2.1.

Gain medium

There are several design aspects concerning the proper choice of the gain medium. When restricting the laser wavelength to roughly one micron, only Neodymium and Ytterbium can efficiently serve as active ions. At a first glance, Ytterbium has an appealing level structure and offers very little intrinsic energy losses. However, if an efficient amplifier is to be built, the fluence of the amplified pulse should be around five times the saturation fluence of the gain medium. Otherwise, complex multipass amplifier geometries have to be employed in order to achieve efficient energy extraction. The saturation flu-ence of Yb:YAG is approximately 10 J/cm², so the pulse should have a fluence of about 50 J/cm². This fluence exceeds typical coating damage thresholds. Therefore, Ytterbium is not a suitable laser ion for a high energy pulse amplifier.

High pulse energy amplifiers with Nd-doped laser media have been built for decades, mainly for inertial confinement fusion experiments. Although the standard for high pulse energy amplifiers, Nd:Glass is also excluded, mainly due to its high saturation fluence and inhomoge-neous level broadening. The intended laser is forced to single-frequency operation by a twisted-mode resonator. Thus, if a zigzag-type geometry is employed, naturally birefringent host materials such as YVO4 have to be excluded. As it turns out, the best choice left over for a space-qualified amplifier is Nd:YAG. It is nevertheless a rather unsuitable medium for high efficiency operation, since roughly 36 % of the pump energy are directly lost due to nonradiative losses inside the ions. The reason why Nd:YAG is usually not used for high energy storage amplifiers is the relatively high small-signal gain. ASE and PO are strongly facilitated in high gain crystals, so measures have to be taken to prevent the onset of parasitic oscillation and to reduce ASE.

2.2.

ASE and PO

To reduce the gain for a single pass of spontaneously emitted photons, the laser beam is folded three times, as shown by the crystal footprint in figure 3. The longest path through the medium is then given by the diagonal, assuming that TIR at the sidefaces is prohibited. When keeping the footprint area constant, the diagonal scales with with x as aspect ratio of the rectangle. The pathlength therefore minimizes for a square footprint (x = 1). It is important to note, that this reduction of ASE works only if all three beams are blocked by the Q-switch.

Figure 3.

PO risk reduction: a) long medium, b) wide medium with triply folded beam

Another important aspect is gain that can not be accessed by the laser mode and therefore helps ASE and PO. This problem can be overcome by doing just a single “zig”, a laser geometry that is commonly referred to as “grazing-incidence” laser (GISL). Figure 5 shows the CTSL (composite thin slab laser) developed by the University of Applied Sciences Muenster [6, 7]. It consists of a thin Nd:YVO₄ plate diffusion-bonded to an undoped YVO₄ crystal. With this geometry, the only inversion left over is in the tips at the crystal edges.

Other than using the introverted trapezoid of figure 5, all side faces should form an outwards directed trapezoid, as shown in figure 4. This geometry prevents ASE from doing a multiple-bounce path along the crystal by reflecting the photons upwards into an undoped crystal.

Figure 4.

Outward directed trapezoid

Figure 5.

Topview of the CTSL

2.3.

Pumping and cooling

According to the requirements above, some general design drivers can be developed. Starting from a zigzag slab design, first of all the cooling principle of thin disk lasers is implemented. Therefore, the crystal is divided into two crystals, a thin Nd:YAG slab and a large undoped YAG crystal, which are bonded together. With the thin disk directly cooled over the large surface, heat extraction will be facilitated and transverse thermal gradients will be reduced. The undoped crystal will prevent the thin disk from bending upwards. A prototype for a laser design including all design aspects made so far is shown in figure 6. The pump light is coupled in over the large top surface. The geometry allows to use diode stacks without any pump beam shaping optics, since the pump light is TIR-guided inside the bulk undoped crystal. A highly reflective coating at the bottom side of the crystal assures a double pass of the pump light.

Figure 6.

Laser prototype design

However, due to the thin doped crystal, the laser becomes sensitive for wavelength drifts of the pump diodes. In order to get an estimate of the percentage of absorbed pump power, a 4-nm (90 % power content) diode bar spectrum is numerically folded with a Nd:YAG absorption spectrum. The peak absorption coefficient of the Nd:YAG spectrum is conservatively assumed to be 9.5 cm^-1. Figure 7 shows that an 8 mm thick crystal is required to achieve more than 80 % pump power absorption, when diode wavelength drifts of 5 nm are allowed. With a 2 mm thick crystal, an effective thickness of 8 mm can be achieved by a 4-pass pumping scheme, as depicted in figure 8(a). To pump the crystal, one can simply put a vertical diode stack (stacked in fast-axis direction) close to the left side of the roof in figure 8(a). Up to date there is no commercial vertical stack available with more than 40-50 diode bars, while an efficient 500-mJ laser requires more than 200 diode bars. Therefore, the diodes are distributed over several stacks with 30 bars each. Since the diodes are clamped inside a metal holder as shown in figure 8(b), the fill factor would be rather low and the pump light distribution would be inhomogeneous. To solve this problem, the diode stacks can be arranged in an interleaved manner. Figure 8(c) shows a top view of the crystal roof. By coating the crystal with a pattern of AR and HR coatings, diode stacks of 30 bars each can be arranged such that the metal holders only cover the HR coated parts of the pattern. Another advantage of this pump geometry would be pump redundancy. Since the geometry is scalable, redundant stacks can be implemented to compensate for the failure of a whole diode stack. To conclude, figure 8(d) shows the complete laser crystal layout with the triply folded laser beam.

Figure 7.

Effective absorption in Nd:YAG

Figure 8.

a) 4-pass pumping scheme, b) diode stack (DILAS), c) interleaved diode arrangement, d) complete laser concept layout

3.1.

Master-oscillator

The master-oscillator is pumped by two diode stacks of 20 bars each. Using 40 bars allows to operate them at 80 % of the nominal output power, in order to compensate for unavoidable degradation. Furthermore, the pump pulse duration can be kept low to avoid excessive spontaneous emission losses.

Our numerical model based on the work of Degnan [8, 9] and Zhang [10, 11] predicts an electro-optical efficiency of about 10 % in the worst case, when roundtrip losses are assumed to be 10 %. This efficiency can only be achieved when ASE and PO are prohibited, which will be the case for our new concept. The model takes the overlap of the gain and beam intensity distribution into account. A FEM model shows that the temperature distribution is smooth and will be averaged out by the grazing-incidence laser beam path, as has been experimentally confirmed for the CTSL [6].

Figure 9.

General MOPA scheme

3.2.

Power-amplifier

The power-amplifier has to add 450 mJ to the pulse. The necessary pump energy is supplied by 8 diode stacks with 30 bars each, which are arranged in the shown interleaved pattern. Numerical calculations of the amplifier indicate an expected e-o efficiency of 14 %. Again, this enormously high efficiency can only be achieved, if ASE and PO are successfully suppressed.