Generation of tunable infrared picosecond pulses at 100 MHz bydifference-frequency mixing in KTiOPO4
Herman VanherzeeleCentral Research & Development Department, E. I. du Pont de Nemours & Company, Inc., Experimental Station, P.O. Box 80356,
Wilmington, Delaware 19880-0356
Received January 27, 1989; accepted April 11, 1989
Picosecond pulses, tunable in the wavelength range of 1.2-4.5 ,um, are generated at a 100-MHz repetition rate inKTiOPO 4 by difference-frequency mixing the output of a synchronously mode-locked dye laser and a Nd:YLF laserthat pumps the dye laser. The typical average infrared output power is in the milliwatt range, which represents animprovement of several orders of magnitude over previously reported results using other nonlinear materials in asimilar setup with Nd:YAG as the driving source. The high average power of this infrared source and its broadtunability make it an attractive alternative for mode-locked color-center lasers.
The generation of continuous trains at high repetitionrates (typically 100 MHz) of tunable IR picosecondpulses by difference-frequency mixing the output of asynchronously mode-locked dye laser and its pumpsource (e.g., a cw mode-locked Nd:YAG or Ar-ion la-ser) is a well-established technique.'-4 The high repe-tition rate allows low-signal lock-in detection and sig-nal-averaging schemes, and the short IR pulse dura-tions permit time-resolved studies. As a result, thistechnique can be used for many applications, includ-ing, e.g., (nonlinear) pulse propagation in optical fi-bers and carrier dynamical studies in semiconductors.Compared with other cw mode-locked laser sources inthe IR, such as color-center lasers, difference-frequen-cy mixing offers a large tuning range, depending on thenonlinear material that is used as the mixing crystal.Because of the inherently low peak power availablefrom both cw input sources, the mixer should have alarge nonlinearity. Typically LiNbO3 and LiIO3 areused for IR radiation tunable from 1.2 to 4.5 Am.However, the conversion efficiency that can be ob-tained in these materials is limited by crystal damage.As a result the average IR power is very low: <0.5 AWin the picosecond domain, and one order of magnitudelarger in the femtosecond regime. It is the purpose ofthis Letter to demonstrate that significantly higheroutput power can be obtained in the range of 1.2-4.5Am by using KTiOPO4 (KTP) (Ref. 5) as the mixingcrystal in an improved laser system.
In contrast to all previous research for which eithermode-locked Ar-ion or Nd:YAG lasers were used asthe high-repetition-rate driving source, our system isbased on a Nd:LiYF4 (Nd:YLF) laser. The advan-tages of Nd:YLF over Nd:YAG are manifold, as wasdemonstrated recently.6 The average mode-lockedoutput power is 18 W, with a typical pulse duration of40 psec (FWHM). Only 8 W of 1.053-Am radiation isused for generating approximately 2 W of green light(in KTP) to pump a synchronously mode-locked dyelaser. The dye laser is tunable with a birefringentfilter from 0.560 to 1.03 Am (using three sets of mirrorsand five different dyes). Over this wavelength range
the output power is 150-250 mW, with pulse durationsof 2-4 psec. The dye laser output is collinearly mixedwith the remaining 10-W beam of the Nd:YLF laser inKTP.
There are many key features that make KTP anexcellent material for nonlinear-optics applications,and for difference-frequency mixing in particular. Acomprehensive review of the superior properties ofKTP has been published recently7 and is not repeatedin this Letter. From the Sellmeier equations in Ref. 8for hydrothermally grown KTP, we have calculatedthe type II phase-matching angles for frequency-dif-ference mixing between the tunable dye laser and the1.053-yim Nd:YLF output (k2'1 - X1-1 = AX3) forcollinear propagation in the x-z plane (where the non-linear interaction is maximum). Henceforth, sub-scripts 1, 2, and 3 refer to the Nd:YLF, the dye, and theIR beams, respectively. The resulting tuning curvesare shown in Fig. 1. The lower curves represent thedye-laser wavelength, and the upper curves representthe IR wavelength. The IR upper limit (4.5 Aim) isdictated by the onset of absorption in KTP. The solid
Fig. 1. Type II phase-matching curves for difference-fre-quency mixing in KTP for collinear propagation in the x-zplane. The circles represent the experimental data.
