Hostname: page-component-7b7567c8c-tnkh4 Total loading time: 0 Render date: 2026-11-17T01:19:47.894Z Has data issue: false hasContentIssue false

Bifurcations in a quasi-two-dimensional Kolmogorov-like flow

Published online by Cambridge University Press:  12 September 2017

Jeffrey Tithof*
Affiliation:
Center for Nonlinear Science, School of Physics, Georgia Institute of Technology, Atlanta, GA 30332-0430, USA
Balachandra Suri
Affiliation:
Center for Nonlinear Science, School of Physics, Georgia Institute of Technology, Atlanta, GA 30332-0430, USA
Ravi Kumar Pallantla
Affiliation:
Center for Nonlinear Science, School of Physics, Georgia Institute of Technology, Atlanta, GA 30332-0430, USA
Roman O. Grigoriev
Affiliation:
Center for Nonlinear Science, School of Physics, Georgia Institute of Technology, Atlanta, GA 30332-0430, USA
Michael F. Schatz
Affiliation:
Center for Nonlinear Science, School of Physics, Georgia Institute of Technology, Atlanta, GA 30332-0430, USA
*
†Email address for correspondence: jtithof@gatech.edu

Abstract

We present a combined experimental and theoretical study of the primary and secondary instabilities in a Kolmogorov-like flow. The experiment uses electromagnetic forcing with an approximately sinusoidal spatial profile to drive a quasi-two-dimensional (Q2D) shear flow in a thin layer of electrolyte suspended on a thin lubricating layer of a dielectric fluid. Theoretical analysis is based on a two-dimensional (2D) model (Suri et al., Phys. Fluids, vol. 26 (5), 2014, 053601), derived from first principles by depth-averaging the full three-dimensional Navier–Stokes equations. As the strength of the forcing is increased, the Q2D flow in the experiment undergoes a series of bifurcations, which is compared with results from direct numerical simulations of the 2D model. The effects of confinement and the forcing profile are studied by performing simulations that assume spatial periodicity and strictly sinusoidal forcing, as well as simulations with realistic no-slip boundary conditions and an experimentally validated forcing profile. We find that only the simulation subject to physical no-slip boundary conditions and a realistic forcing profile provides close, quantitative agreement with the experiment. Our analysis offers additional validation of the 2D model as well as a demonstration of the importance of properly modelling the forcing and boundary conditions.

Information

Type
Papers
Copyright
© 2017 Cambridge University Press 
Figure 0

Figure 1.A schematic diagram of the two-immiscible-layer experimental set-up for generating Kolmogorov-like flow viewed (a) from above and (b) from the side. The vectors $\boldsymbol{J}$, $\boldsymbol{B}$ and $\boldsymbol{F}$ denote, respectively, the directions of the electric current, the magnetic field and the resulting Lorentz force. The flow is bounded by two endwalls, two sidewalls (electrodes) and a no-slip bottom surface, while the top surface is a free electrolyte–air interface. This container is mounted on an aluminium plate that is levelled and submerged in a water bath that is temperature-regulated such that the electrolyte is maintained at $23.0\pm 0.2\,^{\circ }\text{C}$.

Figure 1

Figure 2.The $z$-component of the magnetic field $B_{z}$ (a) at the longitudinal centre of the domain ($x=0$) and (b) along the magnet centrelines at$y=\pm \{0.5,1.5,2.5,3.5,4.5,5.5,6.5\}$. In (a), the experimental measurements at a height $z=0.265$ (just above the dielectric–electrolyte interface) and at $z=0.438$ (just below the electrolyte free surface) are shown, respectively, as open squares and filled circles. In (b), the black symbols indicate the experimental measurements at a height $z=0.438$ along the magnet centrelines. A least-squares fit has been performed using the data in (a) to determine the scaling factor for the dipole summation; the scaled dipole summation magnetic field is shown as the red full lines. The experimental uncertainties are the size of the symbols or smaller.

