Adaptive Pulse Compression of Femtosecond Laser Pulses Using a Low-Loss Pulse Shaper

Transcription

1 Japanese Journal of Applied Physics Vol. 3, No. A, 2, pp #2 The Japan Society of Applied Physics Adaptive Pulse Compression of Femtosecond Laser Pulses Using a Low-Loss Pulse Shaper Kyung-Han HONG ;2 and Chang Hee NAM y Coherent X-ray Research Center and Department of Physics, Korea Advanced Institute of Science and Technology, 373- Guseong-dong, Yuseong-gu, Daejeon 35-7, Korea 2 Advanced Photonics Research Institute, Gwang-Ju Institute of Science and Technology, Orong-dong, Puk-gu, Gwangju 5-72, Korea (Received March 3, 2; accepted April 2, 2; published August, 2) A simple low-loss pulse shaper, employing a micromachined deformable mirror and a Brewster-cut prism, is demonstrated for the adaptive compression of femtosecond laser pulses. Using this pulse shaper, we compress a positively chirped fs pulse down to 23 fs and compensate for the high-order phase distortion of a sub--fs laser pulse. The transmission efficiency of the pulse shaper reaches 93% with dielectric mirrors and 5% with gold-coated mirrors, designed for broad spectrum application. [DOI:.3/JJAP.3.529] KEYWORDS: adaptive pulse compression, micromachined deformable mirror, genetic algorithm. Introduction The adaptive control of the temporal phase of femtosecond laser pulses makes it possible to compress pulses with minimal phase distortion or to generate arbitrarily reshaped femtosecond pulses. Adaptive control techniques for femtosecond pulse shaping have rapidly advanced with the development of active phase control devices and optimization algorithms. Yelin et al. ) demonstrated the adaptive compression of fs pulses down to fs pulses using a spatial light modulator (SLM) in combination with a grating pair and Tournois 2) reported the active compression of 5 fs pulses down to fs pulses using an acoustooptic programmable dispersive filter (AOPDF). On the other hand, Zeek et al. used a micromachined deformable mirror (MMDM) for femtosecond pulse compression. 3) More recently, AOPDF and MMDM have also been utilized for the optimization of chirped-pulse amplification (CPA) laser systems.,5) As for the optimization of active phase control systems, adaptive learning algorithms, such as genetic algorithms (GAs) and evolutionary strategies (ESs), 6) have been widely exploited. These algorithms are useful not only for the adaptive pulse compression/shaping but also for more general applications such as the coherent control of light-matter interactions. 7 3) Even though MMDM is a very low cost device compared with AOPDF and SLM, it has several advantageous features as an active phase control device. The reflectance of a metalcoated MMDM is more than 95% over the entire spectrum spanning the visible and near-infrared (NIR) wavelengths; this is higher than the diffraction efficiency of AOPDF (5% over nm bandwidth ) ) and the transmittance of SLM (<5% in combination with polarizers in visible and NIR regions 5) ). Owing to their broad bandwidth characteristics, both MMDM and SLM are suitable for the active phase control of sub--fs pulses 6) and the white-light continuum. 7,) MMDM has a higher throughput efficiency and does not induce the phase discontinuity as observed in systems with SLM whereas SLM has a higher dynamic range of phase control. Moreover, the use of MMDM in a pulse shaper based on zero-dispersion -f stretcher configuration 3) results in a more compact setup than that using address: khhong@kjist.ac.kr y address: chnam55@kaist.ac.kr 529 SLM. However, the conventional -f pulse shaper employing MMDM does not have high transmission owing to a limitation in the diffraction efficiency of the grating used. For example, as is reported by Armstrong et al., 6) the transmission efficiency of the pulse shaper is %. Even though specially designed diffraction gratings with a firstorder efficiency of 95% were demonstrated, 9,2) they have not been used in pulse shaping techniques thus far. In contrast, transmission efficiency can be improved relatively easily if a Brewster-cut prism is used instead of a diffraction grating. 