762

OPTICS LETTERS / Vol. 30, No. 7 / April 1, 2005

Widely tunable optical parametric generation in a photonic crystal ber

A. Y. H. Chen, G. K. L. Wong, S. G. Murdoch, R. Leonhardt, and J. D. Harvey
Department of Physics, University of Auckland, Private Bag 92019, Auckland, New Zealand

J. C. Knight, W. J. Wadsworth, and P. St. J. Russell

Optoelectronics Group, Department of Physics, University of Bath, Claverton Down, Bath BA2 7AY, UK Received September 21, 2004 We report on the observation of widely tunable optical parametric generation in a photonic crystal ber. The frequency shift of the generated sidebands that arise from modulational instability is strongly dependent on the detuning of the pump from the bers zero-dispersion wavelength. We are able to demonstrate experimentally more than 450 nm of sideband tunability as we tune the pump wavelength over 10 nm. Excellent agreement has been found between the experimentally measured and theoretically predicted shifts. 2005 Optical Society of America OCIS codes: 060.2320, 060.4370, 190.4380.

Modulational instability (MI) refers to the phenomenon by which an initially continuous wave spontaneously breaks up into a periodic wave train as it propagates though a nonlinear dispersive medium. In the spectral domain this effect is characterized by the appearance of a pair of sidebands, one on either side of the pump frequency. It is well known that MI can occur in a single-mode optical ber when a strong pump wave propagates in the anomalous-dispersion regime.1 Furthermore, it was recently shown that near a bers zero-dispersion wavelength MI can also be observed over a range of wavelengths that extends into the normal-dispersion regime. This fact has permitted the observation of MI when pumping in the normal-dispersion regime of a photonic crystal ber (PCF).2 The effect has also been investigated in the context of four-wave mixing for pumping about the zero-dispersion wavelength of standard single-mode bers,3 and has been proposed as a useful form of nonlinear frequency conversion, with discrete tunability of the MI sidebands possible by varying the bers core diameter.4 We show how one can use the limited pump tunability about the zero-dispersion wavelength to obtain a pair of continuously tunable MI sidebands, tunable over a wide wavelength range. To do this we extend the standard MI analysis to allow us to predict the sidebands gain spectrum for a ber with arbitrary dispersion. This analysis shows that for a large air-lling fraction PCF of the type described in Ref. 2, two distinct sets of MI sidebands exist. Near the zero-dispersion wavelength we nd that tuning the pump wavelength over 10 nm results in a tunable range of more than 450 nm for the inner set of sidebands. As a mechanism for frequency conversion, MI has several advantages, one of the most important being the nonresonant nature of the Kerr nonlinearity of an optical ber. This means that, provided that a suitable pump is available, the technique is applicable to any region of the optical frequency spectrum. Thus it is possible to construct widely tunable optical
0146-9592/05/070762-3/$15.00

sources in regions that previously were difcult to reach. In addition, the high optical nonlinear interaction coefcient of a small-core diameter PCF raises the intriguing possibility of a continuous-wave version of this source. The interaction of the pump and the Stokes (frequency downshifted) and anti-Stokes (frequency upshifted) sidebands in a single polarization optical ber with an arbitrary dispersion prole can be described by the three coupled-mode equations presented in Ref. 5: i dAp dz + Ap 2 + 2 As 2 + 2 Aa 2 Ap
L

+ 2 AaAsAp* exp i i dAa,s dz +

z = 0,

1a

Aa,s 2 + 2 As,a 2 + 2 Ap 2 Aa,s

L

2 + ApAs,a* exp i

z = 0.

1b

Here = n2 0 / cAeff, where Aeff is the effective mode area of the ber, and is the linear wave-vector L mismatch of the MI process:
L

is the linear propagation constant of the ber and contains contributions from the bers material and waveguide dispersions. As is such a strongly varying function of wavelength in a PCF, we prefer to express it as an arbitrary function of frequency rather than using the traditional Taylor expansion. For an undepleted pump and two weak sidebands it is simple to show that the gain spectrum of the MI sidebands as a function of , their frequency detuning from the pump, can be written as g = Im
L L

+4 P

1/2

The peak MI gain is 2 P, which occurs at a frequency detuning that satises the following relation:
2005 Optical Society of America

