Single Frequency Lasers

For science the two most important types of lasers are arguably Single Frequency Lasers and Mode Locked Lasers. While both are intrinsically very stable, single frequency lasers operate in continuous wave (continuously emits light) while the mode locked lasers emit very short pulses of light.

Back when I worked at Laser Zentrum Hannover e.V. we were developing laser sources for gravitational wave detectors (specifically LIGO). These Laser systems need to be very stable as they are used to detect very small distance changes caused by gravitational waves. They also need to be continuous wave as pulses would become radiation pressure noise.

A Single Frequency Laser is not really single frequency

Anyone who paid attention in quantum mechanics class knows that a pure mathematical sine wave with a delta peak as the spectrum is not possible in the physical world. This is - of course - also true for lasers and in fact the best possible ‘single frequency’ laser will have a linewidth given by the Schawlow–Townes equation:

This begs the question what is really meant by a single frequency laser? Basically a better name would be single longitudinal mode laser in the sense that only a single longitudinal mode oscillates. Or to be even more precise narrow linewidth laser.

The heart of a single frequency laser: meet the NPRO

From the Schawlow–Townes equation you might assume that single frequency lasers would have a very long resonator length, as linewidth can be reduced by increasing the resonator length. However this is not the case. While the Schawlow–Townes Equation gives a fundamental limit for the quantum limited linewidth, in practice this almost never reached. The most important thing to get the linewidth down is to make sure only one longitudinal mode oscillates in the resonator. What decides what modes oscillate in a resonator? In principle all modes that have positive net gain (gain from the active medium minus losses) can oscillate. The gain from active medium is usually somewhere from few nm up to many nm. The smallest range is usually supported by gases, which is why many early single frequency lasers were gas lasers. Besides the medium bandwidth, the other important factor is the resonator. The resonator - which is typically a Fabry–Pérot resonator can support modes with a spacing given by the free spectral range $\Delta \nu_{fsr} = c/(2L)$. Here you see the length of the resonator in the denominator. If you only want a single mode supported in your laser - hence within the gain bandwidth - you will usually need to make $\Delta \nu_{fsr}$ large - which means you will need a short resonator even though this also means you will increase the Schawlow–Townes linewidth. There are a few more tricks, for example you are not only limited by the material gain but you can also insert filters (gratings etc.) however in practice all such lasers have a resonator length in the order of cm.

More power: let’s amplify

Solid state amplifiers are quite straightforward, but as always suffer from limited gain (naturally this is more or less severe depending on material and seed power) and thermal lenses. Fiber amplifiers on the other hand allow large gain and are very efficient. However they suffer from stimulated Brilloin scattering. Brilloin scattering is an effect where photons scatter with acoustical phonons. In case of stimulated Brilloin scattering this leads to an effective index grating in the fiber which exponentially increases the backscattered light.

There are some more exotic ways of achieving amplification such as injection locked lasers. Here a high power laser is used as the high power source and by coupling the low power master oscillator beam into the high power lasers resonator and locking to it, the single frequency property is copied from the low power master oscillator onto the high power slave oscillator. As this is more complicated then a normal amplifier one could wonder why do this at all. One major motive is often that the quantum limited noise can be lower than for the amplifier approach, however this only matters if quantum noise is really the dominating effect. In frequency regions below 1 MHz where technical noise dominates there is basically no difference between the two.