7) When using a prism, one has to be careful about the positive dispersion of the prism material and the nonlinear effects induced by high-intensity laser pulses. First, material dispersion can be compensated for using MMDM though it will reduce the dynamic range necessary to compensate for the positive chirp of a laser pulse to be shaped. The precompensation of material dispersion using passive elements such as prism pair and chirped mirrors can help the adaptive pulse shaper to have a balanced dynamic range for the positive and negative directions of the laser chirp. Second, the nonlinear effect can be usually ignored for nj pulses generated from a typical femtosecond laser oscillator because the B-integral value of these pulses is small in the prism material. For example, the B-integral of a 5 fs, nj pulse with a center wavelength of nm and a beam diameter of 2 mm is calculated to be. when the pulse propagates through a 2- mm-thick fused silica whose nonlinear index (n 2 )is3 6 cm 2 /W. In the same configuration, an energy up to mj is acceptable to the medium because the B-integral criterion for avoiding the nonlinear effect is usually regarded as unity. Thus, the prism is suitable for the high-transmission pulse shaper for unamplified femtosecond laser pulses. In this study, a low-loss adaptive pulse compression technique 2) is demonstrated using a pulse shaper composed of MMDM and a Brewster-cut prism. We describe an adaptive pulse compression procedure for a chirped fs pulse in 2 and the compensation of the high-order phase distortion of a sub--fs pulse in 3. The conclusions drawn are presented in. 2. Adaptive Compression of Chirped Femtosecond Pulse The experimental setup of our low-loss adaptive pulse shaper is based on a f -zero-dispersion stretcher as

2 529 Jpn. J. Appl. Phys., Vol. 3, No. A (2) K.-H. HONG and C. H. NAM SHG crystal Spectrometer 3 Flipper SPIDER Feedback via GA 2 Uncompensated pulse from laser oscillator Fig.. Blue Red MMDM Experimental setup of low-loss adaptive pulse compressor. 9 illustrated in Fig.. As the first experiment, we attempted to adaptively compress a positively chirped femtosecond pulse without any precompensation. In the pulse shaper, a laser pulse, generated from a prism-dispersion-controlled (PDC) Ti:sapphire oscillator, was dispersed by a Bewster-cut prism and imaged onto MMDM (OKO Technologies) using a focusing mirror with a magnification of. A Brewster-cut SF prism was employed instead of a diffraction grating for a high transmission efficiency, and a dielectric curved mirror having a large focal length (.25 m) was used to overcome the small angular dispersion of the prism and to exploit the full surface of the MMDM. Since prisms are usually more advantageous at short-wavelength region than at longwavelength region in terms of the angular dispersion, i.e., spectral resolution, this setup can be made more compact in short-wavelength applications. The resultant transmission efficiency of the pulse shaper was 93% without any degradation of the spatial beam quality. MMDM was of the linear type, with a clear aperture of inch; it contained 9 actuators to deform the silver-coated membrane mirror surface with a maximum deflection of mm. 3) Each actuator was controlled by a driver supplying a -bit digitized DC voltage from to 255 V and was interfaced to a Pentium PC. For adaptive pulse compression, the second-harmonic (SH) signal generated in a -mm-thick type-i -BaB 2 O (BBO) crystal by the laser pulse was maximized because shorter pulse duration yields a higher SH peak intensity. Unlike the direct feedback method by means of the temporal characterization of laser pulses, the adaptive feedback method by means of the maximization of the SH signal does not require the calibration of the surface influence function of the MMDM with regards to the voltage applied to each actuator. The SH signal was detected using a spectrometer (PC2, Ocean Optics) and its peak value was fed back to MMDM to actively adjust it for maximizing the SH signal. A simple GA was chosen as an adaptive learning algorithm. The first step in GA is to generate a random population, which consists of a number of individuals and is refreshed in the subsequent evolutionary process. The 2 individuals, which are voltage arrays loaded onto MMDM, defines a population. In the second step, the individuals are selected according to fitness, i.e., the peak intensity of the SH signal, where a better individual is selected more frequently. Then, the individuals are crossovered and mutated at a certain rate to improve population fitness. A new generation begins with the evaluation of fitness fs Fig. 2. Spectral and temporal characteristics of uncompressed chirped -fs pulse. (capturing SH signal) and the process repeats until a termination criterion is fulfilled, i.e., there is no more forward evolution. The temporal characterization of the pulse before and after the compression was performed using spectral-phase interferometry for direct electric-field reconstruction (SPIDER) technique. 