betaFeedback

WorksPhysics and Astronomy

Waves beneath a drop levitating over a moving wall

Abstract

In recent experiments, Sawaguchi et al. directly probed the lubrication layer of air beneath a droplet levitating inside a rotating cylindrical drum. For small rotation rates of the drum, the lubrication film beneath the drop adopted a steady shape, while at higher rotation rates, travelling waves propagated along the drop's lower surface with roughly half the wall velocity. We here rationalize the physical origin of these waves. We begin with a simplified model of the lubrication flow beneath the droplet, and examine the linear stability of this base state to perturbations of the Tollmien–Schlichting type. Our developments lead to the Orr-Sommerfeld equation (OSE), whose eigenvalues give the growth rates and phase speeds of the perturbations. By considering wavelengths long relative to the lubrication film thickness, we solve the OSE perturbatively and so deduce the wavelength and phase velocity of the most unstable mode. We find satisfactory agreement between experiment and theory over the parameter regime considered in the laboratory.

Introduction

Sheared interfaces are ubiquitous in nature and technology, arising when wind blows over oceans and when liquid films flow down inclined planes [2, 3, 4]. The question of the stability of such flows has thus received considerable attention [2, 5, 6]. The classic Kelvin-Helmholtz instability rationalizes the waves generated by shear across an interface of two inviscid fluid layers, and has been studied extensively in the context of wind-driven flow over water [7, 8]. Alternatively, for gas flow over a thin liquid film, the film stability may be assessed via the lubrication approximation wherein viscous effects becomes significant [5, 9, 10]. For example, Kapitza [11] showed that when gravity drives a liquid film down an incline, capillary waves may emerge. The stability of thin viscous film flows has been extensively explored by the fluid mechanics community [5, 12, 13, 14].

The approach of two liquid phases, or that of a liquid towards a solid, is resisted by lubrication pressures generated in the intervening fluid [15]. For example, sheared air films extend the lifetime of milk drops levitating on coffee [16], may suppress coalescence of liquid jets impinging on a liquid bath [17, 18], and may preclude contact of droplets impacting on a wet incline [19] or a solid substrate [20, 21]. In the present paper, we examine the lubrication flow beneath a levitating, rolling droplet. We first describe the mechanics of such drops, then assess the stability of the underlying air layer.

It is well-known that a droplet may be suspended indefinitely on a thin layer of its own vapour when placed above a hot substrate — the so-called Leidenfrost effect [22, 23]. However, it is also possible to sustain a droplet above a thin lubrication flow in an isothermal setting. Specifically, relative motion between the drop and its substrate may induce lubrication pressures in the gas flow beneath the drop, thereby providing a lift force that supports its weight. For example, Sreenivas et al. [24] achieved stable isothermal levitation of a drop on the free surface of a hydraulic jump and Lhuissier et al. [25] levitated a drop inside of a cylinder rotating about a horizontal axis (see Figure 1). Gauthier et al. [26] studied droplets levitating atop a table rotating about a vertical axis, and coined the phrase “aerodynamic Leidenfrost effect” to describe this dynamic levitation.

Image not extracted

Figure 1. (a) Schematic of a droplet rolling in a rotating drum. (b) The shape of the droplet is determined by the balance of surface tension and gravity. Viscous drops advance via rigid body apart from the shear experienced in the highlighted dark region[28, 29, 30]. In the regime considered here, the drop deformation is characterized by height-to-radius ratios in the range 0.85<H/R<1.250.85 < H/R < 1.25, where H/R=2H/R = 2 corresponds to the case of an undeformed sphere.

Sawaguchi et al. [1] revisited the rotating cylinder experiments of Lhuissier et al. [25] and investigated the details of the interface shape and pressure distribution beneath levitating drops of 50-5000 cSt Si oil with a typical diameter of 5mm. When the cylinder wall velocity beneath the droplet was just large enough to support levitation, a stable interface was observed below the drop. However, when the wall velocity was increased, unsteady travelling waves were observed along the base of the drop. Typically, these waves had a phase velocity of roughly half the wall velocity and a wavelength on the order of 1-2 mm. Compared to the thickness of the air film, hl∼10 μmh_l \sim10\,\mu\mathrm{m}, these are long waves, a fact that will be exploited in our theoretical developments. While Sawaguchi et al. [1] mentioned the emergence of such waves, and Ayumi and Tagawa [27] suggested that they may be shear driven, these waves have yet to be rationalized theoretically. The objective of the current paper is to do so.

After developing a simple model for the base flow beneath the drop, we build on the analysis of Yih [31] to assess the stability of long wavelength perturbations of the Tollmien-Schlichting type. We note that neither inviscid nor viscous potential flow models [32] capture the waves of interest (see Appendix § 0.3.1). Moreover, the frequency of the observed waves is markedly difference from the natural vibration frequency of the droplet (see Appendix § 0.3.2); thus, the observed waves can not simply be rationalized in terms of advection of the droplet’s natural modes of vibration.

The remainder of this paper is arranged as follows. In §II, we present a physical picture of the levitating droplet system of interest. In §III A, we lay out the framework for analysing the stability of the lubrication flow beneath the droplet. In §III B, we solve the equations of §III A perturbatively via a long wavelength expansion, and predict a fastest growing wavelength and its corresponding phase velocity. In §IV, we compare the predictions of our model with experiments. In §V, we discuss the limitations of our model and discuss other settings where similar instabilities might arise.

Experimental Procedure and Physical Picture

We consider the experimental arrangement employed by Lhuissier et al. [25], Sawaguchi et al. [1], and Ayumi and Tagawa [27] depicted in Figure 1. We placed silicone oil drops (100 cSt, 20.9 mN/m) with radii RR ranging between 1.7 and 3.4 mm on the inner surface of a hollow silica-glass cylinder rotating on its axis of symmetry. The inner diameter of the cylinder was 200 mm and its angular velocity was used to control the wall speed UwU_{\mathrm{w}}, which ranged between 1.05 m/s and 2.10 m/s. In this parameter regime, the droplet rests on an air lubrication layer and equilibrates to a static angle θ\theta on the inner surface of the rotating cylinder. At this relatively large droplet viscosity, the bulk of the drop is in a rigid-body rotation except in a neighborhood adjoining the base of the drop [1 see Supplementary Movie 2]. At sufficiently high cylinder rotation rates, the droplet may develop travelling surface waves on its base [27]. These surface waves were imaged with a high-speed camera (FASTCAM SA-X, Photron Co.) placed outside the rotating cylinder and directed towards the base of the drop. A beam of 630 nm monochromatic light was directed from the camera towards the base of the drop using a coaxial zoom lens (12x Co-axial Ultra Zoom Lens, Navitar Co.). The reflected light from the base of the drop then formed interference fringes, as shown in Figure 2. High-speed imaging of the interference fringes allowed us to measure both wavelength and the wave speed of the surface waves.

