There is a very easy reading essay in Shuji Nakamura's Book "Brilliant" teaches us how to make a LED.
"
The conventional method of making LEDs, liquid phase epitaxial could not grow films for high enough quality and sufficient thinness..
There are two more modern alternatives. Both were capable of growing films of material just a few atoms thick. One molecular beam epitaxial works via vacuum evaporation. MBE is especially popular with academic scientist. It can product small quantities of material, enough for researchers to extract data based on what they can write and published papers. But MBE requires a ultrahigh vacuum, has very slow growth rates, and is difficult to scale up. In the opinion of most people, Shinji included, the method is not suitable for mass production.
That Left MOCVD, which does not need a high vacuum and can be applied to the factory floor. The choice was thus, as they say in Silicon valley, a no-brainer, Shuji selected MOCVD without hesitation, But he had little idea of how MOCVD was done.
By a stroke of good fortune, it just so happened that one of Japan. Leading Experts on the technique was an old acquaintance of his from Tokushima University. Though Shiro Sakai had been two years Sakai’s senior, they had worked together in the same laboratory, and Shuji knew him well.. In the interviewing years, Sakai, now a professor at Tokushima, had become well known for his expertise n MOCVD, Now he was on sabbatical at the University of Florida. During the Summer holiday of 1987, he returned to Japan. Nakamura went to see Sakai to ask hi advice on how to learn MOCVD.
When Sakai returned to Japan for a week at the end of 1987. Nakamura invited hi to Visit Nichia. There, the professor explained to Ogawa the significance of MOCVD as a Crucial tool for the production of state-of-the-Arts LEDs. At this meeting, blue LEDS were not mentioned, Sakai recommended that Nichia should send Nakamura for a year to the engineering school at the University of Florida, where he was currently on sabbatical. A deal was arranged: Nakamura would learn MOCVD under Sakai’s tutelage. In return, Nichia Would donate around $100,000 to fund Sakai’s research.
To make a chip the size of a grain of sand takes a mighty big box. A typical MOCVD system is almost as big as a shipping container and costs well over a million dollars. Seen from outside, MOCVD equipment looks rather dull, like a row of office cabinets. Peek behind the bland-looking doors, however, and you will discover a bewildering assemblage f tanks, pumps and valves connected by what appears t be several miles of thin, stainless steel pipe. At the end of cabinets is a rack containing a computer that runs the recipes for growing LEDs. These are programs that, with exquisite precision, control the pressure and flow of gases, while monitoring the temperature and the rate at which the thin films of crystal grow.
The heart of MOCVD system is through a little window in one of the cabinet doors, it its reactor chamber. This is a cylinder about the size of a cookie jar, made of quartz in some systems, of metal in others. It may, be positioned either horizontally or perpendicularly. The chamber is remarkably small in comparison to the whole. It occupies perhaps 3 or 4 percents of the total space. Inside the jar there is a graphite chuck, mounted on a little pedestal, here sits the wafer on which the thin films are growth. The Chuck is connected via a thermocouple that monitors the growth temperature, to a heater. T grow Gallium Nitride, the wafer is heated to between 1,000 and 1,200 degrees Celsius. At which point it grows bright golden-orange, In case of quartz chambers, the heat comes form copper coils would round the jar. An exhaust system, typically a vacuum pump, completes the process, It sucks the unused gases out of the reactor chamber, flushing them away to a scrubber for disposal.
For more than twenty years LED were grown by one of the two methods, liquid phase epitaxy (LPE) or vapour phase epitaxy (VPE). Epitaxy simply means stacking crystal layer upon crustal exactly the same orientation, like piles of eggs trays, But when it came to growing hg-quality thin films and quantum wells, which require abrupt atomic level transitions from one layer to the next, both processes were too crude. For example, an LPE system consists of a quartz tube in which are .lined up with little graphite dishes called, because of their cigar like shape, “boats, ” Each boat contains a different semiconductor material that is heated until it melts. You slide your wafer along the tube, leaving it to sit a while on top of each boat. Cooling causes some of the materials go precipitate onto the surface of the wafer. LPE produces relatively thick layers, and the boundaries between them are gradually rather than sharply defined. Precise control move thickness almost impossible to achieve.
