1. Introduction

There is increasing interest in multi-watt solid-state, picosecond laser sources for use in industrial applications, as well as basic scientific applications, as the high peak powers attainable in a short pulse format mean they offer advantages for material processing with reduced heat damage, and provide efficient harmonic conversion to the useful visible and ultraviolet spectral regions. Many of these systems are based on passive modelocking with Semiconductor Saturable Absorber Mirrors (SESAM) as they are simple to implement, and modern growth technologies mean that their properties can be tailored for particular implementations [1,2]. Further, many of these SESAM laser systems are based on diode end-pumped solid state geometries [3–5], and whilst this technique offers the possibility of achieving very high beam quality, the power scaling potential is limited by thermal issues arising from the high optical intensities at the pump face. Side-pumped slab geometries have been investigated to try and overcome this issue by distributing the pump power over a larger volume [6,7]. This gives better power scaling potential, but comes at the cost of generally reducing the attainable beam quality and efficiency due to the asymmetry between the pumped volume and the laser mode.

An alternative to these is the bounce geometry [8], which has been shown to simultaneously address the problems associated with both of these schemes, and demonstrate high power scaling potential whilst also maintaining exceptional beam quality. By making use of a total internal reflection at a grazing incidence angle from the pump face of a highly absorbing slab laser crystal, spatial averaging of both the gain non-uniformities and thermal aberrations is possible, and the laser mode is able to extract the inversion close to the pump face very efficiently to experience a high gain. This has been demonstrated in previous work where the bounce geometry has been employed at the 100 W power level in a high power CW TEM₀₀ oscillator and amplifier system [9], high power TEM₀₀ Q-switched operation at high repetition rates >600kHz [10], and as a high power TEM₀₀ self-adaptive system [11]. This body of work demonstrates the ability of the bounce geometry to produce high spatial quality beams at high powers and as such is a very promising candidate for high power modelocked operation. In this work, we present the first implementation of a modelocked oscillator in a diode side-pumped bounce amplifier geometry at 1064 nm, and further, the use of a master oscillator power bounce amplifier to demonstrate the potential for high power scaling of the modelocked source.

2. Modelocked oscillator

The experimental set-up for the modelocked bounce oscillator is shown in Fig. 1. The oscillator consists of a slab crystal of Nd:YVO₄ with 1.1 at.% doping and dimensions 20 × 5 × 2 mm. The crystal was pumped on its 20 × 2 mm face by a single 40 W diode bar at 808 nm and the face was anti-reflection coated at this wavelength. The diode employed fast axis collimation and the emission was brought to a line focus on the crystal face by a vertical cylindrical lens (VCL_D) of focal length 30 mm. The end faces of the slab crystal were anti-reflection coated at the lasing wavelength of 1064 nm and the cavity was formed by a SESAM designed for operation at 1064 nm, and an output coupling plane mirror coated to give a transmission of 50% at the lasing wavelength. The SESAM used is a commercially available device from BATOP [12] and has a modulation depth of 3%, a recovery time of less than 10 ps, non-saturable losses of less than 0.3% and a saturation fluence of 70 μJ/cm². Two spherical mirrors with radii of curvatures 50 cm and 25 cm were positioned to re-image an intracavity beam waist onto the SESAM, with a reduction in the mode area by a factor of 4. Two vertical cylindrical lenses (VCL₁ and VCL₂) of focal length 50 mm and a horizontal cylindrical lens (HCL) of focal length 100 mm were used inside the cavity to facilitate mode-matching between the gain region and the TEM₀₀ cavity mode.

 

Fig. 1. The experimental set-up for the modelocked Nd:YVO₄ bounce oscillator. The dotted box indicates the position of the back mirror when the cavity is operated in a compact, CW-running form.

Download Full Size | PDF

The spherical mirrors re-image a beam waist in the cavity that occurs at the plane that would contain the back mirror should the bounce geometry be used in a compact, CW running, form. This enables us to both extend the cavity and reduce the mode size on the SESAM, whilst still being able to maintain the TEM₀₀ beam quality attainable from the compact cavity. The cavity extension and increased intensity at the SESAM are both desirable functions for obtaining CW modelocking and help to avoid Q-switched modelocking. Previous work has given rise to the now well known inequality E ² _p ≥ E_SL E_SAΔR , where, E_P is the intracavity pulse energy, E_SL the saturation energy of the laser gain, E_SA the saturation energy of the SESAM, and ΔR the modulation depth of the SESAM [13]. This inequality needs to be satisfied if we are to generate CW, in preference to Q-switched, modelocking. With this in mind, the extension of the cavity tends to increase the value of E_P, and the reduction in the mode size reduces E_SA, thus helping to keep the inequality biased in favour of CW modelocking.