Image not extracted

Figure 2. Experimental snapshots of the wave progression beneath the rolling droplet. These images are taken from below using interferometry, for an experiment where U=2 m/sU = 2\,\mathrm{m/s}. AA and BB track the instability extrema between frames.

The pressure, plubp_{\mathrm{lub}}, induced by the lubrication flow in a thin air layer of thickness d1≈10 μmd_1 \approx10\,\mu\mathrm{m} supports the drop’s weight, preventing contact with the substrate. The force balance normal to the cylinder surface requires,

ρgVcos⁡(θ)∼πl2plub,(1)\rho g V \cos(\theta) \sim\pi l^2 p_{\mathrm{lub}}, \tag*{(1)}

where VV is the droplet volume and πl2\pi l^2 is the drop’s contact area.

The angle θ\theta is set by a balance between gravity, drag, and viscous shear forces on the drop [25]. The tangential force balances reads,

ρgVsin⁡(θ)=Fdrag+πl2τlub,(2)\rho g V \sin(\theta) = F_{\mathrm{drag}} + \pi l^2 \tau_{\mathrm{lub}}, \tag*{(2)}

where τlub\tau_{\mathrm{lub}} is the viscous stress in the lubrication layer. Experimentally, the majority of drops sit at an angle not exceeding 15 degrees. The approximation cos⁡(θ)≈1\cos(\theta) \approx1 introduces an error of less than 5% so that the normal force balance (1) can be solved independent of (2) and the equations become effectively decoupled. According to this approximation, since plubp_{\mathrm{lub}} depends on UwU_{\mathrm{w}} and d1d_1, d1d_1 is independent of angle. Since our model will only require d1d_1, we obviate the need to solve the tangential force balance, (2). As the wall speed, UwU_{\mathrm{w}}, is increased beyond a critical threshold, traveling waves with wavelengths λ≈1 mm\lambda\approx1\,\mathrm{mm} appear at the lower droplet interface and move at a speed of roughly Uw/2U_{\mathrm{w}}/2 in the direction of the wall motion. Figure 2 displays interferometry images of these waves.

To simplify the flow geometry, we approximate the interface between the drop and the air film as flat (see Figure 3). The lubrication flow is then modelled by a planar bounded Couette flow with the bottom wall moving at the cylinder speed, UwU_{\mathrm{w}}, and an upper effective wall, a distance d2d_2 inside the oil layer, where the fluid comes to rest. In their experiments, Sawaguchi et al. [1, pp. 273, figure 12] measured the pressure beneath the droplet, demonstrating its constancy away from a small dimple region formed at the edge of the drop. This constancy gives rise to a linear velocity profile within the lubrication layer, since lubrication analysis dictates that quadratic components of velocity only arise in the presence of pressure gradients.

Image not extracted

Figure 3. Simplified model of the lubrication flow beneath the droplet. A linear Couette flow is assumed to arise in both the oil layer of thickness d2d_2, and the air lubrication layer of thickness d1d_1. While d1d_1 is taken as the average lubrication layer beneath the droplet (see §II), d2d_2 is the average shear thickness in the drop which is determined by a single fitting parameter across all experiments (see §IV).

We proceed by outlining how the lubrication film height, d1d_1, and the effective oil shear layer thickness, d2d_2, are determined. Since d1d_1 is independent of θ\theta for small θ\theta, the situation may be related to the flat configuration of Gauthier et al. [26]. In their experiments, they found that the lubrication layer height scales as d1=CRCaair2/3d_1 = CR\mathrm{Ca}_{\mathrm{air}}^{2/3}, with C=0.6±0.2C = 0.6 \pm0.2, when Caair≪1\mathrm{Ca}_{\mathrm{air}} \ll1. Note that the Capillary number in our experiments is small, Caair=μ1Uw/σ∼10−3\mathrm{Ca}_{\mathrm{air}} = \mu_1 U_{\mathrm{w}}/\sigma\sim10^{-3}, suggesting that their scaling should also apply in our setting. When considering small drops (R≪lcR \ll l_c) that are nearly spherical, so that H≈2RH \approx2R, the scaling of Gauthier et al. [26] becomes d1=C(H/2)Ca2/3d_1 = C(H/2)\mathrm{Ca}^{2/3}. Conversely, large tank-treading droplets (R≫lcR \gg l_c) at low Reynolds number conform to the scaling d1∼lcCa2/3d_1 \sim l_c\mathrm{Ca}^{2/3}[33]. Noting that the height of such large drops saturates to H=2lcH = 2l_c, the scaling recast in terms of height becomes d1∼HCa2/3d_1 \sim H\mathrm{Ca}^{2/3}. We thus adopt the following scaling for our problem,

d1=CHHCa2/3,(3)d_1 = C_H H\mathrm{Ca}^{2/3}, \tag*{(3)}

which scales appropriately in both the large and small Bo limits. In the small drop limit of (3), the experiments of Gauthier et al. [26] indicate that CH=0.3±0.1C_H = 0.3 \pm0.1.

What remains is to define d2d_2, the effective shear layer thickness at the base of the drop. Small drops (Bo≪1\mathrm{Bo} \ll1) at low Capillary and Reynolds numbers reach a state of nearly solid-body rotation, and dissipation is confined to a neighborhood adjoining the contact area with vertical extent d2∼Bo1/2Rd_2 \sim\mathrm{Bo}^{1/2}R (see Eq. (3) in [28]). For large tank-treading drops, Hodges et al. [33] found that the shear layer permeates the entirety of the drop, so d2∼Hd_2 \sim H.

The drops considered in the present paper have Bo=O(1)\mathrm{Bo} = O(1), lying awkwardly between the cases analysed by Mahadevan and Pomeau [28] and Hodges et al. [33]. We thus expect a scaling for d2d_2 that is bounded by these two extreme cases. For the sake of simplicity, we introduce the simple scaling law, d2=AHd_2 = AH, where AA is a fitting parameter to model the shear layer thickness. For small drops, the scaling of Mahadevan and Pomeau [28], d2∼Bo1/2Rd_2 \sim\mathrm{Bo}^{1/2}R, leads to A=Bo1/2/2A = \mathrm{Bo}^{1/2}/2 so that A≪1A \ll1. For large drops, the scaling of Hodges et al. [33], d2∼Hd_2 \sim H, implies that A=O(1)A = O(1). For our drops, we thus expect AA to be less than 1/21/2. We discuss the fitting of AA in §IV, where we deduce that A=0.08A = 0.08 adequately describes our experiments.

