betaFeedback

WorksPhysics and Astronomy

Diffusioosmotic flow in a soft microfluidic configuration induces fluid-structure instability

Abstract

Diffusioosmotic flow arises in microfluidic configurations due to solute concentration gradients. In soft microfluidic channels, internal pressure gradients generated by diffusioosmotic flow to conserve mass result in elastic deformation of the channel walls, triggering fluid-structure interaction. In this work, we analyze the fluid-structure interaction between diffusioosmotic flow of an electrolyte solution and a deformable microfluidic channel. We provide insight into the physical behavior of the system by developing a reduced-order model, in which a viscous film is confined between a rigid bottom surface and an elastic top substrate, represented as a rigid plate connected to a linear spring. Considering a slender configuration and applying the lubrication approximation, we derive a set of two-way coupled governing equations describing the evolution of the fluidic film thickness and the solute concentration. Our theoretical predictions show that above a certain concentration gradient threshold, negative pressures induced by diffusioosmotic flow give rise to fluid-structure instability, causing the elastic top substrate to collapse onto the bottom surface. We employ theoretical analysis to elucidate the underlying physical mechanisms for the onset of fluid-structure instability by performing a linear stability analysis of the system and identifying three distinct dynamic regimes. We validate our theoretical results with finite-element simulations and find excellent agreement. The understanding of this instability is of fundamental importance for improving the control and design of microfluidic systems driven by diffusioosmotic flow and containing soft elements.

Introduction

Diffusioosmotic flow or diffusioosmosis is the spontaneous movement of fluid adjacent to a surface, driven by a solute concentration gradient [1, 2, 3, 4]. In electrolyte solutions, diffusioosmotic flow arises from two contributions. The first contribution is chemiosmosis, which is driven by osmotic pressure gradients along the surface [2, 4]. The second contribution is electroosmosis, which can be induced by an interaction of a spontaneously generated electric field with the net charge in the electric double layer, due to unequal ion diffusivities [2, 4]. Prieve et al. [1] developed a classical model for diffusioosmotic flow of electrolyte solutions and provided an analytical expression for the diffusioosmotic slip velocity at a solid surface in the thin-double-layer limit. Recently, Shi et al. [5] presented a generalized, conceptually unified theory of linear osmotic slip velocity over a solid surface, building on Derjaguin’s original approach [6, 7] and reproducing known formulas for linear electroosmosis, chemiosmosis, and diffusioosmosis.

With the advent of microfabrication technology in the 2000s, diffusioosmotic flow has become a common driving mechanism for fluid manipulation in microfluidic and lab-on-a-chip devices [3, 4, 8, 9], alongside pressure-driven and electroosmotic flows [10, 11, 12]. For example, Ault et al. [13] presented a microfluidic approach that leverages diffusioosmotic flow to measure the zeta potentials of electrolyte solutions. Recently, Teng et al. [14] studied solute dispersion driven solely by diffusioosmotic flow. A closely related phenomenon to diffusioosmosis is diffusiophoresis, which refers to the spontaneous phoretic motion of suspended particles in a solute concentration gradient due to slip flow at the surface [1, 2, 6, 7]. The physical origin of diffusiophoresis is analogous to that of diffusioosmotic flow and results from chemiphoretic and electrophoretic contributions in electrolyte solutions [1, 2]. Since its discovery by Derjaguin et al. [6], diffusiophoresis has been widely studied as a driving mechanism for the migration and manipulation of particles [3, 4, 15, 16, 17, 18, 19]. In particular, understanding the coupled effects of diffusioosmosis and diffusiophoresis on particle dynamics in various microfluidic configurations, including dead-end channels and T-junctions, has received much attention in the fluid mechanics and soft matter communities through analytical modeling, numerical simulations, and experiments [9, 20, 21, 22, 23, 24, 25, 26, 27]. We refer the reader to the recent review articles by Shim [3] and Ault and Shin [4] and the references therein, for a further discussion on diffusioosmotic flow and diffusiophoresis.

Microfluidic configurations are often fabricated from soft materials such as poly(dimethylsiloxane) (PDMS) [28], and as a result, viscous flows through such configurations may induce elastic deformation of the channel crosssection [29]. Elucidating the fluid–structure interaction between internal fluid flow and the elastic boundaries of deformable channels plays a central role in understanding hydrodynamic behavior across a wide range of applications, including wearable diagnostics [30], biomedical applications [31, 32], control of viscous fingering [33, 34], fabrication of microfluidic devices [29, 35, 36, 37, 38], soft robotics [39, 40, 41, 42, 43, 44], and flow control in plants’ vasculature [45]. For instance, Elbaz and Gat [40] explored the transient fluid–structure interactions of pressure-driven flows in axisymmetric geometries, demonstrating their relevance to soft robotics applications. Christov et al. [38] employed lubrication theory along with the Kirchhoff–Love bending model to derive the relationship between flow rate and pressure drop in a deformable microchannel, showing that the flow rate–pressure drop relation exhibits a sublinear behavior due to wall compliance. Recently, Boyko et al. [46] demonstrated that the reciprocal theorems for Stokes flow and linear elasticity can be leveraged to derive a closed-form expression for the flow rate–pressure drop relation in deformable channels, thereby bypassing the need to solve the full fluid–structure interaction problem. Martínez-Calvo et al. [47] extended the steady-state analysis of Christov et al. [38] to the start-up flow in deformable microchannels by incorporating fluid and solid inertia into the lubrication and Kirchhoff–Love bending equations, while Guyard et al. [48] studied the elastohydrodynamic relaxation behavior of such configurations. Furthermore, Biviano et al. [49] explored how the compliance of deformable microchannels can be harnessed to attenuate oscillations in peristaltic pump flow using bioinspired passive components.

In tandem with studies on pressure-driven flows through deformable microchannels [29], there has also been growing interest in exploring the role of alternative actuation mechanisms, such as electroosmotic flow, in driving fluid–structure interactions at the microscale [50, 51, 52, 53]. In particular, Rubin et al. [52] and Boyko et al. [53] examined the use of non-uniform electroosmotic flow as a driving mechanism to generate desired dynamic deformations in a lubricated elastic sheet by inducing internal pressure gradients within the fluid. Assuming small deformations and strong pre-stretching of the elastic sheet, Boyko et al. [53] analyzed the linear and weakly nonlinear viscous–elastic interaction driven by non-uniform electroosmotic flow, which showed stable behavior. For a further discussion on low-Reynolds-number fluid–structure interactions, we refer the reader to the overviews given recently by Christov [29] and Rallabandi [54].

It should be noted that in all the studies mentioned above, the fluid–structure interactions exhibited stable behavior. However, the study of low-Reynolds-number fluid–structure interactions is not limited to stable scenarios. Indeed, various fluid–structure instabilities of the fluid–elastic interface have been reported, highlighting the rich dynamics beyond stable behavior [29, 55, 56]. These include viscous flows in collapsible tubes [57], snap-through instabilities arising from the interaction between a buckled elastic arch and viscous flow [58], wrinkling of lubricated elastic sheets under compression [59], underactuated control of multistable elastic membranes using viscous fluids [60], and flow-induced choking in compliant Hele-Shaw cells [61]. Of particular interest are the studies by Boyko et al. [62] and Boyko et al. [63] on fluid–structure interactions driven by non-uniform electroosmotic flow. They showed that above a certain critical value of the imposed electric field, negative pressures formed by non-uniform electroosmotic flow can give rise to the instability of the liquid–elastic interface. Even though diffusioosmotic flow is a common driving mechanism at the microscale, to the best of the authors’ knowledge, neither the fluid–structure interaction nor the fluid–structure instability that may be induced by diffusioosmotic flow in deformable channels has been studied to date, thereby motivating the present study.

In this theoretical work, we analyze the low-Reynolds-number fluid–structure interaction between diffusioosmotic flow and a deformable microchannel. We provide insight into the physical behavior of the system by developing a simplified model, in which a viscous film is lubricated between a rigid bottom surface and an elastic top substrate, modeled as a rigid plate connected to a linear spring. Diffusioosmotic flow, driven by a solute concentration difference at the edges, induces a fluidic pressure exerted on the plate, leading to the fluid–structure interaction. Since the pressure generated by diffusioosmotic flow scales inversely with the fluidic film thickness, as we show, exceeding a certain solute concentration difference between the edges can result in sufficiently negative pressures to trigger fluid–structure instability, leading to the collapse of the elastic top substrate onto the bottom surface.

The paper is organized as follows. In Sec. II, we present the problem formulation and governing equations for the fluid dynamics and the solute transport. In Sec. III, we present a reduced-order model that describes fluid–structure interaction driven by diffusioosmotic flow. We derive a pair of two-way coupled equations governing the evolution of the fluidic film thickness and the solute concentration. We provide their scaling and introduce key non-dimensional parameters. In Sec. IV, we present our theoretical results for the fluid–structure interaction dynamics and demonstrate the onset of fluid–structure instability driven by diffusioosmotic flow. We elucidate the underlying physical mechanisms that lead to the onset of fluid–structure instability and identify three distinct dynamic regimes. We further present a linear stability analysis of the system and rationalize the dynamic predictions of the theoretical model. We validate the results of our theoretical model by performing two-dimensional finite-element numerical simulations, finding excellent agreement. Finally, we conclude with a discussion of the results in § 2.1.

Problem formulation and governing equations

We study the fluid–structure interaction and instability of a viscous fluid with density ρ~\tilde{\rho}, viscosity μ~\tilde{\mu}, and permittivity ε~\tilde{\varepsilon}, driven by diffusioosmotic flow and confined between a rigid bottom surface and a rigid top plate. The top plate has density ρ~s\tilde{\rho}_s, thickness b~\tilde{b}, and length ℓ~\tilde{\ell}, and is connected to a linear spring with stiffness k~\tilde{k}. The initial film thickness, representing the initial gap between the flat surfaces, is given by h~i\tilde{h}_i. We consider a planar two-dimensional configuration and employ a Cartesian coordinate system (x~,z~)(\tilde{x},\tilde{z}), whose x~\tilde{x} axis lies at the bottom flat surface and z~\tilde{z} is perpendicular thereto, as shown in Figure 1.

Image not extracted

Figure 1. Schematic illustration of the modeled configuration, showing the coordinate system and relevant physical and geometric parameters. A thin viscous film of initial thickness h~i\tilde{h}_i is confined between a rigid bottom surface and a rigid top plate, connected to a linear spring of stiffness k~\tilde{k}. A solute concentration difference imposed at the edges generates a local concentration gradient, inducing a diffusioosmotic slip velocity u~DOF(x~,t~)\tilde{\boldsymbol{u}}_{\mathrm{DOF}}(\tilde{x},\tilde{t}) along the bottom surface. This slip velocity drives a diffusioosmotic flow, which generates an internal fluidic pressure distribution acting on the top plate, resulting in fluid–structure interaction and the time evolution of the film thickness h~(t~)\tilde{h}(\tilde{t}).

We assume that, initially, the solute concentration c~\tilde{c} is uniform, such that c~(x~,z~,t~=0)=c~L\tilde{c}(\tilde{x},\tilde{z},\tilde{t}=0)=\tilde{c}_L, where t~\tilde{t} denotes time. We induce a concentration gradient by imposing a solute concentration difference between the left and right edges, c~(x~=0,z~,t~)=c~L\tilde{c}(\tilde{x}=0,\tilde{z},\tilde{t})=\tilde{c}_L and c~(x~=ℓ~,z~,t~)=c~R\tilde{c}(\tilde{x}=\tilde{\ell},\tilde{z},\tilde{t})=\tilde{c}_R, which generates a diffusioosmotic slip velocity at the bottom surface, u~DOF(x~,t~)=−Γ~∇∥ln⁡c~\tilde{\boldsymbol{u}}_{\mathrm{DOF}}(\tilde{x},\tilde{t})=-\tilde{\Gamma}\boldsymbol{\nabla}_{\parallel}\ln\tilde{c} [1, 2, 4], where Γ~\tilde{\Gamma} is the diffusioosmotic mobility and ∇∥=(∂/∂x~)x^\boldsymbol{\nabla}_{\parallel}=(\partial/\partial\tilde{x})\hat{\boldsymbol{x}} is the gradient parallel to the surface. This slip velocity drives a diffusioosmotic flow with pressure distribution p~\tilde{p} and velocity u~=(u~,w~)\tilde{\boldsymbol{u}}=(\tilde{u},\tilde{w}), which in turn causes the time evolution of the gap h~(t~)\tilde{h}(\tilde{t}) and leads to solute advection in addition to diffusion. The characteristic velocities in the x^\hat{\boldsymbol{x}} and z^\hat{\boldsymbol{z}} directions are u~c\tilde{u}_c and w~c\tilde{w}_c, respectively, while the characteristic pressure and time are denoted by p~c\tilde{p}_c and t~c\tilde{t}_c. For simplicity, we assume that the influence of gravity is negligible.

