17-fs pulses from a self-mode-locked Ti:sapphire laser
Chung-Po Huang, Melanie T. Asaki, Sterling Backus, Margaret M. Murnane, and Henry C. Kapteyn
Department of Physics, Washington State University, Pullman, Washington 99164-2814
Howard Nathel
Lawrence Livermore National Laboratory, Livermore, California 94550
Received May 12, 1992
We have generated sub-17-fs-duration pulses directly from a self-mode-locked Ti:sapphire laser. These pulses
are near transform limited, with a wavelength centered at 817 nm, a pulse repetition rate of 80 MHz, and an
average power of 500 mW. By minimizing the amount of material inside the laser cavity and choosing the cor-
rect glass for the intracavity prism pair, third-order dispersion in the laser can be significantly reduced com-
pared with that in previous designs. Extracavity compensation for group-velocity dispersion in the output
coupler and autocorrelator optics is necessary to measure this pulse width. To our knowledge this laser gener-ates pulses substantially shorter than any other laser to date.
There has been much interest of late in the use ofsolid-state titanium-doped sapphire lasers to gener-ate ultrafast laser pulses.`8 In previous research,we demonstrated the generation of 32-fs pulses froma mode-locked Ti:sapphire laser.5 Further researchon this laser reduced the pulse duration to 28 fs, andlater to 25 fs. The 28-fs measurement was achievedby choosing the glass type for the intracavity prismpair in order to minimize third-order dispersion inthe laser cavity, i.e., to make the group delay for lighttraveling in the laser cavity as frequency indepen-dent as possible over a wide bandwidth. For the 25-fsresult, a mirror was used (New Focus Model 5103),which was found to compensate partially for prismthird-order dispersion if the laser center wavelengthwas at 830 nm.
In this Letter we report further progress in mini-mizing intracavity dispersion in a self-mode-lockedTi:sapphire laser. This research has resulted thegeneration of 17-fs-duration pulses. The generationof such short-duration pulses directly from a laseris unprecedented. This laser is reliable, and thereis no degradation in performance, stability, orease of use compared with longer-pulse-durationTi:sapphire lasers. Our result demonstrates thatthe exceptionally broad bandwidth of Ti:sapphire,which is large enough to support <<10-fs pulses,9
can be utilized to generate pulses shorter than hasbeen possible with any other type of laser materialto date.
The laser is a conventional linear cavity laser,consisting of two end mirrors, two 10-cm radius-of-curvature mirrors focusing into a Ti:sapphirerod, and a pair of intracavity prisms for dispersioncompensation. The Ti:sapphire crystal is a 9-mm-long, 0.11%-doped material with a figure of merit>150 (Union Carbide). The 10-cm mirrors areSpectra-Physics 800-900-nm mirrors for a Model3900 cw Ti:sapphire laser. The Ti:sapphire crystalwas pumped by 5 W of all-lines green from an argon-
ion laser (Coherent INNOVA-20) through one of the10-cm mirrors and focused by a 12.5-cm focal-lengthlens (Newport Corporation). The two end mirrorsof the cavity (CVI Laser Corporation and Spectra-Physics) are flat, with transmissions of 6.5% and2.5% and a center wavelength of -790 nm. Thetotal cavity length is 185 cm, with 85 cm in one armof the laser and 90 cm on the side of the dispersion-compensating prisms. The Ti:sapphire crystal is'1 mm off center between the two 10-cm mirrors.No apertures are used in the laser cavity to inducemode locking, although there is some clipping of thebeam at the apexes of the prisms.
The prisms used in this laser are made of SchottLaKL21 glass, cut to an apex angle (63°) such thatthe angle of incidence corresponds to the Brewsterangle at minimum deviation. The prism separationwas 51 cm. This prism glass was chosen based onthree criteria. First, it is a relatively stable, hardglass that can be polished optically flat. Second,the prism separation required for compensation ofthe linear (second-order) dispersion introduced bythe Ti:sapphire rod can be accommodated in a 1.5-2.0-m cavity length. Third, of the glasses that fitthe first two criteria, LaKL21 introduces the mini-mum amount of third-order dispersion in the lasercavity. This fact was determined by calculating themagnitude of third-order dispersion in the laser,given a prism separation that compensates for boththe second-order dispersion of the laser rod at800 nm and the self-phase-modulation present inthe focal region. These calculations were done byusing Sellmeier coefficients tabulated in the Schottcatalog of optical glasses. To compare the magni-tude of third-order dispersion in the new laser designwith that of a standard Ti:sapphire laser (assuminga 2.2-cm Ti:sapphire rod and SF10 glass prismsand n = 1.71125, dn/dA = -0.04958 Am-', d2n/dA2 = 0.1755 Lm 2 , and d3n/dA3 = -0.9972 /tmM3
Fig. 1. Wavelength spectrum of the laser output from theundispersed end of the laser. The FWHM is 42.5 nm,with a center wavelength of 817 nm.
