2184 OPTICS LETTERS / Vol. 29, No. 18 / September 15, 2004

High-efficiency, simple setup for pulse cleaning at themillijoule level by nonlinear induced birefringence

Aurélie Jullien, Frédérika Augé-Rochereau, Gilles Chériaux, and Jean-Paul Chambaret

Laboratoire d’Optique Appliquée, Ecole Nationale Supérieure des Techniques Avancées —Ecole Polytechnique,91761 Palaiseau Cedex, France

Pascal d’Oliveira and Thierry Auguste

Service des Photons, des Atomes et des Molécules, Département de Recherche sur l’Etat Condensé, les Atomes et les Molécules,Commissariat à l’Energie Atomique, 91191 Gif sur Yvette, France

Franck Falcoz

Thales Laser, Route Départementale 128, Domaine de Corbeville, 91400 Orsay, France

Received March 24, 2004

Nonlinear elliptical polarization rotation is used to improve the contrast of femtosecond pulses by several or-ders of magnitude. Using nonlinear induced birefringence in air, we produced cleaned pulses with an energyof a few hundreds of microjoules. This technique presents several major advantages, such as convenienceand stability of the setup. We investigated the phase profile required for obtaining high-energy pulses. Nophase distortion is observed, and the spatial quality of the beam is preserved. © 2004 Optical Society ofAmerica

OCIS codes: 140.7090, 320.2250, 320.7110.

One of the main bottlenecks for applications of ul-trashort and ultraintense lasers in high-field physicsis the temporal contrast of the pulses. In classicalchirped-pulse amplif ication laser chains a nanosecondpedestal of amplif ied spontaneous emission (ASE)and some parasitic pulses are generated at the sametime as the femtosecond pulse. The intensity ratiobetween the ASE or parasite pulses and the mainfemtosecond pulse is called pulse contrast. Forlaser–matter interaction experiments the pulsesare focused on solid targets with an intensity thatusually reaches 1020 1021 W cm22. Even though theintensity level of the ASE pedestal �1013 W cm22� isusually 6 or 7 orders of magnitude lower than themain pulse intensity, it is high above the ionizationthreshold of the target �1010 W cm22�. Therefore theASE completely modifies the experimental conditionsof the laser–matter interaction. To avoid plasmaformation before the femtosecond pulse, it is crucialboth to suppress satellite pulses and to decrease theASE background by at least 3 orders of magnitude.

It has been demonstrated that, in Ti:Al2O3 laserchains, the pulse contrast is higher than 10 orders ofmagnitude at the oscillator output.1 The ASE and thesatellite pulses are generated mainly in a high-gainpreamplifier and then amplif ied in the power ampli-fier stages. Schemes for pulse amplif ication, such ashigh-energy seed pulse injection into the amplif ierchain2 and the use of a saturable absorber,3 have beenelaborated. Another method consists of cleaning thepulse after preamplif ication by use of a nonlinearfilter. Experiments with Sagnac interferometershave already been reported.4

In this Letter we present a new technique basedon elliptical polarization rotation generated by non-

0146-9592/04/182184-03$15.00/0

linear induced birefringence. The application ofnonlinear polarization rotation to femtosecond pulsecontrast improvement was first studied by Ho-moelle et al., using hollow waveguides f illed withxenon.5 However, the setup efficiency has beendemonstrated only for a relatively low level of en-ergy: the energy of the cleaned pulse was limited to afew microjoules. Using higher-intensity pulses wouldbe a significant advantage for further amplif ication,since power amplif iers present low gain and createless ASE than the preamplifier stage. By seedingclean pulses in the power amplif iers, one can expectintense high-contrast pulses at the output of thelaser chain. In comparison, the system that we havedesigned has several advantages: it is simple, robust(the nonlinear effects occur in air), and appropriate forhigh-energy pulses. We have succeeded in producingclean pulses with an energy of a few hundred micro-joules. To our knowledge, this is the highest-energycleaning filter that has been reported.