Stability Analysis

Mathematical formulation and stability analysis

We begin by writing down the base flow as depicted in Figure 3. Let us first define non-dimensional quantities to be used in the subsequent analysis. Let the ratios of densities, viscosities, and fluid depths be denoted by m=μ2/μ1m = \mu_2/\mu_1, r=ρ2/ρ1r = \rho_2/\rho_1, and n=d2/d1n = d_2/d_1, respectively, where subscripts 1 and 2 will denote quantities in the air and oil layers, respectively. For reference, m≈5555m \approx5555 and n≈100n \approx100 for 100cSt silicone oil and depths typical in experiments. We define the Reynolds number characterizing the air flow in the lubrication layer as Re⁡=ρ1Uwd1/μ1\operatorname{Re} = \rho_1 U_w d_1/\mu_1 and that characterizing the air flow around the droplet as Re⁡d=Re⁡R/d1=ρ1UwR/μ1\operatorname{Re}_d = \operatorname{Re} R/d_1 = \rho_1 U_w R/\mu_1. We also define a non-dimensional surface tension, an inverse Weber number, by S≡σ/(ρ1d1Uw2)S \equiv\sigma/(\rho_1 d_1 U_w^2).

We non-dimensionalize the flow velocity by the wall velocity so that U=Udim/UwU = U_{\mathrm{dim}}/U_w, where UdimU_{\mathrm{dim}} is the dimensional base flow. The non-dimensional pressure is given by p=pdim/(ρ1Uw2)p = p_{\mathrm{dim}}/(\rho_1 U_w^2). Lengths are non-dimensionalized with respect to the thin air layer so that x=xdim/d1x = x_{\mathrm{dim}}/d_1, y=ydim/d1y = y_{\mathrm{dim}}/d_1. We will proceed to work exclusively with non-dimensional variables.

The base flow in the upper and lower regions is given by

U(y)={a1y+b,y∈(0,1)a2y+b,y∈(−n,0),(4)U(y) = \begin{cases} a_1 y + b, & y \in(0,1) \\ a_2 y + b, & y \in(-n,0), \end{cases} \tag*{(4)}

where,

a1=m/(m+n);a2=1/(m+n);b=n/(m+n).(5)a_1 = m/(m+n);\quad a_2 = 1/(m+n);\quad b = n/(m+n). \tag*{(5)}

Let us define the total flow as u=(U(y),0)+u′\mathbf{u} = (U(y),0) + \mathbf{u}', where u′=(u,v)\mathbf{u}' = (u,v) defines the perturbation velocity. We now consider the stability of small two-dimensional perturbations to the film. When assessing stability of parallel shear flows in a homogeneous medium, Squire [34] proved that it is sufficient to consider two-dimensional disturbances. Yih [35] extended this proof to the present case of non-uniform density and viscosity. We thus proceed to consider the evolution of Tollmien–Schlichting perturbations following the non-dimensionalization employed by Yih [31].

Consider two-dimensional wave disturbances (Tollmien–Schlichting waves) parameterized by a streamfunction, ψ\psi, and perturbation pressure, pp, as follows,

{ψ,p}={ϕ(y),f(y)}exp⁡iα(x−ct),(6)\{\psi,p\} = \{\phi(y),f(y)\}\exp i\alpha(x-ct), \tag*{(6)}

where (u,v)=(ψy,−ψx)(u,v) = (\psi_y,-\psi_x) and subscripts denote partial differentiation. The non-dimensional wavenumber is related to the dimensional wavenumber kk by the relation, α=kd1\alpha= k d_1. After plugging u=(U(y),0)+u′\mathbf{u} = (U(y),0) + \mathbf{u}' into the Navier–Stokes equations, and using (6) to eliminate pressure, we linearize to reach the standard Orr–Sommerfeld equation governing ϕ\phi,

ϕiv−2α2ϕ′′+α4ϕ=iαRe⁡i((U(y)−c)(ϕ′′−α2ϕ)−U′′(y)ϕ),(7)\phi^{\mathrm{iv}} - 2\alpha^2\phi'' + \alpha^4\phi= i\alpha\operatorname{Re}_i\left((U(y)-c)(\phi''-\alpha^2\phi)-U''(y)\phi\right), \tag*{(7)}

where i=1i=1 in the air layer, i=2i=2 in the oil layer, and cc is an eigenvalue to be determined by solving (7) subject to the appropriate boundary conditions. We note that U′′(y)=0U''(y)=0 in the case of the base flow (4) and ϕ\phi must satisfy (7) in the bulk, for y∈(−d2,0)∪(0,d1)y \in(-d_2,0)\cup(0,d_1). It is thus convenient to express ϕ\phi in terms of the functions Φ\Phi and Ψ\Psi, defined in the air and oil respectively, such that

ϕ(y)={Φ(y),y∈(0,1)Ψ(y),y∈(−n,0),(8)\phi(y) = \begin{cases} \Phi(y), & y \in(0,1) \\ \Psi(y), & y \in(-n,0), \end{cases} \tag*{(8)}

with Φ\Phi and Ψ\Psi both satisfying (7)

Φiv−2α2Φ′′+α4Φ=iαRe⁡(U(y)−c)(Φ′′−α2Φ)Ψiv−2α2Ψ′′+α4Ψ=iαrRe⁡m(U(y)−c)(Ψ′′−α2Ψ).(9)\begin{aligned} \Phi^{\mathrm{iv}} - 2\alpha^2\Phi'' + \alpha^4\Phi&= i\alpha\operatorname{Re}(U(y)-c)(\Phi''-\alpha^2\Phi) \\ \Psi^{\mathrm{iv}} - 2\alpha^2\Psi'' + \alpha^4\Psi&= i\alpha\frac{r\operatorname{Re}}{m}(U(y)-c)(\Psi''-\alpha^2\Psi). \tag*{(9)} \end{aligned}

At the boundaries y=−d2y=-d_2 and y=d1y=d_1, the flow must satisfy no-slip conditions. Since the base flow satisfies the no-slip conditions, so too must the perturbations. Thus,

Φy(1)=0; Φ(1)=0; Ψy(−n)=0; Ψ(−n)=0.(10)\Phi_y(1)=0;\ \Phi(1)=0;\ \Psi_y(-n)=0;\ \Psi(-n)=0. \tag*{(10)}