The time evolution of the gap h~(t~)\tilde{h}(\tilde{t}) is governed by the coupled dynamics of the fluid motion and solute transport within the channel. In this work, we focus on slender configurations, where h~(t~)≪ℓ~\tilde{h}(\tilde{t}) \ll\tilde{\ell}, and assume negligible fluidic inertia, represented by small Womersley and reduced Reynolds numbers,

ϵ=h~iℓ~≪1,Wo=ρ~h~i2μ~t~c≪1,eRe=ϵρ~u~ch~iμ~≪1,(1)\epsilon=\frac{\tilde{h}_i}{\tilde{\ell}}\ll1,\qquad Wo=\frac{\tilde{\rho}\tilde{h}_i^2}{\tilde{\mu}\tilde{t}_c}\ll1,\qquad eRe=\epsilon\frac{\tilde{\rho}\tilde{u}_c\tilde{h}_i}{\tilde{\mu}}\ll1, \tag*{(1)}

corresponding to the assumptions of the classical lubrication approximation [64]. Therefore, the fluid motion is governed by the lubrication equations [64]

∂u~∂x~+∂w~∂z~=0,(2)\frac{\partial\tilde{u}}{\partial\tilde{x}}+\frac{\partial\tilde{w}}{\partial\tilde{z}}=0, \tag*{(2)}
∂p~∂x~=μ~∂2u~∂z~2,(3)\frac{\partial\tilde{p}}{\partial\tilde{x}}=\tilde{\mu}\frac{\partial^2\tilde{u}}{\partial\tilde{z}^2}, \tag*{(3)}
∂p~∂z~=0,(4)\frac{\partial\tilde{p}}{\partial\tilde{z}}=0, \tag*{(4)}

and the solute transport is governed by the advection–diffusion equation

∂c~∂t~+u~∂c~∂x~+w~∂c~∂z~=D~(∂2c~∂x~2+∂2c~∂z~2),(5)\frac{\partial\tilde{c}}{\partial\tilde{t}}+\tilde{u}\frac{\partial\tilde{c}}{\partial\tilde{x}}+\tilde{w}\frac{\partial\tilde{c}}{\partial\tilde{z}}=\tilde{D}\left(\frac{\partial^2\tilde{c}}{\partial\tilde{x}^2}+\frac{\partial^2\tilde{c}}{\partial\tilde{z}^2}\right), \tag*{(5)}

where D~\tilde{D} is the solute diffusivity. From Eq. (2c), it follows that p~=p~(x~,t~)\tilde{p}=\tilde{p}(\tilde{x},\tilde{t}), i.e., the pressure is uniform across the gap and varies only along the axial direction and in time.

These governing equations are subject to the following boundary conditions on the fluid and solute

Diffusioosmotic slip at the bottom surface: u~(x~,z~=0,t~)=u~DOF(x~,t~)=−Γ~∂ln⁡c~∂x~∣z~=0,\tilde{u}(\tilde{x},\tilde{z}=0,\tilde{t})=\tilde{u}_{\mathrm{DOF}}(\tilde{x},\tilde{t})=-\tilde{\Gamma}\left.\frac{\partial\ln\tilde{c}}{\partial\tilde{x}}\right|_{\tilde{z}=0}, (4a)

No fluid penetration at the bottom surface: w~(x~,z~=0,t~)=0,\tilde{w}(\tilde{x},\tilde{z}=0,\tilde{t})=0, (4b)

No slip at the top plate: u~(x~,z~=h~(t~),t~)=0,\tilde{u}(\tilde{x},\tilde{z}=\tilde{h}(\tilde{t}),\tilde{t})=0, (4c)

Kinematic condition at the top plate: w~(x~,z~=h~(t~),t~)=dh~dt~,\tilde{w}(\tilde{x},\tilde{z}=\tilde{h}(\tilde{t}),\tilde{t})=\frac{\mathrm{d}\tilde{h}}{\mathrm{d}\tilde{t}}, (4d)

Fixed pressure at the edges: p~(x~=0,z~,t~)=p~(x~=ℓ~,z~,t~)=0,\tilde{p}(\tilde{x}=0,\tilde{z},\tilde{t})=\tilde{p}(\tilde{x}=\tilde{\ell},\tilde{z},\tilde{t})=0, (4e)

Fixed concentration at the edges: c~(x~=0,z~,t~)=c~L,c~(x~=ℓ~,z~,t~)=c~R,\tilde{c}(\tilde{x}=0,\tilde{z},\tilde{t})=\tilde{c}_{L},\quad\tilde{c}(\tilde{x}=\tilde{\ell},\tilde{z},\tilde{t})=\tilde{c}_{R}, (4f)

No-flux conditions at the bottom and top surfaces: ∂c~∂z~∣z~=0=0,D~∂c~∂z~∣z~=h~(t~)=dh~dt~c~∣z~=h~(t~).\left.\frac{\partial\tilde{c}}{\partial\tilde{z}}\right|_{\tilde{z}=0}=0,\quad\left.\tilde{D}\frac{\partial\tilde{c}}{\partial\tilde{z}}\right|_{\tilde{z}=\tilde{h}(\tilde{t})}=\left.\frac{\mathrm{d}\tilde{h}}{\mathrm{d}\tilde{t}}\tilde{c}\right|_{\tilde{z}=\tilde{h}(\tilde{t})}. (4g)

The slip boundary condition (4a) represents the diffusioosmotic slip velocity, which serves as the driving mechanism for the fluid–structure interaction in our study. This diffusioosmotic flow arises from solute concentration gradients introduced into the system through the boundary conditions (4f), where c~R\tilde{c}_{R} is the prescribed solute concentration at the right edge. The last boundary condition (4g) represents the no-flux conditions at the bottom and top surfaces, i.e., j~⋅n^=0\tilde{\boldsymbol{j}}\cdot\hat{\boldsymbol{n}}=0 at z~=0,h~(t~)\tilde{z}=0,\tilde{h}(\tilde{t}), where j~=c~u~−D~∇c~\tilde{\boldsymbol{j}}=\tilde{c}\tilde{\boldsymbol{u}}-\tilde{D}\nabla\tilde{c} is the molar flux. Noting that n^=±z^\hat{\boldsymbol{n}}=\pm\hat{\boldsymbol{z}}, the latter condition leads to j~⋅n^=w~c~−D~∂c~/∂z~=0\tilde{\boldsymbol{j}}\cdot\hat{\boldsymbol{n}}=\tilde{w}\tilde{c}-\tilde{D}\partial\tilde{c}/\partial\tilde{z}=0 at z~=0,h~(t~)\tilde{z}=0,\tilde{h}(\tilde{t}). Thus, using Eqs. (4b) and (4d), the no-flux boundary conditions take the form given in Eq. (4g).

As initial conditions, we assume a uniform solute concentration throughout the channel,

c~(x~,z~,t~=0)=c~L.(6)\tilde{c}(\tilde{x},\tilde{z},\tilde{t}=0)=\tilde{c}_{L}. \tag*{(6)}

and an initial gap height,

h~(t~=0)=h~i.(7)\tilde{h}(\tilde{t}=0)=\tilde{h}_{i}. \tag*{(7)}

In this work, we assume a constant zeta potential and diffusioosmotic mobility coefficient Γ~\tilde{\Gamma}, which is a reasonable first approximation under many scenarios, as discussed by recent previous studies [27, 65, 66]. This assumption is widely used in the literature (see, e.g., [9, 13, 14, 23]), as it enables the derivation of closed-form analytical results and is typically valid when the ratio of solute concentrations in the system is not too large (i.e., c~R/c~L\tilde{c}_{R}/\tilde{c}_{L} is not too large). In addition, we neglect any dispersion because the effects of diffusioosmotic flow on the dispersion are O(ϵ2)O(\epsilon^{2}). Indeed, recent studies by Alessio et al. [26] and Teng et al. [14] have demonstrated that the effective diffusivity D~eff\tilde{D}_{\mathrm{eff}} induced by diffusioosmotic flow in a two-dimensional channel is given by

D~eff=D~[1+(Γ~D~)2ϵ2210(∂ln⁡c~∂x~)2+O(ϵ4)],(8)\tilde{D}_{\mathrm{eff}}=\tilde{D}\left[1+\left(\frac{\tilde{\Gamma}}{\tilde{D}}\right)^{2}\frac{\epsilon^{2}}{210}\left(\frac{\partial\ln\tilde{c}}{\partial\tilde{x}}\right)^{2}+O(\epsilon^{4})\right], \tag*{(8)}

and therefore, the effective diffusivity can be approximated by the solute diffusivity to the leading order in ϵ\epsilon.

Reduced-order model for fluid–structure interaction driven by diffusioosmotic flow

In this section, we develop a reduced-order model to describe fluid–structure interaction driven by diffusioosmotic flow, and derive the governing equations for the coupled evolution of solute concentration and gap height.

Evolution equation for the film thickness

Applying the lubrication approximation and using Eqs. (2) and (4a)–(4d), the evolution of the film thickness h~(t~)\tilde{h}(\tilde{t}) is related to the fluidic pressure p~(x~,t~)\tilde{p}(\tilde{x},\tilde{t}) by the Reynolds equation (see p. 313 in [64])

dh~(t~)dt~−h~(t~)312μ~∂2p~(x~,t~)∂x~2=−h~(t~)2∂u~DOF(x~,t~)∂x~.(9)\frac{\mathrm{d}\tilde{h}(\tilde{t})}{\mathrm{d}\tilde{t}}-\frac{\tilde{h}(\tilde{t})^{3}}{12\tilde{\mu}}\frac{\partial^{2}\tilde{p}(\tilde{x},\tilde{t})}{\partial\tilde{x}^{2}}=-\frac{\tilde{h}(\tilde{t})}{2}\frac{\partial\tilde{u}_{\mathrm{DOF}}(\tilde{x},\tilde{t})}{\partial\tilde{x}}. \tag*{(9)}

The term on the right-hand side of Eq. (9) represents spatial variations in diffusioosmotic flux 12h~u~DOF\frac{1}{2}\tilde{h}\tilde{u}_{\mathrm{DOF}}, which drives the fluid–structure interaction. Solving the Reynolds equation (9) subject to Eq. (4e) yields the fluidic pressure p~(x~,t~)\tilde{p}(\tilde{x},\tilde{t}),

p~(x~,t~)=6μ~h~3dh~dt~x~(x~−ℓ~)+6μ~h~2(A~(x~,t~)−x~ℓ~A~(ℓ~,t~)),(10)\tilde{p}(\tilde{x},\tilde{t}) = \frac{6\tilde{\mu}}{\tilde{h}^{3}}\frac{d\tilde{h}}{d\tilde{t}}\tilde{x}(\tilde{x}-\tilde{\ell})+\frac{6\tilde{\mu}}{\tilde{h}^{2}}\left(\tilde{A}(\tilde{x},\tilde{t})-\frac{\tilde{x}}{\tilde{\ell}}\tilde{A}(\tilde{\ell},\tilde{t})\right), \tag*{(10)}

where A~(x~,t~)\tilde{A}(\tilde{x},\tilde{t}) is defined as

A~(x~,t~)=∫0x~u~DOF(x~,t~) dx~=−Γ~ln⁡(c~(x~,z~=0,t~)c~L).(11)\tilde{A}(\tilde{x},\tilde{t})=\int_{0}^{\tilde{x}}\tilde{u}_{\mathrm{DOF}}(\tilde{x},\tilde{t})\,d\tilde{x}=-\tilde{\Gamma}\ln\left(\frac{\tilde{c}(\tilde{x},\tilde{z}=0,\tilde{t})}{\tilde{c}_{L}}\right). \tag*{(11)}