Figure 2

Figure 3.Straight flow fields at $Re=8.1$ with $\unicode[STIX]{x1D6FC}=0.064~\text{s}^{-1}$, $\unicode[STIX]{x1D6FD}=0.83$ and $\unicode[STIX]{x1D708}=3.26\times 10^{-6}~\text{m}^{2}~\text{s}^{-1}$ for the (a) DPS, (b) SPS, (c) NPS and (d) experiment. The dashed lines in (d) indicate the locations of velocity profiles in the experiment that are compared to the simulations. The vorticity colour scale plotted for (a) also applies to (b–d). The velocity vectors are downsampled in each direction by a factor of 8 for the simulations and 4 for the experiment.

Figure 3

Figure 4.Profiles of the longitudinal velocity and longitudinal velocity differences at $Re=8.1$ with $\unicode[STIX]{x1D6FC}=0.064~\text{s}^{-1}$, $\unicode[STIX]{x1D6FD}=0.83$ and $\unicode[STIX]{x1D708}=3.26\times 10^{-6}~\text{m}^{2}~\text{s}^{-1}$. (a) Plot of $u_{x}^{exp}$ as a function of $y$ at the longitudinal centre ($x=0$). (b) The difference between the longitudinal velocity in the simulations and the experiment, $u_{x}^{sim}-u_{x}^{exp}$, as a function of $y$ at the longitudinal centre ($x=0$). Note that the curves corresponding to the DPS and SPS are virtually indistinguishable. (c) Plot of $u_{x}^{exp}$ as a function of $x$ at the centreline of a middle magnet ($y=-0.5$). (d) The difference between the longitudinal velocity of the simulations and the experiment, $u_{x}^{sim}-u_{x}^{exp}$, as a function of $x$ at the centreline of a middle magnet ($y=-0.5$). Note that the curves corresponding to the DPS and SPS are virtually indistinguishable in the region $-4, where the DPS is defined. Experimental uncertainties are the size of the symbols or smaller.

Figure 4

Figure 5.Neutral stability curves (4.1) describing the primary instability. The blue dashed line corresponds to $\unicode[STIX]{x1D6FC}=0.064~\text{s}^{-1}$ and $\unicode[STIX]{x1D6FD}=0.83$, while the green dot-dashed line corresponds to $\unicode[STIX]{x1D6FC}=0.064~\text{s}^{-1}$ and $\unicode[STIX]{x1D6FD}=1.0$. The measurement from the experiment (NPS) is plotted as a black dot (red square); note that the uncertainties in $Re_{c}$ are smaller than the size of the symbols. In all the cases, $\unicode[STIX]{x1D708}=3.26\times 10^{-6}~\text{m}^{2}~\text{s}^{-1}$ is held constant.

Figure 5

Figure 6.Modulated flow fields at $Re=14$ with $\unicode[STIX]{x1D6FC}=0.064~\text{s}^{-1}$,$\unicode[STIX]{x1D6FD}=0.83$ and $\unicode[STIX]{x1D708}=3.26\times 10^{-6}~\text{m}^{2}~\text{s}^{-1}$ for the (a) DPS, (b) SPS, (c) NPS and (d) experiment. The vorticity colour scale plotted for (a) also applies to (b–d). The velocity vectors are downsampled in each direction by a factor of 8 for the simulations and a factor of 4 for the experiment.

Figure 6

Figure 7.Primary instability for $\unicode[STIX]{x1D6FC}=0.064~\text{s}^{-1}$, $\unicode[STIX]{x1D6FD}=0.83$ and $\unicode[STIX]{x1D708}=3.26\times 10^{-6}~\text{m}^{2}~\text{s}^{-1}$. (a) A bifurcation diagram and (b) the average wavelength of the pattern, $\bar{\unicode[STIX]{x1D706}}_{x}$, as a function of $Re$ for the modulated flow regime. At each $Re$, wavelength measurements are made in the central region $|y|\leqslant 4$, then averaged; the uncertainty bars indicate one standard deviation in the spatial measurements.