MOCVD (sometimes, confusingly, also known as MOVPD) system because the method of choice for growing high-brightness devices, original red LEDs, in the mid-1980’s MOCVD accomplished the abrupt transition between layers by allowing the crystal grower to run two mixes of gases through the system simultaneously. While using mix A to grow a film in the reactor, you have all the gases for mix B flowing directly to exhaust, Then, at just the right moment, you switch mix A to the moments. All you hear is sound of the compressed air-driven pneumatic valves. They open and close in quick succession- phsst, phsst, phsst, phsst-et voila! You have grown a quantum well.
So much for vapor deposition, Now we come to metal organic chemicals, why it is necessary to sue such fancy-sounding stuff instead of ordinary metal? The answer is that, in their vapor ;phase neither aluminum, gallium, nor Indium- the three most common metals use in growing bright blue (and Red and green) LEDs- can muster sufficient vapor to be picked up the carried there, in organic form. To pump up the vapour pressure, organize chemical such a Methyl groups are attached to the metals Gallium becomes trimethyl gallium; the positive-types dopant magnesium becomes bis(Cyclopentadienyl) magnesium, mercifully abbreviated as CP2MG. The carrier gas in hydrogen. It is kept Flowing through the system at a rate of many litres per minute. During the travels, the hydrogen bubbles through the temperature of the compounds. Which it transports to the reactor. When the com[pound gases get to the hot zone. They lose their methyl groups. The nitrogen or Gallium nitrides arrives at the jar in the form ammonia.
The heat decomposes the gases. Leaving nitrogen atoms hot to trot their gallium Partners.
The process of growing a gallium nitride LED begins by heating the sapphire wafer to a very high temperature. Once hot, you clean the surface by flowing nitrogen over it. Then you drop the temperature way down to maybe 500 degrees Celsius to grow the first layer, the so called nucleation, or buffer, layer. This is a thin film, typically of gallium or aluminum nitride and juts 50 to 100 atoms thick, that is grown directly on the wafer. The buffer layer is amorphous, that is lacking a crystalline structure. When you heat it up, the surface of this amorphous layer becomes very lumpy as nucleation islands oriented to the surface of the sapphire start to form, As you reach higher temperatures, however, these islands grow together laterally, to form a smooth , mirror-like layer of gallium or aluminum nitride., One of the secrets of growing high-quality GaN is being abe to control exactly how this nucleation layer is deposited, how is crystallizes, and how it grows together during the heat-up step.
On top of the nucleation layer, you deposit plain vanilla (i.e., undoped) gallium nitride.
Next comes a layer of negative-types gallium nitride, wit silane as the electron-donaing dopant. This is followed by a layer of negatively doped aluminum gallium nitride, a compound with a wider bandgap than GaN. This layer plus another . positively doped layer of AlGaN o the other side serve to confine the charge carriers within the active _i.e. light emitting) layer of the device, Then you drop the temperature down from 1,000-1,200 degrees Celsius to 750-850 degrees Celsius s that you can grow an indium gallium nitride quantum well. You grow, say ., 20 angstroms of InGaN, then maybe 100 angstronms of GaN, then repeat the process for as many quantum wells as your recipe calls for, adjusting the amount of indium to produce the desired wavelength of light. The more indium you include, the greener the output will be. After growing the last combo of InGaN+GaN, you crank the temperature back up and deposit your other confining layer of positively doped aluminum gallium nitride. Then you cap the whole thing off with a layer of positive-type gallium nitride using magnesium as the hole-donating dopant, That completed the device.
In a typical growth run, the whole process takes anywhere between two and a half hours and hour hours. If you load your wafer first things in the morning just as the coffee is brewing. You will gte the growth sun out around lunchtime. You should schedule another run around two o’clock and have it out before dinner. Between runs, you have to clean the reactor b baking it out at high temperature. In the R&D lab, two runs is not a bad day. On the production line, four growth runs in a twenty-four-hour period is consider pretty good going. A large production –line reactor may contain a platter with a many as one hundred wafer on board.
The growth process itself is not in the least dramatic. You can hear faint hums and hisses from the pumps and the valves, but that’s about it. The only smell MOCVD machine gives off is a subtle whiff of burnt reactants that emanates from inside the jar, if you smell anything else—ammonia, for example- that means there is a leak, This is a good time to leave the lab. Quickly.
"
