A collection of LEDs of different colors

LEDs are very common as displays, as part of digital devices like phones and music players, and as a growing market for white lighting. But what is an LED after all? And how does it change electrical current into light? And why are LEDs such a growing trend? You might ask many even more basic questions as well: what is light and how is it generated in the first place?

LED efficiency and advantages

LEDs have a number of advantages over other sources for lighting applications. A large part of this is that they are more energy efficient: they generate more light per watt of consumed power. In the section below on generation mechanisms for incandescent and fluorescent bulbs you will see why a good deal of heat or other forms of non-light energy is generated by non-LED systems. However, there are other advantages as well. Here is a partial list of LED advantages over more traditional ways of creating light:

  • Efficiency: LEDs produce more light per watt than incandescent bulbs.
  • Color: LEDs can emit light of an intended color without the use of color filters that traditional lighting methods require. This is more efficient and can lower initial costs.
  • Size: LEDs can be very small and are easily populated onto printed circuit boards.
  • On/Off time: LEDs light up very quickly. A traditional red indicator LED will achieve full brightness in microseconds. LEDs used in communications devices can have even faster response times.
  • Cycling: LEDs are ideal for use in applications that are subject to frequent on-off cycling, unlike fluorescent lamps that burn out more quickly when cycled frequently, or HID lamps that require a long time before restarting.
  • Cool light: In contrast to most light sources, LEDs radiate very little heat in the form of infrared (heat) that can cause damage to sensitive objects or fabrics. Whatever wasted heat is created is dispersed through the base of the LED.
  • Slow failure: LEDs mostly fail by dimming over time, rather than the abrupt burn-out.
  • Lifetime: LEDs can have a relatively long useful life. One report estimates 35,000 to 50,000 hours of useful life, though time to complete failure may be longer. Fluorescent tubes typically are rated at about 10,000 to 15,000 hours, depending partly on the conditions of use, and incandescent light bulbs at 1,0002,000 hours.
  • Shock resistance: LEDs, being solid state components, are difficult to damage with external shock, unlike fluorescent and incandescent bulbs which are fragile.
  • Toxicity: LEDs do not contain mercury, unlike fluorescent lamps.

Light and its generation

The term light is normally used for the visible form of energy that allows us to see things and gives them color. But to be more precise and technical, light is really an electromagnetic wave and as such has a much wider range of applicability and usefulness. By electromagnetic wave we mean that electric and magnetic fields, those lines of force that permeate space around electric charges and magnets, are undulating in time very much as the surface of water moves when there are waves in it. Or how plucking a taught string or rope creates a traveling undulation. Sound is also a wave where the air itself is vibrating (more precisely sound is a wave of pressure in the air: high and low pressure patterns that move at the speed of sound; it is this pattern of pressure that pushes your eardrum and makes you hear sounds). Put very simply, waves are disturbances that ``wiggle'' and move around. One of the great discoveries of the end of the 19th century, led by the physicist James Maxwell, was the light was in fact nothing than electric and magnetic fields undulating in tandem. This should come across as remarkable: normally we learn of electric fields as those associated by charges created by batteries or voltage sources or even by friction (rubbing plastics or fur) while magnetic fields are associated with magnets... it seems remarkable that such different things can be unified to create waves and light. But that is in fact what Maxwell discovered through some deep insights and mathematical analysis that was then verified in actual experiments.

Since light is nothing more than electric and magnetic fields that undulate, it would seem that to create it one would have simply make charges move about in an undulatory fashion. And that is exactly what happens: if you move electric charges back and forth, they emit electromagnetic waves. If the frequency of the undulation is appropriate, the waves will be in the visible part of the spectrum and register as visible light. For example, a radio station generates the radio waves (electromagnetic waves) by running currents up and down an antenna: this makes the antenna emit radiation. The frequency of the current pushing current up and down the antenna sets the radio station's frequency (e.g. 100 MHz means the current in the antenna is causing charges to move back and forth at a frequency of 100 million times a second).

