August 1, 1996 / Vol. 21, No.

15 / OPTICS LETTERS

1129

Observation of vortex solitons created by the instability of dark soliton stripes

Vladimir Tikhonenko, Jason Christou, Barry Luther-Davies, and Yuri S. Kivshar
Laser Physics Center, The Australian National University, Canberra ACT 0200, Australia Received February 6, 1996 The first experimental observation to the authors knowledge is reported of the generation of optical vortex solitons from instability and breakup of a dark soliton stripe propagating through a saturable self-defocusing nonlinear medium. 1996 Optical Society of America

Nonlinear phenomena, including beam self-focusing and the stable propagation of solitons, have been subjects of considerable theoretical and experimental research in physics. One of the most remarkable effects produced by nonlinearity is the development of instabilities that can lead to the eventual formation of stable nonlinear structures (see, e.g., Ref. 1), allowing us to investigate the strongest manifestation of a physical systems nonlinear properties. Transverse instability is seen in all types of quasi (1 1 1)-dimensional soliton stripes (plane solitons), as it is a generic property of various nonlinear modes.2 For the case of bright solitons, instability arises from the action of selffocusing, leading to a modulational instability of the soliton stripe along its length. Transverse instability of dark solitons3 is different. As a result of the defocusing nonlinearity, a decrease in the size of a dark solitons intensity dip (increased grayness) increases the solitons transverse velocity. Amplitude or phase noise along the stripe, resulting in local variations of the soliton grayness, will thus impart a nonuniform transverse velocity along the stripes length. This leads to instability of the plane dark soliton and its eventual breakup into stable vortex solitons, an effect that we have seen experimentally, for first the time to our knowledge, in a self-defocusing saturable nonlinear medium. Dark solitons have been observed experimentally in a bulk nonlinear medium as dark stripes or grids with properties similar to those of the (1 1 1)-dimensional dark solitons.4 However, the linear stability analysis developed for a defocusing Kerr medium5 indicates that a plane dark soliton is unstable to transverse long-wavelength modulations. Numerical simulations6,7 and asymptotic analytical theory8 have been used to demonstrate that such an instability leads to the generation of pairs of optical vortex solitons with alternate polarities. These optical vortex solitons are stable structures that were predicted theoretically9 and recently observed experimentally.10,11 In this Letter we report what is to our knowledge the f irst experimental observation of the instability of dark soliton stripes and demonstrate that such an instability is an effective mechanism for the creation of optical vortex solitons, as previously predicted by theory for a Kerr medium.6 8 In accordance with our recent results12 regarding the theory of dark soliton instabilities in saturable self-defocusing media, we discuss two pos0146-9592/96/151129-03$10.00/0

sible mechanisms involved in the development of this instability. Experiments were performed with a continuous wave Ti:sapphire laser and a nonlinear medium composed of atomic-rubidium vapor. The laser output was a linearly polarized slightly elliptical Gaussian beam with a wavelength tuned close to the rubidium atoms resonance line at 780 nm. An expanding telescope imaged the beam waist onto a phase mask, which imposed the p phase jump on the prof ile of the beam. The mask was constructed from a quartz substrate with the phase discontinuity generated by a straight-edged step prof ile of appropriate depth. A demagnifying telescope then imaged the plane of the mask onto the input face of a cell containing the nonlinear medium. The 50-mm-diameter 200-mmlong cylindrical Pyrex cell contained rubidium vapor only, providing a saturated vapor concentration up to 1013 cm23 . The output window of the cell was then imaged onto a screen, the resulting intensity pattern being recorded by a CCD camera. We obtained resonant enhancement of nonlinearity by tuning the laser frequency close to the rubidium atom D2 line. The resonance lines were inhomogeneously broadened with Doppler width 0.75 GHz at the highest cell temperatures. Hyperfine splitting of the 85 Rb ground state was substantially greater than the inhomogeneous line width, allowing the hyperf ine 3 2 5P3/2 F 2, 3, 4 to be used transitions 5S1/2 F as the working resonance, with red detunings producing a self-defocusing nonlinearity. We could manipulate the strength of the nonlinearity by varying vapor concentration, using cell temperature adjustments. The beam power at the input face of the cell was 240 mW with a 1 e2 waist of 0.3 mm. Detunings ranged from 0.4 to 1.0 GHz, which, from previous experiments,13 gives a maximum nonlinear refractive index of 1024 . A sequence of output prof iles, shown in Fig. 1, was captured for a range of cell temperatures, with detuning f ixed at approximately 0.85 GHz. For vanishingly small vapor concentration the beam propagates linearly through the medium, leading to the output intensity prof ile of Fig. 1(a). With increasing concentration (i.e., nonlinearity), the launched prof ile was observed to form a vertically uniform narrow dark soliton stripe [Fig. 1(b)]. Further increase in concentration led to a destruction of the vertical uniformity of
1996 Optical Society of America