April 1, 2005 / Vol. 30, No. 7 / OPTICS LETTERS

763

Fig. 1. MI phase-matching diagram for a silica strand with core diameter 1.556 m and pump power 50 W. Inset, Close up of the sideband wavelengths over the experimental range of interest.
L

+ 2 P = 0.

can be accurately approxiFor a ber for which mated by a second-order Taylor expansion, L reduces to 2 2, and we recover the standard expression for the gain spectrum of the MI sidebands.1 The PCF used in the research reported here consists of a central hexagonal silica rod of 1.5- m diameter held in place by a honeycomb of ne silica struts. For a large air-lling fraction PCF of this kind the dispersive characteristics of the ber can be approximated by those of a bare silica strand. This allows us to calculate the dispersion of the ber from the analytic expression for the waveguide dispersion of a circular step-index waveguide and from knowledge of the material dispersion of bulk fused silica. For a silica strand of 1.556- m diameter, the zerodispersion wavelength of the HE11 mode is shifted from 1.3 m to 650 nm because of the large waveguide dispersion of such a small-diameter ber. In Fig. 1 we plot the wavelength of the MI sidebands calculated from Eq. (4) for a silica strand of 1.556- m diameter and an input pump power of 50 W. From Fig. 1 we can see that MI sidebands occur for all pump wavelengths from 600 nm to 1.6 m. At all pump wavelengths between these two extremes there are two frequency shifts that result in phase-matched MI sidebands. Phase-matching curves of this type have been predicted to occur in dispersion-attened bers also.6,7 Experimentally we were able to observe only the inner set of sidebands. We believe that the reason for this is the short interaction length for the large-frequency-shift outer sidebands as well as the high ber loss in the UV and IR regions where these sidebands would occur. At a pump wavelength of 650 nm the walk-off between the outer Stokes and anti-Stokes sidebands is more than 900 ps/ m. For the 20-ps pump pulse used in our experiment this large walk-off limits the interaction length to only a few centimeters. From the point of view of nonlinear frequency conversion it is the re-

gion about the zero-dispersion wavelength (indicated in Fig. 1 by the dotted rectangle) that is of most interest to us. A close-up of this region is shown in the inset of Fig. 1. It clearly shows that for wavelengths just below the zero-dispersion wavelength, 10 nm of pump tunability results in a tunable sideband range of more than 450 nm. For 50 nm of pump tunability the tunable sideband range is more than 1200 nm and covers almost the entire visible and near-IR regions of the optical spectrum. In addition to the frequency shift of the MI sidebands it is useful to consider the gain bandwidth of the MI process as a function of pump wavelength. Figure 2 is a plot of the half-gain bandwidth of the inner sidebands as a function of pump wavelength for a pump power of 50 W. It shows that near the zero-dispersion wavelength the sidebands have a signicant bandwidth, which decreases rapidly as we move into the normaldispersion regime. To observe MI experimentally in a PCF we use a synchronously pumped mode-locked cavity-dumped DCM-dye laser as our pump source. An intracavity Lyot lter allows us to tune the wavelength of this laser from 600 to 700 nm. The pulse width of the laser is approximately 20 ps, and the peak power is 300 W. Pulses from the dye laser were coupled into a 1.3-m length of the large air lling fraction PCF described above (ber A). In common with many PCFs, this ber exhibited signicant residual birefringence, so a half-wave plate was placed in front of the ber to allow us to pump on either the high or the low groupindex mode. The light exiting the ber was collimated by a 60 microscope objective and then sent either to an optical spectrum analyzer or to a prism that allowed us to spatially separate the pump and sidebands. Thus we could verify that all three waves were propagating in the fundamental HE11 transverse mode, which had an approximately Gaussian mode prole. Figure 3 shows the observed MI sideband shifts as a function of pump wavelength for a pump polarized parallel to both the high and low

Fig. 2. Sideband half-gain bandwidth as a function of pump wavelength near the zero-dispersion wavelength of a silica strand with core diameter 1.556 m and pump power 50 W.

