What Is a Laser Diode? How It Differs From an LED
A laser diode, also called a semiconductor laser, makes light the same way an LED does: electrons meet holes. The difference is that both ends of the chip act as mirrors. Light bounces back and forth, and light triggers more light just like it. The result is a beam of one color that can be focused tightly. Laser diodes carry data over fiber-optic internet links and inside data centers.
What you'll learn
- How laser light differs from LED light
- What "stimulated emission," "population inversion," "resonator" and "threshold" mean
- The two main shapes: edge-emitting lasers and VCSELs
- Why fiber-optic links use 1310 nm and 1550 nm
- Why lasers are hard to make from silicon, and how laser safety classes work
Prerequisites: it helps to know how LEDs make light ("How Do LEDs Work?").
Where you see it
A semiconductor laser is a part that gives off light of one uniform color from a tiny piece of semiconductor. It is also called a laser diode (LD). A diode is a part that lets current flow easily in only one direction. A laser diode is one kind of diode.
The announcement of the 2000 Nobel Prize in Physics listed several of its uses: fiber-optic internet, CD players, barcode readers and laser pointers. CDs went on sale in 1982. A laser beam reads the tiny bumps on the disc. DVDs and Blu-ray Discs work the same way.
Your phone has them too. Face ID, the face unlock used since the iPhone X in 2017, projects thousands of invisible infrared dots onto your face. From the pattern of dots, it measures the shape of your face. The light source is a type of laser diode called a VCSEL, explained later (Apple). The 2020 iPad Pro added a LiDAR Scanner, which measures distance from the time light takes to come back.
Out of sight, data centers use huge numbers of them. Servers and network equipment are linked by optical cables. Parts announced in 2022–2023 send 100 to 200 gigabits per second over a single beam of light.
What makes laser light special
Compared with LED light, laser light has three special features.
- One color: the range of wavelengths (the length of the light wave) is very narrow
- Waves in step: the peaks and troughs of the waves stay lined up as the light travels
- Easy to focus: a lens can gather it into a tiny spot, or turn it into a parallel beam that reaches far
LED light has a spread of colors and goes out in all directions. That is good for lighting a room. But it is poor at entering a thin glass strand and traveling far. Note that laser diode light also spreads right after it leaves the chip, because the opening is so small. So a lens shapes the beam before use. The table "LED vs laser diode" also compares the two.
How it works 1: light calls out light ("stimulated emission")
The starting point is the same as in an LED. Current flows across the boundary between two different kinds of semiconductor, called a PN junction. There, electrons meet holes (empty seats for electrons), and the energy comes out as a particle of light. In an LED, each meeting happens at a random time and in a random direction. This is called spontaneous emission.
Now suppose a particle of light with just the right energy passes an electron that is about to meet a hole. The electron is prompted to give off light too. The new light has the same color, direction and wave timing as the passing light. This is called stimulated emission. Albert Einstein predicted it in theory in 1917.
In stimulated emission, one particle of light becomes two. If those two make other electrons give off light, they become four, and so on. The name LASER itself is an acronym of "Light Amplification by Stimulated Emission of Radiation."
How it works 2: growth needs "population inversion"
A particle of light does not only make electrons give off light. When it hits a low-energy electron, it is absorbed, and its energy lifts that electron up. Normally there are far more electrons on the low side. So the light is absorbed before it can grow.
To make light grow, you need more high-energy electrons, which can give off light, than low-energy ones. This is the reverse of the normal arrangement, so it is called population inversion. In a laser diode, a large current keeps pushing electrons and holes into the glowing layer to create this state. To keep things simple, we skip the detailed spread of electron energies here.
How it works 3: mirrors at both ends bounce the light
Even with population inversion, light grows only a little in a single pass through the chip. So mirrors are placed at both ends of the glowing layer, and light bounces back and forth many times. Each round trip makes the light stronger and its direction more uniform. This pair of mirrors is called a resonator.
In many laser diodes, the mirrors are not separate parts. If you split a crystal in a certain direction, it breaks along the rows of atoms and leaves a flat face. This split face, called a cleaved facet, serves as the mirror. Light travels very differently in a semiconductor and in air. So about 30% of the light reflects at the boundary.
The remaining 70% or so passes through the mirror and leaves. This is the laser light we use. Some products coat one face so that it reflects more strongly, and take most of the light from the other face.