phase with respect to frequency (a3o/acw3 ) of-4.8 X 10-42 s3 for the standard setup for a singlepass through the cavity. The magnitude of thisterm is proportional to the quadratic chirp intro-duced by the prism pair to the laser. The use ofLaFN28 glass with the same 2.2-cm crystal reducesa34/,1aw to -2.2 x 10-42 s3.5 However, reduction ofthe crystal length to 9 mm and use of LaKL21 glassreduces alo/la( 3 to -8.9 x 10-43 s3. The time errorarising from the cubic phase (d#") will be -1/24/"Aw
2,
where Awc is the laser bandwidth, assumed to betransform limited. Assuming that this numberrepresents the fixed time error tolerated by the cav-ity, and that the pulse width is limited by this effect,one predicts a square-root dependence of pulsewidth on third-order dispersion. We can thus ex-trapolate from the 28-fs results. to determine thebandwidth possible in the new cavity configura-tion-use of the LaKL21 glass with a short crystalshould result in a reduction of the observed pulsewidth from 28 to 18 fs. This is close to what weobserve experimentally. Since our calculations donot consider the nonlinear index of refraction, it canbe concluded that the presence of linear cubicdispersion is the primary factor that determinesthe minimum pulse duration from mode-lockedTi:sapphire lasers. Recent theoretical research haspredicted the same scaling dependence of pulsewidth on cubic dispersion.' 0 At a prism separationof 51 cm, we calculate a second derivative of phasewith respect to frequency of -2.1 X 10-23 s2 for asingle pass through the cavity (excluding the contri-bution of self-phase-modulation, which we have notdetermined).
Figure 1 shows the spectrum of the output of thislaser through the spectrally undispersed end of thelaser. These data taken using a Jarrel-AshMonospec 50 monochromator and a silicon photo-detector (United Detector PIN-1OD). The spectrumFWHM is 42.5 nm.
Since the pulses from the laser disperse as theypass through the substrate of the output coupler[3/8-in. (9.5-mm) fused silica] and the autocorrelatorbeam splitter, external compression is necessary toregain the original pulse width. We used a double-
pass prism compressor arrangement using BK7prisms 121 cm apart. Figure 2 shows a representa-tive autocorrelation trace, along with a fit to theautocorrelation of a sech2 pulse. The autocorrela-tion trace was taken using a slow-scanning autocor-relator with one arm driven by an Oriel CorporationEncoder Mike. This fit yields an-autocorrelationFWHM of 26.1 fs, corresponding to a pulse width of16.9 fs, with an estimated total error of <3%. Weestimate that the extracavity prism separation isjust sufficient to compensate for the output couplerand the autocorrelator optics; thus it appears thatthere is little chirp on the pulse at the point where itexits the cavity. A compensating plate was used inone arm of the autocorrelator so that equal glasspath lengths in both arms could be maintained. Areflective focusing optic, a 100-,um-thick KDP crys-tal, and noncollinear phase matching were used forthis measurement.
The time-bandwidth product of the laser output isthus 8v8r = 0.323, compared with the ideal value of0.315 for a sech2 pulse. Thus the time-bandwidthproduct is within -3% of the value expected for atransform-limited sech2 pulse shape. However, thespectrum does not exactly correspond to that of asech' pulse and is asymmetric. The spectral shapefits best to a Gaussian at long wavelengths; on theshort-wavelength side there is missing energy in thespectrum (a gap also roughly Gaussian in shapewith a width of -20 nm). This missing energy maybe due to absorption in the laser cavity. The fit tothe long-wavelength side of the spectrum corre-sponds to that of an 18.4-fs FWHM Gaussian pulse,which would have an autocorrelation width of 26 fs.However, the autocorrelation does not fit so closelyto a Gaussian shape as to a sech2 .
The output power of the laser was 370 mW fromthis end of the laser and 130 mW from the dispersedend, resulting in a total output power of 500 mW for5 W of pump power. This output power could alsobe obtained by using a 10% output coupler at oneend, with a high reflector at the prism side. Theoutput mode was primarily a round TEMoo mode,with <5% light in higher-order modes. The short-term laser output power duplicated intensity fluctu-
-40 -20 0 20 40Time (fs)
Fig. 2. Autocorrelation trace of the 16.9-fs pulse(solid curve). Also shown is a sech2 (t) fit to the data(dotted curve).