But not all charges emit electromagnetic waves equally well. For a given frequency of undulation, the lighter the charged particle, the more it radiates. And electrons are the most ubiquitous light charged particles that are mobile in conductors, so almost all electromagnetic waves in nature and technology are created by electrons moving around. (The contribution of the much heavier nuclei making up atoms in materials is essentially negligible).

The above description is one based on classical physics, but with the advent of quantum mechanics in the early 20th century, it became clear that the more general way to understand light generation is that one requires an electron to change its state and lose some energy --- the lost energy can then be radiated as light. This then explains how atoms can generate light: if the electrons in an atom are excited (i.e. in a higher energy state than their native situation), then can fall back down to lower energy states and emit light --- you can read the glossary entry on the laser for more details. The frequency of the emitted light is directly proportional to the energy lost by the electron. Below is a schematic image showing how atom can be excited and then emit light when de-exciting back to its native (lowest energy) state.

Schematic showing how an atom (like hydrogen) will start with its electron in its lowest energy state (left), the electron will get excited via some mechanism an enter a higher energy state (center), and finally de-excite and "drop" back down to the lowest energy state and emit light with energy equal to the difference (right).

Light, wavelengths, and colors

It is more customary to discuss the wavelength of light instead of the frequency: the relation of wavelength λ to frequency f is simply given by λ=c/f where c=3x108 m/s is the speed of light. Visible light is in the range 400 nm < λ < 760 nm. The longer wavelengths are more reddish and the shorter ones more bluish (you can see a table further below for the various ranges and colors). However, electromagnetic waves are not restricted to this range --- light with wavelengths close to but below 400 nm is ultraviolet (visible by bees for example and what causes one to tan or burn in the summer), light with wavelengths close to but below 760 nm is infrared (heat that you can feel radiated by a fire or any hot object), and much longer wavelengths give microwaves (like those in the microwave oven with typical wavelengths of a few cm) or radio waves with much longer wavelengths. The list goes one: X-rays, gamma rays, etc. with the only physical difference being the wavelength of the electromagnetic wave.

Generation mechanism for incandescent and fluorescent bulbs

Up to know, we have the general principle that when electrons become excited to higher energy states and then drop back down to lower energy states, the liberated energy (the energy loss) can be radiated as light. The most elementary method to achieve this is to heat up a material, and this is how an incandescent bulb works: an electrical current through a thin metal wire causes the wire to heat up to a very high temperature; this thermal energy causes some of the electrons in the metal to become excited; then when these electrons lose energy and drop down to lower energy states, they give off light. In fact, for an incandescent bulb, a large range of wavelengths are emitted with much of the energy in the non-visible infrared region, which is why incandescent bulbs cause so much heating (and are relatively inefficient ways to generate light).

A fluorescent bulb works on a different principle. The lamp is filled with a very low pressure gas of mercury and some noble gasses (argon, krypton, neon, or xenon). When an electrical current passes through this rarefied gas, the electrons comprising the electrical current sometimes collide with the atoms of the gas and can transfer energy to them. This transferred energy, if large enough, can excite an electron in the gas atom; when the excited electron drop back down to its native (ground) state, light is emitted. However, this light is typically in the ultraviolet and thus not visible. This emitted light then hits a phosphor coating the inner glass wall of the fluorescent lamp: a phosphor is a material that will absorb these ultraviolet energy and emit instead visible light.

Diodes versus light-emitting diodes

A diode emitting light due to carrier recombination

With all the above introductory information on light and light generation, we are finally ready to begin discussing LEDs. The acronym LED stands for Light Emitting Diode. You can read more details about what a diode is and how it works by following the previous link. In brief, a diode is a combination of two semiconductors one of which is doped n-type and the other p-type. In a diode, when electrons and holes meet in the junction region between the n-type and p-type material, they can recombine to liberate energy. (As explained in the diode section on photo-generation, the opposite process is possible where light turns into electron and hole and thus current is made to flow, and this is the basis for photodiode detectors as well as semiconductor solar cells.)