22,23) The initial intensity and phase profile of the (uncompressed) pulse is shown in Fig. 2. All the initial voltages on MMDM actuators were set to the same median voltage, i.e., 2 V. Even though this flat voltage array might still impose a small amount of curvature on the MMDM surface, an accurate calibration to set the initial MMDM surface perfectly flat was not necessary for adaptive compression. The positively chirped laser pulse generated in the laser oscillator was further broadened within the pulse shaper mainly by the SF prism material and secondarily by the initial curvature of MMDM. The resultant pulse duration was measured to be fs with a positive group-delay dispersion (GDD) value of fs 2. The instantaneous carrier frequency (!! ) curve in Fig. 2 also indicates the positive chirp. The transform-limited pulse duration determined by the laser spectrum, ranging from 76 nm to 9 nm, was calculated to be 2 fs. The maximum GDD value that MMDM can compensate for in this geometry was estimated to be 27 fs 2, assuming that MMDM ranged from 7 nm to 92 nm. Hence, the GDD of the chirped pulse could be properly compensated for using only this pulse shaper. However, it should be noted again that this pulse shaper is more advantageous in the compression of negatively chirped pulses than in that of positively chirped ones ω-ω (rad/fs)

3 Jpn. J. Appl. Phys., Vol. 3, No. A (2) K.-H. HONG and C. H. NAM 529 Best 3 Fitness/(flat voltage case) 2 Average Generation Fig. 3. Evolution of population vs evolution of generation. Fitness is enhanced with the evolution of generation and then stagnates after the 6th generation. due to the material dispersion of the prism. The estimated path length of an input laser pulse through an SF prism is.7 mm per pass, resulting in a GDD of 626 fs 2 with double pass. Hence, the acceptable range of input laser chirp for this pulse shaper is set between 96 fs 2 and 6 fs 2, considering the compensable GDD limit (27 fs 2 ) of MMDM. In addition, the high-order dispersion of the SF material (TOD of 33 fs 3 and FOD of 9 fs ) will also result in some bias to be compensated for by MMDM itself. In the experiment, the number of individuals was set to 2, the crossover rate to.5 and mutation rate to.. The number of generations needed for convergence was 5 and the computation required about 5 min. The evolution of population fitness with the number of generations is plotted in Fig. 3, which indicates that fitness is enhanced up to.5 times the value generated by loading the initial flat voltage array. Figure shows the electric field of the adaptively compressed pulse measured using the SPIDER technique. The pulse duration was reduced from fs to 23 fs, which is 5% larger than the transform-limited pulse duration of 2 fs. The phase curve shown in Fig. did not get perfectly flattened in the marginal region of the spectrum, but the adaptive pulse compression with a compression ratio of was obtained, indicating a significant compensation of the quadratic phase term, i.e., a GDD of fs Adaptive Compensation of High-Order Phase Distortion of Sub--fs Pulse A similar adaptive control was carried out to compensate for high-order phase distortions, such as third-order dispersion (TOD) and fourth-order dispersion (FOD), in sub- -fs pulses. The optical spectrum from a mirror-dispersioncontrolled (MDC) Ti:sapphire laser is shown in Fig. 5. This covered the spectral region from 65 nm to 95 nm and would support a transform-limited pulse of duration 7.6 fs. The pulse compression using only a pair of chirped mirrors with a negative dispersion of 6 fs 2 (3 bounces) yielded a pulse duration of.7 fs as shown in Fig. 5(c), wherein temporal characterization was performed using an interferometric autocorrelation (IAC) technique in combination with an evolutionary phase retrieval algorithm. 2) The circles and Fig.. Spectral and temporal characteristics of adaptively compressed 23 fs pulse Intensity (normalized) solid line in Fig. 5 represent the measured and retrieved IAC traces, respectively, and the transform-limited pulse is shown by a dotted line in Fig. 5(c). Our SPIDER apparatus discussed in section 2 was not applicable to unamplified (c).7 fs Fig. 5. Optical spectrum of MDC Ti:sapphire laser and.7 fs pulse generated after compression using chirped mirrors. The measured IAC trace and reconstructed one, obtained using an evolutionary phase retrieval algorithm, are represented by circles and solid line, respectively, in. The recontructed electric field is represented by a solid line in (c), and the transform-limited pulse by a dotted line in (c). 5-5