How it works 4: the current where lasing starts ("threshold")
Raise the current through a laser diode slowly from zero. At first it only glows weakly, like an LED. That is because more light is lost on each round trip than is gained. Light is lost when it leaves through the mirrors or is absorbed inside the crystal.
Once the current passes a certain value, the gain beats the loss, and the uniform light suddenly gets strong. The current at this border is called the threshold current. Above it, the light gets stronger in step with each extra bit of current.
The threshold changes with temperature: it rises as the laser warms up. The color of the light also shifts slightly. So lasers for communication are sometimes paired with a part that keeps the temperature constant.
Two shapes: edge-emitting lasers and VCSELs
A laser like the one in the figures so far sends light out from the end of a long chip. It is called an edge-emitting laser. Its glowing layer is long, so it can easily give strong light. It has long been used in long-distance fiber-optic links and optical discs.
The other kind sends light straight up from the surface of the wafer, the round disc on which chips are made. It is called a VCSEL (pronounced "vixel"), short for vertical-cavity surface-emitting laser. Mirrors made of dozens of stacked thin films sit above and below the glowing layer, so light bounces up and down.
| Edge-emitting | VCSEL | |
|---|---|---|
| Light comes out | From the side end of the chip | Upward from the wafer surface |
| Mirrors | Split crystal faces | Stacked thin films |
| Strengths | Strong light, long-distance links | Can be tested before the wafer is cut. Easy to place many side by side |
| Main uses | Fiber-optic links, optical discs | Face unlock, short links in data centers |
The material sets the color
As in an LED, a laser's color (wavelength) is set by the energy step an electron drops down, called the band gap. Each material has its own step. The bigger the step, the shorter the wavelength. Lasers use semiconductors that combine two or more elements, such as gallium arsenide. Most combine elements from groups 13 and 15 of the periodic table. These are called III-V compound semiconductors. The table below lists only the main material families.
| Use | Wavelength | Main material |
|---|---|---|
| Blu-ray | 405 nm (blue-violet) | Gallium nitride (GaN) family |
| DVD | 650 nm (red) | AlGaInP family |
| CD | 780 nm (infrared) | AlGaAs family |
| Short links in data centers | 850 nm (infrared) | Gallium arsenide (GaAs) VCSELs |
| Fiber-optic links | 1310 nm and 1550 nm (infrared) | InP family |
The shorter the wavelength, the smaller the spot light can be focused into. Each step from CD to DVD to Blu-ray shortened the wavelength. So a disc of the same size could hold more data.
Why optical fiber uses 1310 nm and 1550 nm
An optical fiber is a cable that carries light through a glass strand about as thin as a hair ("How Is Data Sent Over the Internet?"). Glass does not pass every color of light equally well. How fast light fades depends on its wavelength, and fading is lowest near 1550 nm.
In standard fiber, light at 1550 nm loses about 0.2 decibels (dB) per kilometer. Major makers list maximum values of about 0.18 to 0.22 dB per km (as of 2024). The decibel is a scale for how many times smaller something has become. A loss of 0.2 dB is a drop of about 5%. After 100 km, about 1% of the light is still left ("Decibels Explained").
Near 1310 nm, the speed of light in the fiber hardly changes with wavelength. Laser light still has a slight spread of colors. So at other wavelengths, the shape of each light pulse slowly blurs. At 1310 nm it blurs less. That is why this wavelength is common inside data centers and for links up to about 10 km.
These two wavelength ranges are called the O-band (around 1310 nm) and the C-band (around 1550 nm). Laser diodes built on an InP base can give off both efficiently.
Blinking light fast to send data
In optical communication, strong and weak light stand for 1 and 0. A laser diode changes its brightness as soon as its current changes. So some lasers, such as VCSELs, are blinked directly by changing the current.
For faster links, the laser stays on, and an "optical shutter" placed right in front of it does the blinking. Another trick splits brightness into four levels, so each blink carries 2 bits. In 2022, Coherent announced a part that puts the laser and the shutter on one chip. It sends 200 gigabits per second over one beam of light. Calling it a "shutter" is a simplification. It is really a section whose light absorption changes with the voltage applied.
Why lasers are hard to make from silicon
The chips in PCs and phones are made of silicon. But silicon is an indirect band gap semiconductor. When electrons meet holes in it, they make hardly any light. Most of the energy turns into heat instead. The page "Direct vs indirect band gap" explains why.