Integrating Eq. (2b) twice with respect to z~\tilde{z} and applying the boundary conditions (4a) and (4c) at the bottom and top surfaces, we obtain the expression for axial velocity u~(x~,z~,t~)\tilde{u}(\tilde{x},\tilde{z},\tilde{t}). Substituting this result into the definition of the flow rate q~(x~,t~)=∫0h~(t~)u~ dz~\tilde{q}(\tilde{x},\tilde{t})=\int_{0}^{\tilde{h}(\tilde{t})}\tilde{u}\,d\tilde{z} and using Eqs. (10)–(11), the flow rate can be expressed as

q~(x~,t~)=−h~312μ~∂p~∂x~+h~2u~DOF=ℓ~dh~dt~(12−x~ℓ~)+h~2ℓ~A~(ℓ~,t~)=ℓ~dh~dt~(12−x~ℓ~)−h~2ℓ~Γ~ln⁡(c~Rc~L).(12)\tilde{q}(\tilde{x},\tilde{t})=-\frac{\tilde{h}^{3}}{12\tilde{\mu}}\frac{\partial\tilde{p}}{\partial\tilde{x}}+\frac{\tilde{h}}{2}\tilde{u}_{\mathrm{DOF}}=\tilde{\ell}\frac{d\tilde{h}}{d\tilde{t}}\left(\frac{1}{2}-\frac{\tilde{x}}{\tilde{\ell}}\right)+\frac{\tilde{h}}{2\tilde{\ell}}\tilde{A}(\tilde{\ell},\tilde{t})=\tilde{\ell}\frac{d\tilde{h}}{d\tilde{t}}\left(\frac{1}{2}-\frac{\tilde{x}}{\tilde{\ell}}\right)-\frac{\tilde{h}}{2\tilde{\ell}}\tilde{\Gamma}\ln\left(\frac{\tilde{c}_{R}}{\tilde{c}_{L}}\right). \tag*{(12)}

We note that, in the case of h~=const.\tilde{h}=\mathrm{const.}, Eq. (12) is consistent with the expression for the diffusioosmotic flow rate in a channel, as derived by Lee et al. [8]. Integrating Eq. (10) with respect to x~\tilde{x} from 00 to ℓ~\tilde{\ell} provides the expression for the fluidic force F~f(t~)=∫0ℓ~p~ dx~\tilde{F}_{f}(\tilde{t})=\int_{0}^{\tilde{\ell}}\tilde{p}\,d\tilde{x},

F~f(t~)=−μ~ℓ~3h~(t~)3dh~(t~)dt~+6μ~B~(t~)h~(t~)2,(13)\tilde{F}_{f}(\tilde{t})=-\frac{\tilde{\mu}\tilde{\ell}^{3}}{\tilde{h}(\tilde{t})^{3}}\frac{d\tilde{h}(\tilde{t})}{d\tilde{t}}+\frac{6\tilde{\mu}\tilde{B}(\tilde{t})}{\tilde{h}(\tilde{t})^{2}}, \tag*{(13)}

where B~(t~)\tilde{B}(\tilde{t}) is defined as

B~(t~)=∫0ℓ~[A~(x~,t~)−x~ℓ~A~(ℓ~,t~)]dx~=∫0ℓ~∫0x~u~DOF(x~′,t~)dx~′dx~−ℓ~2∫0ℓ~u~DOF(x~,t~)dx~.(14)\tilde{B}(\tilde{t})=\int_{0}^{\tilde{\ell}}\left[\tilde{A}(\tilde{x},\tilde{t})-\frac{\tilde{x}}{\tilde{\ell}}\tilde{A}(\tilde{\ell},\tilde{t})\right]d\tilde{x}=\int_{0}^{\tilde{\ell}}\int_{0}^{\tilde{x}}\tilde{u}_{\mathrm{DOF}}(\tilde{x}',\tilde{t})d\tilde{x}'d\tilde{x}-\frac{\tilde{\ell}}{2}\int_{0}^{\tilde{\ell}}\tilde{u}_{\mathrm{DOF}}(\tilde{x},\tilde{t})d\tilde{x}. \tag*{(14)}

The first term on the right-hand side of Eq. (13) represents the viscous resistance and the second term represents the diffusioosmotic force, which can be either attractive or repulsive depending on the sign of B~\tilde{B}.

Substituting the expression for the diffusioosmotic slip velocity (4a) into Eq. (14) and using the boundary conditions (4f), B~(t~)\tilde{B}(\tilde{t}) takes the form

B~(t~)=−Γ~∫0ℓ~ln⁡(c~(x~,z~=0,t~)c~L)dx~+ℓ~2Γ~ln⁡(c~Rc~L).(15)\tilde{B}(\tilde{t})=-\tilde{\Gamma}\int_{0}^{\tilde{\ell}}\ln\left(\frac{\tilde{c}(\tilde{x},\tilde{z}=0,\tilde{t})}{\tilde{c}_{L}}\right)d\tilde{x}+\frac{\tilde{\ell}}{2}\tilde{\Gamma}\ln\left(\frac{\tilde{c}_{R}}{\tilde{c}_{L}}\right). \tag*{(15)}

In addition to the fluidic force, an elastic force F~e\tilde{F}_{e} acts on the top plate, which we model as a linear restoring force proportional to the variation in the film thickness

F~e=k~(h~i−h~(t~)).(16)\tilde{F}_{e}=\tilde{k}(\tilde{h}_{i}-\tilde{h}(\tilde{t})). \tag*{(16)}

Considering the force balance acting on the plate and accounting for the fluidic force Eq. (13) and the elastic force Eq. (16), we obtain the governing equation for the plate motion

ρ~sb~ℓ~d2h~dt~2=F~f+F~e.(17)\tilde{\rho}_{s}\tilde{b}\tilde{\ell}\frac{d^{2}\tilde{h}}{d\tilde{t}^{2}}=\tilde{F}_{f}+\tilde{F}_{e}. \tag*{(17)}

In this study, we focus on the viscous–elastic time scale (ℓ~/h~i)3(μ~/k~)(\tilde{\ell}/\tilde{h}_{i})^{3}(\tilde{\mu}/\tilde{k}) (see, e.g., [40, 62]), which is typically much longer than the characteristic elastic inertial time scale and therefore we neglect solid inertia and any vibrations it may induce. Neglecting the solid inertia, ρ~sb~ℓ~/(k~t~2)≪1\tilde{\rho}_{s}\tilde{b}\tilde{\ell}/(\tilde{k}\tilde{t}^{2})\ll1, and substituting the expressions for F~f\tilde{F}_{f} and F~e\tilde{F}_{e}, given in Eqs. (13) and (16), into Eq. (17), yields the evolution equation for h~(t~)\tilde{h}(\tilde{t})

−μ~ℓ~3h~(t~)3dh~(t~)dt~⏟Viscous resistance+6μ~B~(t~)h~(t~)2⏟Diffusioosmotic forcing+k~(h~i−h~(t~))⏟Elastic forcing=0.(18)-\underbrace{\frac{\tilde{\mu}\tilde{\ell}^{3}}{\tilde{h}(\tilde{t})^{3}}\frac{d\tilde{h}(\tilde{t})}{d\tilde{t}}}_{\text{Viscous resistance}}+\underbrace{\frac{6\tilde{\mu}\tilde{B}(\tilde{t})}{\tilde{h}(\tilde{t})^{2}}}_{\text{Diffusioosmotic forcing}}+\underbrace{\tilde{k}(\tilde{h}_{i}-\tilde{h}(\tilde{t}))}_{\text{Elastic forcing}}=0. \tag*{(18)}

Equation (18) is a nonlinear governing equation for h~(t~)\tilde{h}(\tilde{t}), which accounts for viscous, diffusioosmotic, and elasticity effects, and describes the gap evolution driven by diffusioosmotic flow.

We note that similar nonlinear evolution equations are encountered in various instability problems, such as electrostatic MEMS actuators [67], elastocapillary coalescence [68], and fluid–structure instability driven by non-uniform electro-osmotic flow [62]. For non-uniform electro-osmotic flow, the fluid–structure interaction is described by a single evolution equation for h~(t~)\tilde{h}(\tilde{t}) [62]. However, for diffusioosmotic flow, the evolution of the gap depends on a time-varying function B~(t~)\tilde{B}(\tilde{t}), which is determined by the solute concentration distribution, as shown in Eq. (15). Consequently, the gap dynamics are coupled with solute transport, for which we derive a governing equation in Sec. III C.

Finally, we note that in electrolyte solutions with unequal cation and anion diffusivities, the introduction of a concentration gradient spontaneously induces an electric field, E~induced\tilde{\mathbf{E}}_{\mathrm{induced}}, which arises to maintain the no-current condition [1, 2]. For a symmetric binary (z:zz:z) electrolyte with unequal cation and anion diffusion coefficients, the induced electric field is E~induced=βDϕ~T∇∥ln⁡c~\tilde{\mathbf{E}}_{\mathrm{induced}}=\beta_D\tilde{\phi}_T\nabla_{\parallel}\ln\tilde{c}, where ϕ~T=k~BT~/ze~\tilde{\phi}_T=\tilde{k}_B\tilde{T}/z\tilde{e} is the thermal voltage and βD=(D~+−D~−)/(D~++D~−)\beta_D=(\tilde{D}_{+}-\tilde{D}_{-})/(\tilde{D}_{+}+\tilde{D}_{-}) is the normalized diffusivity difference [1, 2, 4]. Here, k~B\tilde{k}_B is the Boltzmann constant, T~\tilde{T} is the absolute temperature, e~\tilde{e} is the elementary charge, zz is the ion valence, and D~+\tilde{D}_{+} and D~−\tilde{D}_{-} are the diffusivities of the cations and anions, respectively. The induced electric field gives rise to Maxwell stresses acting on the top plate [62]. Assuming that the permittivities of both the rigid plate and the surrounding air are negligible relative to that of the fluid, we neglect their contributions to the Maxwell stress and consider only the dielectric contribution of the fluid, given as 12ε~E~induced2\frac{1}{2}\tilde{\varepsilon}\tilde{E}_{\mathrm{induced}}^2. Therefore, the resulting upward-directed dielectric force is given by

F~d=12ε~∫0ℓ~E~induced2dx~=12βD2ε~ϕ~T2M~withM~(t~)=∫0ℓ~[∂∂x~ln⁡(c~(x~,z~=0,t~)c~L)]2dx~.(19)\tilde{F}_d=\frac{1}{2}\tilde{\varepsilon}\int_{0}^{\tilde{\ell}}\tilde{E}_{\mathrm{induced}}^2\mathrm{d}\tilde{x}=\frac{1}{2}\beta_D^2\tilde{\varepsilon}\tilde{\phi}_T^2\tilde{\mathcal{M}}\quad\text{with}\quad\tilde{\mathcal{M}}(\tilde{t})=\int_{0}^{\tilde{\ell}}\left[\frac{\partial}{\partial\tilde{x}}\ln\left(\frac{\tilde{c}(\tilde{x},\tilde{z}=0,\tilde{t})}{\tilde{c}_L}\right)\right]^2\mathrm{d}\tilde{x}. \tag*{(19)}

While, in principle, the dielectric force F~d\tilde{F}_d, Eq. (19), should be included in the force balance Eq. (17), an evaluation using the physical parameters listed in Table I for a representative microfluidic configuration reveals that the ratio between the dielectric and diffusioosmotic forces scales as βD2ε~ϕ~T2h~2/(μ~Γ~ℓ~2)≈1.8×10−4βD2\beta_D^2\tilde{\varepsilon}\tilde{\phi}_T^2\tilde{h}^2/(\tilde{\mu}\tilde{\Gamma}\tilde{\ell}^2)\approx1.8\times10^{-4}\beta_D^2. Since βD\beta_D is less than unity, we obtain that the dielectric contribution is negligible compared to the diffusioosmotic forcing. Moreover, for a symmetric binary electrolyte such as KCl, where the cation and anion have equal diffusivities [69], βD=0\beta_D=0, and consequently, the dielectric force F~d\tilde{F}_d vanishes. Therefore, in the subsequent analysis, we restrict our attention to the case of βD=0\beta_D=0 and neglect the contribution of dielectric forces.