Across the interface, the velocity must be continuous. The linearized interface height, η(x,t)\eta(x,t), is defined by the kinematic condition,

(∂∂t+U∂∂x)η=v=−iαexp⁡iα(x−ct),(11)\left(\frac{\partial}{\partial t}+U\frac{\partial}{\partial x}\right)\eta=v=-i\alpha\exp i\alpha(x-ct), \tag*{(11)}

so that, after defining c′≡c−U(0)c' \equiv c-U(0), the interface is given by

η=ϕ(0)c′exp⁡iα(x−ct).(12)\eta=\frac{\phi(0)}{c'}\exp i\alpha(x-ct). \tag*{(12)}

The continuity of vertical velocity, at y=0y=0, simply yields

Φ(0)=Ψ(0).(13)\Phi(0)=\Psi(0). \tag*{(13)}

However, the continuity of horizontal velocity must be enforced at the interface defined in (12). A Taylor expansion for small interface perturbations then yields the horizontal velocity continuity condition

Φ′(0)−Ψ′(0)=Φ(0)c′(a2−a1).(14)\Phi'(0)-\Psi'(0)=\frac{\Phi(0)}{c'}(a_2-a_1). \tag*{(14)}

The continuity of shear stress at the interface takes the form

Φ′′(0)+α2Φ(0)=m(Ψ′′(0)+α2Ψ(0)).(15)\Phi''(0)+\alpha^2\Phi(0)=m\left(\Psi''(0)+\alpha^2\Psi(0)\right). \tag*{(15)}

Finally, the continuity of normal stress may be written as

−iαRe⁡(c′Φ′+a1Φ)−(Φ′′′−α2Φ′)+2α2Φ′+irαRe⁡(c′Ψ′+a2Ψ)+m(Ψ′′′−α2Ψ′)−2α2mΨ′=iRe⁡α3SΦ/c′,(16)\begin{aligned} -i\alpha\operatorname{Re}(c'\Phi'+a_1\Phi)-(\Phi'''-\alpha^2\Phi')+2\alpha^2\Phi' +{}\\ i r\alpha\operatorname{Re}(c'\Psi'+a_2\Psi)+m(\Psi'''-\alpha^2\Psi')-2\alpha^2m\Psi'&=i\operatorname{Re}\alpha^3S\Phi/c', \tag*{(16)} \end{aligned}

where all quantities are evaluated at y=0y=0. The eigenvalue problem for cc is now fully specified through equations (9)–(16). In order to make progress in determining the stability of various wavenumber perturbations, α\alpha, we note that in the experiments of interest, the nondimensional wavenumber is small with α=2πd1/λ∼2π(2μm)/(1mm)≈0.03\alpha=2\pi d_1/\lambda\sim2\pi(2\mu\mathrm{m})/(1\mathrm{mm})\approx0.03. Therefore, we are justified in confining our attention to perturbations with α≪1\alpha\ll1. In the following section, we solve for Φ\Phi and Ψ\Psi order by order in the small parameter α\alpha via a perturbation analysis.

Long wavelength perturbation expansion

To solve the system of equations of §III A perturbatively, we begin by writing Φ(y)=Φ0(y)+αΦ1(y)+α2Φ2(y)+α3Φ3(y)+O(α4)\Phi(y)=\Phi_0(y)+\alpha\Phi_1(y)+\alpha^2\Phi_2(y)+\alpha^3\Phi_3(y)+O(\alpha^4) and Ψ(y)=Ψ0(y)+αΨ1(y)+α2Ψ2(y)+α3Ψ3(y)+O(α4)\Psi(y)=\Psi_0(y)+\alpha\Psi_1(y)+\alpha^2\Psi_2(y)+\alpha^3\Psi_3(y)+O(\alpha^4). We also write c′=c0′+αc1+α2c2+α3c3+O(α4)c'=c'_0+\alpha c_1+\alpha^2c_2+\alpha^3c_3+O(\alpha^4). Plugging these expressions into (7–16), we obtain a hierarchy of equations in increasing powers of α\alpha that can be solved sequentially. At each order, an eigenvalue problem emerges for the corresponding wave speed correction; if complex, these corrections correspond to temporal growth or decay. It will become clear that we need to solve to O(α3)O(\alpha^3) to capture the stabiliting effect of surface tension, since the pre-factor of SS in (16) is O(α3)O(\alpha^3).

O(1) Equations

The zeroth order form of (9) simplifies to

Φ0iv=0;Ψ0iv=0,(17)\Phi_0^{\mathrm{iv}}=0;\quad\Psi_0^{\mathrm{iv}}=0, \tag*{(17)}

whose general solution is simply a set of third degree polynomials for Φ0(y)\Phi_0(y) and Ψ0(y)\Psi_0(y). We are free to set the constant term to unity, since the resulting eigenvalue problem will only be specified up to a constant scaling. We can then write Φ0=1+A1y+A2y2+A3y3\Phi_0=1+A_1y+A_2y^2+A_3y^3 and Ψ0=1+B1y+B2y2+B3y3\Psi_0=1+B_1y+B_2y^2+B_3y^3. With this choice, the vertical velocity continuity (13) is automatically upheld so that

Φ0(0)=Ψ0(0)=1.(18)\Phi_0(0)=\Psi_0(0)=1. \tag*{(18)}

Six undetermined coefficients (three in each layer) remain undetermined, along with the wave speed, all to be determined by application of the following boundary conditions.

The four homogeneous wall conditions (10) must be satisfied by Φi\Phi_i and Ψi\Psi_i at each order of α\alpha. At O(1)\mathcal{O}(1), the horizontal velocity continuity (14) becomes

Φ0′(0)−Ψ0′(0)=Φ0(0)c0′(a2−a1).(19)\Phi_0'(0)-\Psi_0'(0)=\frac{\Phi_0(0)}{c'_0}(a_2-a_1). \tag*{(19)}

The continuity of shear stress at the interface then simplifies to,

Φ0′′(0)=mΨ0′′(0).(20)\Phi_0''(0)=m\Psi_0''(0). \tag*{(20)}

Finally, the continuity of normal stress may be written as,

−Φ0′′′+mΨ0′′′=0,(21)-\Phi_0''' + m\Psi_0''' = 0, \tag*{(21)}

where all quantities are evaluated at y=0y=0. The problem for c0′c'_0 is now fully specified, through the following seven equations: the four equations in (10), along with (19), (20), and (21). The seven unknowns are the three undetermined polynomial coefficients of each of Φ0\Phi_0 and Ψ0\Psi_0, and the zeroth order wave speed c0′c'_0. The linear algebra problem for the coefficients can be solved symbolically in Mathematica or numerically in MATLAB. We implemented both approaches and cross-checked results at all orders and find the approaches to be in agreement. Yih [31] also solved the system up to O(α)\mathcal{O}(\alpha); as a third check, we compared our results to his and found favourable agreement. We proceed to outline the O(α)\mathcal{O}(\alpha), O(α2)\mathcal{O}(\alpha^2) and O(α3)\mathcal{O}(\alpha^3) equations. We suppress the symbolic solutions for the wave speed at each order for brevity, and report the final results of the perturbation analysis in the predictions given in figure 4.