1130

OPTICS LETTERS / Vol. 21, No. 15 / August 1, 1996

jump) caused the f inal stages in evolution of the instability, wherein optical vortex soliton pairs are created, to disappear. Only the snakelike bending of the soliton stripe was observed, corresponding to the predicted lower growth rate of instability for gray soliton stripes.5,12 The stripes average transverse velocity then carried it away from the beam center, where nonlinear effects were rapidly suppressed by the increased absorption and lower background intensity. Similar behavior was observed for a pair of gray soliton stripes launched with an even mask. To describe numerically the dynamics of dark soliton breakup, we used the standard model for the slowly varying envelope E of the electric field E E exp 2ik0 n0 z , leading to the well-known generalized nonlinear Schr dinger equation o 2i 1 E D E 1 k0 nNL I E 1 z 2k0 n0 ig I E , (1)

where D is the transverse Laplacian, k0 is the freespace wave number, n0 is the linear refractive index, and I jEj2 is the beam intensity. We estimated real nNL I and imaginary g I parts of the intensitydependent refractive index, using 1 , ns Is1 1 2 nNL I 1 1 I Is1 p gI gs 1 1 I Is2
p

(2)

Fig. 1. Output beam intensity prof iles demonstrating the instability of a dark soliton stripe as the nonlinearity is increased. The vapor concentration increases from vanishingly small in (a) to the order of 1013 cm23 in (f ). The cell temperatures (C) were (a) 40, ( b) 72, (c) 82, (d ) 90, (e) 112, and (f ) 125. The inset shows the intensity distribution along a line through the center of the upper vortex of (f ). Laser detuning was held constant at 0.85 GHz, and the power of the beam at the cell input was 240 mW, corresponding to a maximum intensity of approximately 170 W cm2 .

the stripe [Fig. 1(c)]. As concentration was further increased, the breakup of the stripe began, initially appearing as progressive, snakelike bending [Fig. 1(d )], then as breaking, with the field coalescing into dark spots around points of inf lection in the bend. Regions connecting the stripe and the dark spots were no longer black, thus destroying continuity of the stripe across the background [Fig. 1(e)]. Separation between the dark spots and the broken stripe became more defined as the self-defocusing nonlinearity peaked, with the dark spot assuming a circular symmetry consistent with the predicted formation of the vortex soliton [Fig. 1(f )]. The process of breakup was found to depend on the depth of the phase step in the input beam distribution. Misalignment of the phase mask from optimum orientation (providing the closest to a p phase

1026 cm2 W , Is1 with the parameters ns 1 cm21 , Is2 0.01 W cm2 , and 10 W cm2 , gs p 0.5 corresponding to a laser frequency detuning of 0.85 GHz. (See Ref. 13 for methods used to estimate nonlinear parameters.) Numerical simulations of the model of Eqs. (1) and (2) were performed by a 2 1 1 dimensional split-step beam propagation method. The experimentally recorded input prof ile, shown in Fig. 2(g), with a p phase jump imposed over its central intensity dip, was propagated on a 1024 3 1024 mesh for 20 cm, corresponding to the length of the nonlinear medium. This initial condition was renormalized such that its computed power matched the measured power (240 mW) of the experimental input prof ile. A good correlation with experimental results was found [see Figs. 2(a) 2(f )]. The calculated output intensity distributions revealed the same dynamics as observed experimentally for increasing vapor concentration, from the initial stage of the dark soliton stripe formation through to its perturbation, bending, and f inal breakup into pairs of optical vortex solitons. The beam spreading, power depletion, and instability growth rates seen in the experiments appear to be well approximated by those achieved in the simulations. However, the period of transverse perturbation corresponding to the maximum growth rate was smaller by a factor of 1.5 2 than that seen in experiment, as can be estimated from the characteristic period of bending or the number of vortex pairs. These discrepancies are most likely attributable to (i) the physically complicated nonlinear response of rubidium vapor, for which a more comprehensive yet unwieldy model might prove effective, and (ii) the difficulty in obtaining sufficiently