Scaling analysis and non-dimensionalization of the evolution equation

Scaling by the characteristic dimensions, we introduce the non-dimensional variables

h=h~h~i,x=x~ℓ~,z=z~h~i,t=t~t~c,p=p~p~c,uDOF=u~DOFu~c,c=c~c~L,cR=c~Rc~L,(20)h=\frac{\tilde{h}}{\tilde{h}_i},\qquad x=\frac{\tilde{x}}{\tilde{\ell}},\qquad z=\frac{\tilde{z}}{\tilde{h}_i},\qquad t=\frac{\tilde{t}}{\tilde{t}_c},\qquad p=\frac{\tilde{p}}{\tilde{p}_c},\qquad u_{\mathrm{DOF}}=\frac{\tilde{u}_{\mathrm{DOF}}}{\tilde{u}_c},\qquad c=\frac{\tilde{c}}{\tilde{c}_L},\qquad c_R=\frac{\tilde{c}_R}{\tilde{c}_L}, \tag*{(20)}

where u~c=Γ~/ℓ~\tilde{u}_c=\tilde{\Gamma}/\tilde{\ell} is the characteristic axial velocity based on the diffusioosmotic slip, p~c=μ~u~cℓ~/h~i2\tilde{p}_c=\tilde{\mu}\tilde{u}_c\tilde{\ell}/\tilde{h}_i^2 is the characteristic lubrication pressure scale, and t~c=(ℓ~/h~i)3(μ~/k~)\tilde{t}_c=(\tilde{\ell}/\tilde{h}_i)^3(\tilde{\mu}/\tilde{k}) is the characteristic time scale. We note that since the diffusioosmotic mobility Γ~\tilde{\Gamma} scales inversely with the fluid viscosity μ~\tilde{\mu} [1, 2], the characteristic pressure based on the diffusioosmotic slip velocity is independent of the viscosity.

Substituting Eq. (20) into Eq. (18), we obtain the non-dimensional governing equation for the gap evolution

1h(t)3dh(t)dt⏟Viscous resistance=427FDOFB(t)h(t)⏟Diffusioosmotic force+1−h(t)⏟Elastic forcewithB(t)=−∫01ln⁡c dx+12ln⁡cR,(21)\underbrace{\frac{1}{h(t)^3}\frac{\mathrm{d}h(t)}{\mathrm{d}t}}_{\text{Viscous resistance}}=\underbrace{\frac{4}{27}\mathcal{F}_{\mathrm{DOF}}\frac{B(t)}{h(t)}}_{\text{Diffusioosmotic force}}+\underbrace{1-h(t)}_{\text{Elastic force}}\quad\text{with}\quad B(t)=-\int_{0}^{1}\ln c\,\mathrm{d}x+\frac{1}{2}\ln c_R, \tag*{(21)}

where we have introduced a non-dimensional parameter, referred to as the elasto-diffusioosmotic number,

FDOF=812t~cℓ~/u~c=(81/2)μ~Γ~ℓ~/h~i2k~h~i/ℓ~,(22)\mathcal{F}_{\mathrm{DOF}}=\frac{81}{2}\frac{\tilde{t}_c}{\tilde{\ell}/\tilde{u}_c}=\frac{(81/2)\tilde{\mu}\tilde{\Gamma}\tilde{\ell}/\tilde{h}_i^2}{\tilde{k}\tilde{h}_i/\tilde{\ell}}, \tag*{(22)}

which represents the ratio between the diffusioosmotic driving forces and the elastic restoring forces. In addition, the elasto-diffusioosmotic number FDOF\mathcal{F}_{\mathrm{DOF}} can be interpreted as the ratio of the viscous–elastic to advection time scales.

The evolution equation (21) for the gap h(t)h(t) represents the force balance between the viscous resistance, the diffusioosmotic force, and the restoring effect of the elastic force. For βD=0\beta_D=0, corresponding to diffusioosmotic flow driven purely by the chemiosmotic contribution, both Γ~\tilde{\Gamma} and the elasto-diffusioosmotic number FDOF\mathcal{F}_{\mathrm{DOF}} are strictly positive. In contrast, the function B(t)B(t) may take either positive or negative values, depending on the solute concentration profile. Therefore, the direction of the diffusioosmotic forcing may vary over time. Since the function B(t)B(t) depends on the solute concentration, we derive an additional governing equation for solute transport in the following subsection. Together, this set of equations provides a complete description of the time evolution of the gap.

Advection–diffusion equation governing the solute transport

In this subsection, we derive a governing equation for solute transport, which should be solved together with Eq. (21) to determine h(t)h(t). Using the non-dimensionalization Eq. (20), the advection–diffusion equation (5) takes the form

ϵ2ℓ~2/D~t~c∂c∂t+ϵ2ℓ~2/D~ℓ~/u~c(u∂c∂x+w∂c∂z)=ϵ2∂2c∂x2+∂2c∂z2.(23)\epsilon^{2}\frac{\tilde{\ell}^{2}/\tilde{D}}{\tilde{t}_{c}}\frac{\partial c}{\partial t}+\epsilon^{2}\frac{\tilde{\ell}^{2}/\tilde{D}}{\tilde{\ell}/\tilde{u}_{c}}\left(u\frac{\partial c}{\partial x}+w\frac{\partial c}{\partial z}\right)=\epsilon^{2}\frac{\partial^{2}c}{\partial x^{2}}+\frac{\partial^{2}c}{\partial z^{2}}. \tag*{(23)}

We recognize three relevant time scales in Eq. (23). These time scales involve the axial diffusive time scale ℓ~2/D~\tilde{\ell}^{2}/\tilde{D}, the axial advection time scale ℓ~/u~c\tilde{\ell}/\tilde{u}_{c}, and the viscous–elastic time scale (ℓ~/h~i)3(μ~/k~)(\tilde{\ell}/\tilde{h}_{i})^{3}(\tilde{\mu}/\tilde{k}), which we choose as the characteristic time scale for non-dimensionalization. Using the definition of the elasto-diffusioosmotic number FDOF\mathcal{F}_{\mathrm{DOF}} in Eq. (22), Eq. (23) can be expressed as

ϵ2α∂c∂t+2ϵ2FDOF81α(u∂c∂x+w∂c∂z)=ϵ2∂2c∂x2+∂2c∂z2,(24)\frac{\epsilon^{2}}{\alpha}\frac{\partial c}{\partial t}+\frac{2\epsilon^{2}\mathcal{F}_{\mathrm{DOF}}}{81\alpha}\left(u\frac{\partial c}{\partial x}+w\frac{\partial c}{\partial z}\right)=\epsilon^{2}\frac{\partial^{2}c}{\partial x^{2}}+\frac{\partial^{2}c}{\partial z^{2}}, \tag*{(24)}

where we have introduced a non-dimensional parameter, referred to as the elasto-diffusion number α\alpha,

α=t~cℓ~2/D~=μ~D~ℓ~k~h~i3,(25)\alpha=\frac{\tilde{t}_{c}}{\tilde{\ell}^{2}/\tilde{D}}=\frac{\tilde{\mu}\tilde{D}\tilde{\ell}}{\tilde{k}\tilde{h}_{i}^{3}}, \tag*{(25)}

which represents the ratio of the viscous–elastic time scale (ℓ~/h~i)3(μ~/k~)(\tilde{\ell}/\tilde{h}_{i})^{3}(\tilde{\mu}/\tilde{k}) to the axial diffusive time scale ℓ~2/D~\tilde{\ell}^{2}/\tilde{D}. Recalling that 2FDOF/812\mathcal{F}_{\mathrm{DOF}}/81 is the ratio of the viscous–elastic time scale (ℓ~/h~i)3(μ~/k~)(\tilde{\ell}/\tilde{h}_{i})^{3}(\tilde{\mu}/\tilde{k}) to the advection time scale ℓ~/u~c\tilde{\ell}/\tilde{u}_{c}, we note the ratio 2FDOF/81α2\mathcal{F}_{\mathrm{DOF}}/81\alpha appearing in the second term on the left-hand side of Eq. (24) can be interpreted as the Péclet number Pe=2FDOF/81α=u~cℓ~/D~=Γ~/D~Pe=2\mathcal{F}_{\mathrm{DOF}}/81\alpha=\tilde{u}_{c}\tilde{\ell}/\tilde{D}=\tilde{\Gamma}/\tilde{D}, representing the ratio of the axial diffusive time scale ℓ~2/D~\tilde{\ell}^{2}/\tilde{D} to the advection time scale ℓ~/u~c\tilde{\ell}/\tilde{u}_{c}. We note that the Péclet number is typically an order-one parameter, i.e., Pe=O(1)Pe=O(1) [14, 23].

The governing equation (24) is supplemented by no-flux boundary conditions at the bottom and top surfaces, given in Eq. (4g), which, using Eq. (20), can be expressed in a non-dimensional form as

∂c∂z∣z=0=0and∂c∂z∣z=h(t)=ϵ2αdh(t)dtc∣z=h(t).(26)\left.\frac{\partial c}{\partial z}\right|_{z=0}=0 \quad\text{and} \quad\left.\frac{\partial c}{\partial z}\right|_{z=h(t)}=\frac{\epsilon^{2}}{\alpha}\frac{dh(t)}{dt}\left.c\right|_{z=h(t)}. \tag*{(26)}

Assuming that α=O(1)\alpha=O(1) and Pe=O(1)Pe=O(1), we are interested in the leading-order dynamics through a formal expansion with the small parameter ϵ=h~i/ℓ~≪1\epsilon=\tilde{h}_{i}/\tilde{\ell}\ll1. As can be seen in the non-dimensional governing equation (24) and the boundary conditions (26), only ϵ2\epsilon^{2} appears. Thus, we seek the solution in the form

c=c0+ϵ2c1+O(ϵ4).(27)c=c_{0}+\epsilon^{2}c_{1}+O(\epsilon^{4}). \tag*{(27)}

At the leading order, O(1)O(1), we obtain

∂2c0∂z2=0subject to∂c0∂z∣z=0=0and∂c0∂z∣z=h(t)=0.(28)\frac{\partial^{2}c_{0}}{\partial z^{2}}=0 \quad\text{subject to} \quad\left.\frac{\partial c_{0}}{\partial z}\right|_{z=0}=0 \quad\text{and} \quad\left.\frac{\partial c_{0}}{\partial z}\right|_{z=h(t)}=0. \tag*{(28)}

Integrating the governing equation and applying the no-flux boundary conditions, we find that c0=c0(x,t)c_{0}=c_{0}(x,t), i.e., the solute concentration is independent of zz to the leading order in ϵ\epsilon. At the next order, O(ϵ2)O(\epsilon^{2}), the governing equation (24) takes the form

