Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
6. Pulse Amplification. 6.1. Introduction. Output pulse energies from femtosecond lasers typically do not exceed a few nanojouls, and peak powers of megawatts. For many applications, higher energies or peak powers are required. Researchers are seeking methods to shorten pulses, to increase peak powers and peak intensities on targets. Given the limits on further trimming of the pulse duration, further increases in peak power and peak intensities can only be obtained by increasing output energy. Amplification of energy from the femtosecond lasers makes possible terawatt peak powers. Amplification, combined together with the initial pulse stretching and compression as a final stage can convert terawatt systems into petawatt lasers with subpicosecond pulses. So far the highest energies, peak powers and irradiance can be achieved in Nd:glass amplifiers, not those based on Ti:sapphire. The most powerful laser in the world (in 2003) is “Vulcan” in Rutherford Appleton Laboratory, United Kingdom delivering 2.5 kJ in two 150 nm beams, 1 pW, 1021 W/cm2 and Nova system at Lawrence Livermore National Laboratory delivering 1.3 kJ pulse at 800 ps that can be compressed to 430 fs to achieve 1.3 pW and 1021W/cm2. When a laser pulse passes through an optically active material in which the population inversion is maintained by a pumping source, it gains energy from the stimulated emission generated by itself in the medium. As a result the output pulse is amplified (fig. 6.1). pumping input pulse active medium Fig. 6.1 Illustration of amplification 181 amplified pulse 6.2. Theoretical Background. Let us consider a mechanism of amplification in the three-level system (Fig. 6.2) 3 τ 32 Wp τ 21 2 (m) laser transition 1 (n) Fig. 6.2 Three-level system. Let assume that the relaxation time τ32 for the 3 → 2 transition is short in comparison with the lifetime of the E2 state τ21, which is a good approximation in solid-state lasers. This denotes that the number of molecules N3 occupying the level E3 is negligible compared with the number of molecules in N1 and N2 (N3 ≈ 0, and N1 + N2 = N0) because the molecules are pumped almost immediately into the metastable level E2 with only a momentary stay in E3. Based on this assumption, the change in the population is N dN 2 = W ( N1 − N 2 ) − 2 + W p N1 , τ 21 dt (6.1) dN 1 dN 2 =− dt dt (6.2) W = B21 ρ . (6.3) where Neglecting spontaneous emission and pumping Wp during the pulse duration, which is usually much shorter, the eq. (6.1) can be written in the form dn = −γWn , dt (6.4) n = N1 − N 2 (6.5) where 182 denotes the population inversion, γ = 1 + g2 (for three-level system, where g2, g1 are the g1 degeneration numbers). One can show 6.1 that W = σI hν (6.6) J ,ν where σ [cm2] is the stimulated emission cross section, I is the radiation intensity 2 s cm is the frequency of the 2 → 1 transition. Substituting (6.6) into (6.4) and expressing the intensity in terms of the photon density φ [photons/cm3] I =φ, (6.7) dn = −γ cσ nφ . dt (6.8) hν c one gets The rate at which the photon density changes in a small volume at x of an active medium is equal to ∂φ ∂φ = Wn − c ∂t ∂x (6.9) where the first term describes the number of photons generated by the stimulated emission and the second term – the flux of photons which flows out from that region. Using once again (6.6) and (6.7) in (6.9) one gets ∂φ ∂φ = cnσφ − c. ∂t ∂x (6.10) Using the differential equations (6.8) and (6.10) one can solve them for various types of the input pulse shapes 6.2-6.3. For a square pulse of duration tp and the initial photon density φ0 one obtains: −1 φ ( x, t ) x = 1 − [1 − exp(− σ n x )]exp − γσφ0 c (t − . φ0 c (6.11) After passing through an active medium of length x = L the amplification is given by ∞ G= ∫ φ (L,t ) dt −∞ φ0 t p . (6.12) After substituting (6.11) into (6.12) and integrating one gets G= 1 c γ σ φ0 t p { } ln 1 + [exp (γ σ φ 0 τ 0 c ) − 1] e n σ L . 