O(α)\mathcal{O}(\alpha) Equations

The O(α)\mathcal{O}(\alpha) form of (9) simplifies to

Φ1iv=iRe⁡(a1y−c0′)Φ0′′,\Phi_1^{\mathrm{iv}}=i\operatorname{Re}(a_1y-c'_0)\Phi_0'',
Ψ1iv=irRe⁡m(a2y−c0′)Ψ0′′.(22)\Psi_1^{\mathrm{iv}}=i\frac{r\operatorname{Re}}{m}(a_2y-c'_0)\Psi_0''. \tag*{(22)}

whose homogeneous solution is again a set of third degree polynomials for Φ1(y)\Phi_1(y) and Ψ1(y)\Psi_1(y). The particular solution is obtained through the direct integration of the zeroth order solution (cubic polynomial) as obtained in §III B 1. All constants of integration can be set to zero since this corresponds to the homogeneous part of the solution. We can then write the homogeneous part of the solution as Φ1=1+A1(1)y+A2(1)y2+A3(1)y3\Phi_1=1+A_1^{(1)}y+A_2^{(1)}y^2+A_3^{(1)}y^3 and Ψ1=1+B1(1)y+B2(1)y2+B3(1)y3\Psi_1=1+B_1^{(1)}y+B_2^{(1)}y^2+B_3^{(1)}y^3, where we have once again set the constant term to unity. Again, and at all subsequent orders, this choice ensures the vertical velocity continuity (13) is automatically satisfied. The six undetermined coefficients and the wave speed correction c1c_1 are determined by the boundary conditions.

The four homogeneous wall conditions (10) must be satisfied by Φi\Phi_i and Ψi\Psi_i at each order of α\alpha. At this order, the horizontal velocity continuity (14) becomes

Φ1′(0)−Ψ1′(0)=−c1Φ0(0)+c0′Φ1(0)(c0′)2(a2−a1).(23)\Phi_1'(0)-\Psi_1'(0)=\frac{-c_1\Phi_0(0)+c'_0\Phi_1(0)}{(c'_0)^2}(a_2-a_1). \tag*{(23)}

The continuity of shear stress again simplifies to the form,

Φ1′′(0)=mΨ1′′(0).(24)\Phi_1''(0)=m\Psi_1''(0). \tag*{(24)}

Lastly, the continuity of normal stress may be written as,

−iRe⁡(c0′Φ0′+a1Φ0)−Φ1′′′+irRe⁡(c0′Ψ0′+a2Ψ0)+mΨ1′′′=0,(25)-i\operatorname{Re}(c'_0\Phi_0'+a_1\Phi_0)-\Phi_1''' + ir\operatorname{Re}(c'_0\Psi_0'+a_2\Psi_0)+m\Psi_1'''=0, \tag*{(25)}

where all quantities are evaluated at y=0y=0. The problem for c1c_1, along with the six undetermined coefficients, is now fully specified. The simple linear algebra problem can again be solved symbolically in Mathematica or numerically in MATLAB. We now move to the O(α2)\mathcal{O}(\alpha^2) and O(α3)\mathcal{O}(\alpha^3) equations. At this point, we mention that c1c_1 is purely complex, as can be seen by examination of the perturbation equations at O(α)\mathcal{O}(\alpha). Thus, this correction corresponds to the growth rate of the perturbation. As the corrections to the wave speed alternate between being purely real and purely imaginary, it is required to solve to O(α3)\mathcal{O}(\alpha^3) to get the next imaginary correction to the growth rate. This is consistent with the fact that the effects of surface tension, which should be stabilizing, first enter at that order.

O(α2)O(\alpha^2) Equations

The O(α2)O(\alpha^2) form of (9) simplifies to

Φ2iv−2Φ0′′=iRe⁡((a1y−c0′)Φ1′′−c1Φ0′′),Ψ2iv−2Ψ0′′=irRe⁡m((a2y−c0′)Ψ1′′−c1Ψ0′′).(26)\begin{aligned} \Phi_2^{\mathrm{iv}}-2\Phi_0''&=i\operatorname{Re}\left((a_1y-c_0')\Phi_1''-c_1\Phi_0''\right),\\ \Psi_2^{\mathrm{iv}}-2\Psi_0''&=i\frac{r\operatorname{Re}}{m}\left((a_2y-c_0')\Psi_1''-c_1\Psi_0''\right). \tag*{(26)} \end{aligned}

We can then write the homogeneous part of the solution as Φ2=1+A1(2)y+A2(2)y2+A3(2)y3\Phi_2=1+A_1^{(2)}y+A_2^{(2)}y^2+A_3^{(2)}y^3 and Ψ2=1+B1(2)y+B2(2)y2+B3(2)y3\Psi_2=1+B_1^{(2)}y+B_2^{(2)}y^2+B_3^{(2)}y^3, where have once again set the constant term to unity to ensure vertical velocity continuity. The particular solution is obtained through direct integration of the terms in (26) involving the zeroth and first order solutions. Integration constants are again set to zero. Six undetermined coefficients, and the wave speed correction c2c_2, are determined by application of the boundary conditions as follows.

