How Do LEDs Work? Turning Electricity Into Light in a Chip
An LED is a part that sends current across the boundary between two different kinds of semiconductor. There, electrons meet "empty seats for electrons," called holes. The energy released when they meet comes out as light. The color of that light is set by the size of an energy "step" (the band gap), which depends on the material. Most white LEDs use a yellow-glowing material to convert part of a blue LED's light, then mix the two colors.
What you'll learn
- How an LED makes light, and how it relates to a solar cell
- Why the material sets the color, and how to calculate the color
- Why LEDs can hardly be made from silicon
- Why the blue LED was so hard to make and so important, and how white LEDs work
- Why LEDs get hot even though they save energy, and how to read lifetime numbers
- How LEDs relate to displays and laser diodes
Prerequisites: it helps to know n-type and p-type, holes (empty seats for electrons) and the energy "step" called the band gap ("What Is a Semiconductor?").
Where you see it
LED bulbs, traffic lights, car lights and the flashlight on your phone all glow with LEDs. The tip of a TV remote control also holds an LED. It gives off infrared light, which the eye cannot see.
LED stands for light-emitting diode. A diode is a part that lets current flow easily in only one direction. An LED is a diode built to give off light when current flows through it. The part that glows is a tiny piece of semiconductor inside, called the chip.
How it works 1: electrons and holes meet at a boundary
At the heart of an LED, an n-type and a p-type semiconductor are joined together. The n-type side has many electrons that can move freely. The p-type side has many holes, the empty seats for electrons. This boundary is called a PN junction.
Connect the p-type side to the plus end of a battery and the n-type side to the minus end, and current flows easily. Applying voltage in this direction is called forward bias. Electrons are then pushed toward the boundary from the n-type side, and holes from the p-type side, one after another.
An electron that can move freely has more energy than the state that holds a hole. The size of this difference is fixed for each material. This difference is called the energy step, or band gap.
When an electron meets a hole at the boundary, the electron drops into the empty seat, and both disappear. This is called recombination. The electron gives up energy equal to the step. In LED materials, this energy comes out as a particle of light.
How it works 2: the reverse of a solar cell or image sensor
What happens in an LED is the reverse of what happens in a solar cell or a camera sensor. In a solar cell or a sensor, light creates pairs of electrons and holes. The electric force at the boundary pulls them apart and turns them into current ("How Do Solar Cells Work?", "How Do Image Sensors Work?"). In an LED, current brings electrons and holes to the boundary, where they meet and become light.
But which direction works well depends on the material. Camera sensors can be made from silicon, but LEDs can hardly be made from silicon. We explain why later.
How it works 3: the size of the step sets the color
The color of light depends on the length of its waves, called the wavelength. The human eye sees roughly 380 to 780 nanometers (nm; 1 nm is a billionth of a meter). Shorter waves look blue or violet, and longer waves look red. Waves longer still are infrared, which we cannot see.
A shorter wavelength means more energy in each particle of light. So a material with a bigger step gives off light closer to blue. The step is measured in eV (electron volts). One eV is the energy one electron gains when a voltage of 1 volt pushes it.
There is a handy rough rule: divide 1240 by the step in eV, and you get the wavelength in nm. Working backward, blue light (450 nm) needs a step of 1240 ÷ 450, or about 2.76 eV. Red light (630 nm) needs 1240 ÷ 630, or about 1.97 eV. You can try other values with the band gap and wavelength calculator.
To get different colors, makers change the combination of materials and the mix. Blue and green use InGaN, which is gallium nitride mixed with indium. Red and yellow use AlGaInP, made of aluminum, gallium, indium and phosphorus. Infrared uses materials such as gallium arsenide (GaAs). All of these are III-V compound semiconductors, which combine elements from groups 13 (formerly III) and 15 (formerly V) of the periodic table.
The voltage needed to light an LED is also close to the size of its step. In typical values for ROHM products, a red LED needs 1.9 V and a blue LED needs 3.2 V (at 20 mA).
Why silicon doesn't glow
Silicon's step is about 1.1 eV. By the rule above, it should give off infrared light at about 1,100 nm. Yet when current flows through silicon, it gives off only a tiny amount of light. That is because, besides energy, one more thing must match when an electron meets a hole.
An electron in a crystal has energy, and it also has a kind of "push" in its motion. Physicists call this quantity momentum; here we simply call it "push." A particle of light can carry almost none of this push. So if the electron and the hole do not already have matching push, recombination that gives off light rarely happens.