1α∂c0∂t+2FDOF81αu∂c0∂x=∂2c0∂x2+∂2c1∂z2,(29)\frac{1}{\alpha}\frac{\partial c_{0}}{\partial t}+\frac{2\mathcal{F}_{\mathrm{DOF}}}{81\alpha}u\frac{\partial c_{0}}{\partial x}=\frac{\partial^{2}c_{0}}{\partial x^{2}}+\frac{\partial^{2}c_{1}}{\partial z^{2}}, \tag*{(29)}

and the boundary conditions become

∂c1∂z∣z=0=0and∂c1∂z∣z=h(t)=1αdh(t)dtc0∣z=h(t).(30)\left.\frac{\partial c_1}{\partial z}\right|_{z=0}=0 \quad\text{and}\quad \left.\frac{\partial c_1}{\partial z}\right|_{z=h(t)} =\frac{1}{\alpha}\frac{dh(t)}{dt}c_0\big|_{z=h(t)}. \tag*{(30)}

Recalling that c0=c0(x,t)c_0=c_0(x,t), we integrate Eq. (29) with respect to zz from 0 to h(t)h(t) and use the boundary conditions (30) to eliminate the term involving c1c_1 in Eq. (29) and obtain

∂c0∂th(t)+281FDOF∂∂xq(x,t)=α∂2c0∂x2h(t)+dh(t)dtc0.(31)\frac{\partial c_0}{\partial t}h(t)+\frac{2}{81}\mathcal{F}_{\mathrm{DOF}}\frac{\partial}{\partial x}q(x,t) =\alpha\frac{\partial^2c_0}{\partial x^2}h(t)+\frac{dh(t)}{dt}c_0. \tag*{(31)}

Here, q(x,t)=∫0h(t)u dzq(x,t)=\int_0^{h(t)}u\,dz is the dimensionless flow rate obtained from Eq. (12) as

q(x,t)=ℓ~/u~ct~cdh(t)dt(12−x)−12Γ~ℓ~u~ch(t)ln⁡cR=812FDOFdh(t)dt(12−x)−12h(t)ln⁡cR,(32)q(x,t)=\frac{\tilde{\ell}/\tilde{u}_c}{\tilde{t}_c}\frac{dh(t)}{dt}\left(\frac{1}{2}-x\right)-\frac{1}{2}\frac{\tilde{\Gamma}\tilde{\ell}}{\tilde{u}_c}h(t)\ln c_R =\frac{81}{2\mathcal{F}_{\mathrm{DOF}}}\frac{dh(t)}{dt}\left(\frac{1}{2}-x\right)-\frac{1}{2}h(t)\ln c_R, \tag*{(32)}

where in the last equality in Eq. (32) we have used the fact that 2FDOF/81=t~c/(ℓ~/u~c)2\mathcal{F}_{\mathrm{DOF}}/81=\tilde{t}_c/(\tilde{\ell}/\tilde{u}_c) and u~c=Γ~/ℓ~\tilde{u}_c=\tilde{\Gamma}/\tilde{\ell}.

Finally, substituting Eq. (32) into Eq. (31), the advection–diffusion equation reduces to

∂c∂th(t)+∂c∂x[dh(t)dt(12−x)−181FDOFh(t)ln⁡cR]=α∂2c∂x2h(t)+dh(t)dtc,(33)\frac{\partial c}{\partial t}h(t)+\frac{\partial c}{\partial x}\left[\frac{dh(t)}{dt}\left(\frac{1}{2}-x\right)-\frac{1}{81}\mathcal{F}_{\mathrm{DOF}}h(t)\ln c_R\right] =\alpha\frac{\partial^2c}{\partial x^2}h(t)+\frac{dh(t)}{dt}c, \tag*{(33)}

where we dropped the subscript “0” in c(x,t)c(x,t) for simplicity.

The evolution equations (21) and (33) form a set of two-way coupled nonlinear equations, governing the fluid–structure interaction driven by diffusioosmotic flow, that should be solved at once to determine both the gap height h(t)h(t) and solute concentration c(x,t)c(x,t). The governing equations (21) and (33) are supplemented by the boundary conditions, corresponding to fixed solute concentrations at the left and right edges,

c(x=0,t)=1andc(x=1,t)=cR,(34)c(x=0,t)=1 \quad\text{and}\quad c(x=1,t)=c_R, \tag*{(34)}

and the initial conditions,

c(x,t=0)=1andh(t=0)=1.(35)c(x,t=0)=1 \quad\text{and}\quad h(t=0)=1. \tag*{(35)}

We observe that the governing equations (21) and (33) and the boundary conditions (34) depend on three key dimensionless parameters: FDOF\mathcal{F}_{\mathrm{DOF}}, α\alpha, and cRc_R. The parameters FDOF\mathcal{F}_{\mathrm{DOF}} and α\alpha remain constant for a given configuration, whereas cRc_R may vary by changing the solute concentration at the right edge. Therefore, in the next section, we mainly present the results for fixed values of FDOF=40.5\mathcal{F}_{\mathrm{DOF}}=40.5 and α=1\alpha=1, while varying cRc_R. These values of FDOF\mathcal{F}_{\mathrm{DOF}} and α\alpha correspond to the Péclet number Pe=2FDOF/81α=Γ~/D~=1Pe=2\mathcal{F}_{\mathrm{DOF}}/81\alpha=\tilde{\Gamma}/\tilde{D}=1.

Results

In this section, we present our theoretical results for the fluid–structure interaction dynamics and the onset of fluid–structure instability driven by diffusioosmotic flow. We further validate the predictions of our reduced-order theoretical model against the two-dimensional numerical simulations with the finite-element software COMSOL Multiphysics (version 6.2, COMSOL AB, Stockholm, Sweden).

Steady-state results

First, in this subsection, we examine the steady-state response of the system. The governing equations for the gap height evolution, Eq. (21), and the solute transport, Eq. (33), at steady state take the form

427FDOFBsshss2+1−hss=0,Bss=−∫01ln⁡css(x) dx+12ln⁡cR,(36)\frac{4}{27}\mathcal{F}_{\mathrm{DOF}}\frac{B_{\mathrm{ss}}}{h_{\mathrm{ss}}^2}+1-h_{\mathrm{ss}}=0, \qquad B_{\mathrm{ss}}=-\int_0^1\ln c_{\mathrm{ss}}(x)\,dx+\frac{1}{2}\ln c_R, \tag*{(36)}
αd2cssdx2+181FDOFln⁡cRdcssdx=0⇒d2cssdx2+12Peln⁡cRdcssdx=0.(37)\alpha\frac{d^2c_{\mathrm{ss}}}{dx^2}+\frac{1}{81}\mathcal{F}_{\mathrm{DOF}}\ln c_R\frac{dc_{\mathrm{ss}}}{dx}=0 \quad\Rightarrow\quad \frac{d^2c_{\mathrm{ss}}}{dx^2}+\frac{1}{2}Pe\ln c_R\frac{dc_{\mathrm{ss}}}{dx}=0. \tag*{(37)}

where the subscript ss denotes the steady-state solution. At steady state, the governing equations (35) become one-way coupled, allowing for an analytical solution of the advection–diffusion equation (37), given as

css(x)=1cRPe/2−1[(1−cR)cR(Pe/2)(1−x)+cR1+(Pe/2)−1].(38)c_{\mathrm{ss}}(x) = \frac{1}{c_R^{Pe/2}-1}\left[(1-c_R)c_R^{(Pe/2)(1-x)}+c_R^{1+(Pe/2)}-1\right]. \tag*{(38)}

Knowing the steady-state concentration css(x)c_{\mathrm{ss}}(x), we can find the parameter BssB_{\mathrm{ss}} and the steady-state gap height hssh_{\mathrm{ss}} using Eq. (35a). We note that the steady-state concentration distribution and the parameter BssB_{\mathrm{ss}} are independent of the steady-state gap height hssh_{\mathrm{ss}}.

Once css(x)c_{\mathrm{ss}}(x) and hssh_{\mathrm{ss}} are determined, the corresponding steady-state pressure distribution in a non-dimensional form can be calculated using Eq. (10) as

pss(x)=−6hss2(ln⁡css(x)−xln⁡cR).(39)p_{\mathrm{ss}}(x) = -\frac{6}{h_{\mathrm{ss}}^2}\left(\ln c_{\mathrm{ss}}(x)-x\ln c_R\right). \tag*{(39)}

Figure 2(a,b) presents the steady-state concentration and pressure distributions for cR=3/2c_R = 3/2 and cR=2/3c_R = 2/3. Due to the non-negligible advective contribution with a Péclet number of Pe=1Pe = 1, the concentration distributions shown in Figure 2(a) slightly deviate from a linear behavior. We observe in Figure 2(b) that the diffusioosmotic flow induces a negative fluidic pressure, which acts as a destabilizing force. This negative diffusioosmotic pressure reduces the gap height until it reaches a stable steady-state value, hssh_{\mathrm{ss}}, where the elastic restoring force balances the diffusioosmotic driving force. Furthermore, consistent with the inverse ratio of the cRc_R values, the pressure distributions exhibit mirror symmetry about the channel center, and the steady-state gap height hss=0.8155h_{\mathrm{ss}} = 0.8155 is identical in both cases.

Image not extracted

Figure 2. Steady-state concentration and pressure distributions in the fluid–structure interaction driven by diffusioosmotic flow. (a) The concentration distribution at steady state, given by Eq. (38). (b) The corresponding pressure distribution at steady state, given by Eq. (39). Black solid and gray dashed curves represent the results for cR=3/2c_R = 3/2 and cR=2/3c_R = 2/3, respectively. The pressure profiles exhibit mirror symmetry across the channel center, consistent with the inverse ratios of cRc_R. The steady-state gap height for both cases is hss=0.8155h_{\mathrm{ss}} = 0.8155. All calculations were performed using α=1\alpha= 1 and FDOF=40.5\mathcal{F}_{\mathrm{DOF}} = 40.5.

Scaling analysis based on lubrication theory shows that the negative fluidic pressure generated by diffusioosmotic flow, represented by the first term on the right-hand side of Eq. (21), scales inversely with the gap height, hh. Therefore, as the gap height decreases due to negative fluidic pressure, the pressure becomes increasingly negative, further amplifying the reduction in gap height. In the next subsection, we show that above a certain critical threshold of the concentration at the right edge, cR=cR,CRc_R = c_{R,\mathrm{CR}}, this mechanism gives rise to the onset of the fluid–structure instability, ultimately leading to the collapse of the top plate onto the bottom surface.

Dynamic results

In this subsection, we investigate the fluid–structure interaction dynamics and the onset of fluid–structure instability by solving numerically the gap evolution equation (21) coupled with the advection–diffusion equation (33) using MATLAB’s routine ode15s. We provide the details of the numerical procedure in Appendix A.

First, we present in Figure 3(a,Figure 3b) the time evolution of FDOFB(t)\mathcal{F}_{\mathrm{DOF}}B(t) and the corresponding gap height h(t)h(t) for different values of cRc_R. The sign of FDOFB(t)\mathcal{F}_{\mathrm{DOF}}B(t) determines the direction of the diffusioosmotic forcing (see Eq. (21)). Positive values of FDOFB(t)\mathcal{F}_{\mathrm{DOF}}B(t) correspond to a rise of the rigid top plate, whereas negative values of FDOFB(t)\mathcal{F}_{\mathrm{DOF}}B(t) indicate its descent. Figure 3(a) clearly shows that at early times, FDOFB(t)\mathcal{F}_{\mathrm{DOF}}B(t) is positive, but at late times, it becomes negative for all values of cRc_R. Thus, except for the small increase in the gap height observed at early times, we identify in Figure 3(b) three distinct dynamic behaviors at late times, all characterized by a decrease in the gap height h(t)h(t). (i) Below the threshold value of cR,CR=1.5617c_{R,\mathrm{CR}}=1.5617, corresponding to FDOFB(t)=−1\mathcal{F}_{\mathrm{DOF}B}(t)=-1 (see Sec. IV C and Figure 6), the top plate approaches a stable steady-state height. (ii) However, a small increase in cRc_R above the threshold value (here, we show an increase by 0.1%0.1\%) triggers the onset of the fluid–structure instability, leading to a drastic change in the behavior of the system. In this regime, the gap height evolution exhibits a bottleneck, during which the descent of the top plate slows down. However, the top plate eventually accelerates and collapses onto the bottom surface. (iii) As cRc_R is further increased, the bottleneck behavior disappears, and the top plate collapses immediately onto the bottom surface. Therefore, we refer to these three distinct dynamic behaviors as the stable, bottleneck, and immediate collapse dynamic regimes.