The four homogeneous wall conditions (10) must be satisfied by Φ2\Phi_2 and Ψ2\Psi_2. At this order, the horizontal velocity continuity (14) becomes

Φ2′(0)−Ψ2′(0)=c12Φ0−c0′c2Φ0−c0′c1Φ1+(c0′)2Φ2(c0′)3(a2−a1).(27)\Phi_2'(0)-\Psi_2'(0)=\frac{c_1^2\Phi_0-c_0'c_2\Phi_0-c_0'c_1\Phi_1+(c_0')^2\Phi_2}{(c_0')^3}(a_2-a_1). \tag*{(27)}

The continuity of shear stress simplifies to the form

Φ2′′(0)+Φ0(0)=m(Ψ2′′(0)+Ψ0(0)).(28)\Phi_2''(0)+\Phi_0(0)=m\left(\Psi_2''(0)+\Psi_0(0)\right). \tag*{(28)}

Finally, at this order, the continuity of normal stress may be written as,

3Φ0′−iRe⁡(a1Φ1+c0′Φ1′+c1Φ0′)+irRe⁡(a2Ψ1+c0′Ψ1′+c1Ψ0′)−Φ2′′+m(Ψ2′′−3Ψ0′)=0,(29)3\Phi_0'-i\operatorname{Re}\left(a_1\Phi_1+c_0'\Phi_1'+c_1\Phi_0'\right)+ir\operatorname{Re}\left(a_2\Psi_1+c_0'\Psi_1'+c_1\Psi_0'\right)-\Phi_2''+m\left(\Psi_2''-3\Psi_0'\right)=0, \tag*{(29)}

where all quantities are evaluated at y=0y=0. The problem for c2c_2, along with the six undetermined coefficients, is now fully specified. The linear algebra problem can again be solved symbolically in Mathematica or numerically in MATLAB. We now outline the O(α3)O(\alpha^3) equations and then compare our theoretical results to experiment.

O(α3)O(\alpha^3) Equations

The O(α3)O(\alpha^3) form of (9) simplifies to

Φ3iv−2Φ1′′=iRe⁡((a1y−c0′)(−Φ0+Φ0′′)−c1Φ1′′−c2Φ0′′),Ψ3iv−2Ψ1′′=iRe⁡rm((a2y−c0′)(−Ψ0+Ψ0′′)−c1Ψ1′′−c2Ψ0′′).(30)\begin{aligned} \Phi_3^{\mathrm{iv}}-2\Phi_1''&=i\operatorname{Re}\left((a_1y-c_0')(-\Phi_0+\Phi_0'')-c_1\Phi_1''-c_2\Phi_0''\right),\\ \Psi_3^{\mathrm{iv}}-2\Psi_1''&=i\operatorname{Re}\frac{r}{m}\left((a_2y-c_0')(-\Psi_0+\Psi_0'')-c_1\Psi_1''-c_2\Psi_0''\right). \tag*{(30)} \end{aligned}

By writing a general homogeneous solution in each layer, and integrating (30) to obtain a particular solution, we again reach six undetermined coefficients, and the wave speed correction c3c_3, to be determined via application of the following boundary conditions.

The four homogeneous wall conditions (10) must be satisfied by Φ3\Phi_3 and Ψ3\Psi_3. At this order, the horizontal velocity continuity (14) becomes

Φ3′(0)−Ψ3′(0)=(−c12+2c0′c1c2−(c0′)2c3)Φ0+c0′(c12−c0′c2)Φ1−(c0′)2c1Φ2+(c0′)3Φ3(c0′)4(31)\Phi_3'(0)-\Psi_3'(0)=\frac{(-c_1^2+2c_0'c_1c_2-(c_0')^2c_3)\Phi_0+c_0'(c_1^2-c_0'c_2)\Phi_1-(c_0')^2c_1\Phi_2+(c_0')^3\Phi_3}{(c_0')^4} \tag*{(31)}

The continuity of shear stress simplifies to the form,

Φ3′′(0)+Φ3(0)=m(Ψ3′′(0)+Ψ1(0)).(32)\Phi_3''(0)+\Phi_3(0)=m\left(\Psi_3''(0)+\Psi_1(0)\right). \tag*{(32)}

Finally, at this order, the continuity of normal stress may be written as,

iRe⁡(sc0′Φ0+a1Φ2+c2Φ0′+c1Φ1′+c0′Φ2′−r(a2Ψ2+c2Ψ0′+c1Ψ1′+c0′Ψ2′))=3Φ1′−Φ3′′′+m(Ψ3′′′−3Ψ1′)(33)i\operatorname{Re}\left(\frac{s}{c_0'}\Phi_0+a_1\Phi_2+c_2\Phi_0'+c_1\Phi_1'+c_0'\Phi_2'-r\left(a_2\Psi_2+c_2\Psi_0'+c_1\Psi_1'+c_0'\Psi_2'\right)\right)=3\Phi_1'-\Phi_3'''+m\left(\Psi_3'''-3\Psi_1'\right) \tag*{(33)}

where all quantities are evaluated at y=0y=0. The linear algebra problem for c3c_3 and the undetermined coefficients is again solved symbolically in Mathematica.

The Fastest Growing Mode

With cc known, the phase speed of a given wave is given by Re⁡{c}=c0+α2c2\operatorname{Re}\{c\} = c_0 + \alpha^2 c_2. Moreover, the growth rate of the wave is given by G=αIm⁡{c}=α2c1+α4c3G = \alpha\operatorname{Im}\{c\} = \alpha^2 c_1 + \alpha^4 c_3 (see (12)). The fastest growing mode is then given by solving G′(αmax⁡)=0G'(\alpha_{\max}) = 0, so that the fastest growing wavenumber is given by

αmax⁡=−c12c3(34)\alpha_{\max} = \sqrt{-\frac{c_1}{2c_3}} \tag*{(34)}

This result is valid provided the computed wavenumber conforms to our long-wavelength assumption, so that αmax⁡≪1\alpha_{\max} \ll1; this assumption requires c3≫c1c_3 \gg c_1, which we find to be true, as seen in Figure 4. The fastest growing mode is selected by surface tension since c3c_3 depends on SS. Redimensionalizing, the fastest growing wavelength is

Comparison of predictions from perturbation analysis with experiments

Figure 4. Comparison of the predictions from the perturbation analysis of §III A with experiments. a) Red dots give the fastest growing wavelength as predicted by our model. The lubrication layer thickness is given by the scaling d1=0.34HCa2/3d_1 = 0.34H\mathrm{Ca}^{2/3}, while d2d_2 is the effective shear length in the drop, prescribed by the scaling d2=AHd_2 = AH, with A=0.08A = 0.08 representing the only fitting parameter in the model. The region where α>0.1\alpha> 0.1 is shaded, to illustrate the range of validity of our long wavelength approximation, α≪1\alpha\ll1. b) Red lines give theoretical predictions of the wave speed according to our theory to O(α2)O(\alpha^2) plotted versus the drop Reynolds number, Red=ρ1UwR/μ1\mathrm{Re}_d = \rho_1 U_w R/\mu_1. The dotted line is computed using the average radius of the droplets across all experiments. The upper and lower solid red lines represent theoretical predictions for radius values one standard deviation above and below this mean, respectively.