764

OPTICS LETTERS / Vol. 30, No. 7 / April 1, 2005

group-index modes. The peak power of the pump pulses coupled into the ber was 60 W. Superimposed on top of the experimentally measured shifts are the predictions of Eq. (4). The only free parameter in this t is the effective core diameter of the silica strand, which we found to be 1.537 m for the high group-index mode and 1.556 m for the low group-index mode. The agreement between the experimental and theoretical shifts is extremely good, giving a strong conrmation of the validity of the silica strand model. In addition, Fig. 3 gives a clear experimental verication of our claim that 10 nm of pump tunability near the zero-dispersion wavelength will result in more than 450 nm of sideband tunability. The experimentally observed spectra at the four pump wavelengths 639, 641.5, 644.3, and 660.5 nm (pump polarized parallel to the high group-index mode) are shown in the inset of Fig. 3. They show behavior in qualitative agreement with the predictions of Fig. 2. For pumping well inside the normaldispersion regime we observe sharp MI sidebands, with the widths of these sidebands increasing as we tune the pump toward the zero-dispersion wavelength. The conversion efciency is modest (1%, dropping to 0.1% for the large-frequency-shift sidebands) owing to the different group velocities of the sidebands and the pump. We believe that much higher conversion efciencies and a wider tuning range would have been possible were it not for our relatively short-duration 20-ps pump pulses, which result in a rapid walk-off between the pump and the sidebands. A longer ber length and longer pump pulses, or even a continuous-wave pump, would remove this limitation. From the same ber preform that we used to draw the PCF described above (ber A) we were able to draw two more PCFs (bers B and C), each with a

Fig. 4. Experimentally measured sideband wavelengths as a function of pump wavelength for the high group-index mode (circles) and the low group-index mode (squares) of bers A, B, and C. The pump power used was 60 W. Solid curves are the theoretical sideband wavelengths predicted by Eq. (4).

successively larger core diameter. In Fig. 4 we plot the observed sideband shifts for the high and low group-index modes of all three bers. The peak power of the pump pulses coupled into each ber was 60 W. The experimentally measured shifts and the theoretical predictions are in excellent agreement for all three bers. Thus we can determine that for the high and low group-index modes of ber A the effective core diameters were 1.537 and 1.556 m; for ber B, 1.650 and 1.668 m; and for ber C, 1.786 and 1.796 m. In conclusion, we have demonstrated that modulational instability about the zero-dispersion wavelength of a photonic crystal ber offers the possibility of widely tunable nonlinear frequency conversion. Experimentally we observed a 450-nm sideband-tunable range when we tuned the pump wavelength over 10 nm. Our results dramatically conrm the validity of the silica strand approximation as a model for the dispersive properties of a large air lling fraction photonic crystal ber. S. G. Murdochs e-mail s.murdoch@auckland.ac.nz. References
1. G. P. Agrawal, Nonlinear Fiber Optics, 3rd ed., Optics and Photonics Series (Academic, San Diego, Calif., 2001). 2. J. D. Harvey, R. Leonhardt, S. Coen, G. K. L. Wong, J. C. Knight, W. J. Wadsworth, and P. St. J. Russell, Opt. Lett. 28, 2225 (2003). 3. C. Lin, W. A. Reed, A. D. Pearson, and H. T. Shang, Opt. Lett. 6, 493 (1981). 4. C. Lin, W. A. Reed, A. D. Pearson, H. T. Shang, and P. F. Glodis, Electron. Lett. 18, 87 (1982). 5. G. Cappellini and S. Trillo, J. Opt. Soc. Am. B 8, 824 (1991). 6. M. Yu, C. J. McKinstrie, and G. P. Agrawal, Phys. Rev. E 52, 1072 (1995). 7. R. A. Sammut and S. J. Garth, J. Opt. Soc. Am. B 6, 1732 (1989).

address

is

Fig. 3. Experimentally measured sideband wavelengths as a function of pump wavelength for the high and the low group-index modes of ber A. The pump power used was 60 W. Solid curves, theoretical sideband wavelengths predicted by Eq. (4) for silica strands with effective core diameters of 1.537 m (high group-index mode) and 1.556 m (low group-index mode). Inset, spectra of light exiting the ber for pump wavelengths (i) 639, (ii) 641.5, (iii) 644.3, and (iv) 660.5 nm (pump polarized parallel to the high group-index mode).