4 5292 Jpn. J. Appl. Phys., Vol. 3, No. A (2) K.-H. HONG and C. H. NAM laser pulses with a very broad spectrum because the SPIDER signal was not sufficiently strong for the characterization. We replaced 5 dielectric mirrors in the non-dispersive arm of the SPIDER apparatus with 5 metallic ones to support the full spectral bandwidth of the MDC laser pulse, which caused an energy loss of about 23%. In addition, the pulse stretching ratio of the dispersive arm was about 2 times larger than the original value due to the realization of a spectrum 2 times broader than the original (from the comparison of spectral edges), making the pulse intensity at the dispersive arm also 2 times weaker. These effects made the sum-frequency mixing between the pulses from the nondispersive and dispersive arms significantly weaker than before, which did not allow the application of SPIDER to the analysis of unamplified sub--fs pulses. The broad MDC laser pulse requires modifications in the pulse shaper for proper compression. We, first, precompressed the laser pulse by bouncing the pulse off the chirped mirrors times to compensate for the pulse broadening due to GDD in the output coupler and the prism of the pulse shaper. The precompensation allowed MMDM to sensitively work for the compensation of high-order phase distortions. A less dispersive Brewster-cut prism made of fused silica was utilized instead of the SF prism because the spectrum was broader than that of the PDC Ti:sapphire laser. The transmission efficiency decreased to 5% due to the 2 reflections at the gold-coated curved mirror ( f ¼ : m), which was used instead of the dielectric curved mirror to cover the entire spectrum of the MDC laser pulse. A SH optimization process was performed by choosing genetic parameters similar to those of section 2. In this case, for the optimization, a 3-mm-thick BBO crystal was substituted for the -mm-thick one and the area of the SH spectrum was maximized rather than the peak value because the laser spectrum had two peaks. (In our previous experiment, a higher peak value of the SH spectrum always resulted in a larger area because of the single-peaked spectral shape.) The number of generations and computation time for convergence were nearly the same as those in our previous experiment. The precompressed pulse with a flat voltage array, characterized using the IAC measurement and the phase retrieval algorithm, is shown in Fig. 6. The measured and retrieved IAC traces of the pulse are represented by circles and a solid line, respectively, in Fig. 6, and the corresponding intensity and phase profiles of the retrieved pulse are shown in Fig. 6. Residual GDD, TOD, and FOD in the spectral phase are estimated to be 6 fs 2, 5 fs 3, and 2 fs, respectively. The pulse broadening resulted from the high-order dispersion terms from chirped mirrors, prism material, and uncalibrated small curvature of MMDM in the flat voltage array. Due to the large TOD and FOD, the pulse had a duration of.5 fs and a strong wing structure. On the other hand, the IAC trace of the adaptively compressed pulse is shown in Fig. 7 and the reconstructed intensity and phase profiles are shown in Fig. 7. The wing structure was significantly suppressed, compared with that of the precompressed.5 fs pulse, because TOD and FOD significantly decreased. The GDD, TOD, and FOD of this pulse were 2 fs 2,9fs 3, and 5 fs, respectively, and the corresponding pulse duration was. fs, which was % fs Fig. 6. Precompressed.5 fs laser pulse. The measured IAC trace and reconstructed one are represented by circles and solid line, respectively, in. The recontructed electric field is shown in, and the transformlimited pulse by a dotted line in. larger than the transform-limit value of 7.6 fs. The duration was slightly shorter than that of the pulse compressed using the chirped mirrors shown in Fig. 5. Although the improvement in pulse duration was small, it was not due to the convergence error of the phase retrieval algorithm used. The direct comparison between the two IAC traces in Figs. 5 and 7 shows that the IAC trace in Fig. 7 has a slightly shorter pulse duration and a weaker wing structure than that in Fig. 5. Hence, the reduction in pulse duration really occurred. Considering that the GDD, TOD, and FOD of the.7 fs pulse were 3 fs 2, 2 fs 3, and 6 fs, respectively, we can conclude that the adaptive compression achieved using the pulse shaper has clearly