λmax⁡=22πd1−c3c1.(35)\lambda_{\max} = 2\sqrt{2}\pi d_1 \sqrt{-\frac{c_3}{c_1}}. \tag*{(35)}

Generally, if there is instability, we have that c1>0c_1 > 0, corresponding to the viscous shear instability that was observed by Yih [31]. We note that c3<0c_3 < 0, indicating the stabilizing effect of surface tension arising at O(α3)O(\alpha^3). The non-dimensional wave phase speed, corresponding to the fastest growing mode, is then simply given by cmax⁡=c0+α2c2=c0′+b+α2c2c_{\max} = c_0 + \alpha^2 c_2 = c'_0 + b + \alpha^2 c_2.

Determination of d2d_2 and Comparison with Experiments

We now fit d2d_2, the only free parameter in our model, which defines the effective shear layer thickness in the droplet. As described in §II, we assume the scaling d2=AHd_2 = AH, where AA is to be determined. We determine AA by fitting to the wavelength data across our experiments. Taking A=0.08A = 0.08, we achieve the theoretical predictions to the data shown in 4(a).

Experimental phase speeds are plotted in Figure 4(b). Our theory at O(α2)O(\alpha^2) is presented in red lines, yielding fair agreement with experiments. The vertical spread in the model prediction at O(α2)O(\alpha^2) comes from a weak dependence of the theory on variations of dimensionless groups, other than the Reynolds number, between experiments. The dotted line represents the theoretical prediction with the drop radius fixed at the average value of the drops across all experiments, Ravg.=2.57mmR_{\mathrm{avg.}} = 2.57\mathrm{mm}. Then UU is varied, for this fixed radius, in order to attain all values of Re reported in the plot. The upper and lower solid red lines are theoretical predictions corresponding to the radii R−σRR-\sigma_R and R+σRR+\sigma_R, respectively, where σR=0.49mm\sigma_R = 0.49\mathrm{mm} is the standard deviation of the spread of radii across all experiments. Overall, there is a fair agreement between the theory and experiment.

Discussion and Conclusion

We have examined a propagating wave instability that develops beneath a levitating droplet as reported by Sawaguchi et al. [1]. By making assumptions consistent with experimental observations, we have developed a simplified model of the lubrication flow beneath the droplet. We then analysed the stability of the idealized base state to small perturbations of the Tollmien-Schlichting type. Our analysis builds on the work of Yih [31], who examined a long wavelength instability due to viscosity stratification. Notably, Yih [31] considered a small wavenumber (α≪1\alpha\ll1) expansion and solved to first order in α\alpha to assess the onset of instability. Conversely, we were obliged to solve to third order in α\alpha in order to capture the effect of surface tension. We find that surface tension selects a most unstable mode, whose wavelength and phase speed are in adequate agreement with available experimental data.

The main inputs to our model are d1d_1, the thickness of the air film beneath the drop, and d2d_2, an effective shear layer thickness inside the droplet. The former input, d1d_1, is governed by the scaling presented by Gauthier et al. [26]. The only fitting parameter in our model is thus A=d2/H≈0.08A = d_2/H \approx0.08, which was found by fitting to our experimental data for the wavelength of instability indicated as red points in Figure 4(a). With A≈0.08A \approx0.08 fixed, the theory predicts the phase speed of waves evident in Figure 4(b). The data trend of phase speed increasing with the drop Reynolds number is captured well by the theory. Note that the small vertical spread in theoretical predictions evident in 4(b) indicates that the phase speed depends weakly on parameters other than the drop Reynolds number, Red\mathrm{Re}_d, including the drop radius. While it would have been preferable to plot the data and theory in Figure 4 against the lubrication Reynolds number, as the lubrication thickness d1d_1 was not measured directly in experiments, we were obliged to plot against the droplet Reynolds number, Red\mathrm{Re}_d.

While our analysis applies to the idealized geometry of Figure 3(b), it only serves as an approximate model for the levitating droplet under consideration. We believe the slight discrepancy between theory and experiment, evident in Figure 4(b), is due to both geometric simplifications (see §II) as well as our assumed scaling for d2d_2. For example, weak curvature of the drop’s lower surface [25] might augment phase velocities relative to those on a flat interface. Likewise, three dimensional aspects of the flow reported by Sawaguchi et al. [1], Supplementary Movie 3 suggest that our assumption of purely unidirectional flow may have limited validity. Finally, one does not expect our inferred shear layer scaling, d2∼0.08Hd_2 \sim0.08H, to remain valid outside the experimental regime considered in the present paper.

Overall, we have found adequate agreement between our idealized theoretical model and experimental observations. We thus believe that we have elucidated the key mechanism of instability. Specifically, the waves observed by Sawaguchi et al. [1] emerge due to a viscous shear instability of the lubricating air flow beneath the droplet. While the problem involves a variety of dimensionless groups, the unstable mode’s wavelength and phase speed depend primarily on the Reynolds number. We conclude by noting that similar wave instabilities might appear in other systems with high interfacial shear, including the rolling drop system of Gauthier et al. [26], as well as other high-speed noncoalescence events [15, 17, 18, 21]. Similar instabilities might also appear in inverse Leidenfrost systems [36] with high viscosity drops.

Funding. K.M was generously funded by a Mathworks fellowship during this work. JB thanks the NSF for financial support through grant CMMI-2154151.

Acknowledgments

The authors thank Ayumi Matsuda for help in obtaining the image in Figure 2. The authors are grateful for valuable discussions with T.R. Akylas, and K.J. Burns.

Inviscid and Viscus Potential Flow Models

At first glance, the waves appear to be of the classic Kelvin–Helmholtz (KH) type: the air lubrication layer has velocities near 1-2 m/s while the drop is almost at rest, giving rise to a shear flow. We now crudely approximate the flow in the potential flow limit and show the analysis to be lacking through comparison with experimental data. We first consider an inviscid potential flow model using parameter values typical in the experiments of Sawaguchi et al. [1].

