August 1, 1989 / Vol. 14, No. 15 / OPTICS LETTERS 797

Chirped-pulse amplification of 100-fsec pulses

Maurice Pessot, Jeff Squier, and Gerard Mourou*

Laboratory for Laser Energetics, University of Rochester, 250 East River Road, Rochester, New York 14623-1299

Donald J. Harter

Allied Signal, Inc., Corporate Technology, P.O. Box 1021R, Morristown, New Jersey 07960

Received February 7, 1989; accepted May 9, 1989

Chirped-pulse amplification is used to generate 2-mJ pulses of 106-fsec duration in an alexandrite amplifier.Compression of the optical pulse is achieved by using a sequence of intracavity prisms in conjunction withdiffraction gratings. This allows for the compensation of both linear and quadratic contributions to the dispersionfrom the amplifier.

The advent of chirped-pulse amplification1 had led toa renewed interest in the development of solid-stateamplifiers for ultrashort pulses. The broad band-width and high energy-storage capabilities of solid-state materials such as alexandrite and Ti:sapphirepoint to the potential for these materials in the pro-duction of high-energy femtosecond pulses. Recent-ly, the generation of 300-fsec pulses at the millijoulelevels in an alexandrite regenerative amplifier 2 wasreported. This was achieved by using diffractiongratings in both positive 3'4 and negative 5 group-veloci-ty-dispersion configurations to manipulate the pulseboth before and after amplification. It was also point-ed out that the presence of dispersive elements withinthe amplifier leads to significant contributions to thedispersion from cubic phase shifts that would limit thepulse width from such a system. In this Letter weshow that this limitation can be overcome by using acombination of diffraction gratings in conjunctionwith a sequence of prisms6 to compensate for both thequadratic and cubic phase shifts arising from the dis-persive elements. In this way we are able to generatepulses as short as 106 fsec with peak powers of -20GW.

The basic system used here is similar to that of Ref.2, although some modifications have been made.First, the dye laser used in these experiments (765 nm,300 fsec) does not provide a short enough pulse. Inorder to provide sufficient bandwidth for the genera-tion of 100-fsec pulses, the dye-laser output was am-plified in a conventional dye amplifier and coupledinto a 20-cm-long section of polarization-preserving,single-mode fiber. The spectrum could be broadenedout to -25 nm through self-phase modulation withinthe fiber,7 with the pulse width broadening to -1 psec.The pulse was then preamplified and double passedthrough a grating expander before it was injected intoan alexandrite regenerative amplifier. Grating ex-pansion has been shown to easily provide sufficientdispersion to broaden femtosecond pulses by a factorof 102-103.2,3 The grating expander was identical tothat used previously,2 consisting of two antiparallel

1800-line/mm gratings separated by 145 cm, with apair of 500-mm lenses between the gratings forming aunit magnification telescope. The injected energywas typically in the 5-10-nJ range inma pulse '120 pseclong.

Streak-camera measurements of the expandedpulse are shown in Fig. 1(b). The pulse has acquired asevere modulation that appears to mirror the pulsespectrum [Fig. 1(a)]. This is in fact generally true andwas observed to hold as we adjusted the input powercoupled into the optical fiber. As the power was in-

814

-360

751

nm

688

0 360

Picoseconds

50 0 50

Picoseconds

Fig. 1. (a) Self-phase-modulated spectrum from the fiber,(b) streak-camera image of the expanded pulse, and (c) nu-merical simulation of the expanded pulse.

0146-9592/89/150797-03$2.00/0 © 1989 Optical Society of America

I (a)

798 OPTICS LETTERS / Vol. 14, No. 15 / August 1, 1989

Fig. 2. Spectrum of (a) the injected pulse and (b) the am-plified pulse for a detuning of <1 nm. The amplified pulseenergy is 5 mJ.

creased the pulse spectrum would broaden, and theexpanded pulse would also broaden (owing to the larg-er bandwidth) and acquire a temporal structure thatmirrored the spectral features. This can be under-stood to be a result of the non-transform-limited na-ture of the self-phase-modulated input pulse. Theinput pulse can be pictured as consisting of a numberof overlapping spectral wave packets centered at dif-ferent carrier frequencies. These wave packets willspread apart as the pulse propagates through the ex-pansion system, giving rise to the structure observed.This structure also appears in our simulations of thepulse, as evidenced in Fig. 1(c). Only a qualitativeagreement is reached, as we have not taken into con-sideration the asymmetry of the input spectrum or thenonlinear propagation of the pulse within the opticalfiber.