The setup that we designed is shown in Fig. 1. Theinput pulse passes through a first achromatic zero-order l�4 plate to generate an elliptically polarizedbeam. The angle between one of the axes of this plateand the input polarization direction is 22.5± to optimizethe setup efficiency. Then the beam is focused by anf 0 � 500 mm lens. The polarization rotation of thehigh-intensity part of the pulse occurs during propa-gation, mostly at the vicinity of the focus point. Thepolarization state of the less intense part of the pulse(pedestal and satellite pulses) is unchanged. Then thebeam is collimated � f 0 � 800 mm� and passes throughanother l�4 plate oriented with its fast axis perpen-dicular to the fast axis of the first wave plate. Afterthis second plate the low-intensity part of the pulse has

© 2004 Optical Society of America

September 15, 2004 / Vol. 29, No. 18 / OPTICS LETTERS 2185

Fig. 1. Setup of the f ilter at the millijoule level based onnonlinear induced birefringence.

returned to its initial state, whereas the polarizationof the intense part has rotated. A polarizer placed af-ter the collimation lens transmits a part of the intensepulse and rejects all the other components. Using airas a nonlinear medium leads to a simple setup. Thesame experimental behavior would be expected withnoble gas, when one takes into account the pressureand the nonlinear index value. With this convenientmethod we succeeded in filtering an 850-mJ pulse witha transmission of 25%.

Experiments were performed with a chirped-pulseamplification laser including regenerative and multi-pass amplifiers. It produced 42-fs, 850-mJ pulses ata 1-kHz repetition rate. The spectrum of the pulse,shown in Fig. 2, was centered at 808 nm. To esti-mate the polarizing system extinction ratio, we usedthe nanosecond regime of the regenerative amplif ierand measured a value of 7 3 1024.

To optimize the setup efficiency we investigated theeffect of the spectral phase of the pulse on filter trans-mission. When the pulse was perfectly compressed,its power (20 GW) exceeded the critical power in the air(3 GW). To avoid filamentation near the focus point,we reduced the pulse intensity by introducing second-�f2� and third-order �f�3 phases. Experimental inves-tigations were undertaken to estimate thoroughly theconsequences of spectral phase on the filtering, usingan acousto-optic programmable dispersive filter (Daz-zler by Fastlite) in the chain. Phase defaults couldalso be added by changing the compressor adjustment.Spectral phase and intensity profiles were accuratelyevaluated from spectral phase interferometry for directelectric-f ield reconstruction (SPIDER) measurement.

It was observed that the sign of the third-orderphase is crucial. For positive values of f3, from0 to 250, 000 fs3, the f ilter transmission is a fewpercent. However, a balance between negative f3

and f2 led to good results. For a fixed value of f2,by tuning the negative cubic phase, experiments haveshown the existence of a threshold of efficiency nearf3 � 2200, 000 fs3. This value corresponds to asignificant change of the temporal shape of the pulse,with more energy in the main pulse, as shown inFig. 3. The highest transmission of the f ilter wasobtained for f3 � 2200, 000 fs3 and f2 � 7000 fs2.As the diameter of the beam is 100 mm at the focuspoint, in this case the intensity of the focused pulse isI � 4.5 3 1013 W cm22. At this intensity the ioniza-tion ratio is only 1025 and plays no role with respectto the beam propagation.6 Figure 2 shows the outputspectrum obtained with this phase profile. Theredshift of the spectrum is generated by the delayed

Kerr effect: self-phase modulation in asymmetricmolecular gases, such as nitrogen and oxygen, gener-ates a delayed change of the nonlinear refractive indexthat causes the observed shift.7 It was also observedthat a positive chirp allowed spectrum broadening,whereas a negative chirp narrowed it, as predicted bythe spectral properties of self-phase modulation.

The very high level of energy that we obtained withthis setup will allow further amplification and use ofthe cleaned beam. Furthermore the spatial profile ofthe beam is not distorted, as shown in Fig. 4.

Figure 5 shows the measured spectral phasesof the pulse before and after the nonlinear effect.By considering the phase of the cleaned pulse, we

Fig. 2. Pulse spectrum before (solid curve) and after(dashed curve) filtering.

Fig. 3. Calculated temporal shape of the pulse before fil-tering for different values of f3 �f2 � 7000 fs2�. The pulseis normalized to the intensity of the pulse with f3 � f2 � 0.Calculations are achieved by taking into account the ex-perimental spectrum and second- and third-order phasevalues.

Fig. 4. (a) Horizontal and (b) vertical cuts of the spatialbeam profile before (solid curves) and after (dotted curves)filtering (f3 �2200, 000 fs3 and f2 � 7000 fs2).

