We have developed a high-contrast, high-beam-quality Ti:sapphire amplifier producing pulses of 10 mJ in a single stage with 19% efficiency. The amplifier has a double-confocal multipass ring configuration that allows for a large mode volume by use of a collimated beam in the gain medium. We have designed the amplifier optics to correct for aberrations and for spatial gain narrowing. The compressed output beam has an M2 of 1.15. The use of an internal saturable absorber in the amplifier results in an intensity contrast of ~109. We anticipate this design will be useful to extend the multipass architecture to low-gain media and to still higher output energy.
© 2004 Optical Society of America
Chirped pulse amplification (CPA) technology for efficient amplification of femtosecond, laser pulses in solid-state materials such as Ti:sapphire is now fairly well-developed . For applications of high peak-power systems, such as x-ray generation , laser-ion-beam acceleration  and hollow fiber frequency conversion [4, 5], high performance in intensity contrast and/or beam quality are especially important. Systems requiring more than a few mJ per pulse typically rely on multistage amplifiers. We report here a compact Ti:sapphire amplifier that produces over 10 mJ with a single stage, with both high contrast and high beam quality. This multipass design is scalable to higher energy and repetition rates.
2. High energy multipass design
The central issue in high-energy amplification is the efficient extraction of the stored energy without damage to the mirrors or crystal. The Ti3+ saturation fluence of ~1 J/cm2 sets a rough scale for the final beam diameter for a desired energy (e.g., ~1.4 mm diameter for 15 mJ output). However, given a desired level of energy extraction efficiency, the actual output fluence is fixed fraction of the absorbed pump fluence, which in turn determines the small signal gain. For low-energy (~1 mJ) Ti:sapphire amplifiers, the small beam waist in the crystal diverges to give a reduced fluence on the mirrors. However, as the energy and the beam size increase, a collimated beam must be used to keep the amplifier within reasonable dimensions. In the design presented here, a collimated beam can take many passes through the gain: this permits high extraction at a reduced level of gain that is compatible with the mirror damage threshold. With the use of a collimated beam, increasing the beam size (at constant small-signal gain) will increase the output energy.
Figure 1 shows the optical layout of our double-confocal design. The optics shown in blue represent the basic layout. A beam entering near CM1 passes through the crystal and is directed by CM2 to mirrors CM3 and CM4. In the basic layout, the mirrors are equally spaced in the confocal position, with equal tilt angles adjusted to increment the incident angle on the crystal on each pass. The confocal placement of the mirrors ensures overlap in the crystal for all passes. Our design is similar to a three-mirror configuration by Backus et al, which produced amplified pulse energies of up to 1–3 mJ . The four-mirror system is optically unfolded, which allows a greater number of passes to be taken without the restriction of the length of the central mirror (at the location of the dashed line in Fig. 1). The unfolded design also gives the option of directing to the crystal either a focused beam (as in the original design) or a collimated beam (which focuses along the diagonal between mirrors CM2 and CM3). In this way, a larger mode size for higher energy extraction can be employed. We note that it is possible to inject a beam with a divergence that places the focus in front of the crystal. However, this produces a smaller beam on the previous curved mirror. Only in the case of very high gain is this advantageous over using a collimated beam.
3. Aberration corrections in the amplifier
Figure 1 also shows optics that add additional round-trips and correct for aberrations. Multiple sets of passes can be taken with this design, multiplexed by vertical offset on the curved mirrors. Note that no additional alignment of the ring is required for the second set. For maximum gain, beams in the first set are focused in the crystal, and collimated for maximum energy extraction in the second set. Lenses L1 and L2 approximate the relay imaging internal to the ring, maintaining the pointing stability. The returning beam enters the ring lower than the seed at M1 and the output emerges above the first set at M2 (see Fig. 1). In between the two sets of 9 passes for the amplifier presented here, the µJ-level beam is focused through a saturable absorber to improve contrast (see below). Spherical lenses L3 and L4 were added to optimize the output beam quality by compensating for aberrations associated with the use of a collimated beam in the crystal and a relatively short mirror focal length. Although the pump light is relayed to the crystal, the pump profile is not completely flat and the high net gain results in substantial spatial gain narrowing. The resulting reduction in the beam size in the crystal causes saturation to occur earlier and at lower energy. This effect is illustrated in Figure 2(a) (thin line), which shows the results of a model of the spatially dependent saturated amplification experienced by an input pulse with a Gaussian profile amplified in a medium pumped with a laser with a super-Gaussian profile I(r)∝exp[-2r6 ]. Without correction, the beam diameter initially shrinks, then after saturation is reached in the center, the mode begins to widen. The expansion of the mode to fill in the pumped area is gradual: during this process, the output energy is limited and the fluence is near the damage threshold of the mirrors. To compensate this effect, we shortened the focal length of one of the mirrors slightly (with spherical lens L1, f=3.5 m, diameter=2”) so that the beam increased in size by 6% per pass (see Fig. 2(a), thick line). The mirror/lens combination was moved in to preserve the crossing point. This arrangement also has the effect of decreasing the beam spacing per pass, though this was found to be manageable for 9 passes.