In contrast to previous research in which the spec-trum of the input pulse was significantly narrower (-3nm), the spectrum of the amplified pulse from thealexandrite regenerative amplifier was found to besensitive to any mismatch between the peak gain ofthe alexandrite (-760 nm) and the operating wave-length of the dye laser. When the dye-laser operatingwavelength was -3.5 nm to the red of the amplifiergain center, a severe spectral narrowing and reshapingof the self-phase-modulated input spectrum was ob-served. Broad spectra were narrowed considerably,from 23 to 8 nm, after amplification to an energy ofonly -100 AJ, with the red side of the spectrum beingcompletely depleted.

In order to match the dye/amplifier wavelengthsprecisely, the alexandrite rod was run at an elevatedtemperature of 150°C. Alexandrite is known to havea temperature-dependent emission cross section thatshifts to longer wavelengths at higher temperatures. 8

This was preferred to insertion of any bandwidth-limiting tuning elements in either the dye laser oralexandrite amplifier to bring the two into resonance.With the wavelengths matched to within 1 nm, thespectral reshaping is reduced substantially. In Fig. 2we show a 13-nm input spectrum that is amplified toan energy of 5 mJ. The spectrum of the amplified

pulse has narrowed only slightly, to 12 nm. This re-sidual narrowing seems to be a true spectral narrow-ing, as broader input spectra were also narrowed inthis situation to approximately the same width. Theobserved spectral narrowing can arise from the fre-quency-dependent cross section of the material and/orany other optical component with a wavelength de-pendence. The observed output is approximately halfof the expected effective bandwidth for an alexandriteamplifier with a fluorescence bandwidth of 1800 cm-'(Ref. 9) and a gain of 106,10 indicating that other cavitycomponents may also be contributing to the observedspectral reshaping.

The presence of dispersive material elements withinthe regenerative amplifier cavity leads to a furtherbroadening of the input pulse and to limitations on themaximum compression that can be achieved. Thislimitation arises from the need to adjust the compres-sion-grating separation to compensate for the disper-sion accumulated during the 50-55 round trips thatthe pulse makes within the cavity. Adjustment of thecompression-grating separation to compensate forboth the expansion gratings and the accumulated lin-ear dispersion from the material elements results in asubstantial residual quadratic dispersion that limitsthe compressed pulse width. We wish to stress thatthe bulk of the residual (quadratic) dispersion is notcontributed by the material elements but from thegratings. It is a consequence of not using the expan-sion and compression gratings at identical effectiveseparations and angles of incidence. The expansion/compression gratings are then no longer matched dis-persive devices and make a contribution to the residu-al dispersion that is approximately three times thecontribution from the material elements. In Ref. 2 itwas shown that the residual dispersion would limit thepulse width to -200 fsec.

If we wish to compress the pulse beyond this, weneed to compensate for both the linear and quadraticcontributions to the dispersion. The necessary meansfor higher-order compensation come from the obser-vation that the sign of the cubic phase shift, a301/dW 3, ispositive for diffraction gratings and negative for aBrewster-angled prism sequence.'" We have modi-

TFP

FR

PD

Ml PH M2

Fig. 3. Alexandrite regenerative amplifier with four Brew-ster prisms for cubic phase compensation. PD, photodiode;TFP's, thin-film polarizers; Ml, M2, high reflectors; PH,pinhole, FR, Faraday rotator and half-wave plate; PC, Pock-els cell.

1psec

- -~s 106-fsec FWHM

l l I I l l I l I

Fig. 4. Single-shot autocorrelation of the compressedpulse. The FWHM is 106 fsec, assuming a sech2 pulse shapefor deconvolution.

fied our alexandrite regenerative amplifier to incorpo-rate a prism sequence to provide the necessary higher-order compensation (Fig. 3). We have placed fourBrewster-angled prisms of SF-10 glass within the cavi-ty with a separation of 43 cm. Two of the prisms wereplaced on stages so that the prism could be translatedin a direction perpendicular to its base. Placing theprisms internal to the cavity allows us to take advan-tage of the large number of transits that the pulsemakes to accumulate the required amount of compen-sation. External to the cavity the required prismspacing was not feasible.

The prism separation was chosen on the basis ofcalculations of the contributed phase shift from alldispersive elements both internal and external to thecavity. The prism and grating separations were usedas the two degrees of freedom to find some point atwhich the quadratic and cubic phase shifts would can-cel. Strictly, this cancellation holds only for a uniquenumber of transits within the cavity. This could becontrolled by varying the level of the signal injectedinto the cavity and setting the cavity dumper to fire ata specific time interval after the Q switch, rather thancavity dumping on the basis of a threshold detector'smonitoring the energy buildup of the pulse. With theprism sequence included, the compression gratingswere then varied to optimize the compressed pulse.Further optimization should have been possible bytranslating the prisms into the beam path. In prac-tice, it was found that mechanical instabilities pre-vented us from making significant translations of theprisms.

Single-shot, background-free autocorrelationl2measurements were used to monitor the compressedpulse width. Figure 4 shows a 106-fsec pulse, assum-ing a sech2 pulse shape for deconvolution. The 40%(double-pass) diffraction efficiency of the gratings re-duced the pulse energy from 5 to 2 mJ, which yielded apeak power of almost 20 GW. When the prism se-quence was not included within the cavity, the short-est compressed pulse observed was 170 fsec. This is inreasonable agreement with the minimum width ex-pected from the limitations described above, indicat-ing that we have achieved a partial cancellation of thecubic phase. We believe that the incomplete compen-sation is a result of a mismatch of a few degrees in the

August 1, 1989 / Vol. 14, No. 15 / OPTICS LETTERS 799

angle of incidence of the beam between the expansionand compression-grating stages, resulting in a largercubic phase than our design could compensate. Theoverall expansion/compression ratio of over 1100 iscomparable with that seen in multiple-stage fiber-grating compression systems.'3' 14

In summary, we have shown that chirped-pulse am-plification techniques can be used for the productionof amplified femtosecond optical pulses from broad-band solid-state systems. We have demonstratedthat a combination of intracavity prisms and diffrac-tion gratings can be used to provide compensation ofcubic phase shifts, circumventing limitations in thecompressed pulse width that arise from dispersive ele-ments within the cavity. Simple extensions of theresults shown here indicate the potential of chirped-pulse amplification for the production of terawattfemtosecond pulses in the near future.

This research was supported by the Laser FusionFeasibility Project at the Laboratory for Laser Ener-getics, University of Rochester, which is sponsored bythe Empire State Electric Energy Research Corpora-tion, the New York State Energy Research and Devel-opment Authority, Ontario Hydro, and the Universityof Rochester. Additional support was provided by theU.S. Air Force Office of Scientific Research undercontract F49620-87-C-0016 to the Ultrafast OpticalElectronics Center at the Laboratory for Laser Ener-getics. Such support does not imply endorsement ofthe content by any of the above parties.

* Present address, Department of Electrical Engi-neering, University of Michigan, Ann Arbor, Michigan48109.

References

1. D. Strickland and G. Mourou, Opt. Commun. 56, 219(1985).

2. M. Pessot, J. Squier, P. Bado, G. Mourou, and D. Harter,IEEE J. Quantum Electron. QE-25, 61 (1989).

3. M. Pessot, P. Maine, and G. Mourou, Opt. Commun. 62,419 (1987).

4. 0. E. Martinez, IEEE J. Quantum Electron. QE-23, 59(1987).

5. E. B. Treacy, IEEE J. Quantum Electron. QE-5, 454(1969).

6. R. L. Fork, C. H. Brito Cruz, P. C. Becker, and C. V.Shank, Opt. Lett. 12, 483 (1987).

7. R. H. Stolen and C. Lin, Phys. Rev. A 17, 1448 (1978).8. J. C. Walling, 0. G. Peterson, H. P. Jenssen, R. C. Mor-

ris, and E. W. O'Dell, IEEE J. Quantum Electron. QE-16,1302 (1980).

9. L. F. Mollenauer and J. C. White, eds., Tunable Lasers(Springer-Verlag, New York, 1987), p. 352.

10. A. E. Siegman, Lasers (University Science, Mill Valley,Calif., 1986), p. 282.

11. 0. E. Martinez, J. P. Gordon, and R. L. Fork, J. Opt. Soc.Am. A 1, 1003 (1984).

12. F. Salin, P. Georges, G. Roger, and A. Brun, Appl. Opt.26, 4528 (1987).

13. B. Zysset, W. Hodel, P. Beaud, and H. P. Weber, Opt.Lett. 11, 156 (1986).

14. A. S. Gouveia-Neto, A. S. L. Gomes, and J. R. Taylor,Opt. Lett. 12, 395 (1987).