2186 OPTICS LETTERS / Vol. 29, No. 18 / September 15, 2004

Fig. 5. SPIDER measurements of the phase profile of thepulse before filtering (solid curve, f3 �2200, 000 fs3 andf2 � 7000 fs2) and after f iltering (dotted curve).

Fig. 6. Third-order correlation curves before (gray curve)and after (black curve) f iltering (f3 � 2200, 000 fs3 andf2 � 7000 fs2).

calculated a fourth-order f it of experimental data�f2 � 10, 400 fs2, f3 � 219, 000 fs3, f4 � 190 fs4).These values can be explained by taking into accountthe nonlinear phase generated by the Kerr effect andthe dispersion added by plates and a Glan polarizer.After filtering, the pulse will be able to be recom-pressed close to the Fourier-transform limit.

To estimate the contrast enhancement we character-ized the input and output beams with a high-dynamic3v cross correlator (Sequoia by Amplitude Technolo-gies). To obtain comparable cross-correlation curveswe attenuated the energy of the nonfiltered pulse downto the energy of the filtered pulse. The correlationcurves are shown in Fig. 6. Whereas the pulse be-fore f iltering presents several satellite prepulses, nodefect is visible in the filtered pulse. Temporal phaseaberrations as a result of compression defects havealso been suppressed. One can observe a significantdecrease of the ASE background, since the measure

for the filtered beam is limited by the correlator noise3 3 1029. The residual postpulses on the two curvesare actually generated by the l�4 plates and are notproblematic for applications. This indicates that thecontrast improvement is better than 3 orders of mag-nitude, limited only by the extinction ratio of the po-larizing system.

To conclude, we have designed a nonlinear filterthat is highly efficient for cleaning femtosecond pulsesat the millijoule level. This filter, based on nonlinearbirefringence induced directly in the air, presentsgood energy transmission and significant pulse con-trast enhancement. The spatial beam quality ispreserved. We accurately determined the second-and third-order spectral phase profiles required forbalancing self-focusing of the beam and ionizationof the medium for optimal filter transmission. Wethink that the introduced phase defaults can be easilycompensated for further utilization. Furthermore, aneasy way to improve the eff iciency of this setup wouldbe to increase the interaction length by refocusingthe pulse several times in a multipass configura-tion.8 Simulations have been undertaken to validatethe experimental behavior. The simplicity of thesetup and its performance allow us to contemplatea promising development of future ultraintense,high-contrast laser chains.

This work was supported by the EuropeanCommission through the SHARP project (Sup-pression over High dynamic range of ASE at theRising edge of ultra-intense femtosecond Pulses,HPRI-CT-2001-50037). Funding for A. Jullien (Aure-lie.Jullien@ensta.fr) from Thales Laser is gratefullyacknowledged.

References

1. G. Cheriaux, T. Planchon, F. Auge, G. Mourou, andJ. P. Chambaret, in Proceedings of Ultraffast Optics2001 (Springer-Verlag, Berlin, 2001), p. 16.

2. J. Itatani, J. Faure, M. Nantel, G. Mourou, andS. Watanabe, Opt. Commun. 148, 70 (1998).

3. M. Nantel, J. Itatani, A. C. Tien, J. Faure, D. Kaplan,M. Bouvier, T. Buma, P. Van Rompay, P. Pronko,D. Umstadter, and G. Mourou, IEEE J. QuantumElectron. 4, 449 (1998).

4. G. Cheriaux, M. Traff, T. Planchon, F. Auge, J. P. Cham-baret, and G. Mourou, in Conference on Lasers andElectro-Optics (CLEO), Vol. 56 of OSA Trends in Opticsand Photonics Series (Optical Society of America, Wash-ington, D.C., 2001), p. 45.

5. D. Homoelle, A. L. Gaeta, V. Yanovsky, and G. Mourou,Opt. Lett. 27, 1646 (2002).

6. C. Cornaggia and Ph. Hering, Phys. Rev. A 62, 023403(2000).

7. E. T. J. Nibbering, G. Grillon, M. A. Franco, B. S. Prade,and A. Mysyrowicz, J. Opt. Am. B 14, 650 (1997).

8. P. Nickles and M. Kalashnikov, presented at the fifthSHARP Meeting, Lund, Sweden, November 28–29,2003.