Emergence of tripoles in nonlinearly perturbed planar vortices: a numerical study

L. A. BARBA
Department of Mathematics, University of Bristol, Bristol U.K. BS84TW Summary Since the observation of spontaneous generation of vorticity concentrations from random 2D turbulence, a huge interest has been stimulated on so-called coherent structures. The tripole is a much less common one than the monopole and dipole in numerical and experimental observations. It was rst observed in both domains in the mid-late 1980s, and only in 1991 was a tripolar structure detected in the oceans. Laboratory tripoles occur as a result of the growth of a perturbation of azimuthal wavenumber 2 in unstable axisymmetric shielded vortices (a vortex surrounded by a ring of opposite sign vorticity). For this reason, shielded monopoles have also been the subject of several numerical studies. In the present study, in contrast, the tripole is seen to emerge from a non-shielded, Gaussian monopole (which is stable), with a large nonlinear quadrupolar perturbation (which destabilizes it). In this ow, the tripole was rst observed by Rossi et al. [7]. We extend the result by performing a parameter study, spanning values of the amplitude of perturbation and Reynolds number, with numerous simulations using a meshless vortex method that was introduced by Barba et al.[1]. One of the goals is to determine whether there is a threshold amplitude that separates two asymptotic states (axisymmetric and tripolar). The possible relationship between this threshold amplitude and Reynolds number is sought, and several observations are made regarding the nonlinear, long-time evolution of the structure. Initial condition. The ow under study is initiated by a Gaussian vortex with a superposed m = 2 (quadrupolar) perturbation. The initial condition is given by o (x) = 1 exp 4 |x|2 4 , (x) = |x|2 exp 4 |x|2 4 cos 2, (1)

where o stands for the base vorticity, for the perturbation, and = arg x. The main case discussed in Rossi et al.[7] corresponds to = 0.25 and Re = 104 , with Re = / (total circulation divided by the viscosity). Figure 1 shows the initial vorticity for this case, (c), as well as plots of the base vortex, (a), the perturbation, (b), and the tripole obtained after the ow self-organizes for approximately 5 turn over times, (d). Threshold for emergence of the tripole. In Figure 2, the logarithm of the vorticity is plotted in a grey colormap, thus accentuating the region of the domain where the vorticity changes sign. It can be seen that the self-organization of the ow consists in the zero-contour pinching, leaving the satellites of negative vorticity in place. For the case of a smaller initial

t=0
6 6

t=0
6 6

t = 800

6 6

6 6

6 6

6 6

(a) Base vortex

(b) Perturbation

(c) Initial condition.

(d) t = 800

Figure 1: A perturbed Gaussian monopole self-organizes to a quasi-steady rotating tripole. This simulation: Re = 104 , = 0.25. In (b), the dotted lines represent negative contours.

Figure 2: Plot of the logarithm of ||, emphasizing the level zero of vorticity. Black in the
grey scale is saturated at level 106 . This simulation: Re = 104 , = 0.25. Left to right: t = 0, 300, 500, 600, 800.

Figure 3: Same as Figure 2, but this simulation: Re = 104 , = 0.10. perturbation, the plot of log || shows that the zero-contour of vorticity folds at onto itself, not allowing negative satellites (Figure 3). The ow in this case will tend to axisymmetry. In between the two cases discussed above, there will be one value of the initial perturbation amplitude for which the zero-contour of vorticity just pinches, leaving very small satellites that are promptly diused away by viscosity. By performing numerous simulations with dierent values of for each of ve values of Reynolds number, Re = 500, 103 , 3 103 , 104 , 3 104 , we nd this critical initial state, separating the tripole and the monopole solutions. The critical initial amplitude varies approximately as Re1/3 in the middle of this range, but clearly the slope lessens as the Reynolds number increases. See Figure 4. This

suggests that there might be an inviscid limit, where the ow always tends to the tripole solution above a threshold perturbation amplitude. Le Diz`s[3] performed an asymptotic study e of non-axisymmetric vortices, and he predicted a Re2/3 behavior for the amplitude of the nonRe2/3 slope axisymmetric correction. This amplitude, however, refers to the streamfunction of the nal state. Thus, it is not linearly related to the parameter that we have used, corresponding to the initial non-axisymmetric component (which is very much controlled by our type of initial 10 condition). 1/3
1