Figure 7

Figure 8.The effect of the variation in model parameters. (a) A bifurcation diagram and (b) the average wavelength of the pattern in the modulated flow regime. The numerical results correspond to either a 7 % increase in $\unicode[STIX]{x1D708}$, a 22 % increase in $\unicode[STIX]{x1D6FC}$, or a 6 % decrease in $\unicode[STIX]{x1D6FD}$ compared with the depth-averaged values for the straight flow ($\unicode[STIX]{x1D6FC}=0.064~\text{s}^{-1}$, $\unicode[STIX]{x1D6FD}=0.83$ and $\unicode[STIX]{x1D708}=3.26\times 10^{-6}~\text{m}^{2}~\text{s}^{-1}$). The uncertainty bars in (b) are only shown for every other data point for clarity. In all cases, $Re$ is defined using $\unicode[STIX]{x1D708}=3.26\times 10^{-6}~\text{m}^{2}~\text{s}^{-1}$.

Figure 8

Table 1.Critical transition parameters characterizing the stable periodic regime for the experiment and the NPS with different sets of parameters.

Figure 9

Figure 9.A schematic showing the bifurcations corresponding to the primary instability: (a) circle pitchfork in the DPS, (b) sequence of pitchfork bifurcations in the SPS, (c) imperfect pitchfork bifurcation in the NPS. Solid (dashed) lines indicate stable (unstable) solution branches. The vertical and out-of-plane axes correspond to deviations of the flow from straight that are invariant under ${\mathcal{R}}_{x}{\mathcal{R}}_{y}$ and ${\mathcal{R}}_{x}{\mathcal{T}}_{y}^{w}$, respectively.

Figure 10

Figure 10.Modulated flow fields (a) $\boldsymbol{u}_{m}^{1}$, (b) $\boldsymbol{u}_{m}^{3}$, (c) $\boldsymbol{u}_{m}^{2}$ and (d) $\boldsymbol{u}_{m}^{4}$ at $Re=14$ in the DPS. The vorticity colour scale is the same as that in figure 6.

Figure 11

Figure 11.Modulated flow fields (a) $\boldsymbol{u}_{m}^{1}$, (b) $\boldsymbol{u}_{m}^{3}$, (c) $\boldsymbol{u}_{m}^{2}$ and (d) $\boldsymbol{u}_{m}^{4}$ at $Re=14$ in the SPS. Vertical black lines indicate the central region, which is analogous to the flow fields shown in figure 10. The vorticity colour scale is the same as that of figure 6.

Figure 12

Figure 12.Modulated flow fields at $Re=11.6$ in the NPS beyond the imperfect pitchfork bifurcation shown in figure 9(c). The flow fields shown here are (a) $\boldsymbol{u}_{m}^{1}$, which emerges smoothly from the straight flow, and (b) $\boldsymbol{u}_{m}^{2}$, which is formed through a saddle-node bifurcation.

Figure 13

Figure 13.Sensitivity to the magnetic field profile for $\unicode[STIX]{x1D6FC}=0.078~\text{s}^{-1}$, $\unicode[STIX]{x1D6FD}=0.83$ and $\unicode[STIX]{x1D708}=3.26\times 10^{-6}~\text{m}^{2}~\text{s}^{-1}$. (a) A bifurcation diagram for the primary instability and (b) the average wavelength of the pattern in the modulated regime. The simulations were performed with $\unicode[STIX]{x1D6FC}$ increased by 22 % relative to the depth-averaged value for the straight flow and used either $B_{da}$, $B_{b}$, or $B_{t}$.

Tithof et al. supplementary movie

A side-by-side animation comparing the time-periodic flows observed in the experiment and the NPS (with depth-averaged parameters). In each case, the Reynolds number was chosen above the onset of the secondary instability so that the oscillations are clearly visible.

Download Tithof et al. supplementary movie(Video)
Video 24.1 MB