Gain switching

Gain switching produces a short optical pulse by modulating the gain itself. A pump pulse drives the inversion (in a semiconductor, the carrier density) far above threshold before the photon population has had time to build up; the photons then grow exponentially, deplete the gain below threshold in one burst, and the pump is removed before a second burst can form. The result is the first spike of the relaxation oscillation, isolated. Applied to a laser diode with sub-nanosecond current pulses on top of a DC bias just below threshold, it typically yields pulses of 10–100 ps, peak powers from hundreds of milliwatts to a few watts, and repetition rates set freely by the electronics, from single shot to several GHz.

Mechanism and time scales

The pulse cannot be shorter than a few cavity photon lifetimes. For a Fabry–Pérot diode 300 µm long with facet reflectance 0.3, internal loss 10 cm⁻¹ and group index 3.6, the mirror loss is

$$\alpha_m = \frac{1}{L}\ln\frac{1}{R} \approx 40\ \text{cm}^{-1},$$

and the photon lifetime is

$$\tau_p = \frac{1}{v_g(\alpha_i + \alpha_m)} \approx 2.4\ \text{ps}.$$

Real gain-switched pulses are several times longer than this, because the carrier density takes a finite time to be pushed above threshold and the photon number needs many photon lifetimes to grow from spontaneous emission. The shortest pulses come from driving hard, with a steep current edge, and from ending the current pulse before the second relaxation spike appears. High differential gain and short cavities help.

Solid-state and dye lasers can also be gain-switched by a short optical pump pulse, for example a Q-switched frequency-doubled Nd:YAG pulse pumping a Ti:sapphire crystal, or a Q-switched thulium or holmium laser near 2 µm pumping Cr:ZnSe. The output is then a nanosecond pulse delayed from the pump by the photon build-up time, with a duration set by the cavity length and pump intensity.

Chirp

In a semiconductor, the carrier density falls sharply during the pulse, and through the linewidth enhancement factor $\alpha_H$ that changes the refractive index. The transient part of the instantaneous frequency shift is

$$\Delta\nu(t) = \frac{\alpha_H}{4\pi}\frac{d\ln P}{dt}.$$

With $\alpha_H = 5$ and power rising by a factor of e in 10 ps, $\Delta\nu \approx 40$ GHz, about 0.32 nm at 1550 nm. The frequency sweeps from high to low across the pulse, so gain-switched pulses carry a largely downward frequency chirp. A 20 ps Gaussian pulse at 1550 nm would need only 22 GHz (0.18 nm) if transform-limited, since its time-bandwidth product would be 0.441; measured gain-switched spectra are commonly several times broader than that.

Because the red components come last, a length of fiber with normal dispersion (such as dispersion-compensating fiber at 1550 nm) lets them catch up and compresses the pulse, often by a factor of several, to a few picoseconds.

Measurement

A fast photodiode on a sampling oscilloscope resolves pulses down to roughly 10–20 ps, depending on the combined bandwidth; below that, an intensity autocorrelation or a streak camera is used. The spectrum on an optical spectrum analyzer, compared with the measured pulse duration, gives the time-bandwidth product and so the degree of chirp.

Where it is used

Gain-switched diodes are the standard pulsed source for time-correlated single-photon counting and fluorescence lifetime measurement, where 405–1550 nm diode heads give 30–100 ps pulses at user-selected rates. They also seed fiber amplifiers in MOPA systems for picosecond micromachining, drive optical time-domain reflectometers, and serve as pulse sources in quantum key distribution transmitters. The main advantage over mode locking is that the repetition rate and timing follow the electrical trigger with no cavity-length constraint.

Pitfalls

A Fabry–Pérot diode switches on in several longitudinal modes, and the mode partition changes from pulse to pulse; a DFB laser gives a single mode, and injection seeding from a CW laser gives a single mode and lower jitter. Without injection seeding, each pulse starts from spontaneous emission, so the phase between successive pulses is random and the turn-on time fluctuates by a few picoseconds. Too long or too strong a drive pulse produces a second spike or a long tail that dominates an autocorrelation trace.

Common questions

How is gain switching different from Q-switching?

Q-switching holds the cavity loss high while the gain medium stores energy, then drops the loss; gain switching leaves the loss fixed and raises the gain quickly. Q-switching needs a storage lifetime much longer than the pulse (about 230 µs in Nd:YAG), which semiconductors lack, so diodes are gain-switched.

What limits the shortest gain-switched pulse?

The photon lifetime and the speed of the gain rise. Directly gain-switched diodes commonly give 20–50 ps; compensating the chirp can shorten this to a few picoseconds.

Can a gain-switched laser have a narrow linewidth?

Each pulse is chirped, so its spectrum is broad compared with the transform limit. Injection seeding narrows the spectrum, reduces the chirp and fixes the phase, but some transient chirp from the carrier dynamics remains.

References: Siegman, Lasers (University Science Books, 1986); Saleh & Teich, Fundamentals of Photonics 3rd ed. (Wiley, 2019).