In standard silicon diodes, the energy liberated in carrier recombination is mainly lost as heat instead of turning into light. It takes a particular choice of materials to make a good light emitter. Strange as it sounds, the reason for this is something as elementary as conservation of momentum as explained below.

In any semiconductor, whether n-type or p-type, the mobile charges have some characteristic momentum which is either zero or non-zero depending on the details of the bonding between the atoms making up the semiconductor. For example, for an n-type semiconductor, if one has the material sitting alone with no external forcing, then no currents are flowing so if one were to look at all the mobile electrons in the material on average they have no net speed or momentum (zero average momentum). This net zero momentum, however, could come about in two very different ways: case (a) it could be that each electron individually has very little momentum that is close to zero (so we're average a large set of very small numbers all close to zero and getting zero), or case (b) it could be that each electron has a large momentum but because the different electrons have different directions of motion in the net the average is zero. The analogy is the following: say someone gives you a box inside of which there are many marbles that may or may not be moving; all you can see is that the box as whole is still. Without looking inside, however, you can not know whether the box is still because all the marbles are at rest or whether in fact they are all moving in all directions but that on average as many bang into the walls going left as going right so in the net nothing happens.

Whether we have case (a) or case (b) for a given semiconductor depends on the type of bonding involved for that material so no general rule can be given. But for most common semiconductors like silicon, germanium, gallium arsinide or gallium nitride, it turns out that when the semiconductor is p-type, the holes have net zero momentum as they obey case (a) (i.e. the holes all have very small momenta individually). However, when they are n-type, there is a difference: elemental semiconductors such as silicon or germanium have electrons with momenta of type (b) while the compound semiconductors such as GaAs or GaN or InAs will have electrons distributed according to case (a). When both holes and electrons are of type (a) we have a direct gap material where as the mixed case of (a)-(b) is called an indirect gap material.

How does this relate to light emission? As explained in the glossary note on the laser, light is actually made up of little packets of energy called photons. They are the elementary, indivisible parts of light just as electrons are the elementary and indivisible parts of charge and current. Each time an electron loses energy to a lower energy state, it could in principle emit one photon. But like all physical processes, energy as well as momentum has to be conserved. And for visible light, the photon carries very little momentum compared to the momenta carried by typical electrons or holes in a semiconductor: to put things in perspective, if the typical momentum involved in case (b) above is one unit, the photon has only 1/1000 units of momentum. What this means is that for the indirect gap materials, a hole with very small momentum recombining with an electron with large momentum can not give rise to one photon with tiny momentum and conserve total momentum --- the process must be more involved and complicated and thus it is either very unlikely or will involve energy loss as heat. On the other hand, for direct gap materials where both electron and hole have small momenta, there will be pairs of electrons and holes with small momenta so that their net momentum can be carried away by the emitted photon. In short, due to conservation of momentum and the small numerical value of photon momenta for visible light, only direct gap materials can emit light effectively.

Direct gap materials used in LEDs

So what we have so far is that: (a) when making a diode we put a p-type and n-type material in proximity, (b) when a current is made to flow in the diode, holes end up entering the n-type region and electrons enter the p-type region, (c) in both cases electrons can lose energy to fill the holes, and (d) for direct gap materials the lost energy can be directly converted to photons (light).

Typical materials used for LEDs are compound semiconductors. The basic working principle is to have an element from the third column of the periodic table (Al, Ga, or In) and another element from the fifth column (N, P, As) and to combine them or alloy them to tune the wavelength emitted. So, for example, starting from GaAs which gives the longest wavelength (in the red or infrared), by adding Al one gas AlGaAs which emits at shorter wavelengths. Similarly for making GaP instead of GaAs. And then one can even make more complex combinations like AlGaP or AlGaInP, and so on. The basic idea is to change the chemical composition which will in turn change the emitted wavelength from infrared through visible and onto the ultraviolet. All these alloys turn out to be direct band gap materials because the basic binary parent compounds (GaAs, GaN, InN, AlN, GaP, InP, etc.) are direct gap as well and one is just "interpolating" between their properties.