August 1, 1996 / Vol. 21, No. 15 / OPTICS LETTERS

1131

the soliton velocity (grayness parameter); v 0 corresponds to a black soliton. Analysis of the growth rate for the absorptionless saturation model [Eq. (2)] shows that transverse instability vanishes for smaller values of the saturation intensity Is1 (or larger intensities of the input beam). However, in the regime of strong saturation a black soliton always becomes unstable because of its purely one-dimensional instability, with this instability leading to a transformation of the black soliton v 0 into a moving (gray) soliton of lower intensity (v 0). This kind of instability was recently revealed in Ref. 14 for the particular case p 2 in the model of Eq. (2), but we found that this instability exists for any p, and the corresponding growth rate depends strongly on the slope of the function nNL I . In experiments we have excited the dark soliton stripe within a background beam of a varying (Gaussian) prof ile. Near the beam center, where the amplitude is largest, we expect growth of the one-dimensional instability, whereas lower amplitudes off center f it the region of a standard transverse instability. Thus it seems that the phenomenon of dark soliton breakup observed experimentally in a strongly saturable medium can be understood as a combined effect of two physical mechanisms, the transverse (or two-dimensional) instability and the one-dimensional instability of a dark soliton stripe. In conclusion, we have presented what we believe is the f irst experimental demonstration of vortex soliton generation by means of the instability of dark soliton stripes in a self-defocusing saturable medium. References
Fig. 2. Results for numerical simulations of dark soliton instability, as modeled by Eqs. (1) and (2). Output profiles (a) (f ) were obtained after 20 cm of propagation in the nonlinear medium with nonlinear refraction and absorption parameters corresponding to the experimental range in Figs. 1(a) 1(f ). Inset (g) shows the experimentally obtained intensity distribution at the input of the medium (23 magnif ication with respect to other prof iles) used as the initial condition for these simulations. Other parameters correspond to the experimental situation of Fig. 1. 1. E. Infeld and G. Rowlands, Nonlinear Waves, Solitons and Chaos (Cambridge U. Press, Cambridge, 1990). 2. M. J. Ablowitz and H. Segur, Solitons and the Inverse Scattering Transform (Society for Industrial and Applied Mathematics, Philadelphia, Pa., 1981), Chap. 3.8. 3. See, for example, Yu. S. Kivshar, IEEE J. Quantum Electron. 29, 250 (1993). 4. G. A. Swartzlander, Jr., D. R. Andersen, J. J. Regan, H. Yin, and A. E. Kaplan, Phys. Rev. Lett. 66, 1583 (1991). 5. E. A. Kuznetsov and S. K. Turitsyn, Zh. Eksp. Teor. Fiz. 94, 119 (1988) [Sov. Phys. JETP 67, 1583 (1988)]. 6. G. S. McDonald, K. S. Syed, and W. J. Firth, Opt. Commun. 94, 469 (1992). 7. C. T. Law and G. A. Swartzlander, Jr., Opt. Lett. 18, 586 (1993). 8. D. E. Pelinovsky, Yu. A. Stepanyants, and Yu. S. Kivshar, Phys. Rev. E 51, 5016 (1995). 9. A. W. Snyder, L. Poladian, and D. J. Mitchell, Opt. Lett. 17, 789 (1992). 10. G. A. Swartzlander, Jr., and C. T. Law, Phys. Rev. Lett. 69, 2503 (1992). 11. B. Luther-Davies, R. Powles, and V. Tikhonenko, Opt. Lett. 19, 1816 (1994). 12. Yu. S. Kivshar, D. E. Pelinovsky, J. Christou, B. Luther-Davies, and V. Tikhonenko, Inf luence of nonlinearity saturation on the instabilities of dark solitons, to be submitted to Phys. Rev. E. 13. V. Tikhonenko, J. Christou, and B. Luther-Davies, J. Opt. Soc. Am. B 12, 2046 (1995). 14. Yu. S. Kivshar and W. Krolikowksi, Opt. Lett. 20, 1527 (1995).

accurate measures of the initial f ield, in combination with high sensitivity of the simulated dynamics to small changes in the initial conditions, a sensitivity not seen in experiments. This suggests that the action of some stabilizing physical mechanisms (e.g., diffusion) may need to be incorporated into our model for nonlinear response. We have undertaken further analytic and numerical investigation of soliton stripe instabilities, using a saturating nonlinear model.12 Our results indicate that, unlike in the self-focusing case, saturation of nonlinearity leads to a strong modif ication in the mechanism, leading to instability. First, the maximum growth rate q of the transverse instability of a dark soliton stripe in a non-Kerr medium is modif ied to become C dPs dv v 0 , where C is a positive coefficient q calculated through invariants of the generalized nonlinear Schr dinger equation, Ps is the renormalized o momentum (def ined, for example, in Ref. 14), and v is