The Semiconductor Laser

The same p-n junction, run in a different regime, becomes a light source — and with the right structure, a laser.

The three requirements for lasing

Any laser needs three things, and the value of the semiconductor version is how compactly it supplies all three.

1. Stimulated emission. A photon of energy $h\nu$ meeting an already-excited system triggers emission of a second photon that is identical in energy, phase, direction and polarisation. Two coherent photons leave where one arrived. This is what makes laser light coherent, and it is the amplification step.

2. Population inversion. Stimulated emission must outrun absorption, which requires more excited systems than unexcited ones — a state that never occurs in thermal equilibrium, since the Fermi function guarantees lower states are more populated. It must be forced.

3. Optical feedback. Photons must pass through the active medium repeatedly, so a resonant cavity is needed.

Construction

The active element is a heavily doped p-n junction, degenerate on both sides — the same condition as the tunnel diode. Both regions have doping around $10^{19}\ \text{cm}^{-3}$, so $E_F$ lies inside the conduction band on the n side and inside the valence band on the p side.

The material must be a direct band gap semiconductor, and this is not optional. In gallium arsenide the conduction band minimum sits directly above the valence band maximum in momentum space, so an electron can drop across the gap emitting only a photon. In silicon, an indirect gap material, the minimum is displaced in momentum, so the transition additionally needs a phonon to conserve momentum — a three-body process, far less probable. That is why there is no silicon laser diode, and why optoelectronics uses III–V compounds such as GaAs, GaAsP and InGaAsP while computing uses silicon.

The cavity is elegantly simple: the crystal is cleaved along natural planes at each end. Semiconductor refractive index is high — about 3.6 for GaAs — so the semiconductor-air interface reflects around 32% by Fresnel reflection alone. That is sufficient feedback, so no mirrors need be added. The two remaining faces are roughened to suppress lasing sideways.

Typical dimensions are a few hundred micrometres, making the whole cavity smaller than a grain of sand.

Working

Forward bias the junction heavily.

1. Electrons are injected from n into the junction region, holes from p.
2. In a narrow active region at the junction, both are present in large numbers simultaneously — many electrons in the conduction band and many empty states in the valence band at the same place. That is population inversion, achieved here by carrier injection rather than by optical pumping. No external lamp or second laser is needed, which is the semiconductor laser's great practical advantage.
3. Electrons recombine with holes, emitting photons of energy approximately $E_g$.
4. At low current, recombination is spontaneous, incoherent and broad-spectrum — the device is simply an LED.
5. Above a threshold current, the injected carrier density is high enough that stimulated emission overtakes absorption and spontaneous emission. Photons bouncing between the cleaved faces trigger more identical photons on each pass.
6. Output becomes coherent, monochromatic, sharply directional, and rises steeply with current.

The threshold current is the defining specification. Below it the device is an LED; above it, a laser. Modern heterostructure designs, which confine both carriers and light to a thin layer, have pushed thresholds down to a few milliamps — early homojunction devices needed such large currents that they ran only in pulses, or cooled in liquid nitrogen.

Wavelength is set by the band gap, $\lambda \approx 1240/E_g$ nm with $E_g$ in eV. GaAs at 1.43 eV emits near 870 nm in the infrared. Alloying shifts it — adding phosphorus widens the gap towards red, InGaAsP is tuned to 1310 nm and 1550 nm where optical fibre loss is lowest, and gallium nitride reaches blue and violet.

Comparison and applications

Against a gas laser: far smaller, far more efficient (tens of per cent against under 1% for helium-neon), directly modulated at gigahertz by simply varying the drive current, and cheap. The costs are a wider spectral linewidth, and a beam that diverges strongly because it emerges from an aperture only a wavelength or so across — which is why a collimating lens is always fitted.

Applications: optical fibre communication, the dominant use by value; CD, DVD and Blu-ray, whose progression from 780 to 650 to 405 nm is exactly the pursuit of a shorter wavelength to store more per disc; laser printers; barcode scanners; optical mice; laser pointers; materials processing; and medical and surgical instruments.