Figure 3. Dynamic results of the fluid–structure interaction driven by diffusioosmotic flow. (a) The time evolution of FDOFB(t)\mathcal{F}_{\mathrm{DOF}B}(t) for various values of cRc_R. (b) The corresponding time evolution of the gap height h(t)h(t) for various values of cRc_R, indicating three different dynamic behaviors. Below the threshold value of cR,CR=1.5617c_{R,\mathrm{CR}}=1.5617, corresponding to FDOFB(t)=−1\mathcal{F}_{\mathrm{DOF}B}(t)=-1, the top plate approaches a stable steady-state height. Increasing cRc_R slightly above the threshold value (we show an increase by 0.1%0.1\%) leads to the onset of the fluid–structure instability characterized by a bottleneck. For larger values of cRc_R (e.g., cR=1.1cR,CR=1.7179c_R=1.1c_{R,\mathrm{CR}}=1.7179), the bottleneck disappears and the top plate immediately collapses onto the bottom surface. Note that in panel (a) the curves of FDOFB(t)\mathcal{F}_{\mathrm{DOF}B}(t) for cR=cR,CRc_R=c_{R,\mathrm{CR}} and cR=1.001cR,CRc_R=1.001c_{R,\mathrm{CR}} are almost indistinguishable. All calculations were performed using α=1\alpha=1 and FDOF=40.5\mathcal{F}_{\mathrm{DOF}}=40.5.

Image not extracted

Figure 6. Steady-state analysis of the fluid–structure interaction driven by diffusioosmotic flow. (a) The steady-state gap height hh as a function of FDOFB\mathcal{F}_{\mathrm{DOF}}B. The black solid curve represents the stable steady-state solution, while the red dashed curve represents the unstable result obtained from the linear stability analysis. Black dots represent the results of the dynamic simulation for different values of cRc_R, where we report the minimum value of BB in the simulation. At FDOFB=−1\mathcal{F}_{\mathrm{DOF}}B=-1, the stable steady-state solution intersects an unstable solution and disappears at a saddle-node bifurcation. (b) The growth rate σ\sigma as a function of FDOFBss\mathcal{F}_{\mathrm{DOF}}B_{\mathrm{ss}} obtained from the linear stability analysis. The black solid curve corresponds to the stable steady-state solution and the red dashed curve corresponds to the unstable result in panel (a). (c) The parameter FDOFBss\mathcal{F}_{\mathrm{DOF}}B_{\mathrm{ss}} as a function of cRc_R. The intersection of the curve with FDOFBss=−1\mathcal{F}_{\mathrm{DOF}}B_{\mathrm{ss}}=-1 (red dashed line) occurs at cR,CR=1.5617c_{R,\mathrm{CR}}=1.5617 (gray cross), indicating the threshold value of cRc_R for the onset of the fluid–structure instability. All calculations were performed using α=1\alpha=1, FDOF=40.5\mathcal{F}_{\mathrm{DOF}}=40.5, and cR>1c_R>1.

It is evident from Figure 3(b) that in both the bottleneck and immediate collapse regimes, the top plate significantly slows down just before approaching the bottom surface. We rationalize this slowdown by noting that at late times, when h≪1h \ll1, the diffusioosmotic force balances the viscous resistance in Eq. (21), while the elastic force becomes negligible. Since the parameter FDOFB(t)\mathcal{F}_{\mathrm{DOF}B}(t) remains approximately constant at late times, exhibiting only weak time dependence as shown in Figure 3(a), the diffusioosmotic force scales as h−2h^{-2}, while the viscous resistance scales as h−3(dh/dt)h^{-3}(dh/dt). This balance leads to an exponential approach of the top plate toward the bottom surface.

We note that Boyko et al. [62], who studied the fluid–structure interaction driven by non-uniform electroosmotic flow, reported three qualitatively similar dynamic behaviors. This similarity in behavior is due to the similar scaling of the electroosmotic and diffusioosmotic forcings, both of which scale as h−2h^{-2}. However, there is an important difference. While for non-uniform electroosmotic flow the dynamics of the gap height is governed by a single evolution equation for h(t)h(t), in the case of diffusioosmotic flow, the dynamics of the gap height is governed by coupled evolution equations for both h(t)h(t) and the solute concentration c(x,t)c(x,t).

To provide further insight into the dynamic behavior of the system, we present in Figure 4 the time evolution of the non-dimensional pressure distribution for three different values of cRc_R. Specifically, Figure 4(a), Figure 4(b), and Figure 4(c) correspond to the stable (cR=0.95cR,CRc_R=0.95c_{R,\mathrm{CR}}), bottleneck (cR=1.001cR,CRc_R=1.001c_{R,\mathrm{CR}}), and immediate collapse (cR=1.1cR,CRc_R=1.1c_{R,\mathrm{CR}}) dynamic regimes, respectively. Using Eqs. (10), (11), (20), and (22), the dimensionless pressure distribution p(x,t)p(x,t) reads

Image not extracted

Figure 4. Time evolution of the pressure distribution in the fluid–structure interaction driven by diffusioosmotic flow. (a,b,c) The time evolution of the pressure distribution corresponding to the (a) stable, (b) bottleneck, and (c) immediate collapse dynamic regimes. Black, dark gray, and light gray curves represent times t=3.54×10−4t = 3.54 \times10^{-4}, 1.32×10−21.32 \times10^{-2}, and 279.26279.26, respectively, except for panel (c), where the light gray curve corresponds to t=27.45t = 27.45. All calculations were performed using α=1\alpha= 1 and FDOF=40.5\mathcal{F}_{\mathrm{DOF}} = 40.5.

p(x,t)=243FDOF1h(t)3dh(t)dtx(x−1)−6h(t)2(ln⁡c(x,t)−xln⁡cR),(40)p(x,t)=\frac{243}{\mathcal{F}_{\mathrm{DOF}}}\frac{1}{h(t)^3}\frac{dh(t)}{dt}x(x-1)-\frac{6}{h(t)^2}\left(\ln c(x,t)-x\ln c_R\right), \tag*{(40)}

where h(t)h(t) and c(x,t)c(x,t) are obtained from the solution of Eqs. (21) and (33).

At early times, the pressure distributions are qualitatively similar for all dynamic regimes. However, at later times, we observe a distinct shift in behavior. While the pressure distribution in the stable regime generates a moderate downward fluidic force that is balanced by the elastic restoring force at late times (Figure 4(a)), the pressure distributions in the bottleneck and immediate collapse regimes create significantly stronger downward fluidic forces (Figure 4(b,Figure 4c)), ultimately driving the collapse of the top plate. In particular, within the immediate collapse regime, the late-time pressure distribution generates an even stronger downward fluidic force than that observed in the bottleneck regime, resulting in a more rapid collapse of the top plate.

In addition to the time evolution of the gap height, it is of particular interest to study the time evolution of the solute concentration, which drives the diffusioosmotic flow and the resulting fluid–structure interaction. Figure 5(a) presents the time evolution of FDOFB(t)\mathcal{F}_{\mathrm{DOFB}}(t) for the bottleneck regime with cR=1.001cR,CRc_R = 1.001c_{R,\mathrm{CR}}, where gray-scale squares mark specific times. In Figure 5(b), we present the corresponding time evolution of the concentration distribution. Although in the bottleneck regime the gap height h(t)h(t) continues to decrease until the top plate collapses onto the bottom surface, without satisfying the steady-state equation (36), it is evident from Figure 5(b) that the concentration distribution approaches its steady-state profile, given by Eq. (38). More importantly, the concentration distribution reaches a steady state when t≈O(1)t \approx O(1), which is significantly earlier than the time at which the top plate approaches the bottom surface (see Figure 3(b)). This result is consistent with Figure 3(a), which shows that the parameter FDOFB(t)\mathcal{F}_{\mathrm{DOFB}}(t), directly related to the concentration (see Eq. (21)), exhibits only weak time dependence once t≳O(1)t \gtrsim O(1).

Image not extracted

Figure 5. Dynamic results of the fluid–structure interaction driven by diffusioosmotic flow for the bottleneck regime with cR=1.001cR,CR=1.5632c_R = 1.001c_{R,\mathrm{CR}} = 1.5632. (a) The time evolution of FDOFB(t)\mathcal{F}_{\mathrm{DOFB}}(t), where the gray-scale squares mark specific times. (b) The corresponding time evolution of the concentration distribution, illustrated by gray-scale curves. The purple dashed line represents the analytical steady-state solution, given by Eq. (38). All calculations were performed using α=1\alpha= 1 and FDOF=40.5\mathcal{F}_{\mathrm{DOF}} = 40.5.

Linear stability analysis and investigation of the instability onset

In the previous subsection, we demonstrated the existence of fluid–structure instability driven by diffusioosmotic flow, which occurs when the concentration cRc_R exceeds a critical threshold cR,CRc_{R,\mathrm{CR}}. In this subsection, we perform a linear stability analysis of the system to gain deeper insight into the physical mechanisms underlying the fluid–structure instability.

We recall that, as shown in Figure 5(b), the concentration rapidly evolves towards the steady-state distribution, given

by Eq. (38), which is independent of the gap height. As a result, the parameter FDOFB(t)\mathcal{F}_{\mathrm{DOF}}B(t) remains approximately constant once t≳O(1)t \gtrsim O(1), exhibiting weak time dependence, as shown in Figure 3(a). In particular, the curves of FDOFB(t)\mathcal{F}_{\mathrm{DOF}}B(t) for cR=cR,CRc_R=c_{R,\mathrm{CR}} and cR=1.001cR,CRc_R=1.001c_{R,\mathrm{CR}} are almost indistinguishable and remain constant for t≳O(1)t \gtrsim O(1).

Therefore, we consider a small perturbation only of the gap height from its equilibrium value hssh_{\mathrm{ss}}, while assuming that the value of BB remains constant and equal to its steady-state value BssB_{\mathrm{ss}}, thereby eliminating the need to introduce a perturbation to the concentration. To this end, we let h(t)=hss(1+ϵ0eσt)h(t)=h_{\mathrm{ss}}\left(1+\epsilon_0e^{\sigma t}\right), where ϵ0≪1\epsilon_0\ll1 is some small perturbation and σ\sigma is the non-dimensional growth rate. Substituting this expression into Eq. (21), the leading order provides the equation for the steady-state solution hssh_{\mathrm{ss}}

427FDOFBsshss2+1−hss=0.(41)\frac{4}{27}\mathcal{F}_{\mathrm{DOF}}\frac{B_{\mathrm{ss}}}{h_{\mathrm{ss}}^{2}}+1-h_{\mathrm{ss}}=0. \tag*{(41)}

At the first order, O(ϵ0)O(\epsilon_0), we obtain the equation for the growth rate σ\sigma

σ=−827FDOFBss−hss3=hss2(2−3hss),(42)\sigma=-\frac{8}{27}\mathcal{F}_{\mathrm{DOF}}B_{\mathrm{ss}}-h_{\mathrm{ss}}^{3}=h_{\mathrm{ss}}^{2}(2-3h_{\mathrm{ss}}), \tag*{(42)}

indicating that the system may become unstable for certain negative values of FDOFBss\mathcal{F}_{\mathrm{DOF}}B_{\mathrm{ss}}. From Eq. (42), it follows that the critical steady-state height hss,CRh_{\mathrm{ss,CR}}, below which the system becomes unstable, and the corresponding threshold value FDOFBss,CR\mathcal{F}_{\mathrm{DOF}}B_{\mathrm{ss,CR}}, are given by

hss,CR=23andFDOFBss,CR=−1.(43)h_{\mathrm{ss,CR}}=\frac{2}{3} \quad\text{and}\quad \mathcal{F}_{\mathrm{DOF}}B_{\mathrm{ss,CR}}=-1. \tag*{(43)}

Figure 6(a) presents the steady-state gap height as a function of FDOFB\mathcal{F}_{\mathrm{DOF}}B. The black solid curve represents the linearly stable steady-state solution of Eq. (41), whereas the red dashed curve represents the linearly unstable steady-state solution. These two solutions coincide and disappear at a saddle-node bifurcation (see, e.g., [62, 70]), occurring at FDOFBss,CR=−1\mathcal{F}_{\mathrm{DOF}}B_{\mathrm{ss,CR}}=-1 with hss,CR=2/3h_{\mathrm{ss,CR}}=2/3, consistent with Eq. (43). Black dots represent the predictions of the dynamic theoretical model, showing excellent agreement with the results of the linear stability analysis. In the results presented here, we use the minimum value of BB obtained in the simulation. Figure 6(b) illustrates the growth rate σ\sigma as a function of FDOFBss\mathcal{F}_{\mathrm{DOF}}B_{\mathrm{ss}}, given by Eq. (42). As expected, the growth rate, represented by the black solid curve and corresponding to the linearly stable steady-state solution for hssh_{\mathrm{ss}}, is negative. Similarly, the growth rate, represented by the red dashed curve and corresponding to the linearly unstable steady-state solution for the gap height, is positive. For FDOFBss=−1\mathcal{F}_{\mathrm{DOF}}B_{\mathrm{ss}}=-1, the growth rate is zero, indicating the threshold of fluid–structure instability.