183 (6.13) The equation (6.13) can be cast in a different form using the input energy Ein Ein = c φ0 t p hν , (6.14) and the saturation fluence that we defined in chapter 1 (eq. 1.39) Es = hν γσ = E st , γ g0 (6.15) where Est = hνn is the energy stored per volume, g0 = nσ is the small signal gain coefficient (see chapter 1, eq. 1.23). Using the expressions for Ein and Es in (6.13) one obtains the energy gain G G= E s Ein ln 1 + exp Ein E s − 1G0 , (6.16) where G0 = exp(g0L) is the small-signal single-pass gain. Consider limit cases: 1. The input signal is low, Ein E s << 1 . The eq. (6.16) can be approximated to the exponential dependence on the active medium length L G ≅ G0 = exp( g 0 L ) . (6.17) 2. The high input signal, Ein E s >> 1 . Then, eq. (6.16) becomes G ≅ 1+ Es g0 L Ein (6.18) with the linear dependence of the energy gain. 6.3. Design Features of Amplifiers. For many scientific and practical applications we need powerful pulses 6.4-6.6 . The pulses emitted from the oscillator can be amplified in devices called amplifiers. No single amplifier configuration has emerged because one configuration can be more suitable than other for a given application. Usually, the gain G which can be achieved and the energy which can be extracted from the amplifier is the main parameter in the design of amplifiers. There are four most important configurations for laser light amplification: 1. multistage power amplifier (fig. 6.3), 2. multipass amplifier (MPA) (fig. 6.4), 3. regenerative amplifier (RGA) (fig. 6.5), 4. chirped pulse amplifier (CPA) (fig. 6.6). 184 oscillator rear mirror output mirror active medium laser amplifier gain medium output pumping output pumping pump trigger pump trigger firing signal delay Fig. 6.3 Oscillator-amplifier configuration The multistage power amplifier consists of a laser oscillator and a pumped active medium called amplifier. The amplifier is driven by the oscillator, which generates an initial pulse of moderate power and energy. The pulse from the oscillator passes the active medium of the amplifier and its power grows, in extreme cases, up to 100 times. The pulse can be amplified further by adding the next amplifier in the configuration presented in fig. 6.3. These configurations are used for longer pulses. The double-pass amplifiers are used for typical picosecond system and can deliver 125 mJ pulses of less than 60 fs at 20 Hz with a solid-state saturable-absorber-based oscillator. pump pulse ga me in diu m se pul t u inp output pu ls e Fig. 6.4 Multipass amplifier Multipass amplifiers (fig. 6.4) are used when extremely short pulses are required, shorter than 35 fs at 1 kHz and energy of 1.5 mJ. This configuration is simple and less 185 sensitive to thermal lensing, therefore provides excellent beam quality, and is relatively easy u outp t pul inpu t pul se to calculate and compensate dispersion in the system. se λ/4 PC1 A M1 Pockels cell gain medium Ti:sapphire PC2 P thin layer polarizer Pockels cell M2 pump pulse Second Harmonic Generation (Nd : YLF, Q-switch) Fig. 6.5 Regenerative amplifier Regenerative amplifiers (fig. 6.5) are generally employed for ultrafast system with relatively low output energies from femtosecond oscillators (a few nJ) to achieve the output energy from the amplifier of around 1 mJ at 1 kHz. Higher energy can be obtained from combination RGA/MPA methods that can deliver energy greater than 3.5 mJ per pulse and higher at repetition rates of 1 kHz. Regenerative amplifiers are also used to produce powerful picosecond pulses from a train of low-energy pulses of modelocked lasers. The regenerative amplifiers have originally been pumped by flash lamps or by cw arc lamps 6.7-6.12 . More recently, the Nd:YAG, Nd:YLF or Nd:glass amplifiers are pumped by diode arrays 6.13-6.17. 186 grating grating Stretcher seed pulse from oscillator Compresor grating grating outpu λ/4 PC1 A Pockels cell M1 t puls e PC2 gain medium Ti:sapphire thin layer polarizer Pockels cell M2 pump pulse Fig. 6.6 Chirped pulse amplifier Amplification of ultrashort pulses leads to enormously large peak powers that may be far above the damage threshold of the active medium employed in the amplifier. To avoid this problem the chirped-pulse amplification (fig. 6.6) was developed. 6.4. Regenerative Amplifier. Now we will explain how the regenerative amplifier operates. One can see from fig. 6.5 that the typical commercially available regenerative amplifier consists of an active medium (Ti:sapphire crystal), two Pockels cells (PC1 and PC2), λ/4 plate and thin-film polarizer placed between two mirrors M1 and M2. The gain medium is pumped with the second harmonic of the solid-state laser Nd:YLF, Q-switched with pulses durations of 250 ns and the average energy of 8 W at the Brewster’s angle. The regenerative amplifier selects an individual pulse from a train of modelocked pulses (called a seed pulse). The seed pulse is trapped in an amplifier where it is amplified at each pass in the crystal. The pulse passes many times (usually about 10-20 times) through the gain medium to get more energy. Once its energy increases by as much as 106, the amplified pulse is removed from the amplifier as the output pulse that can be used for further applications. 