We found astigmatism and divergence control to be important in our configuration. Geometrically, the automatic beam overlap in the crystal derives from the 1-f spacing from the curved mirrors to the crossing point. The diagonal distance is necessarily slightly longer, resulting in a gradual focusing of the collimated beam from pass to pass. This focusing is much stronger than the natural divergence of the Gaussian beam. We placed a negative correction spherical lens (L2, f=-3 m, diameter=0.5”) at the second crossing point to correct for this. The mirror focus was optimized for horizontal beam overlap in the crystal by imaging the crossing point to a CCD camera. (The sensitivity of the crossing to the focal position was ~1 mm.) The astigmatism resulting from the longer vertical focal length of the tilted mirrors not only changed the beam shape but also affected the vertical crossing of the second pass. A tilt around the horizontal of lens L1 corrected for astigmatism, yielding a collimated final output beam with a small ellipticity. Figure 2(b) shows the evolution of the horizontal and vertical beam divergence in the amplifier, calculated with the ABCD Gaussian beam propagation technique. The two pairs of lines show that lenses L1 and L2 correct for both the divergence and astigmatism of the beam. We note that these correction lenses make possible the compact configuration of the amplifier and allow for tight pump focusing for low energy systems. A second amplifier with f=500 mm is currently under development - the reduction in mirror tilt angle eliminates the requirement for astigmatism and divergence correction.
4. Amplifier intensity contrast
Regenerative amplifiers have some disadvantages in maintaining high contrast: the intracavity polarizer provides a surface for prepulses to be reflected in the output direction, and the cavity allows the buildup and temporal concentration of ASE preceding the pulse. The multipass architecture circumvents these limitations and is more favorable for high contrast. Pockels cells are frequently used, but the finite risetime allows a window for ASE near the pulse. In the 3-mirror design, mask with a series of holes is used to counteract both ASE buildup and spatial gain narrowing . In addition to the energy loss, this technique introduces a hard edge on the beam that decreases the amplified beam quality. In our system, the pulse is amplified to the microjoule level in 9 passes, extracted from the ring, then focused through a saturable absorber  (2-mm Schott RG850 color filter, see Fig. 1). The filter passes 75% of the seed pulse energy without any noticeable spectral change. The contrast is improved by the ratio of the linear to nonlinear transmission: ~102. Maintaining (not reversing) the direction of the second set further reduces the ASE in the output direction.
5. Amplifier performance and characterization
Seed pulses (1-nJ, 15-fs) from a Kerr-lens mode-locked Ti:sapphire oscillator (KMLabs)  pumped with 3 W from a Spectra Physics Millennia laser were stretched to ~150 ps by an all-reflective pulse stretcher [9, 10] with a 1200 gr/mm grating. The pulses passed through a Pockels cell pulse selector before being injected into the amplifier. The amplifier crystal was pumped from both sides at 10 Hz by a frequency doubled Nd:YAG laser (Continuum Model 681C-10), which provided up to 100 mJ per pulse at 10 Hz. In our pumping arrangement, the pump beam was focused through the dielectric mirrors CM1 and CM2 to a point in front of the crystal, giving a pump spot size of ~1.3 mm, relay imaged from inside the pump laser. The Brewster crystal was doped to absorb 95% of the incident pump light. The round trip passive losses were approximately 5%.
The seed beam was focused in the crystal for the first set of 9 passes, then collimated to ~600 µm for the second set. Without gain the round trip magnification expanded the beam to 1.4 mm by the final pass. With 46 mJ of pump energy absorbed by the crystal, the measured peak single pass gain in the Brewster crystal was 3.8 and the output energy from the amplifier was typically 10 mJ, corresponding to a conversion efficiency of 19%. Figure 3(a) shows a lineouts of an images of the crystal fluorescence at 680nm with and without pump depletion. The output energy is currently limited by our relay design for the pump: the tight focus in front of the crystal results in air breakdown at higher pump energy. Redesigning this relay system to enlarge the pump spot should increase output energy. The CCD camera monitoring the beams in the crystal serves as monitor for the day-to-day alignment of the seed direction. The ring mirrors rarely need adjustment.
With the output pulse energy at 8.6 mJ the seed pulse was blocked and the output energy decreased to 80 µJ. The majority of this ASE emerges after the output pulse. Figure 3(b) (solid line) shows the time dependence of the ASE without a seed but with the saturable absorber present, measured with a high-speed photodiode. The integral of this curve was normalized to the measured ASE energy. The amount energy preceding the output pulse was then estimated by integration of the diode signal up to the output pulse arrival time. The majority of the ASE was within 25-ns of the main pulse, containing ~0.8 µJ, yielding an intensity contrast of ~109. The dashed line in Fig 3b shows the trace without the saturable absorber. In this case the total ASE energy was 3.5 mJ and the contrast was reduced to 107.
