafer is coated with a light-sensitive chemical called photoresist. A pattern is burned into the resist with high-intensity light (or lasers, if the feature sizes are small). A soap wash removes the unexposed resist, then an acid wash etches away the p-type layer and most of the n-type layer. This leaves a pattern of islands with an active junction surrounded by empty space.
Next, the electrical contacts are built by depositing an even layer of metal. This is usually aluminum. Another coat of photoresist is applied, then a different pattern is burned in. After a wash and acid etch, lines of aluminum run along the top of each island and along the etched lower level of the n-type layer.
The wafer is tested and cut into dies or chips with a diamond saw. Each die is anchored to a support, the contacts are wired to leads (with thermally assisted vibration welding or via tin dots), a high refractive index plastic is applied and the whole assembly is cast in a plastic lens shape. This produces the typical 3mm LED bulbs. Other form factors are possible, including surface-mount packages and high-power packages with integrated heatsinks.
A modification of this process starts the same way by depositing a thick n-type layer and thin p-type layer, then an aluminum conductor layer. Next, the wafer is flipped over and bonded to a 'carrier' wafer or substrate; this gives the method it's name of 'flip chips'. Since a layer of metallic aluminum is between the two wafers, the carrier can be made of anything that will survive the rest of the processing steps. Preferably it will be something with good thermal conductivity and a similar coefficient of thermal expansion. Usually the carrier is pre-coated with an aluminum layer; the two wafers are acid-stripped or polished to remove any aluminum oxide, then stacked in a vacuum chamber and heated until the two aluminum layers weld.
After flipping and bonding, a powerful laser is applied; sapphire is transparent to high UV (as from an excimer laser), but gallium nitride absorbs those wavelengths easily. Done properly, each laser pulse causes a tiny layer of gallium nitride to dissociate into nitrogen gas and metallic gallium. The whole wafer is kept above gallium's melting point, so a layer of liquid gallium is formed. Once the whole wafer is exposed to the laser the sapphire wafer can be slid off, leaving just the layers of semiconductor bonded to the carrier. This technique is called laser lift-off and is also useful for making MEMS devices. The sapphire can be polished and reused several times if necessary.
A second layer of aluminum is deposited on what is now the top of the wafer. A single photoresist and acid etch cycle leaves thin aluminum contacts spaced evenly. As an alternative, a transparent conductive layer of ITO or AZO can be deposited instead. The wafer is tested, cut, mounted and packaged much like standard LED chips, though only one lead wire to the surface contact is required. Since there are no islands, a reflector is often placed around the outside of the chip to direct light from the sides out the front of the package.
This flip chip process is sometimes used in multijunction PV chips to reduce their mass. Using a highly-reflective carrier can improve efficiency, as can the improved heat transport.
Photoresist is the breakthrough process that allows more complex devices to be manufactured. It requires spin coating for very accurate thickness, which means wafers need to be attached to a carrier, spun at high speed and sprayed with resist. The chemical coating is exposed to light, causing some of it to resist washing. Laying in contacts on chips that will be around 1mm across can be done with high-intensity visible light, but later devices will require incredibly tiny details to be burned into the resist which in turn requires high-UV lasers and complex optical arrangements. Acid etching removes material at predictable rates, so the resist must be at least as thick as the desired depth of etching.
The exposure step uses a machine called a stepper. The resist-coated wafer is loaded into a workspace. The pattern is formed with a mask, which is often made of chrome; this mask has to completely block UV light. The mask is precisely positioned above the wafer and the laser scans through the mask to burn in the resist. The mask is 'stepped' to the next position and the process is repeated. In some cases the wafer is moved instead of the mask. Once the entire wafer is exposed it can be washed and etched.