curves in Fig. 1 correspond to the calculated values,
while the circles represent the experimental data.The excellent mutual agreement validates the Sell-meier equations in Ref. 8. Two interactions are possi-ble, denoted (A) and (B) in Fig. 1. For scheme (A) the
1.053-gm beam is an extraordinary beam and the gen-erated IR beam is an ordinary beam; for scheme (B)the opposite is true. In both schemes the dye laser isan ordinary beam. The experimental data are ob-tained with two KTP crystals. For phase-matchingscheme (A) we use a crystal cleaved and polished alongthe natural (201) faces. The normal to the entranceface is in the x-z plane and makes an angle of 58.8°with the z axis. On the other hand, for scheme (B) a
crystal cleaved and polished along the natural (100)plane is used. Each configuration has its particularadvantages. The preferred one therefore depends ona number of practical considerations. Clearly, scheme(A) offers a larger tunability than scheme (B). For a
dye laser tunable from 560 to 853 nm, scheme (A)covers a broad IR spectrum: 1.2-4.5 gm. For scheme
(B) the upper limit of the tuning range is 2.3 ,um. Inscheme (A) both input beams have a different polar-ization. Therefore one can use a polarizer as the
beam-combining element, which is in general moreconvenient than a dichroic mirror. On the otherhand, because of the larger phase-matching angles in
scheme (B), this one has a larger efficiency (larger deff
and smaller walkoff). Moreover, group-velocity mis-match between the IR and dye-laser pulses, whichusually is the dominant pulse-broadening mechanism,can be compensated by crystal birefringence only in
scheme (B). This is an important consideration if thetechnique is applied in the femtosecond regime (seebelow).
In the small-gain regime, the expressions for sum
and difference mixing are identical. If we neglectdouble-refraction effects for the Gaussian inputbeams and assume perfect phase matching, the IRpeak power is given by (in mks units)
P = 16rdeff 212 PlP2 (1)
E0 cX32nln 2 n 3 (wl2 + w22 )
where 1 is the crystal length, w is the beam waist, and Pis the peak power. If we assume that the pulse widthsof beams 2 and 3 are the same (see below), Eq. (1)predicts an average IR power in the milliwatt range fora 1-mm KTP crystal in which our input beams arefocused to a spot size of approximately 30 gim. Ex-perimental results have confirmed this prediction. Asa specific example, let us consider a dye-laser input at610 nm (X3 = 1.45 gim) using phase-matching scheme(B). The phase-matching angle for this particularconfiguration is 78.80. The spot size (HWI/e2M) of
the Nd:YLF beam and the dye-laser beam in the sam-ple is 32 and 18 gim, respectively. The calculated
walkoff angle p for the IR beam is 20 mrad, leading to
an aperture length (la - w;r/p) of approximately 1.6mm, which is somewhat shorter than the actual crys-tal length that we use (2 mm). For average input
powers of 10 W (Nd:YLF) and 240 mW (dye) we ob-
tain an average IR power of 7 mW (data not correctedfor Fresnel losses). This is approximately three or-
ders of magnitude larger than the results reported inRefs. 1 and 2 and clearly reveals the potential formaking this IR source competitive with color-centerlasers. From the observation of a linear dependenceof P3 with respect to both Pi and P2 [as predicted byEq. (1)], it was verified that we still are in the small-gain regime. Hence a further increase of IR outputcan be expected by increasing the peak power of ei-ther one of the input beams. Optical damage in KTPdoes not present a problem for this application. Wemeasured a damage threshold well over 10 GW/cm 2
for picosecond pulses.8
The temporal characteristics of the IR pulse trainwere also examined. Obviously, the jitter of the dye-
laser pulse train is much smaller than the Nd:YLFpulse duration. Therefore the temporal profile ofthe IR pulses should (at least in principle) be nearlyidentical to that of the dye-laser pulses (since thelatter are much shorter than the Nd:YLF pulses).However, as the pulses propagate through the crystal,group-velocity mismatch between the dye-laser pulseand the IR pulse will in general lengthen the latter.This broadening can be estimated from the differencein transit time in the crystal,
At l[(9k/,w) 3 - (ak/&) 2].
In practice, a At of 50% of the dye-laser pulse width isstill acceptable and will lead to only a 10% broadeningof the IR pulse. From the Sellmeier equations in Ref.8 we calculated At for both interactions over their
respective tuning ranges. The results (Fig. 2) clearlyindicate the possibility of using KTP for the genera-tion of IR pulses in the femtosecond regime. Over thewhole tuning range of phase-matching scheme (B), IAtI
< 100 fsec/mm. In particular, for the interaction dis-cussed above (X3 = 1.45 ,im), tlti = 40 fsec for a 1-mmKTP crystal. Thus in the wavelength range of 1.2-2.3gim one should use scheme (B). To generate wave-lengths longer than 2.3 gim, one has to use scheme (A)since phase matching is not possible for scheme (B).In this case, lAti < 100 fsec/mm only for wavelengthsgreater than 3 gnm.
For diagnostic purposes, the IR beam can be upcon-verted in KTP either by second-harmonic generationor by sum-frequency mixing (cross correlation) with
2.0
QWc)'A
cI
1.0
0.0
- 1.0
- 2.0
3.0
4.01.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
Wavelength (Mm)
Fig. 2. Group-velocity mismatch between the dye-laserpulse and the IR pulse in KTP as a function of wavelengthfor both phase-matching schemes shown in Fig. 1.
