How does an LED make light?

How the parts work together

1. A chip in a cup

   This red LED runs at 20 mA, the current its datasheet is tested at. The supply is set to 2.1 V but limits the current to 20 mA, so 2.00 V sits across the LED. Seen close up, the glowing chip sits on the floor of the reflective cup coined into the cathode's anvil, and a gold wire joins its top to the anode post. Orbit to look into the cup, or pick the blue LED to see a chip with two wires.

2. Electrons and holes

   The slice on the right magnifies the chip's layers: n-type at the bottom, the active layer in the middle and p-type on top. At 0 V no current flows. The n-type layer holds free electrons (blue balls) and the p-type layer holds holes (orange rings), but each kind keeps back from the active layer: an electric barrier at the junction leaves a gap either side of it. A forward voltage lowers that barrier and closes the gap. With none, nothing meets and the chip is dark.

3. They meet and make light

   At 1.9 V about 6.2 mA flows. Electrons drift up from the cathode side and holes drift down from the anode side. Where an electron and a hole meet in the active layer, the electron drops into the hole and can give off its spare energy as a photon of about 1.97 eV: red light at 630 nm. Each electron gets 1.9 eV from the supply, a little less than that; heat taken from the crystal can make up the difference. The model draws a photon for every meeting on the datasheet curve; in a real chip only some meetings give light. The glow is drawn in proportion to the light, here 31% of the light at 20 mA, so at 1 mA (5%) the dome looks almost dark: read the light readout as well.

4. A blue photon needs more voltage

   Here the blue InGaN LED gets 2.1 V, enough to drive the red LED to the supply's 20 mA limit. The blue one passes under 0.1 mA, so little that its datasheet curve still sits on zero, and it stays dark. Its photons carry 2.70 eV, about 0.6 eV more than each electron gets at 2.1 V, and the crystal's heat cannot supply that much. On the graph the blue curve leaves zero only near 2.4 V and reaches 1 mA at about 2.57 V. Raise the voltage slowly towards 2.6 V and the first meetings appear. At first most are dull brown flashes with no photon: meetings that give only heat. Below 1 mA the curve cannot be read, so the readout gives only an upper bound, and the model assumes that the extra current there makes no light.

5. Then the current takes off

   At 3.2 V the blue LED passes 16.4 mA and gives 82% of the light it gives at 20 mA. Past the bend the current rises steeply with voltage, and the light rises in proportion to the current. The photons stay at 460 nm: more current gives more photons each second, not bluer ones. At 20 mA this LED needs 3.30 V, 0.60 V more than its photons carry; that extra energy per electron ends up as heat in the chip. So heat flows both ways: below the photon energy the crystal's heat tops up the electrons that make light, while above it the extra voltage, and the whole energy of any meeting that gives no photon, warms the chip.

6. Photon energy is only a guide

   Photon energy sets only roughly where a curve rises. Green photons carry 2.38 eV, 0.31 eV less than blue ones, yet at 2.9 V the green LED passes 4.9 mA while the blue one passes 7.4 mA: on the graph the bold green curve lies to the right of the faint blue one below about 13 mA. The green part needs 2.61 V for 1 mA, 0.23 V more than its photons carry. The red and yellow parts have almost the same curve, though yellow photons carry 0.13 eV more. So read each LED's own curve rather than predicting it from the colour.

7. Light you cannot see

   The infrared GaAlAs chip passes 19.3 mA at only 1.3 V. Its photons carry 1.41 eV, a little more than each electron gets, and again heat taken from the crystal makes up the difference. Its 880 nm light is invisible to the eye, so the photons are drawn grey and dashed and the dome stays dark while the current flows. Remote controls are one use for LEDs like this.

8. Connected backwards

   Turn the supply round and the LED blocks it: the anode now goes to the supply's negative side and the cathode to the positive. Its datasheet allows at most 10 µA at 5 V backwards (the green and blue parts allow up to 50 µA), so the model shows no current and no light. The supply now pulls the electrons and holes further away from the active layer instead of pushing them in. Current through it one way only: that is the diode in light-emitting diode.

The main parts

Epoxy lens

The clear plastic body: a 5 mm cylinder with a domed top and a rim at the bottom, about 8.7 mm tall, with a flat on the rim beside the cathode. It holds the metal parts in place and its dome bends the light into a narrow beam (the datasheets give 20 degrees between the half-brightness directions). It is drawn see-through.

Anode lead and post (+)

The longer leg, 0.5 mm square, which goes to the positive side of the supply. Inside the lens it ends in a small flat post, and a gold bond wire runs from the post to the top of the chip. Conventional current enters here (the arrow on the leg).

Cathode lead and anvil (−)

The shorter leg, on the side of the flat, which goes to the negative side of the supply. Inside the lens it widens into the anvil, whose head carries the reflector cup and the chip. Electrons enter the LED here; conventional current leaves (the arrow on the leg). Each leg has a short wider section about 4 mm below the rim, as the package drawings show. The legs are drawn shortened; real ones are at least 27 mm long.

Reflector cup

A shallow cone coined into the top of the anvil's head, on the lens's axis. The chip sits on its floor, and its sloping wall turns sideways light upward into the lens.

LED chip

The chip (die) where the light is made, about 0.3 mm across in this drawing, smaller than a grain of sand. AlGaInP chips are grown on conducting gallium arsenide, so the bottom of the chip sits on the cup as one contact and one wire joins the top; the infrared chip is drawn the same way. The green and blue datasheets do not say what their InGaN chips are grown on. On sapphire, which does not conduct, both contacts are on top and two wires are needed, as drawn here; Kingbright's WP7113VGC/A is an InGaN chip on conducting silicon carbide, which needs only one.

Bond wire

Gold wire joining the top of the chip to the anode post; InGaN chips have a second wire from their n-side contact to the cathode's cup. Real bond wire is about 25 µm thick; it is drawn thicker so you can see it.

n-type layer

In the magnified slice: semiconductor doped so that it has spare free electrons, shown as blue balls. It connects to the cathode. The slice is a cut through the chip's layers, enlarged far more than the rest of the drawing and not to scale.

Active layer

In the magnified slice: the thin layer between the n-type and p-type layers where electrons and holes meet. Its material decides the photon energy and so the colour. Real active layers are only nanometres thick; it is drawn much thicker.

p-type layer

In the magnified slice: semiconductor doped so that it is short of electrons. Each gap behaves like a positive carrier called a hole, shown as an orange ring. It connects to the anode.

Electrons

Blue balls. With enough forward voltage they drift up out of the n-type layer into the active layer. More current means more of them arrive each second. Their drawn speed is symbolic, not a real speed.

Holes

Orange rings: places where an electron is missing. With enough forward voltage they drift down out of the p-type layer into the active layer, where each can meet an electron.

Photons

When an electron meets a hole it drops into the hole and can give off its spare energy as one photon, drawn as a short wave. The spacing of the wiggles is in proportion to the real wavelength, so blue waves are shorter than red ones. Infrared photons are drawn grey and dashed because the eye cannot see them. Only light heading up out of the chip is drawn. The drawn rate of meetings is in proportion to the current: 8 per shown second at 20 mA, when about 1.2 × 10¹⁷ electrons a second really cross. Below 1 mA a growing share of the drawn meetings give only a small dull brown flash of heat and no photon: the model assumes the extra current there makes no light.

Questions worth exploring