On the other hand, silicon absorbs almost no light at 1310 nm or 1550 nm, so the light passes through well. That is why silicon photonics is spreading: it builds optical parts on silicon chips. These parts include light paths, shutters and photodetectors (parts that turn light into electricity). Only the light source must come from elsewhere. An InP or similar laser is bonded on, or light is fed in from a separate part.
Safety classes
Laser light travels as a narrow, strong beam, so it can damage the retina if it enters the eye. The international standard IEC 60825-1 sorts laser products into classes by how dangerous they are. In Japan, for example, the matching national standard is JIS C 6802.
- Class 1: safe in normal use. The iPhone 15 Pro, for example, is in this class (Apple)
- Class 2: visible light, up to 1 milliwatt. You are protected by the reflex of shutting your eyes at the glare
- Class 3R and 3B: can harm the eyes if you look straight into the beam
- Class 4: even scattered reflections are dangerous, and it can cause burns or fires
There are also Classes 1M and 2M, which are dangerous only when viewed through a magnifying glass or binoculars. In Japan, for example, laser pointers are specially regulated by law. Those that meet the standard carry the PSC mark (National Consumer Affairs Center of Japan). Infrared lasers are invisible, so there is no glare to warn you of danger. Take extra care with them.
A short history of the laser diode
In the fall of 1962, four groups in the United States each made a semiconductor laser on their own. They worked at two GE labs, IBM and MIT Lincoln Laboratory. Most used gallium arsenide, and Nick Holonyak's team at GE made visible red light. But these lasers had to be cooled, for example with liquid nitrogen (about −196 °C). Current could flow only for very short moments.
To keep a laser shining at room temperature, it had to reach population inversion with less current. The answer was to sandwich the glowing layer between other semiconductors with a larger step. Electrons, holes and light then gather in the middle layer. This is called a double heterostructure ("hetero" means "different").
In 1970, two teams made lasers that shine continuously at room temperature, at almost the same time. One was Zhores Alferov's team in the Soviet Union. The other was Izuo Hayashi and Morton Panish at Bell Labs in the US. Alferov's team reported slightly earlier, and Hayashi's team confirmed their result in June of the same year. The structure had been proposed separately in 1963 by Alferov's team and by Herbert Kroemer. Alferov and Kroemer won the 2000 Nobel Prize in Physics for this "heterostructure" research. The other half of the prize went to Jack Kilby, for the integrated circuit.
Common misconceptions
Myth: Laser light is visible, like red or green
In fact, many laser diodes give off infrared light, which is invisible. The dots for face unlock and the 1310 nm and 1550 nm light in optical fiber are all infrared. Familiar laser pointers are red or green, so people tend to think lasers are always visible.
Myth: Push enough current through an LED and it becomes a laser
In fact, without a resonator to bounce light back and forth, stimulated emission hardly builds up uniform light. An LED only gets brighter. Its spread of colors and the way its light spreads out stay the same. Both glow at a PN junction, so the difference can look like a matter of brightness alone.
Myth: Silicon cannot handle light at all
In fact, silicon lets the infrared light used in communication pass well. Light paths, optical shutters and photodetectors can be built on a silicon chip. What silicon is bad at is giving off light. So silicon photonics leaves only the light source to compound semiconductors.
By the numbers
| Item | Number | Note |
|---|---|---|
| First laser diodes | 1962 | Four groups; cooled with liquid nitrogen |
| Continuous at room temperature | 1970 | Double heterostructure |
| Light reflected at a cleaved facet | About 30% | The rest leaves the chip |
| Fiber loss (1550 nm) | About 0.2 dB per km | About 1% left after 100 km |
| Speed over one beam | 200 gigabits per second | Example of a part announced in 2022 |
| Class 2 output | Up to 1 milliwatt | Visible-light lasers |
200 gigabits per second means sending 200 billion 0s and 1s every second. With four brightness levels, each step sends 2 bits. So the brightness switches about 100 billion times a second. A tiny chip, inside a part the size of a fingertip, changes its brightness this fast.
Learn more
- Compare in a table: LED vs laser diode, Single-mode vs multimode fiber, NRZ vs PAM4 (how many brightness levels to use)
- Terms for the next step: Quantum well (an even thinner glowing layer), WDM (wavelength division multiplexing) (combining light of different colors in one fiber)
- Light and chips in the near future: CPO (co-packaged optics) (placing optical parts right next to the chip)
- The basics of how chips make light: "How Do LEDs Work?"
- The receiving side: "How Do Image Sensors Work?"