With this cavity design, we obtain a CW modelocked pulse-train with a repetition rate of 78 MHz, and a background free intensity auto-correlation, based on a Michelson interferometer, was used to determine the pulse duration as 30ps. The spectral bandwidth of the laser emission could be determined only to be greater than 10 GHz by use of an available Fabry-Perot etalon with a free spectral range of 10 GHz. Figure. 2 shows the pulse train on two different timescales, one short, demonstrating clean pulses, the other significantly longer, demonstrating CW modelocking with no large scale variations in the pulse amplitude.

 

Fig. 2. The pulse train from the modelocked oscillator. On the left, a short timescale showing artefact-free pulses. On the right, a longer timescale showing no modulation in the pulse amplitude.

Download Full Size | PDF

The output average power for the cavity as a function of diode pump power is shown in Fig. 3. We obtain a maximum output power of 16.7 W with 38 W of optical pumping, representing an overall optical-optical efficiency of 44%. The slope efficiency was 55% and the output beam is TEM₀₀ at all power levels.

 

Fig. 3. Average output power vs input pump power for the modelocked oscillator (red triangles), and the compact CW-running cavity (black squares).

Download Full Size | PDF

To assess how well the oscillator is performing, we performed a comparison of the modelocked cavity with a bounce oscillator configured for CW running in a compact form. In this case, the back section of the laser cavity which contains the SESAM, is replaced by a 100% reflecting mirror placed in the position indicated in Fig. 1, and we compare the power curves obtained from each. The resultant output average power for the compact CW cavity is also shown in Fig. 3. For the CW cavity, we obtain a maximum output of 18.1 W in a TEM₀₀ beam from 38 W of optical pumping. Considering the insertion losses of the SESAM, this demonstrates that the modelocked cavity is operating very efficiently close to ideal. The two cavities operate with very similar slope efficiencies and it is observed that the output beam profiles are also almost identical, further evidence of the integrity of the modelocked cavity. It is noted further that the lasing threshold for both cavities is similar (~3 W optical pump power), and that the threshold for the onset of modelocking is at approximately 7 W of optical pumping.

3. Modelocked master oscillator power amplifier

With this result and configuration we are approaching the damage limits of the SESAM structure and so a further increase in the attainable modelocked power would require a redesign of the cavity. This would then have to be the case at another power level when the damage limit is once again reached. A fixed cavity design is thus not scaleable to an arbitrarily high power and a more flexible approach to achieving high power is to use a Master Oscillator - Power Amplifier (MOPA) arrangement (Fig. 4.) [14,15].

 

Fig. 4. Experimental set-up for modelocked MOPA

Download Full Size | PDF

For this, we use another high gain bounce geometry module as the power amplifier, with the previously described modelocked cavity as the master oscillator. For the amplifier, we use a 25 × 5 × 2 mm, 1.1 at.% doped Nd:YVO₄ crystal and an internal bounce angle of 7°. The crystal is pumped by a 100 W diode bar brought to a line focus by a 25 mm focal length VCL. Again, we use two VCLs with focal lengths 50 mm for mode matching in the vertical. With a master oscillator average power of 10.4 W, and varying the pump power of the amplifier, we obtain the power curve shown in Fig. 5. With maximum amplifier pumping of 104 W, we obtain a maximum modelocked output power of 60 W, which represents an amplifier extraction efficiency of 48%. The beam quality is maintained as TEM₀₀ over the full pumping range of the amplifier which demonstrates the high power scaling potential of the bounce geometry as an amplifier solution. The MOPA power curve shows no sign of saturation and so it should be possible to scale the output power further without degrading the beam, by simply increasing the amplifier pumping. With this further power scaling potential we expect to obtain a modelocked output of greater than 100 W.

 

Fig. 5. MOPA average output power vs amplifier pump power.

Download Full Size | PDF

4. Conclusion

In conclusion, we have demonstrated CW modelocked operation of the bounce amplifier geometry and attained a modelocked oscillator that gives a constant amplitude pulse train with pulses of duration 30 ps. Modelocking can be maintained over a wide range of pump powers and we achieve a maximum output power of 16.7 W with overall optical-optical conversion efficiency of 44% and a slope efficiency of 50%. We then implemented a further bounce power amplifier in a MOPA configuration, and using a 10.4 W master oscillator achieved an amplified modelocked output power of 60 W, which represents an extraction efficiency from the amplifier of 48%. No sign of saturation was visible and we thus anticipate that further power scaling of the modelocked output should be possible to the 100 W level by simply increasing the amplifier pump power.