Answer a) is wrong: the saturation intensity is relevant only under steady-state conditions, which have not been reached here. Answer b) is correct: as the pulse builds up in a time well below the upper-state lifetime, saturation occurs when the incident energy reaches the saturation energy. Answer c) is not quite right, but would usually not be too far off; the stored energy is typically several times the saturation energy.

Answers b), c) and d) are correct. The upper-state lifetime is not relevant – it governs decay via spontaneous emission, while the pulse reduces the gain via stimulated emission.

Answer a) is correct. One will usually design a passively mode-locked laser so that the saturable absorber will be sufficiently saturated by the pulses under steady-state conditions. (This means that stronger focusing on the absorber will be applied for short resonators.) A large ratio of peak power to average power then means that in the initial phase of pulse build-up there is hardly any saturation of the absorber, so that self-starting mode locking is difficult to achieve. Indeed this is consistent with experimental experience.

Answer c) is correct. Dispersion compensation is only approximate (if at all used), and also there are usually significant nonlinear effects in the laser resonator. Exact equidistance of the lines in the Fourier spectrum arises from the periodicity of the generated pulse train, and that is ensured by the action of the saturable absorber or active mode locker: such a device prevents the pulses e.g. from progressive broadening.

Answer b) is correct. The idler wave is essential for the parametric amplification process: additions to the complex phase of the signal wave (as required for amplification) depend on both pump and idler amplitude.

Answer b) is correct. The losses cannot be different for the two propagation directions; such non-reciprocal behavior requires some other effects such as Faraday rotation. This is true although intuition may suggest that the losses are smaller when the light comes from the fiber with the smaller core. This intuition, however, is at least partly based on ray optics, which breaks down in this regime, where full wave optics is required.

Answer c) is correct. Answer b) is wrong: it looks like two rings, but actually it is a single ring with an inserted nonlinear loop mirror. See the article on mode-locked fiber lasers for more details.

Answer a) is correct: a high numerical aperture combined with a large mode area would lead to multimode guidance (→ multimode fibers). As an unfortunate side effect, a low NA leads to high bend losses even for moderate bend radii. It is true that this can be used for attenuating higher-order modes of a fiber, since bend losses tend to be higher for those than for the fundamental mode, but for a single-mode fiber this effect can of course not occur.

Answer b) is correct, although the term "superfluorescence" is sometimes wrongly used instead of "superluminescence".

Answer b) is correct. A DFB laser (= distributed feedback laser) is a laser where single-frequency operation is achieved with a periodic structure within the gain medium. This principle can be applied to laser diodes, but also to fiber lasers and other types of lasers.

Answer a) is correct. Fiber lenses are often used for fast axis collimation of diode bars.

Answers a) and c) are correct. Parametric fluorescence is indeed similar to amplified spontaneous emission (ASE), and as ASE it is a quantum effect. However, while spontaneous emission in a laser gain medium generates a power loss which is responsible for a finite threshold pump power of a laser, parametric fluorescence does not lead to a significant power loss, since it is emitted only into a single spatial mode.

Answer c) is correct, answer b) at least not totally wrong. ASE is weak for the typical gains (at most a few dB) encountered in bulk lasers, but it is actually also quite weak in most fiber lasers, even though their laser gain may on average be somewhat higher. A waveguide effect is not relevant here.

Answer a) is correct. The vacuum field has no photons at all, even though it has fluctuating electric and magnetic fields. Photons are associated with the excitation of the electromagnetic field above its ground state.

The "one photon per mode" rule, as sometimes applied e.g. for estimating the output of optical parametric generators or ASE sources, is nevertheless no nonsense. It can be used for estimating the magnitude of the vacuum field fluctuations in a semiclassical picture, even though there are actually no photons in the vacuum field.

Only answer c) is correct. Ytterbium-doped fiber amplifiers are usually more efficient than those based on erbium-doped fibers, and particularly neodymium-doped ones can have a lower noise figure. Erbium-doped fiber amplifiers (EDFAs) do attenuate any input signals when not being pumped (particularly for signal wavelengths around 1535-1550 nm), because they are based on a quasi-three-level gain medium. Nevertheless they are suitable for transparent optical networks; that kind of transparency has a different meaning.

Answer b) is correct. A frequency comb structure occurs only when there is a well-defined phase relationship between subsequent pulses. Such a phase relationship usually exists for pulse trains generated from mode-locked lasers, at least when there is a single circulating pulse in the resonator (→ fundamental mode locking) from which all emitted pulses are derived. In a gain-switched laser diode, each pulse is generated with a random phase, determined by noise processes, so the spectrum of the pulse train is continuous in that case and does not exhibit a comb structure.

However, there is a bit of truth in answer c): in principle, one could operate the laser diode so that some small optical power remains between the pulses, and coherence is not lost. However, I am not aware that anybody has tried that, and the achievable coherence would probably be very poor compared to that of a mode-locked laser.