187 The number of passages in the regenerative amplifier depends on the round-trip time between the M1 and M2 mirrors in the amplifier. If the round-trip time is 10 ns and the pumping pulse duration is 250 ns, a typical time evolution of the pulse energy in a regenerative amplifier is given in fig. 6.7. output energy 1 100 200 300 t [ns] 10 20 30 number of steps Fig. 6.7 Time evolution of the amplified pulse energy versus time and the number of passages in the regenerative amplifier The amplified pulse shape follows the shape of the pumping pulse originating from the depletion of the excited-state in the gain medium. It is obvious from fig. 6.7 that the pulse should be switched out from the amplifier after around 20 passages where there is a gain maximum and the energy of the amplified pulse is the highest. Now we have to understand how the input pulse is trapped in the amplifier. We will show that it is trapped on the polarization basis. Let us recall that: 1. λ/4 plate changes the polarization from linear to circular, 2. λ/2 plate rotates the linear polarization by 2α, where α is the angle between the polarization vector and the optic axis, 3. thin layer polarizer acts as a reflective polarizer (see chapter 1) that at the Brewster’s angle reflects the polarization perpendicular to the plane of reflection, and transmits the polarization parallel and perpendicular to the plane of reflection, 4. Pockels cell acts like a λ/4 or λ/2 plate depending on the external voltage applied.. 188 6.4.1. Pockels Cell. The Pockels cell plays an important role as an electro-optic Q-switch 6.18. Now we will explain its operation. The Pockels cell consists of a nonlinear optical material with the voltage applied to the material. The electric field can be applied along the direction of the optical beam (fig. 6.8a) or perpendicular to it (fig. 6.8b). a) b) z nonlinear material (KD*P) y laser beam x nonlinear material z optic axis E V E V laser beam Fig. 6.8 The Pockels cell, the electric field can be applied along the direction of the optical beam (a) or perpendicular to it (b) The crystal becomes birefringent under the influence of the applied electric field. The crystals used for parallel configuration (fig. 6.8a) are uniaxial in the absence of an electric field with the optic axis along z direction. The ellipsoid of the refraction index projects as a circle on a plane perpendicular to the optic axis (fig. 6.9a). 189 a) b) y’ y y z optic axis x z optic axis x x’ Fig. 6.9 Change of the refractive index in a KD*P crystal, x, y, z – the crystalographic axes a) without an electric field, b) when an electric field is applied (E ≠ 0), x’, y’, z – the electrically induced axes So, the laser beams having polarizations along x or y directions propagate with the same velocities along the z axis because the crystal is not birefringent in the direction of the optic axis. For the situation presented in fig. 6.8a, the laser beam linearly polarized along y passes through the KD*P crystal unchanged when no voltage is applied. When an electric field is applied parallel to the crystal optic axis z, the ellipsoid of the refraction index projects as an ellipse, not a circle, on the plane perpendicular to the optic axis with the axes x’ and y’ rotating by 45° with respect to the x and y crystallographic axes (fig. 6.9b). The angle of 45° is independent of the magnitude of the electric field. Therefore, when a voltage is applied, the KD*P crystal becomes birefringent along the z axis, and divides the laser beam into two components (along x’ and y’) that travel through the crystal at different velocities. The polarization of the output beam dependents on the phase difference between the two ortogonally polarized components – ordinary and extraordinary rays, which depends on the applied voltage δ= 2π λ l∆n , (6.19) where ∆n is the difference in the indices of refraction for the ordinary and extraordinary beams, l is the crystal length. It has been shown 6.19 that ∆n can be expressed as ∆n = n03 r63 E z , (6.20) where r63 is the element of the electro-optic tensor of the third rank that is the response