Fig. 3. Type II phase-matching curve for IR upconversionwith the dye laser for collinear propagation in the x-z plane.The upper curve represents the dye-laser signal, and thelower curve represents the upconverted signal by sum-fre-quency mixing the dye and IR wavelengths shown in Fig. 1.The circles represent the experimental data.
the remaining dye-laser beam (X2-1 + X3-1 = Xs-l).The phase-matching curve for cross correlation isshown in Fig. 3. The upper curve represents the dye-laser wavelength, and the lower curve represents thesum-frequency signal. The corresponding IR wave-length has been omitted for clarity. The solid curvesare calculated from the Sellmeier equations, while thecircles represent the experimental data obtained witha KTP crystal cleaved along the (201) natural faces.Cross-correlation measurements of the (picosecond)dye-laser pulses with the generated IR pulses haveconfirmed the absence of pulse broadening within thelimit of the experimental resolution (- 200 fsec).
Further enhancement in IR output power can bereadily obtained in several ways. First, by adding afiber/grating pulse compressor for the 1.053-,gm pulsetrain, the average IR power can be boosted by a factorof 2-3. Addition of a pulse compressor for the dye-laser pump allows one to extend the technique wellinto the femtosecond regime, without losing the tun-ability. By the same token, the IR peak power willincrease correspondingly. As another possibility, thedye laser can be cavity dumped (at megahertz repeti-tion rates) to increase the IR peak power further. Fi-nally, improvements are expected by using KTiOAsO4(KTA) instead of KTP. The linear optical propertiesof KTA are similar to those of KTP, but KTA has a60% larger deff.9 Experiments to demonstrate theseimprovements are currently under way in our labora-tory.
In its present state, our system generates picosecondpulses tunable from 0.56 to 4.5 gim (with a gap from 1to 1.2 gim at the high end of the dye-laser tuningrange). Amplification (at a lower repetition rate) overthis entire range is possible in a KTP parametric am-plifier, pumped at either 1.053 or 0.526 ,im. For this
purpose we have added a Nd:YLF regenerative ampli-fier to the system. The output of the regenerativeamplifier, which is seeded by the nonconverted 1.053-,gm pulse train from the KTP doubler, is hardly syn-chronized with both the dye-laser and IR pulse trains.A single-pass power gain of the order of 105-106 in a 2-cm KTP crystal has been demonstrated in this config-uration.7"l0 A more detailed description of this part ofour laser system and its performance will be publishedin a future paper.
In conclusion, we have demonstrated efficient dif-ference-frequency mixing at high repetition rates inKTP using a Nd:YLF laser system. The large nonlin-earity, which is temperature insensitive, as well as thelarge tuning range combined with a high damagethreshold make KTP a superior material for this ap-plication. The use of KTP, combined with the higherpeak power from Nd:YLF (compared with Nd:YAG),leads to average IR power in the milliwatt range. Thisis several orders of magnitude larger than in previous-ly reported data with a similar system using a Nd:YAGlaser and LiIO3 as the mixing crystal. Additional im-provements have been outlined in this Letter to makethis IR source competitive with color-center lasers inthe picosecond and femtosecond time domains. Thepossibility of strong pulse amplification at a lowerrepetition rate in a KTP parametric amplifier adds tothe attractiveness of our system.
This research was presented at the 1988 AnnualMeeting of the Optical Society of America in SantaClara, California." I thank J. Bierlein for providingthe KTP samples, J. Kelly for valuable technical assis-tance during the course of this research, and G. Mer-edith for supporting this research.
References
1. D. Cotter and K. I. White, Opt. Commun. 49,205 (1984).2. L. Reekie, I. S. Ruddock, and R. Illingworth, Opt. Quan-
tum Electron. 8, 169 (1984).3. A. Mokhtari, L. Fini, and J. Chesnoy, Opt. Commun. 61,
421 (1987).4. A. G. Yodh, H. W. K. Tom, and G. D. Aumiller, in Digest
of Conference on Lasers and Electro-Optics (OpticalSociety of America, Washington, D.C., 1988), paperTHX5.
5. F. C. Zumsteg, J. D. Bierlein, and T. E. Gier, J. Appl.Phys. 47, 4980 (1976).
6. H. Vanherzeele, Appl. Opt. 27, 3608 (1988).7. J. D. Bierlein and H. Vanherzeele, J. Opt. Soc. Am. B 6,
622 (1989).8. H. Vanherzeele, J. D. Bierlein, and F. C. Zumsteg, Appl.
Opt. 27, 3314 (1988).9. J. D. Bierlein, H. Vanherzeele, and A. A. Ballman, Appl.
Phys. Lett. 54, 783 (1989).10. H. Vanherzeele, "Recent advances in the generation of
picosecond tunable infrared radiation," Proc. Soc. Pho-to-Opt. Instrum. Eng. (to be published).
11. H. Vanherzeele, in Digest of Annual Meeting of theOptical Society of America (Optical Society of America,Washington, D.C., 1988), paper TuN3.