In some materials, the states where electrons and holes gather have matching push. These are called direct band gap materials. Gallium nitride and gallium arsenide are of this type. When electrons and holes meet in them, they easily recombine and give off light.
Silicon is an indirect band gap material: the push of its electrons and holes does not match. To give off light, a vibration of the crystal's atoms must take up the difference in push. Such meetings rarely happen. While electrons wait, most recombine through flaws in the crystal and give up their energy as heat. The table "Direct vs indirect band gap" compares the two types.
When it comes to taking in light, though, silicon works well. Silicon absorbs light less readily than a direct band gap material. But a particle of light keeps traveling through the crystal until it is absorbed. With a thicker layer of silicon, most of the light can be taken in. Here, there is no fast path to heat competing with it, as there is when giving off light.
Why the blue LED was so hard, and so important
In 1962, Nick Holonyak of General Electric (GE) in the US made an LED that gave off visible red light. This was the starting point of practical LEDs. Red, yellow and green LEDs then spread into uses such as indicator lights. But for a long time, no one could make a bright blue one.
Blue light needs a large step of around 2.7 eV. The candidate material, gallium nitride (GaN), was very hard to grow as a clean crystal. No base crystal had a matching spacing between atoms, so the crystals came out full of flaws. Making p-type gallium nitride, with many holes, was also very hard.
In 1986, Isamu Akasaki and Hiroshi Amano grew a good-quality crystal by placing a thin layer between it and a sapphire base. In 1989, they made p-type material and built a PN-junction LED. Shuji Nakamura of Nichia Corporation improved the crystal-growing machine in his own way. In 1993, he made a bright blue LED.
The three received the 2014 Nobel Prize in Physics. The official citation reads: "for the invention of efficient blue light-emitting diodes which has enabled bright and energy-saving white light sources." With blue in hand, LEDs now covered all three primary colors of light, and white LED lighting became possible. Today, gallium nitride is also used in power semiconductors for power supplies.
How white LEDs work
No single step gives off white light. White light is a mixture of many colors. So most white LEDs cover a blue-glowing LED chip with a material that glows yellow.
This yellow-glowing material is called a phosphor. A phosphor takes in light of one color and gives off light of another color with a longer wavelength. Part of the blue light passes straight through. The rest is turned into yellow light by the phosphor. When blue and yellow mix, the human eye sees white. A common yellow phosphor is a crystal called YAG with cerium added (YAG:Ce).
Nichia developed this method in 1996. There are other ways too. One mixes three LEDs: red, green and blue. Another uses an LED close to ultraviolet to make several phosphors glow. Still, in lighting, a blue LED plus a phosphor is the main approach. When an LED is off, its glowing part often looks yellowish. That is the color of this phosphor.
Why LEDs save energy compared with incandescent bulbs
An incandescent bulb makes a thin metal wire, the filament, extremely hot so that it glows. Hot objects give off light, but most of it is infrared, which we cannot see. In an incandescent bulb, about 90% of the energy used becomes heat (ENERGY STAR, the US energy-efficiency label program).
An LED does not glow from heat. It turns the electrons' energy straight into light. Lighting efficiency is compared by brightness per watt of power, in lm/W (lumens per watt).
A typical 60 W incandescent bulb gives about 16 lm/W (US Energy Information Administration, 2014). According to the International Energy Agency (IEA), LEDs on sale in 2026 average close to 100 lm/W. Some products exceed 200 lm/W. That means the same brightness takes roughly one sixth of the power.
Why LEDs still get hot
LED light contains almost no infrared, so the things it shines on barely warm up. But inside the LED, much of the electricity still becomes heat. Some electrons recombine without making light. Some energy also turns into heat when the phosphor changes the color, and in the power supply circuit.
Comparing with the upper limit gives a sense of how much becomes heat. A 2020 report from the US Department of Energy (DOE) gives 325 lm/W as the theoretical upper limit for white LED light. A 100 lm/W lamp reaches less than a third of that. The energy that does not become light becomes heat. This estimate leaves out many details, but more than half of the electricity is likely turning into heat.
An incandescent bulb's heat spreads out into its surroundings as infrared. An LED's heat stays trapped inside its small chip. As heat builds up, the light gets weaker and the lifetime gets shorter. That is why the base of an LED bulb contains metal or other parts that carry heat away. These are called heat sinks.
LEDs have another trait. The more current you send through one chip, the smaller the share that becomes light. This is called efficiency droop. Bright lights use larger chips or many chips side by side, to keep the current in each one low.