Considering an air layer with thickness 10 µm that moves at velocity 2 m/s and a liquid layer of 100 cSt silicone oil of thickness 1 mm at rest, and taking the surface tension value σ=21\sigma= 21 mN/m, inviscid theory predicts maximum growth of the wavelength λi=2.14\lambda_{\mathrm{i}} = 2.14 mm with phase speed Vi=0.087V_{\mathrm{i}} = 0.087 m/s. While the predicted wavelength is within a factor of 2 of the experimental value, λexp=1.3±0.1\lambda_{\mathrm{exp}} = 1.3 \pm0.1 mm, the phase speed is far below the observed value of Vexp≈0.72V_{\mathrm{exp}} \approx0.72 m/s. Considering now the viscous potential flow framework of Funada and Joseph [32], where viscous effects are included only in the normal stress interfacial conditions, the wavelength of maximum growth is worse, λvis=3.59\lambda_{\mathrm{vis}} = 3.59 mm, while the wave speed is still underestimated with Vvis=0.095V_{\mathrm{vis}} = 0.095 m/s. It is thus necessary to incorporate the effects of shear stress to obtain predictions for the wavelength and phase speed.

Analysis of Natural Vibration Modes

We here demonstrate that the frequency of the observed waves is an order of magnitude different from the first natural vibration modes of the droplet.

The upper portion of the droplet is approximately spherical. It should thus vibrate naturally roughly according to the inviscid natual oscillation formula given by Lamb [37, §253], where the nnth mode of vibration has a frequency of

ωn=n(n−1)(n+2)σρR3(36)\omega_n = \sqrt{\frac{n(n-1)(n+2)\sigma}{\rho R^3}} \tag*{(36)}

In the smallest drops in our experiments (R=1.7R = 1.7 mm), the first three vibrational modes according to (B1) are given by ω2=188\omega_2 = 188/s, ω3=364\omega_3 = 364/s, ω4=563\omega_4 = 563/s. Note that droplet non-sphericity, as well as viscosity, have been shown experimentally to decrease these frequencies further [38].

By way of comparison, the frequency of the waves observed beneath drops, as measured by counting the number of wave crests passing a given point per unit time is ωexp=2772\omega_{\mathrm{exp}} = 2772/s. The experimental frequency was computed by observing 13 wave crests passing by a given point over a video of length 30 ms.

The experimental value corresponds roughly to the n=12n = 12 mode of the smallest drop. For larger drops, the value of nn for which ωn=2775\omega_n = 2775/s would be even higher. Due to the increased damping experienced by higher modes, it is unlikely that the excitation of such high-nn modes is responsible for the observed waves.

References

References

  1. [1]E. Sawaguchi, A. Matsuda, K. Hama, M. Saito, and Y. Tagawa, Journal of Fluid Mechanics 862, 261 (2019).
  2. [2]C. Caulfield, Annual Review of Fluid Mechanics 53, 113 (2021).
  3. [3]N. J. Balmforth and S. Mandre, Journal of Fluid Mechanics 514, 1 (2004).
  4. [4]S. J. Weinstein and K. J. Ruschak, Annu. Rev. Fluid Mech. 36, 29 (2004).
  5. [5]A. Oron, S. H. Davis, and S. G. Bankoff, Reviews of modern physics 69, 931 (1997).
  6. [6]P. G. Drazin and W. H. Reid, Hydrodynamic stability (Cambridge university press, 2004).
  7. [7]F. Ursell, Surveys in mechanics , 216 (1956).
  8. [8]O. M. Phillips, Journal of fluid mechanics 2, 417 (1957).
  9. [9]A. D. Craik, Journal of Fluid Mechanics 26, 369 (1966).
  10. [10]J. W. Miles, Journal of Fluid Mechanics 8, 593 (1960).
  11. [11]P. Kapitza, Zh. Eksp. Teor. Fiz. 18, 3 (1948).
  12. [12]T. B. Benjamin, Journal of Fluid Mechanics 2, 554 (1957).
  13. [13]C.-S. Yih, The physics of Fluids 6, 321 (1963).
  14. [14]R. V. Craster and O. K. Matar, Rev. Mod. Phys. 81, 1131 (2009).
  15. [15]H. P. Kavehpour, Annual Review of Fluid Mechanics 47, 245 (2015).
  16. [16]M. Geri, B. Keshavarz, G. H. McKinley, and J. W. Bush, Journal of Fluid Mechanics 833, R3 (2017).
  17. [17]N. Wadhwa and S. Jung, Physics of Fluids 23, 091105 (2011).
  18. [18]M. Thrasher, S. Jung, Y. K. Pang, C.-P. Chuu, and H. L. Swinney, Physical Review E 76, 056319 (2007).
  19. [19]T. Gilet and J. W. Bush, Physics of Fluids 24 (2012).
  20. [20]Y. Liu, P. Tan, and L. Xu, Proceedings of the National Academy of Sciences 112, 3280 (2015).
  21. [21]J. M. Kolinski, L. Mahadevan, and S. M. Rubinstein, Europhysics Letters 108, 24001 (2014).
  22. [22]J. G. Leidenfrost, Impensis Hermanni Ovenni, Univers. bibliopolae (1756).
  23. [23]D. Quéré, Annual Review of Fluid Mechanics 45, 197 (2013).
  24. [24]K. Sreenivas, P. De, and J. H. Arakeri, Journal of Fluid Mechanics 380, 297 (1999).
  25. [25]H. Lhuissier, Y. Tagawa, T. Tran, and C. Sun, Journal of Fluid Mechanics 733, R4 (2013).
  26. [26]A. Gauthier, J. C. Bird, C. Clanet, and D. Quéré, Physical Review Fluids 1, 084002 (2016).
  27. [27]M. Ayumi and Y. Tagawa, Japanese J. Multiphase Flow 35, 176 (2020).
  28. [28]L. Mahadevan and Y. Pomeau, Physics of fluids 11, 2449 (1999).
  29. [29]P. Aussillous and D. Quéré, Nature 411, 924 (2001).
  30. [30]P. Aussillous and D. Quéré, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 462, 973 (2006).
  31. [31]C.-S. Yih, Journal of Fluid Mechanics 27, 337 (1967).
  32. [32]T. Funada and D. Joseph, Journal of Fluid Mechanics 445, 263 (2001).
  33. [33]S. Hodges, O. Jensen, and J. Rallison, Journal of Fluid Mechanics 512, 95 (2004).
  34. [34]H. B. Squire, Proceedings of the Royal Society of London. Series A, Containing Papers of a Mathematical and Physical Character 142, 621 (1933).
  35. [35]C.-S. Yih, Quarterly of Applied Mathematics 12, 434 (1955).
  36. [36]A. Gauthier, C. Diddens, R. Proville, D. Lohse, and D. van Der Meer, Proceedings of the National Academy of Sciences 116, 1174 (2019).
  37. [38]M. Perez, Y. Brechet, L. Salvo, M. Papoular, and M. Suery, Europhysics Letters 47, 189 (1999).

Paper details

vlnc:W_42wfv9rd
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents