Hostname: page-component-7b7567c8c-tnkh4 Total loading time: 0 Render date: 2026-11-17T15:47:00.849Z Has data issue: false hasContentIssue false

Instability of radially spreading extensional flows. Part 2. Theoretical analysis

Published online by Cambridge University Press:  25 October 2019

Roiy Sayag*
Affiliation:
Department of Environmental Physics, Blaustein Institutes for Desert Research, Ben-Gurion University of the Negev, Sede Boqer Campus, 8499000, IsraelDepartment of Mechanical Engineering, Ben-Gurion University of the Negev, Beer-Sheva 8410501, Israel
M. Grae Worster
Affiliation:
Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Cambridge CB3 0WA, UK
*
†Email address for correspondence: roiy@bgu.ac.il

Abstract

The interface of a strain-rate-softening fluid that displaces a low-viscosity fluid in a circular geometry with negligible drag can develop finger-like patterns separated by regions in which the fluid appears to be torn apart. Such patterns were observed and explored experimentally in Part 1 using polymeric solutions. They do not occur when the viscosity of the displacing fluid is constant, or when the displacing fluid has no-slip conditions along its boundaries. We investigate theoretically the formation of tongues at the interface of an axisymmetric initial state. We show that finger-like patterns can emerge when circular interfaces of strain-rate-softening fluids displace low-viscosity fluids between stress-free boundaries. The instability, which is fundamentally different from the classical Saffman–Taylor viscous fingering, is driven by the tension that builds up along the circular front of the propagating fluid. That destabilising tension is a geometrical consequence and is present independently of the nonlinear properties of the fluid. Shear stresses stabilise the growth either along extended circumferential streamlines or through a street of vortices. However, such stabilising processes become weaker, thereby allowing the instability to develop, the more strain-rate-softening the fluid is. The theoretical model that we present predicts the main experimental observations made in Part 1. In particular, the patterns we predict using linear-stability theory are consistent with the strongly nonlinear experimental patterns. Our model depends on a single dimensionless number representing the power-law exponent, which implies that the instability we describe could arise in any extensional flow of strain-rate-softening material, ranging from suspensions that rupture in squeeze experiments to rifts formed in ice shelves.

Information

Type
JFM Papers
Copyright
© 2019 Cambridge University Press 
Figure 0

Figure 1.Selected plan view snapshots from a laboratory experiment with constant source flux ($Q=2.64~\text{gm}~\text{cm}^{-3}$), showing polymer solution (blue) displacing a denser salt solution (transparent) in a circular geometry (Sayag & Worster 2019).

Figure 1

Figure 2.Diagram of the flow geometry considered in the mathematical model.

Figure 2

Figure 3.The base state and a perturbed state for $n=6$, $R_{0}=2$. (a) The base state is axisymmetric with radially growing viscosity $\unicode[STIX]{x1D707}_{0}$ (colour). (b) The perturbation flow for $k=3$, showing the perturbation viscosity (colour), velocity field (arrows) and streamlines (blue and red curves), where ${\mathcal{G}}=0.1$. In both panels (, brown) marks the inner boundary $r_{G}$, and (, green) marks the leading front $r_{N}$.

Figure 3

Figure 4.(a) Results of the stability analysis, showing the neutral curves (colour) of the set of wavenumbers $k=1,2,3,4,5,10,20,30,50,100,200$ as a function of the power-law exponent $n$ and of the normalised thickness of the annular layer of fluid $\unicode[STIX]{x1D6FF}/(\unicode[STIX]{x03C0}/2)$. The unstable region is always above the neutral curves. The envelope of these neutral curve is marked by (blue colour). When more modes are included (e.g. figure 5) that envelope converges to a global neutral curve, which is flatter at the left edge (smaller $\unicode[STIX]{x1D6FF}$). (b) The same modal neutral curves presented in panel (a), but as a function of the dimensionless quantity $K\equiv k\unicode[STIX]{x1D6FF}$ converge with the growth of $k$ to a universal neutral curve near $K\sim \unicode[STIX]{x03C0}/2$. Modes with $n<1$ are always stable and therefore not shown.

Figure 4

Figure 5.Results of the stability analysis, showing the most-unstable wavenumbers $k$ (a) and the corresponding growth rates (b) as a function of the power-law exponent $n$ and of the thickness of the annular layer of fluid $\unicode[STIX]{x1D6FF}/(\unicode[STIX]{x03C0}/2)$, among a denser (but not complete) set of wavenumbers than that shown in figure 4. Therefore, the envelope of these neutral curves () (blue colour) is more similar to the global neutral curve than the one shown in figure 4. Modes with $n<1$ are stable and therefore not shown.

Figure 5

Figure 6.One-wavelength sections of the two-dimensional flow for $k=4,8,12$ ($R_{0}=1.23$) and for $n=0.3$ (a) and $n=3$ (b), showing the perturbation velocity field ($\rightarrow$), streamlines (, red (clockwise) and , blue (counterclockwise) curves), perturbed front ( green) and viscosity field (colour). All modes are stable apart from the $k=8,n=3$ mode. The specific growth rates are (from low to high wavenumbers) ${\mathcal{G}}\approx -0.54,-0.53,-0.55$ ($n=0.3$) and ${\mathcal{G}}\approx -0.03,+0.1,-0.41$ ($n=3$). Streamlines absolute values: 0.002, 0.016, 0.064, 0.256, 0.512.

Figure 6

Figure 7.The growth rate for $n=1$ as a function of the width of the fluid layer normalised by $\unicode[STIX]{x03C0}/2$ for $k=2,3,4,10$, evaluated by the analytic solution (3.22) (colour), and validated numerically by the full model solution (). The base-state rate-of-strain contribution to the growth rate $-1/R_{0}^{2}$ is marked by ().

Figure 7

Table 1.Asymptotic solutions and distinguished limits considered.

Figure 8

Figure 8.Comparison of the predicted unstable modes between the full perturbation model (a,b) and the thin-film model (c,d), showing the most-unstable wavenumbers (a,c) and the corresponding growth rate (b,d) as a function of $n$ and the annulus width $\unicode[STIX]{x1D6FF}/(\unicode[STIX]{x03C0}/2)$, among $k=10,20,30,50,100,200$. In all panels,  (green colour) mark the global neutral curve (${\mathcal{G}}=0$) predicted by the thin-film model for that specific set of wavenumbers. The details of the secondary flow for $n=6$ and $\unicode[STIX]{x1D6FF}=(\unicode[STIX]{x03C0}/2)/100$ ($+$ red colour) are shown in figure 9.

Figure 9

Figure 9.One-wavelength sections of the two-dimensional flow for $\unicode[STIX]{x1D6FF}/(\unicode[STIX]{x03C0}/2)=1/100$ and $n=6$, comparing the perturbation flow (arrows and streamlines) and viscosity (colour) obtained by the thin-film approximation (a) and by the full solution (b), for the wavenumbers $k=50,100,150$. The perturbed front is marked with  (green colour). The values of $\unicode[STIX]{x1D6FF}$ and $n$ correspond to the $+$ (red colour) in figure 8. The specific growth rates are (from low to high wavenumbers) ${\mathcal{G}}\approx -0.14,+0.40,-0.32$ (thin film) and ${\mathcal{G}}\approx -0.12,+0.42,-0.27$ (full model). Streamlines absolute values: 0.0005, 0.002, 0.008, 0.032, 0.128, 0.512.

Figure 10

Figure 10.(a) Neutral curves (colour) predicted by the thin-film approximation for a set of wavenumbers $k=10,20,30,50,100,200$ as a function of the power-law exponent $n$ and of the normalised thickness of the annular layer of fluid $\unicode[STIX]{x1D6FF}/(\unicode[STIX]{x03C0}/2)$. The unstable region is always above the neutral cures. (b) The same neutral curves as in panel (a), but as a function of the dimensionless wavenumber $K\equiv k\unicode[STIX]{x1D6FF}$ collapse into a universal curve. The value of $K$ that corresponds to the maximum growth rate for each $n$ is marked with ( blue colour).

Figure 11

Figure 11.The instantaneous most-unstable wavenumber for $n=3$ as a function of the thickness of the fluid annulus $\unicode[STIX]{x1D6FF}$, evaluated by the full model solution (, grey), and by the thin-film and $n\gg 1$ limit (, red) according to the round value of the wavenumber predicted by (3.38).

Figure 12

Figure 12.The growth rate at the thin-film approximation and the limit $n\ll 1$ (- - -) compared with the complete thin-film solution (——), for $n$ ranging between $1/100$ and $9/10$.