How to read lifetime numbers
LEDs rarely burn out suddenly. Instead, they slowly get dimmer. So LED lifetime is usually given as the operating time until brightness falls to 70% of the starting value. This is written L70, also described as 70% lumen maintenance.
LED bulb boxes often show a number such as "rated life 40,000 hours." This is an average guide, not a guaranteed time. At 8 hours a day, 40,000 hours is about 13.7 years. The power supply circuit may wear out first. The Japan Lighting Manufacturers Association suggests replacing light fixtures after 8 to 10 years.
LEDs run on a steady current
Above a certain voltage, a small rise in voltage makes an LED's current jump sharply. As the LED warms up, current flows even more easily. Connected straight to a battery, an LED can draw too much current and break.
So LEDs are driven by setting the current, not the voltage. For a small LED such as an indicator light, one resistor in series limits the current. For a large current, as in lighting, a circuit that holds the current steady is used. This is called a constant-current circuit.
The base of an LED bulb also holds a small circuit. It turns the AC from a wall socket into a steady DC current that suits the LED. To dim an LED, a common method is to switch it on and off faster than the eye can notice. Changing the share of time it is on changes the brightness.
LEDs and displays
| Type | What gives off light | In short |
|---|---|---|
| LCD (LED backlight) | White LEDs behind the screen | The liquid crystal passes or blocks the light from behind, pixel by pixel |
| OLED (organic LED) | A film of organic material in each pixel | Each pixel glows by itself. The material contains carbon and glows when electrons meet holes |
| MicroLED | A tiny LED in each pixel | Each pixel glows by itself. Hard to make |
Most products sold as "LED TVs" are LCD screens lit from behind by LEDs. A 4K MicroLED screen needs about 24.9 million LEDs (3,840 × 2,160 pixels × 3 colors; our calculation). Placing them all with none missing is a major challenge.
A cousin: the laser diode
A laser diode also makes light by bringing electrons and holes together at a PN junction, just like an LED. The difference is that both ends of the chip act like mirrors, so the light bounces back and forth inside. This bouncing light triggers more light of the same color and direction. It comes out as a narrow, uniform beam.
LED light has a slight spread of colors and goes out in many directions. That suits lighting and displays. Laser diode light has a narrow range of colors and travels in one direction. It is used in fiber-optic communication and for reading Blu-ray Discs. The table "LED vs laser diode" shows the differences.
Common misconceptions
Myth: LEDs don't give off heat
In fact, more than half of the electricity fed to an LED still becomes heat. LED light has almost no infrared, so the lit side does not get hot. That is why LEDs feel as if they make no heat. The heat stays in the chip, so a heat sink must carry it away.
Myth: An LED's color comes from its colored plastic cover
In fact, the color is set by the step of the chip's material. Even with a clear cover, a red LED glows red and a blue LED glows blue. Many older indicator lights had covers the same color as their light. That made it look as if the cover created the color.
Myth: A white LED gives off white light directly
In fact, most white LEDs combine a blue LED with a yellow-glowing phosphor. They mix the colors so that the light looks white. When the LED is on, you see only white. That makes it hard to notice the colors mixing inside.
By the numbers
| Item | Number | Note |
|---|---|---|
| Step of gallium nitride | About 3.4 eV | Direct band gap. Its light is ultraviolet |
| Step of silicon | About 1.1 eV | Indirect band gap; gives off almost no light |
| Efficiency of a 60 W incandescent bulb | About 16 lm/W | US Energy Information Administration, 2014 |
| Average LED on sale | Close to 100 lm/W | IEA, 2026. High-performance products exceed 200 |
| Brightness used to define lifetime | 70% of the start | L70, 70% lumen maintenance |
| Bright blue LED | 1993 | Nobel Prize in Physics in 2014 |
The Nobel Prize announcement (2014) states that about a quarter of the world's electricity is used for lighting. Using one sixth of the power adds up to a big difference.
Learn more
- Compare in a table: Direct vs indirect band gap, LED vs laser diode, Forward vs reverse bias
- See a diagram: PN junction and depletion layer
- Terms for the next step: Quantum well (a structure that traps electrons and holes in a thin layer), Auger recombination (one cause of efficiency droop)
- Try the numbers: the band gap and wavelength calculator works out how a material relates to the color of its light
- The reverse process: "How Do Solar Cells Work?", "How Do Image Sensors Work?"