to the applied field E in the z direction (Ez), n0 is the index of refraction for the ordinary ray. 190 Employing the relation between the voltage V and the applied electric field Ez, V z = E z l and inserting (6.20) in eq. (6.19) one gets δ= 2π λ n03 r63V z . (6.21) When the applied voltage Vz is adjusted to generate the phase difference δ = π/4 or δ = π/2 the Pockels cell operates as a quarter-wave or half-wave plate. The Pockels cell belongs to the fastest all-optical switching devices, and is highly reliable. Typical commercially available Pockels cells employ KD*P crystals with λ/4 voltage between 3.5 and 4 kV at 0.69 µm and 5 to 6 kV at 1.06 µm 6.1 . As a particular example, Pockels cells are used to select and retain high-peak-power pulses in regenerative amplifiers. Knowing how the Pockels cell operates let us return to fig. 6.5 to explain the mechanisms of trapping the seed pulse from a train of modelocked pulses inside the amplifier. The operation of a regenerative amplifier presented in fig. 6.5 can be divided into the following steps: 1. A seed pulse arrives in the amplifier at the Brewster’s angle by reflecting from the surface of the active medium (Ti+3:Al2O3). The polarization of the seed pulse is perpendicular to the plane of drawing. The polarization of the pulse passing through the Pockels cell PC1, which is initially switched off is unchanged, but double passing through λ/4 (before and after reflection from M1) changes the polarization by 90°. So, the beam reflected from M1 is not reflected from the crystal surface at A point but is refracted and passes the crystal leading to its amplification by the stimulated emission induced in the crystal (Ti+3:Al2O3) in which the population inversion is generated by the pumping source. The pulse passes through the thin-layer polarizer and the Pockels cell (PC2), inactive at the moment. The pulse is reflected from the mirror M2. 2. When PC1 (and PC2) is still switched off, the beam passes twice λ/4 plate and the pulse is removed out of the amplifier by reflection from the crystal surface at the point A. 3. When PC1 is switched on (as a λ/4) before arriving the seed pulse, the total effect of the λ/4 plate and PC1 is λ/4 + λ/4 + λ/4 + λ/4 = λ, which means that the polarization is unaffected and the pulse is trapped inside, not reflected from the surface of the active medium. 4. Each pass of the pulse through the active medium results in the pulse amplification. When the pulse has been amplified to the desired level (~106 times), the voltage λ/4 is applied to the Pockels cell PC2. The pulse going to the mirror M2 and returning through 191 PC2 (λ/4 + λ/4 = λ/2) changes its polarization and is removed from the amplifier at the P point by reflection from the thin layer polarizer. 6.5. Chirped Pulse Amplification (CPA). The configuration of the typical chirped pulse amplifier is presented in fig. 6.6. One can see that it consists of the regenerative amplifier described in chapter 6.4 and the stretcher and the compressor. The stretcher placed before the regenerative amplifier expands the pulse duration by many orders of magnitude (from femtosecond to hundreds picosecond) and thereby reduces peak power. Employing the stretcher solves the problem of high intensities in the amplifiers for ultrashort pulses that are above the damage threshold of the active medium of the amplifier. For example, Ti:sapphire crystal has a high saturation fluence and a large gain-bandwidth needed to generate relatively high energies for sub-picosecond pulses. Self-focusing in the crystal described in chapter 3 is a desirable effect when employed in modelocking, but it has also its dark sides. Beam focusing due to the Kerr lens focusing can lead to catastrophic beam collapse that may destroy the crystal, which makes it necessary to limit the intensity present in amplifiers to reasonable magnitude less than 10 GW/cm2. This obstacle can be removed by the technique called chirped pulse amplification (CPA). We have discussed in chapter 5 the factors that lead to stretching the temporal pulse duration. The group velocity dispersion (GVD) is the most important factor affecting the temporal pulse broadening. Due to the GVD each frequency component that comprises the spectrum of the pulse travels through the medium with different group velocity. For so called positive GVD materials, the red components travel faster than the blue components. As we have shown a pair of prisms (fig. 5.28) can be used to compensate for the excess GVD that results in the pulse compression. The same can be achieved with the diffraction gratings 6.20. 