improved the compensation of the phase distortion due to TOD.. Conclusions An adaptive pulse compression scheme has been demonstrated using a low-loss adaptive pulse shaper comprising an MMDM and a Brewster-cut prism. GA was used as a feedback control algorithm to maximize the SH signal of the laser pulse, thereby minimizing pulse duration. In the first experiment, a positively chirped fs pulse produced in a PDC Ti:sapphire laser was adaptively compressed to a 23 fs pulse, demonstrating the GDD compensation capability of the pulse shaper. In the second experiment, a laser pulse generated in an MDC Ti:sapphire laser was adaptively compressed down to. fs, which is. times the transformlimited pulse duration. The compensation of TOD was

5 Jpn. J. Appl. Phys., Vol. 3, No. A (2) K.-H. HONG and C. H. NAM fs Fig. 7. Adaptively compressed. fs pulse. The measured IAC trace and reconstructed one are represented by circles and solid line, respectively, in. The recontructed electric field is shown in, and the transformlimited pulse by a dotted line in. observed to be enhanced using the pulse shaper as compared with the case of chirped-mirror compression. Transmission efficiency was considerably increased by use of a Brewstercut prism instead of a diffraction grating: Efficiencies up to 93% were obtained with dielectric mirrors and up to 5% with gold-coated mirrors for sub--fs pulses. The high transmission efficiency of this pulse shaper is helpful for the application of unamplified, but shaped, femtosecond laser pulses to the coherent control of laser-matter interactions. In addition, a prism-based pulse shaper is a good candidate for the adaptive pulse compression of high-order harmonics, which have an intrinsic negative chirp, 25) in the vacuum ultraviolet (VUV) region 26) because some prisms such as CaF 2 and LiF have good transmission and dispersion characteristics in this wavelength region, assuming that properly coated deformable mirrors are developed Acknowledgment This work was supported by the Ministry of Science and Technology of Korea through the Creative Research Initiative Program. The help of Mr. Yong Hoon Kang in the SPIDER measurement is highly appreciated. The authors thank Professor Umesh for helpful discussion. ) D. Yelin, D. Meshulach and Y. Silberberg: Opt. Lett. 22 (997) ) P. Tournois: Opt. Commun. (997) 25. 3) E. Zeek, K. Maginnis, S. Backus, U. Russek, M. Murnane, G. Mourou, H. Kapteyn and G. Vdovin: Opt. Lett. 2 (999) 93. ) F. Verluise, V. Laude, Z. Cheng, Ch. Spielmann and P. Tournois: Opt. Lett. 25 (2) ) E. Zeek, R. Bartels, M. M. Murnane, H. C. Kapteyn and S. Backus: Opt. Lett. 25 (2) 57. 6) Z. Michalewicz: Genetic Algorithms + Data Structures = Evolution Programs (Springer, Berlin Heidelberg, 999). 7) C. J. Bardeen, V. V. Yakovlev, K. R. Wilson, S. D. Carpenter, P. M. Weber and W. S. Warren: Chem. Phys. Lett. 2 (997) 5. ) R. Bartels, S. Backus, E. Zeek, L. Misoguti, G. Vdovin, I. P. Christov, M. M. Murnane and H. C. Kapteyn: Nature 6 (2) 6. 9) B. J. Pearson, J. L. White, T. C. Weinacht and P. H. Bucksbaum: Phys. Rev. A 63 (2) 632. ) A. Assion, T. Baumert, M. Bergt, T. Brixner, B. Kiefer, V. Seyfried, M. Strehle and G. Gerber: Science 22 (99) 99. ) T. Brixner, N. H. Damrauer, P. Niklaus and G. Gerber: Nature (2) 57. 2) D. G. Lee, J.-H. Kim, K.-H. Hong and C. H. Nam: Phys. Rev. Lett. 7 (2) ) S. H. Lee, K.-H. Jung, J. H. Sung, K.-H. Hong and C. H. Nam: J. Chem. Phys. 7 (22) 95. ) See the data sheet of DAZZLER. 5) See the data sheet of SLMs produced by CRI Inc. or JENOPTIK GmbH. 6) M. R. Armstrong, P. Plachta, E. A. Ponomarev and R. J. D. Miller: Opt. Lett. 26 (2) 52. 7) L. Xu, N. Nakagawa, R. Morita, H. Shigekawa and M. Yamashita: IEEE J. Quantum Electron 36 (2) 93. ) B. Schenkel, J. Biegert, U. Keller, C. Vozzi, M. Nisoli, G. Sansone, S. Stagira, S. De Silvestri and O. Svelto: Opt. Lett. 2 (23) 97. 9) H. J. Gerritsen and M. L. Jepsen: Appl. Opt. 37 (99) ) M. D. Perry, D. Pennington, B. C. Stuart, G. Tiethbohl, J. A. Britten, C. Brown, S. Herman, G. Golick, M. Kartz, J. Miller, H. T. Powell, M. Vergino and V. Yanovsky: Opt. Lett. 2 (999) 6. 2) K.-H. Hong and C. H. Nam: CLEO/PR 23, Taipei, Taiwan, 23, ThP(6)-5. 22) C. Iaconis and I. Walmsley: Opt. Lett. 23 (99) ) J. H. Sung, K.-H. Hong, Y. H. Cha and C. H. Nam: Jpn. J. Appl. Phys. (22) L93. 2) K.-H. Hong, Y. S. Lee and C. H. Nam: Ultrafast Optics IV, Vienna, Austria, ) J.-H. Kim, D. G. Lee, H. J. Shin and C. H. Nam: Phys. Rev. A 63 (2) ) T. Sekikawa, T. Ohno, T. Yamazaki, Y. Nabekawa and S. Watanabe: Appl. Phys. B 7 (2) S233.