Aiming to rationalize the dynamic predictions of the theoretical model and the results of the linear stability analysis for the onset of instability, we present in Figure 6(c) the parameter FDOFBss\mathcal{F}_{\mathrm{DOF}}B_{\mathrm{ss}} as a function of cRc_R. We observe that intersection between the black curve and the red dashed line FDOFBss=−1\mathcal{F}_{\mathrm{DOF}}B_{\mathrm{ss}}=-1 occurs at cR,CR=1.5617c_{R,\mathrm{CR}}=1.5617 (gray cross), indicating the onset of instability, consistent with the result of the linear stability analysis, given by Eq. (43). This finding provides a theoretical justification for the threshold value cR,CR=1.5617c_{R,\mathrm{CR}}=1.5617, previously identified through dynamic simulations shown in Figure 3.

To better understand the behavior of the system and its transition to fluid–structure instability, we define the function f(cR,h)f(c_R,h), representing the combination of the diffusioosmotic and elastic forces,

f(cR,h)=427FDOFBss(cR)h2⏟Negative diffusioosmotic force+1−h⏟Positive elastic force,(44)f(c_R,h)=\underbrace{\frac{4}{27}\mathcal{F}_{\mathrm{DOF}}\frac{B_{\mathrm{ss}}(c_R)}{h^2}}_{\text{Negative diffusioosmotic force}}+\underbrace{1-h}_{\text{Positive elastic force}}, \tag*{(44)}

where its zero contour, f(cR,h)=0f(c_R,h)=0, corresponds to the equation for the steady-state solution hssh_{\mathrm{ss}}, as given by Eq. (41). We note that from Eq. (35a), it follows that the parameter BssB_{\mathrm{ss}} solely depends on the concentration at the right edge, cRc_R, and the Péclet number, Pe\mathrm{Pe}, but is independent of the gap height.

We present in Figure 7 a stability diagram showing the contour of f(cR,h)f(c_R,h) for different values of hh and cRc_R. Black solid and red dashed curves represent the zero contour of the function f(cR,h)f(c_R,h), where the negative diffusioosmotic force balances the positive restoring elastic force. The black solid curve represents the linearly stable steady-state solution for hssh_{\mathrm{ss}}, whereas the red dashed curve represents the linearly unstable steady-state solution, consistent with the results shown in Figure 6(a). The black cross marks the point cR,CR=1.5617c_{R,\mathrm{CR}}=1.5617 and hss,CR=2/3h_{\mathrm{ss,CR}}=2/3, indicating the threshold values for the onset of fluid–structure instability. As the diffusioosmotic force at steady state is always negative (see Figure 6(c)) and the elastic force remains positive for h<1h<1, the positive (gray) region of ff indicates the dominance of the restoring elastic effect, whereas the negative (white) region reflects the dominance of the diffusioosmotic force.

Image not extracted

Figure 7. Stability diagram showing the contour of f(cR,h)f(c_R,h) for different values of hh and cRc_R, given by Eq. (44). The positive (gray) region of ff represents the dominance of the restoring elastic effect, whereas the negative (white) region represents the dominance of the diffusioosmotic force. The black solid curve represents the zero contour of the function f(cR,h)f(c_R,h), which corresponds to the linearly stable steady-state solution for hssh_{\mathrm{ss}}. The red dashed curve represents the zero contour of the function f(cR,h)f(c_R,h), which corresponds to the linearly unstable steady-state solution. The black cross marks the point cR,CR=1.5617c_{R,\mathrm{CR}}=1.5617 and hss,CR=2/3h_{\mathrm{ss,CR}}=2/3, indicating the threshold values of concentration and gap height for the onset of fluid–structure instability. All calculations were preformed using α=1\alpha=1 and FDOF=40.5\mathcal{F}_{\mathrm{DOF}}=40.5.

Furthermore, we identify several distinct behaviors depending on the values of cRc_R, hh, and the sign of ff. When cR>cR,CRc_R>c_{R,\mathrm{CR}}, the function ff is always negative, indicating that the diffusioosmotic downward force consistently dominates the elastic restoring force, leading to the collapse of the top plate onto the bottom surface for any gap height hh. This behavior corresponds to the bottleneck and immediate collapse regimes, shown in Figure 3(b).

However, for cR<cR,CRc_R<c_{R,\mathrm{CR}}, the behavior of the system depends strongly on the value of hh and the sign of ff: (i) When h>hss,CRh>h_{\mathrm{ss,CR}} and f<0f<0, the dominant diffusioosmotic downward force causes the top plate to descend until it reaches a stable steady-state height, represented by the black solid curve. This behavior corresponds to the stable regime, shown in Figure 3(b). (ii) Similarly, for f>0f>0, the dominant restoring elastic force causes the plate to rise toward the stable steady-state height, even when the gap height is below the critical value hss,CRh_{\mathrm{ss,CR}}. (iii) In contrast, when h<hss,CRh<h_{\mathrm{ss,CR}} and f<0f<0, fluid–structure instability exists, and the dominant diffusioosmotic downward force causes the collapse of the top plate onto the bottom surface.

Finite-element numerical validation

To validate the predictions of our theoretical model, we carry out finite-element numerical simulations with the commercial software COMSOL Multiphysics (version 6.2, COMSOL AB, Stockholm, Sweden). In Appendix B, we provide detailed descriptions of the governing equations, boundary conditions, domain discretization, and physical parameters used in the finite-element simulations.

Figure 8 presents a comparison of finite-element simulation results and theoretical model predictions. Figure 8(a) shows the steady-state gap height, hssh_{\mathrm{ss}}, as a function of cRc_R, showing excellent agreement between the theoretical predictions (black dots) and the finite-element simulation results (red crosses). Specifically, the threshold value of cRc_R for the onset of fluid–structure instability, cR,CRc_{R,\mathrm{CR}}, is also well predicted. In the finite-element simulations, the threshold value is cR,CR,FE=1.5628c_{R,\mathrm{CR,FE}} = 1.5628 (purple square), which is slightly higher than the theoretical prediction of cR,CR=1.5617c_{R,\mathrm{CR}} = 1.5617.

Image not extracted

Figure 8. Comparison of finite-element simulation results and theoretical model predictions of the fluid–structure interaction driven by diffusioosmotic flow. (a) The steady-state gap height, hssh_{\mathrm{ss}}, as a function of cRc_R. Black dots represent the theoretical model predictions. Red crosses and a purple square represent the finite-element simulation results. (b) The time evolution of the gap height, h(t)h(t), for different values of cRc_R. Gray solid lines represent the theoretical model predictions and black dashed lines represent the finite-element simulation results. All calculations were performed using the values from Table I, with α=1\alpha= 1 and FDOF=40.5\mathcal{F}_{\mathrm{DOF}} = 40.5.

Next, in Figure 8(b) we present the time evolution of the gap height h(t)h(t) for different values of cRc_R. Gray solid lines represent the theoretical model predictions and black dashed lines represent the finite-element simulation results. We observe that the theoretical model accurately captures the transient dynamics of the gap height, showing excellent agreement with the finite-element simulation results, except in the close vicinity of the threshold value of cRc_R. For cR=cR,CR=1.5617c_R = c_{R,\mathrm{CR}} = 1.5617 and for a slightly higher value such as cR=1.002cR,CRc_R = 1.002c_{R,\mathrm{CR}}, a small discrepancy arises between the theoretical model predictions and the finite-element simulation results at late times. This discrepancy is due to the small difference in the threshold value cR,CRc_{R,\mathrm{CR}}. As a result, for cR=1.002cR,CRc_R = 1.002c_{R,\mathrm{CR}}, the theoretical model underestimates the total collapse time of the top plate.

Concluding remarks

In this work, we studied the fluid–structure interaction and instability of a thin electrolyte film confined between a rigid surface and an elastic substrate driven by diffusioosmotic flow. Applying the lubrication approximation and modeling the elastic substrate as a rigid plate connected to a linear spring, we derived a pair of two-way coupled nonlinear governing equations for the evolution of the film thickness h(t)h(t) and the solute concentration c(x,t)c(x,t). Our theoretical analysis revealed that, upon application of a certain solute concentration difference between the left and right edges, negative pressures induced by diffusioosmotic flow lead to the onset of fluid–structure instability. Through dynamic simulations of the coupled governing equations, we identified the critical physical parameters for the onset of fluid–structure instability, along with three distinct dynamic regimes: (i) a stable steady state, (ii) a bottleneck, and (iii) an immediate collapse. Furthermore, we performed a linear stability analysis of the system to rationalize the critical values of the film thickness hh and the right-edge concentration cRc_R for the onset of instability predicted by the dynamic simulations. Finally, we validated our theoretical predictions with finite-element numerical simulations and found excellent agreement.

It is worth noting that throughout this work, we presented results for a positive diffusioosmotic mobility Γ~\tilde{\Gamma}, corresponding to diffusioosmotic flow arising solely from the chemiosmotic contribution. In this case, we demonstrated the existence of fluid–structure instability above a certain threshold of cRc_R. However, when both chemiosmotic and electroosmotic contributions drive diffusioosmotic flow, the diffusioosmotic mobility Γ~\widetilde{\Gamma} can be negative. In such a case, diffusioosmotic flow generates positive gauge pressures that raise the rigid plate above the initial gap height to a stable steady state, thus preventing fluid–structure instability.

One interesting extension of the present work, which accounts only for the temporal dynamics of the gap height h(t)h(t), is to study the viscous–elastic interaction and the spatiotemporal evolution of the deformable walls of a soft microchannel driven by diffusioosmotic flow. As a future research direction, it would also be interesting to investigate the coupled dynamics of particle diffusiophoresis and diffusioosmotic flow in soft microfluidic channels, exploring the interplay between diffusiophoresis, diffusioosmosis, and the elasticity of confining boundaries. Diffusioosmotic flow is also encountered as a driving mechanism in nanochannels, where the thin-double-layer assumption may no longer be valid due to comparable Debye length and channel height [8]. In such channels, diffusioosmotic flow can generate pressures large enough to deform even relatively rigid walls (e.g., glass covers), making the resulting fluid–structure interaction an interesting direction for further investigation. We believe that the present study establishes a theoretical foundation for understanding soft microfluidic configurations driven by diffusioosmotic flow, providing key insights for the control and design of fluid–structure interactions and instabilities.

Acknowledgments

N.M. and E.B. acknowledge the support by grant no. 1942/23 from the Israel Science Foundation (ISF). E.B. acknowledges the support from the Israeli Council for Higher Education Yigal Alon Fellowship.