192 grating 1 re d e u bl input pulse mirror M grating 2 output pulse (compressed) Fig. 6.10 Illustration of the principle of the pulse compression The pair of parallel gratings shows a simplified pulse compressor, which demonstrates the concept of compression (fig. 6.10). One can see that the optical path through the grating pair is longer for the longer wavelengths than for the shorter ones. The mirror M reflects the beam back into the grating pair recovering the spatial distribution, but still increasing the difference in the optical paths for the red and the blue components. Thus, the pair of the parallel gratings provides the negative group-velocity. If the input pulse that is positively chirped travels through the gratings, the output pulse becomes shorter due to partial cancellation of the positive GVD effect by the negative GVD of the gratings in the configuration presented in Fig. 6.10. Pulse stretching is essentially the reverse of the pulse compression. The gratings can be configured in such a way so the bluer components have to travel longer path through the stretcher than the redder components. The result is that the stretcher generates a positive GVD effect that results in that the redder components travel even faster than for the input pulse and exit the stretcher with the temporal pulse significantly longer. One of many possible configurations of the pair of gratings that generates the positive GVD is presented in fig. 6.11. One can see that the main difference between the compressor (fig. 6.10) and the stretcher (fig. 6.11) is that a telescope is added between the gratings to invert the sign of the dispersion from the negative in fig. 6.10 to the positive in fig. 6.11. 193 1 mirr or M red bl u output pulse grati ng grating input pulse e lens reverse GVD 2 Fig. 6.11 Illustration of the principle of pulse stretching In practical applications some other configurations are used. For example, instead of using a pair of gratings, a single grating combined with the curved mirror can be used as presented in fig. 6.12. One can see that this multi-pass configuration reduces the complexity and ensures four-pass the grating that is necessary for the spatial reconstruction of the stretched beam. a gr 3 1 tin 2 g input pulse 4 output stretched pulse curved mirror Fig. 6.12 The pulse stretched with a single grating – curved mirror combination The recent decade have witnessed the dramatic improvement in amplification of ultrashort pulses. Amplification of the femtosecond pulses using the chirped pulse amplification (CPA) technique 6.21 is now commercially available. Amplification of ultrashort pulses to mJ, corresponding to multiterawatt peak powers becomes a routine task. CPA technique is used both in the multipass configuration 6.25 6.22 and regenerative amplification 6.23- . Quite often these configurations are combined together to reach terawatt powers. Usually the regenerative amplifier is used as a first stage of amplification followed by multi-pass configuration used for the output power-boost stage. 194 Generally, the multipass amplifiers are used for very short pulses (<50 fs). They need much lower amount of optical material for amplification that reduce nonlinear distortion of the pulse that usually comes at the expense of rather complicated alignment procedure and higher level of ASE (amplified spontaneous emission). The regenerative amplifiers are usually used for pulses longer then 50 fs. They have many advantages including much smaller ASE and a simple operation and maintenance. No realignment is necessary to adjust the number of passages through the amplifier since the beam follows the some path in the cavity. So far we discussed the conventional configurations based on bulk-optic-solid-state amplifiers. The development of optical communication created a growing market for higherpower optical amplifiers based entirely on the fiber technology. The development in doublecladding fibers technology and multi-mode diode pumping has increased cw fiber lasers output to the level comparable with the solid-state lasers. However, achieving a few watts of average power from fiber-based ultrashort laser systems, routinely produced by the solid-state femtosecond systems, is not so straightforward. Limitations