Re

slope

Stability of the tripole. Once formed, the tripole is very robust and can live for a very long Figure 4: Threshold amplitude of the initial time. For example, with = 0.25, the satellites perturbation, , versus Reynolds number: for survive for 6 turn-over times for Re = 3 103 , pinching of zero-contour (solid line), or for and more than 25 turn-over times for Re = 104 half-turn or full-turn satellite survival (dash- (after the pinching of the level of zero vorticity). dot lines above). To test the robustness of the structure, we take the output of one simulation at t = 800 and introduce a random perturbation to the location of the vortex particles (the computational elements in the method), consisting in a random walk with step /2. Performing a continuation run from this perturbed state, one nds that the ow relaxes back to the unperturbed tripole solution. Figure 5 shows the contours of the perturbation, , for = 0 and = 0.1 which is of the order of the initial inter-particle spacing (a resolution parameter of the numerical method). After about 5 turn-over times, there is practically no trace of the random disturbance.
10
3

10

t = 800
6 6 4 2 0 2 4 6 6 6

t = 800
4 2 0 2 4 6 6 6

t = 1600
4 2 0 2 4 6 6 6

t = 1600
4 2 0 2 4 6

(a)

=0

(b)

= 0.1

(c)

=0

(d)

= 0.1

Figure 5: Perturbation vorticity, , for tripole continuation runs with perturbed output; Re = 104 , = 0.25; (a) and (b) initial conditions; (c) and (d) after 5 turn-over times. Contour levels [0.04 : 0.02 : 0.14] (higher contours not present due to decay of perturbation vorticity).

Time scales of the self-organization and relaxation. In the previous work of Rossi et al. [7], the authors were preoccupied that they were unable to detect shear-diusion time scales. Shear-enhanced diusion homogenizes perturbations on a time scale of O(Re1/3 ); this was proved for the case of passive scalars in [6] and for weak vorticity perturbations in [4, 2]. The results for vorticity, however, rely on the assumption that perturbations become rapidly varying in the radial direction (due to dierential rotation), thus cancelling the coupling of streamfunction and vorticity at leading order. This in fact occurs only for suciently small perturbations, and thus when the perturbation is large there is no reason to expect an O(Re1/3 ) time scale. By plotting adequate diagnostics of the decay of the tripolar structure, we nd that indeed an O(Re1/3 ) time scale can be extracted for small, but for larger values the scaling does not work. One can, however, educe quite well the mixing time given from the theory of Meunier and Villermaux[5], when looking at the decay of the positive part of the perturbation. This is the time when the perturbation just starts to decay, and in the theory it follows an error function behavior. It is suggested that mixing processes are active in the early stages of the relaxation, after which nonlinear eects become dominant. These results cannot be shown here due to lack of space, but will be shown in the talk. In conclusion, this is a parametric study of the ow produced by superposing a quadrupole and a Gaussian monopole, spanning a range of Reynolds number and amplitude of the quadrupole. It was partly motivated by previous work suggesting a threshold amplitude separating the domains of attraction of the monopole and the tripole. This threshold has been determined quite precisely for several Reynolds numbers, and it is suggested that an inviscid limit might exist. Furthermore, it was observed that the tripole is very robust, quickly recovering from a random disturbance, and it decays slowly due to diusion.

References
[1] L. A. Barba, A. Leonard, and C. B. Allen. Advances in viscous vortex methods meshless spatial adaption based on radial basis function interpolation. Int. J. Num. Meth. Fluids, 47(5):387421, 2005. [2] A. J. Berno and J. F. Lingevitch. Rapid relaxation of an axisymmetric vortex. Phys. Fluids, 6(11):37173723, 1994. e [3] S. Le Diz`s. Non-axisymmetric vortices in two-dimensional ows. J. Fluid Mech., 406:175198, 2000. [4] T. S. Lundgren. Strained spiral vortex model for turbulent ne structures. Phys. Fluids, 25(12):21932203, 1982. [5] P. Meunier and E. Villermaux. How vortices mix. J. Fluid Mech., 476:213222, 2003. [6] P. B. Rhines and W. R. Young. How rapidly is a passive scalar mixed within closed streamlines? J. Fluid Mech., 133:133145, 1983. [7] L. F. Rossi, J. F. Lingevitch, and A. J. Berno. Quasi-steady monopole and tripole attractors for relaxing vortices. Phys. Fluids, 9(8):23292338, 1997.