Below is a table showing wavelength ranges of emission, their color, and which materials are used in LEDs in each case.

ColorEmitted wavelength λ (nm)Materials
Infraredλ > 760Gallium arsenide (GaAs), Aluminium gallium arsenide (AlGaAs)
Red610 < λ < 760AlGaAs, Gallium arsenide phosphide (GaAsP), Aluminium gallium indium phosphide (AlGaInP), GaP
Orange590 < λ < 610GaAsP, AlGaInP, GaP
Yellow570 < λ < 590GaAsP, AlGaInP, GaP
Green500 < λ < 570Indium gallium nitride (InGaN), GaN, GaP, AlGaInP, AlGaP
Blue450 < λ < 500Zinc selenide (ZnSe), InGaN
Violet400 < λ < 450InGaN
Ultravioletλ < 400diamond, Boron nitride (BN), AlN, AlGaN, AlGaInN
WhiteBroad spectrumBlue/UV diode with yellow phosphor

White light LEDs

As you may know, the color we experience as white does not corresponds to a specific wavelength but rather means that the light has almost equal amounts of all visible wavelengths comprising it. As the table above shows, a given LED made from a particular material will generally emit a particular wavelength --- and thus a single color on the color spectrum (the rainbow ROYGBIV spectrum we learn of in school). How does one get white light from an LED system?

There are two basic approaches. One is to use a combination of LEDs each producing one primary color (red, green, and blue) and then to mix their output (i.e. have their output come out together in the same direction). The other is to use a phosphor material that converts monochromatic light form a blue or ultraviolet LED to broad-spectrum white light, which is what we much like a fluorescent bulb. (If you've been paying attention you may notice that these two differing physical mechanisms will not produce them same spectrum --- i.e. combination of different wavelengths: how can they produce the same white light? In fact two very different spectra can appear to have the same color, a physiological/psychological phenomenon known as metamerism. It is a very interesting topic but not relevant to materials science so you can read it about it elsewhere such as Wikipedia's article on metamerism.) The phosphor method is much more common for reasons we discuss below.

The color mixing approach (RGB LEDs) are in principle more energy efficient and can achieve any desired color due to the generality of the method. However, the design of the devices is complicated in order to ensure proper blending of the constituent colors. As a separate problem, present RGB LEDs systems have poor temperature dependence: changing the temperature slightly changes the power output very dramatically (exponential decrease with increasing temperature). A great deal of active research exists on RGB LEDs, but at present, they are not mass produced for white lighting applications.

The phosphor method uses the monochromatic (single wavelength) output light from a single LED and converts it to a broad spectrum. There a few variants involving the wavelength of the LED and the number and types of phosphors used. One common method is to use a blue InGaN LED and then a yellow phosphor such as yttrium aluminium garnet (Y3Al5O12) with trace cerium doping: the blue light excites the garnet which produces yellowish light and the proper combination of blue and yellow appears approximately as white to the human eye. Another design uses an ultraviolet LED with a mixture of europium (Eu) based red and blue emitting phosphors along with green emitting Cu and Al doped ZnS: the light has better whiteness but the use of many phosphors is not as efficient as the blue+yellow method above. Regardless of the details, light of a shorter wavelength is being converted to one of longer wavelength in the phosphor approach. For photons, change of wavelength means change of energy so some energy is always wasted as heat during the down-conversion. Therefore, phosphor LEDs will always have lower efficiency than RGB or normal LEDs due to the down-conversion inherent in the approach. In addition, there are phosphor-related degredation problems.

However, as you can imagine, making a single LED and coating with some materials is simpler and cheaper than the RGB LED approach. Therefore, the majority of high intensity and mass-produced white LEDs presently on the market use the phosphor approach.

Further additions:

  • OLEDS and other LED schemes in future?
  • Describe lifetime benefits and comparisons in more detail (halogens, sodium, etc.)
  • Switching speed of LEDs (on/off)?
  • Dimming