come from large effective nonlinearities (SPM and GVD) described in chapter 5 that can destroy short femtosecond pulses produced in the fiber core. The obstacles related to pulse amplification can be reduced by using fibers with larger core. The typical core of a single-mode fibers are less than 10 µm, larger cores operate usually in multimode regime. However, it has been found that it is possible to obtain a single mode operation for a careful large-core fiber design 6.26. Increasing an active area of the core enables an efficient pumping at shorter length of the fiber reducing nonlinearities. Moreover, fibers exhibiting a positive GVD rather than negative GVD are used for high-power amplifications. The soliton-like behavior (described in chapter 5) obtained when the SPM and negative GVD offset each other at the properly chosen fiber lengths does not apply at high powers. The typical optical amplifier is presented in fig. 6.13 6.27. 195 pump II Yb-fiber pulse source pump I Yb-fiber amplifier chirp-control fiber (GVD>0) compressor Fig. 6.13 Fiber laser amplification system The system consists of a fiber-based ultrashort laser (Yb-doped fiber laser) that emits a seed pulse for further amplification. The seed pulse (1050 nm, 2 ps, 300 mW, 50 MHz) passes through a piece of a single-mode fiber exhibiting a positive GVD to stretch the seed pulse. Then, the stretched linearly positively chirped seed pulse is amplified in a fiber amplifier. The fiber amplifier (Yb-doped fiber, 4.3 m length, 25 µm-diameter core) is pumped from both ends with two fiber-bundle-coupled laser-diode operating at 976 nm and 14 W. The seed pulse is amplified to 13 W and the width increases from 2 ps to 5 ps. Then, the pulse is recompressed down to 100 fs passing through a conventional diffraction grating compressor (negative GVD designed) achieving 5 W output at 1050 nm. 6.1 W. Koechner, Solid-State Laser Engineering, 5th Edition, vol. 1, Springer-Verlag (1999) L.M. Frantz, J.S. Nodvik, J. Appl. Phys. 34 (1963) 2346 6.3 R. Bellman, G. Birnbaum, W.G. Wagner, J. Appl. Phys. 34 (1963) 780 6.4 M.V. Ortiz, Amplified Ultrafast lasers move out of the laboratory, Laser Focus World 39(7) (July 2003) 69-74 6.5 Andrew T. Jay, Jason S. Robins, Interventional Neuroradiology, Stroke and Carotid Artery Stenting, First Union Securities, Inc. and Estimates Company Reports (Aug. 10, 2000) 6.6 S.K. Gayen, M. Alrubaiee, H.E. Savage, S.P. Schantz, R.R. Alfano, Parotid Gland Tissues investigated by Picosecond Time-Gated and Spectroscopic Imaging Techniques, J. Select. Topics Quantum Electron 7(6) (Nov./Dec. 2001) 906 6.7 W.H. Lowdermilk, J.E. Murray, J. Appl. Phys. 51 (1980) 2436 6.8 J.E. Murray, W.H. Lowdermilk, J. Appl. Phys. 51 (1980) 3548 6.9 I.N. Duling, P. Bado, S. Williamson, G. Mourou, T. Baer, CLEO’84 (Anaheim, CA) paper PD3 6.2 196 6.10 I.N. Duling III, T. Norris, T. Sizer II, P. Bado, G.A. Mourou, J. Opt. Soc. Am. B 2 (1985) 616 P. Bado, M. Bouvier, J. Scott, Opt. Lett. 12 (1987) 319 6.12 M. Saeed, D. Kim, L.F. DiMauro, Appl. Opt. 29 (1990) 1752 6.13 M.D. Selker, R.S. Afzal, J.L. Dallas, A.W. Yu, Opt. Lett. 19 (1994) 551 6.14 M. Gifford, K.J. Weingarten, Opt. Lett. 17 (1992) 1788 6.15 L. Turi, T. Juhasz, Opt. Lett. 20 (1995) 154 6.16 D.R. Walker, C.J. Flood, H.M. Van Driel, U.J. Greiner, H.H. Klingenberg, CLEO/Europe’94 (Amsterdam) paper CThM1 6.17 T.E. Dimmick, Opt. Lett. 15 (1990) 177 6.18 S.J. Matthews, Modulation in all things, Laser Focus World 39(8) (August, 2003) 110-115 6.19 R.O’B. Carpenter, J. Opt. Soc. Am. 40 (1950) 225 6.20 E. Treacy, IEEE J. QE-5 (1969) 454 6.21 D. Strickland, G. Mourou, Compression of amplified chirped optical pulses, Optics Communications 56(3) (1985) 219-221 6.22 Sterling Backus, Charles G. Durfee, Gerard Mourou, Henry C. Kapteyn, Margaret M. Murnane, 0.2-TW laser system at 1kHz, Optics Letters 22(16) (1997) 1256-1258 6.23 M.D. Perry, D. Strickland, T. Ditmire, F.G. Patterson, Cr:LiSrAlF6 regenerative amplifier, Optics Letters 17(8) (1992) 604-606 6.24 F. Salin, J. Squier, G. Mourou, Multikilohertz Ti:Al2O3 amplifier for high-power femtosecond pulses, Optics Letters 16(24) (1991) 1964-1966 6.25 C.P. Barty, C.L. Gordon, B.E. Lemoff, Multiterawatt 30-fs Ti-sapphire laser, Optics Letters 18 (1994) 14421444 6.26 M.E. Ferman, Opt. Lett. 23(1) (1998) 52 6.27 G. Sucha, H. Endert, Femtosecond fiber lasers hit power highs, Laser Focus World 36(8) (August, 2000) 133-136 6.11 197