Fig. 3. Wavelength spectrum of the laser output from thedispersed (prism) end of the laser. The FWHM is 48 nm.
ations in the argon-ion laser (the spectra of Figs. 1and 3 are representative of stability over an -5-minperiod). The long-term stability of the laser waslimited by the beam-pointing stability of the argonlaser. However, after -1 h of warm-up time, thelaser would run for >3 h without adjustment, eventhough the argon-ion laser lacks active beam stabi-lization. This readjustment consists of optimizingthe argon-ion output, then readjusting the end mir-rors of the Ti:sapphire cavity for maximum outputpower. Once the prism configuration is optimizedfor wide-bandwidth operation, the laser consistentlygenerates pulses of bandwidth >40 nm, and thepulses can always be recompressed to near thetransform limit. Thus the laser is always opera-tional at sub-20-fs pulse widths. As is usual withthis type of laser, it does not self-start in a pulsedmode when lasing is interrupted by a substantialmechanical shock.
Figure 3 shows the spectrum of the laser outputfrom the dispersed (prism) end of the cavity. Thisspectrum is somewhat wider than from the undis-persed end, with a FWHM of 48 nm. This wouldcorrespond to a transform-limited pulse of <15 fs,assuming a sech2 shape. Experimentally, we wereonly able to measure pulses of 17-fs duration afterrecompression. We suspect that our autocorrela-tion measurement may be limited by the geometryused (noncollinear), the 100-gm KDP crystal, orthird-order dispersion in the extracavity prisms.Thus the laser may be producing sub-15-fs pulses, orthe different spectra observed may be due to differ-ent pulse shapes. Further research is under way toimprove our measurement resolution. Other recentexperimental and theoretical research has also beenreported on the possibility of different pulse widthsfrom different ends of the same laser cavity."
In conclusion, we have demonstrated a simple andreliable laser capable of generating pulses of un-
precedented short duration. Our results confirmthe fact that linear third-order dispersion is the pri-mary factor limiting pulse duration in these lasers.Further improvement in pulse duration is likelygiven that this third-order dispersion is readilycomputable.
Given the stability and high repetition rate of thislaser, high signal-to-noise measurements with un-precedented time resolution should now be possible.The broad bandwidth of these laser pulses shouldalso find application as a high-repetition-rate white-light source for ultrafast studies in materials. Formany applications, this laser should be a desirable(i.e., much lower cost and much higher stability andpulse repetition rate) substitute for ultrashort lasersystems based on amplification, continuum genera-tion, and pulse compression.'2
The authors acknowledge useful conversationswith Wayne Knox and Colin Seaton. We also thankDennis Garvey for help with the laser diagnostics.This project was supported by National ScienceFoundation grants ECS-9204033 and PHY-9158373,by the U.S. Department of the Air Force, by fundsfrom Washington State University, by generous dona-tions from Tektronix, Inc., Union Carbide, Thorlabs,National Instruments, and Newport Corporation,and through a collaboration with Lawrence Liver-more National Laboratory (project W-7405-ENG-48).
References
1. D. E. Spence, P. N. Kean, and W Sibbett, Opt. Lett.16, 42 (1991).
2. K. Naganuma and K. Mogi, Opt. Lett. 16, 738 (1991).3. J. P. Likforman, G. Grillon, M. Joffre, C. L. Blanc,
A. Migus, and A. Antonetti, Appl. Phys. Lett. 58, 2061(1991).
4. U. Keller, G. W 'tHooft, W H. Knox, and J. E.Cunningham, Opt. Lett. 16, 1022 (1991).
5. C.-P. Huang, H. C. Kapteyn, J. W McIntosh, and M. M.Murnane, Opt. Lett. 17, 139 (1992).
6. F Krausz, C. Spielmann, T. Brabec, E. Wintner, andA. J. Schmidt, Opt. Lett. 17, 204 (1992).
7. J. M. Jacobson, A. G. Jacobson, K. Naganuma, H. A.Haus, and J. G. Fujimoto, in Digest of Conference onLasers and Electro-Optics (Optical Society of Amer-ica, Washington, D.C., 1992), paper CTuU2.
8. B. E. Lemoff and C. P. Barty, in Digest of Conferenceon Lasers and Electro-Optics (Optical Society ofAmerica, Washington, D.C., 1992), paper CDP11.
9. P. F. Moulton, J. Opt. Soc. Am. B 3, 125 (1986).10. H. A. Haus, J. Jacobson, and J. G. Fujimoto, "Effect of
third-order dispersion on passive mode locking," sub-mitted to Opt. Lett.
11. T. Brabec, C. Spielmann, and F Krausz, Opt. Lett. 16,1961 (1991).
12. P. C. Becker, H. L. Fragnito, R. L. Fork, F A. Beisser,and C. V Shank, Appl. Phys. Lett. 54, 411 (1989).
![Untitled-2 [ ] ?· FS 78 FS 68 , FOCUS ÉkJ ËFOCUS FS 78 FS 68 FS 68 , , , FS 68 Foundation FS 68 ,…](https://static.documents.pub/doc/80x56/5b75f9b67f8b9a3b7e8b5e04/untitled-2-fs-78-fs-68-focus-ekj-efocus-fs-78-fs-68-fs-68-fs-68.jpg)