Figure 13

Figure 13.(a) The leading front of the base state (, grey) is under tension due to non-zero normal stress in the $\hat{\unicode[STIX]{x1D703}}\hat{\unicode[STIX]{x1D703}}$ direction, while the other stress components are zero. (b) The base-state normal stress contributes to tangential stress at the perturbed interface (, green). Therefore, the condition of zero tangential stress at the perturbed front implies a perturbation shear stress at the $\hat{r}\hat{\unicode[STIX]{x1D703}}$ direction of the base front that is proportional to the slope of the perturbed front. (c) The secondary flow following a geometric perturbation ($\downarrow$ green colour) consists of two distinct patterns through which momentum dissipates and stabilises the perturbation. At $K\lesssim \unicode[STIX]{x03C0}/2$, the streamlines stretch circumferentially from frontal depressions to frontal bulges through a relatively long path along which energy is dissipated. At $K\gtrsim \unicode[STIX]{x03C0}/2$, vortices form near the inner boundary and screen the interior from the front, so that less fluid is available to sustain the growth of tongues. The volume covered with vortices grows with $K$.

Figure 14

Figure 14.The base-flow tension at the leading front is represented by the hoop stress as a function of the fluid exponent.

Figure 15

Figure 15.The growth rate in the thin-film limit (solid lines) as a function of the effective wavenumber $K$ for fluid exponents $n=0.5,1,5$ (colour). The asymptotic behaviour of the growth rate in the small-wavenumber limit $(K\ll 1)$ is independent of $n$ to leading order (dash line). In the large-wavenumber limit $(K\gg 1)$ the growth rate has a leading-order dependence on the fluid exponent (dash–dot).

Figure 16

Figure 16.The exact perturbation vorticity $\unicode[STIX]{x1D714}_{1}/k$ (colour) and streamlines (vortices are clockwise  (red) or counterclockwise , blue) in a one-wavelength sections of the two-dimensional flow ( green colour marks the perturbed front) for (rows) $K=1.5,1.75,2$ ($\unicode[STIX]{x1D6FF}=0.25;k=6,7,8$), and for (columns) $n=0.3,1,3$. The volume covered with vortices grows the larger $K$ is and the smaller $n$ is, and the corresponding growth rate declines. Streamlines absolute values: $0.001,0.002,0.004,0.008,0.016,0.032,0.064,0.128,0.256,0.512$.

Figure 17

Figure 17.The normalised vorticity $\unicode[STIX]{x1D714}/ki$ in the limit $n\gg 1$ (4.10) along a radius, for several values of $K=1.5,3,6$.

Figure 18

Figure 18.Vorticity and vortex street: the exact perturbation vorticity $\unicode[STIX]{x1D714}_{1}/k$ (colour) and streamlines (vortices are clockwise  (red) or counterclockwise , blue) in a one-wavelength sections of the two-dimensional flow ( green colour marks the perturbed front) for $K=30$ ($\unicode[STIX]{x1D6FF}=0.75;k=40$) and for (a) $n=1$ (streamlines absolute value: $0.5,0.1,10^{-2},10^{-3},10^{-4},10^{-5},10^{-6},10^{-7},10^{-8}$), and (b) $n=100$ (streamlines absolute values: $0.75,0.5,0.25,0.05$).

Figure 19

Figure 19.Comparison of the predicted most-unstable modes and the time evolution of the dominant wavenumbers measured in laboratory experiments for $n=6$ (Sayag & Worster 2019). The experimental measurements (colour) represent the dominant Fourier modes that the front is comprised of at each instant for a range of source fluxes $0.7\lesssim Q\lesssim 13.4~\text{g}~\text{s}^{-1}$. Also shown are the corresponding instantaneous most-unstable modes predicted by the full perturbation model (, grey), and by the perfectly plastic ($n\gg 1$) thin-film model (5.1) (, red). (a,b) Experiments that converge to an integer wavenumber. (c,d) Experiments whose late-time pattern is stochastic. Results are presented in log–linear scale (a,c) and log–log scale (b,d), in which the initial transition to high wavenumber is truncated.

Sayag et al. supplementary movie

Emergence of vortices in the secondary flow as the wavenumber k grows (n=100, δ=0.75).

Download Sayag et al. supplementary movie(Video)
Video 559.6 KB