A blue photon needs more than a blue-looking crystal
A blue LED needs a semiconductor that can release enough energy in one electronic transition, and a junction that repeatedly brings electrons and holes together. A crystal that glows under an electron beam has passed only the first of these tests. A useful lamp must accept electrical current, turn a substantial fraction of it into photons, and let those photons escape.
A semiconductor has a valence band of normally occupied electronic states and a conduction band in which electrons can move through the crystal. Their energy separation is the band gap. A missing electron in the valence band behaves as a positive mobile carrier, called a hole. When a conduction electron fills a hole, the released energy can become light. For a photon, energy is Planck's constant times the speed of light divided by wavelength. Blue light near 450 nanometres carries about 2.76 electronvolts per photon. A large enough gap is necessary, but the transition must also have an appreciable probability of emitting light.
Sources: [1]
A thin buffer changed how GaN could grow
Gallium nitride offered a promising wide-gap material, but growing an electrically useful layer was difficult. The available sapphire substrate survives the growth conditions, yet its atomic spacing differs greatly from GaN. Direct growth readily creates rough, defective material. A junction made in such a layer can lose injected carriers at defects or suffer poor electrical contact before its nominal band gap becomes useful.
Akasaki and Amano's group developed a low-temperature aluminium-nitride buffer before the hotter GaN growth step, reported in 1986. The thin initial layer changes the way crystallites nucleate and merge. It does not simply erase the lattice mismatch. Amano's lecture compares the rough surface obtained without the buffer with the much smoother film grown with it. Nakamura later developed GaN buffer growth and reactor methods. These were experimentally distinct routes through the same obstacle: the emitting structure had to be built on a layer whose morphology, impurity content and electrical properties could be controlled.
Sources: [3][1]
Electrical measurements had to establish holes
An LED normally joins an electron-rich n-type region to a hole-rich p-type region. Forward bias pushes electrons and holes towards an emitting region. GaN could readily conduct with electrons, but reliable hole conduction resisted repeated attempts. Adding magnesium was not sufficient: an acceptor atom must be electrically active before it can leave a mobile hole in the valence band.
In 1989, Amano and colleagues obtained p-type behaviour after low-energy electron-beam treatment of magnesium-doped GaN. The important evidence was electrical. A hot-probe indication alone could be ambiguous; the group also used Hall measurements. A current-carrying specimen in a perpendicular magnetic field develops a transverse voltage whose sign identifies the dominant carrier under the appropriate single-carrier interpretation. Together with resistivity and mobility, that measurement supported hole conduction. A brighter luminescence spot would not by itself establish the same result: optical excitation can create electrons and holes even in material that cannot supply useful p-type current.
Sources: [3][1]
Hydrogen explained why magnesium could remain inactive
Nakamura's work clarified a major reason for the inactive magnesium. Metalorganic growth uses ammonia as a nitrogen source, and hydrogen can enter the growing film. Hydrogen binds with magnesium to form complexes that passivate the acceptor: the magnesium is present chemically but does not supply the intended hole electrically. Annealing in a suitable hydrogen-free environment allows hydrogen to leave and activates p-type conduction. The thermal-annealing report appeared in 1992.
This explanation connected a processing intervention with a carrier measurement. Changing the annealing atmosphere and temperature changed electrical behaviour without requiring a new nominal magnesium concentration. That is stronger evidence for activation than simply finding more light after adding more dopant. Thermal treatment also offered a practical way to process a whole layer. The distinction matters in a device: a small activated patch and a uniformly conducting p-type layer can give very different current spreading and operating voltage. Contact resistance and incomplete acceptor ionization still require attention after hydrogen passivation has been relieved.
Sources: [4]
An emitting layer made the junction a useful blue source
Improved GaN growth and p-type conduction opened the way to a designed emitting region. An InGaN layer has a different gap from its surrounding nitride layers. Appropriate band offsets help keep electrons and holes near that layer, where they can recombine. The 1994 candela-class report used a zinc-doped InGaN active layer in an InGaN/AlGaN double heterostructure. It should not be silently replaced by a modern multiple-quantum-well design when describing that experiment.