Using a zero-order waveplate was inserted to pass the beam through the pulse selector, we characterized the phase compensation with the unamplified oscillator beam. Figure 4(a) shows the second-harmonic frequency-resolved optical gating spectrogram. With 55 nm of unamplified bandwidth passing through the system, narrowed by the dielectric mirrors (Fig. 4(b), dashed) adjustment of the grating compressor gives fourth-order-limited phase (Fig. 4(b), red) and a compressed pulsewidth of 26 fs (not shown). The amplified spectrum of 38 nm FWHM (Fig. 4(b), solid black) yielded an amplified pulse duration of 35 fs (Fig. 4(c). Preshaping the input spectrum or adding an intracavity spectral filter  will allow the amplified bandwidth to be increased. The additional material in the ring (now 18 mm of BK7) can be reduced to 6 mm with a thinner lens L1 and shaping the crystal to eliminate L2, or to zero by eliminating L1 and using a longer cavity and a shorter-radius CM4. Our calculations indicate that the addition of a prism pair  to allow compensation of fourth-order phase will allow compression down to 22 fs.
After focusing the compressed output beam with an f~1250 mm concave mirror we recorded images of the spot with a CCD camera at several longitudinal positions. By fitting the measured beam diameter to the form expected for an aberrated beam (Fig. 5), we measured an M2=1.15, with an ellipticity ratio of 1.1. Both the horizontal and vertical foci were in the same location indicating that the astigmatism is well corrected. An image of the compressed amplified beam with lineouts is shown in fig 6. We attribute the high quality of the amplified beam to the lack of any aperture in the system and to the aberration corrections in the ring.
We have developed a compact, single stage laser amplifier that generates pulses of 10-mJ energy at 10 Hz (compressed to 7 mJ) with good contrast, beam quality, and stability. We anticipate that still higher energy can be obtained from this amplifier by changing the pump image relay to give a larger pump volume. At somewhat higher energy levels self-phase modulation and air breakdown at the internal focus will become important. This aberration-compensated short-cavity design is applicable to low pulse energy, high-repetition rate amplifiers that require tight pump focusing. A 20-pass, 1-mJ kilohertz amplifier being developed uses f=500 mm mirrors and has less need for aberration correction. Finally, the large number of passes possible in this design should make it useful for amplification with higher saturation fluence gain media.
We would like to acknowledge support from the NSF, equipment support from M. Murnane and H. Kapteyn and helpful discussions with J. Squier.
References and Links
1. S. Backus, C. Durfee, M. M. Murnane, and H. C. Kapteyn, “High power ultrafast lasers,” Rev. Sci. Instrum. 69, 1207–1223 (1998). [CrossRef]
2. A. Rousse, C. Rishel, and J.-C. Gauthier, “Colloquium: Femtosecond x-ray crystallography,” Reviews of Modern Physics 73, 17–31 (2001). [CrossRef]
3. M. Roth, T. E. Cowan, C. Brown, M. Christl, W. Fountain, S. Hatchett, J. Johnson, M. H. Key, D. M. Pennington, M. D. Perry, T. W. Phillips, and T. C. Sangster, “Intense ion beams accelerated by petawatt-class lasers,” Nuclear Instruments & Methods in Physics Research A 464, 201–205 (2000). [CrossRef]
4. C. G. Durfee, S. Backus, M. M. Murnane, and H. C. Kapteyn, in Applications of High Field and Short wavelength Sources, ed. L. DiMauro, M. M. Murnane, and A. l’Huillier (Plenum, New York, 1998), p. 71–78.
6. S. Backus, J. Peatross, C. P. Huang, H. C. Kapteyn, and M. M. Murnane, “Ti:Sapphire Amplifier Producing Millijoule-Level, 21 fs Pulses at 1 kHz,” Opt. Lett. 20, 2000–2002 (1995). [CrossRef] [PubMed]
11. C. Barty, G. Korn, F. Raksi, C. Rose-Petruck, J. Squier, A. Tian, K. Wilson, V. Yakovlev, and K. Yamakawa, “Regenerative pulse shaping and amplification of ultrabroadband optical pulses,” Opt. Lett. 21, 219–221 (1996). [CrossRef] [PubMed]
12. C. G. Durfee, S. Backus, M. M. Murnane, and H. C. Kapteyn, “Design and implementation of a TW-class kilohertz amplifier system,” IEEE J. Sel. Top. Quantum Electron. 4, 395–406 (1998). [CrossRef]