The evidence included more than the photograph of a bright lamp. Nakamura, Mukai and Senoh reported light-output power against forward current, an electroluminescence spectrum, and a current–voltage curve. At 20 milliamperes, the reported output was 1.5 milliwatts, the forward voltage 3.6 volts, and external quantum efficiency 2.7 percent. The spectrum peaked at 450 nanometres and had substantial width. These measurements answer different questions: the spectrum identifies the emitted colour, the electrical curve shows how injection is driven, and calibrated optical power establishes how much useful light leaves the device.
Sources: [5][1]
Counting photons and counting watts give different efficiencies
External quantum efficiency counts escaping photons per electron supplied by the external circuit. Wall-plug efficiency compares optical power with electrical power. For a narrow spectrum, the following relation uses optical power, current, elementary charge, photon energy and forward voltage. For a broad spectrum, photon rate must be integrated over wavelength; using only its peak is an estimate.
The reported 1994 power and voltage give a wall-plug efficiency of about 2.1 percent: 1.5 milliwatts of light divided by 72 milliwatts of electrical input. Using the 450-nanometre photon energy with the reported quantum efficiency gives a consistent estimate. The remaining input energy goes chiefly into heat in the operating device. Internal radiative efficiency addresses an earlier step: among carriers recombining in the active region, how many recombinations produce photons? Injection losses occur before that step, and extraction losses occur afterwards. A change in the external curve therefore cannot be assigned to one microscopic loss process without additional measurements.
Sources: [5][6]
A rising light output can accompany falling efficiency
A compact recombination model explains one possible shape of the efficiency curve. Assume a uniform active region with equal electron and hole densities, described by one common carrier density. A term proportional to density represents defect-assisted loss, the quadratic term represents radiative electron–hole recombination, and a cubic term represents a high-density loss channel. The coefficients have different units because every term must be a recombination rate per volume per time. In steady state, injection balances their sum.
At low density, the linear loss wins; increasing injection lets the quadratic radiative process catch up. At high density, a positive cubic loss can reduce the radiative fraction even though the absolute number of emitted photons continues rising. This is an efficiency droop. Such a curve is useful for interpreting competition, but fitting its cubic coefficient does not uniquely establish Auger recombination in a real LED. Carrier leakage, nonuniform density and heating can also affect measured efficiency. Identifying a microscopic cause requires evidence beyond a successful three-coefficient fit, such as independent carrier or energy-resolved measurements.
Sources: [6][4]
A lamp’s lumens are not the light on the desk
The Nobel committee’s background note rates lamps in lumens per watt. A lumen measures light flux weighted by how sensitive the eye is to each colour. Edison’s bulb gave about 16 lm/W and a fluorescent tube 70; white LEDs, in which a blue LED excites a phosphor, had passed 300 by 2014. A lamp of 800 lm therefore draws about 50 W as a bulb and under 3 W as such an LED.
Lumens count the light leaving the lamp. A reader at a desk sees the light arriving: the illuminance, in lux, or lumens per square metre. Hang one 800 lm LED 0.6 m above a 1.2 m desk and, counting direct light only, the desk gets 707 lx straight below but 135 lx at its corners, because light spreads with the square of the distance and reaches the edges at a slant. Only 40 % of the lumens land on the desk at all. A reading desk is usually lit to about 500 lx, and fairly evenly.
Getting there depends on where the light comes from. Raising the lamp evens the desk but dims it. Sending the same light from a wide diffusing face makes the lamp 900 times less bright to look at and softens shadows, yet hardly evens the desk. Three 800 lm lamps spread along it 0.85 m up, bare LEDs or panels, give 640 to 660 lx on average, with the dimmest point about two thirds of the average.
Sources: [1][7]
From crystal growth to bright blue light
Buffer-layer growth in 1986, p-type conduction in 1989 and subsequent bright devices addressed the linked problems of material quality, injection and emission. The 2014 Nobel Prize recognized Akasaki, Amano and Nakamura’s efficient blue LEDs and the route they opened to bright, energy-saving white light. Wavelength fixes the photon-energy scale; a useful lamp requires every stage to deliver more input energy as light that actually escapes the chip.
Sources: [6][2]