Details of a numerical method used in the theoretical model

To solve numerically the gap evolution equation (21) coupled with the advection–diffusion equation (33) subject to the boundary conditions (34) and the initial conditions (35), we first discretize the spatial derivatives in Eq. (33) using a second-order central difference approximation with uniform grid spacing Δx\Delta x. The typical value of the grid size are Δx=0.002\Delta x = 0.002. This discretization leads to a series of ordinary differential equations for the evolution of ci(t)=c(xi,t)c_i(t) = c(x_i,t). We then integrate forward in time the resulting set of ordinary differential equations for the evolution of [h(t),ci(t)][h(t),c_i(t)] using MATLAB’s routine ode15s. As shown in Figure 3(b) and Figure 8(b), since the top plate approaches the bottom surface exponentially slowly, we terminate our numerical integration once the gap height hh falls below 10−210^{-2}.

Details of finite-element numerical simulations

In this appendix we provide the details of finite-element numerical simulations. We perform two-dimensional finite-element simulations with the commercial software COMSOL Multiphysics (version 6.2, COMSOL AB, Stockholm, Sweden).

To capture the coupled dynamic behavior of the system, we employ three different modules in the simulations. The first module is the Solid Mechanics module, which we use to describe the dynamics of the plate. The plate is modeled as a rigid body constrained to vertical motion without rotation and is attached to a spring foundation. The second module is the Laminar Flow module, which we employ to describe the fluid mechanics between a rigid bottom surface and a rigid top plate. The fluid motion is governed by the Navier–Stokes equations, subject to the boundary conditions on the velocity at the bottom surface and top plate, and fixed pressure at the left and right edges, as given by Eqs. (4a)–(4e).

The Solid Mechanics module and Laminar Flow module are fully coupled by the Fluid–Solid Interaction interface, which includes predefined conditions for the interaction at the fluid–solid boundary, z^=h~(t)\hat{z}=\tilde{h}(t). For our case of a rigid top plate attached to a spring foundation, the boundary conditions at the fluid–solid boundary are the kinematic boundary condition, Eq. (4d), and the governing equation for the plate motion, Eq. (17), where the fluidic force F~f\tilde{F}_f is given by F~f=z^⋅∫0ℓn^⋅σ~ dx~=∫0ℓ[p~−2μ~(∂w~/∂z~)]dx~\tilde{F}_f=\hat{z}\cdot\int_0^\ell\hat{n}\cdot\tilde{\sigma}\,d\tilde{x}=\int_0^\ell[\tilde{p}-2\tilde{\mu}(\partial\tilde{w}/\partial\tilde{z})]d\tilde{x}. Here, n^=−z^\hat{n}=-\hat{z} is the unit normal vector to the fluid–solid boundary and σ~=−p~I+μ~(∇u~+(∇u~)T)\tilde{\sigma}=-\tilde{p}I+\tilde{\mu}(\nabla\tilde{u}+(\nabla\tilde{u})^{\mathsf{T}}) is the stress tensor.

The third module we use is the Transport of Diluted Species module. This module describes the spatiotemporal evolution of the solute concentration, which drives diffusioosmotic flow through the slip boundary condition at the bottom surface and, in turn, leads to the fluid–structure interaction. The evolution of the solute concentration is governed by the advection–diffusion equation (5), subject to no-flux boundary conditions at the bottom surface and top plate, and prescribed different concentrations at the left and right edges, as given by Eqs. (4f)–(4g).

Physical propertyNotationValueUnits
Initial gap heighth~i\tilde{h}_i20μm\mu\mathrm{m}
Initial concentrationc~L\tilde{c}_L1mol m−3\mathrm{mol}\ \mathrm{m}^{-3}
Length of the rigid top plateℓ~\tilde{\ell}1mm\mathrm{mm}
Thickness of rigid top plateb~\tilde{b}10μm\mu\mathrm{m}
Density of rigid top plateρ~s\tilde{\rho}_s965kg m−3\mathrm{kg}\ \mathrm{m}^{-3}
Spring stiffness (per unit depth)k~\tilde{k}0.125Pa\mathrm{Pa}
Density of fluidρ~\tilde{\rho}10310^3kg m−3\mathrm{kg}\ \mathrm{m}^{-3}
Viscosity of fluidμ~\tilde{\mu}10−310^{-3}Pa s\mathrm{Pa}\ \mathrm{s}
Permittivity of fluidε~\tilde{\varepsilon}7.08×10−107.08 \times10^{-10}F m−1\mathrm{F}\ \mathrm{m}^{-1}
Solute diffusivityD~\tilde{D}10−910^{-9}m2 s−1\mathrm{m}^2\ \mathrm{s}^{-1}
Pressure at the edgesp~a\tilde{p}_a0Pa\mathrm{Pa}
Concentration at the left edgec~L\tilde{c}_L1mol m−3\mathrm{mol}\ \mathrm{m}^{-3}
Concentration at the right edgec~R\tilde{c}_R1−21-2mol m−3\mathrm{mol}\ \mathrm{m}^{-3}
Diffusioosmotic mobilityΓ\Gamma10310^3μm2 s−1\mu\mathrm{m}^2\ \mathrm{s}^{-1}
Characteristic diffusioosmotic slip velocityu~c=Γ~/ℓ~\tilde{u}_c = \tilde{\Gamma}/\tilde{\ell}1μm s−1\mu\mathrm{m}\ \mathrm{s}^{-1}
Characteristic pressurep~c=μ~Γ~/h~i2\tilde{p}_c = \tilde{\mu}\tilde{\Gamma}/\tilde{h}_i^22.5×10−32.5 \times10^{-3}Pa\mathrm{Pa}
Characteristic viscous–elastic time scalet~c=(ℓ~/h~i)2(μ~/k~)\tilde{t}_c = (\tilde{\ell}/\tilde{h}_i)^2(\tilde{\mu}/\tilde{k})10310^3s\mathrm{s}
Axial diffusive time scaleℓ~2/D~\tilde{\ell}^2/\tilde{D}10310^3s\mathrm{s}
Advection time scaleℓ~/u~c\tilde{\ell}/\tilde{u}_c10310^3s\mathrm{s}

caption: TABLE I. Representative values of physical parameters used in the two-dimensional finite-element numerical simulations of fluid–structure interaction and fluid–structure instability in microfluidic configurations driven by diffusioosmotic flow.

:::

The three modules are fully coupled: the concentration difference between the left and right edges induces fluid motion through the diffusioosmotic slip boundary condition, which exerts a fluidic force on the plate, resulting in its motion and the evolution of the gap height. At the same time, the change in gap height modifies the fluid velocity, the solute concentration distribution, and the magnitude of the applied fluidic force.

We discretize the fluid domain using a rectangular mesh with 500 uniformly distributed elements in the longitudinal dimension and 10 uniformly distributed elements in the transverse dimension. Due to the time evolution of the gap height, we employ a moving mesh with a fixed bottom surface. The left and right edges of the domain are constrained to prevent deformation in the normal (longitudinal) direction. We use the third-order (cubic) discretization for the velocity field and the second-order (quadratic) discretization for the concentration and pressure distributions.

We obtained all finite-element simulation results presented in this study through dynamic simulations. For the unstable cases, we stopped the simulations when the gap height fell below 0.2 μm\mu\mathrm{m}, corresponding to 1% of the initial gap height. For the stable cases, we stopped the simulations after 107 s10^7\ \mathrm{s}, once the system reached a steady state with negligible plate motion.

To assess the grid sensitivity, we performed tests by considering seven different mesh resolutions ranging from 100 ×\times 2 to 700 ×\times 14 elements across 16 values of the right-edge concentration c~R\tilde{c}_R between 1.561 mol m−3\mathrm{mol}\ \mathrm{m}^{-3} and 1.564 mol m−3\mathrm{mol}\ \mathrm{m}^{-3}, encompassing the critical threshold for the onset of instability. We established grid independence using our

Non-dimensional numberDefinitionValue
Aspect ratioϵ=h~i/ℓ~\epsilon= \tilde{h}_i/\tilde{\ell}2×10−22 \times10^{-2}
Reduced Reynolds numberϵRe=ρ~u~ch~i2/(μ~ℓ~)\epsilon Re = \tilde{\rho}\tilde{u}_c\tilde{h}_i^2/(\tilde{\mu}\tilde{\ell})4×10−74 \times10^{-7}
Womersley numberWo=ρ~h~i2/(μ~t~c)Wo = \tilde{\rho}\tilde{h}_i^2/(\tilde{\mu}\tilde{t}_c)4×10−74 \times10^{-7}
Smallness of rigid plate’s inertiaρ~sb~/(k~t~c2)\tilde{\rho}_s\tilde{b}/(\tilde{k}\tilde{t}_c^2)7.7×10−117.7 \times10^{-11}
Smallness of dielectric effect (KCl with βD=0\beta_D = 0)βD2ε~φ~2h~i2/(μ~Γ~2)\beta_D^2\tilde{\varepsilon}\tilde{\varphi}^2\tilde{h}_i^2/(\tilde{\mu}\tilde{\Gamma}^2)0
Smallness of dielectric effect (NaCl with βD=−0.208\beta_D = -0.208)βD2ε~φ~2h~i2/(μ~Γ~2)\beta_D^2\tilde{\varepsilon}\tilde{\varphi}^2\tilde{h}_i^2/(\tilde{\mu}\tilde{\Gamma}^2)7.7×10−67.7 \times10^{-6}
Elasto-diffusioosmotic numberFDOF=81μ~c~Lℓ~2/(2k~h~i3)\mathcal{F}_{\mathrm{DOF}} = 81\tilde{\mu}\tilde{c}_L\tilde{\ell}^2/(2\tilde{k}\tilde{h}_i^3)40.5
Elasto-diffusion numberα=μ~D~ℓ~/(k~h~i3)\alpha= \tilde{\mu}\tilde{D}\tilde{\ell}/(\tilde{k}\tilde{h}_i^3)1
Péclet numberPe=ℓ~u~c/D~Pe = \tilde{\ell}\tilde{u}_c/\tilde{D}1
Non-dimensional right-edge concentrationcR=c~R/c~Lc_R = \tilde{c}_R/\tilde{c}_L1−21-2

caption: TABLE II. Representative values of non-dimensional numbers corresponding to the physical parameters in Table I, indicating that the assumptions of the theoretical model are well satisfied. The values of βD=(D~+−D~−)/(D~++D~−)\beta_D = (\tilde{D}_+ - \tilde{D}_-)/(\tilde{D}_+ + \tilde{D}_-) were calculated using D~+=1.33×10−9 m2 s−1\tilde{D}_+ = 1.33 \times10^{-9}\ \mathrm{m}^2\ \mathrm{s}^{-1} for Na+^+, D~+=2.03×10−9 m2 s−1\tilde{D}_+ = 2.03 \times10^{-9}\ \mathrm{m}^2\ \mathrm{s}^{-1} for K+^+, and D~−=2.03×10−9 m2 s−1\tilde{D}_- = 2.03 \times10^{-9}\ \mathrm{m}^2\ \mathrm{s}^{-1} for Cl−^- [22, 69].

::: selected mesh, with the maximum relative error in the steady-state gap height remaining below 0.15% compared to the finest mesh across all tested concentration values.

We summarize in Table I the representative values of physical and geometrical parameters used in the finite-element simulations. Table II summarizes the corresponding non-dimensional numbers, showing that our theoretical model’s assumptions are well satisfied.

References

References

  1. [1]D. C. Prieve, J. L. Anderson, J. P. Ebel, and M. E. Lowell, Motion of a particle generated by chemical gradients. Part 2. Electrolytes, J. Fluid Mech. 148, 247 (1984).DOI
  2. [2]J. L. Anderson, Colloid transport by interfacial forces, Annu. Rev. Fluid Mech. 21, 61 (1989).
  3. [3]S. Shim, Diffusiophoresis, diffusioosmosis, and microfluidics: surface-flow-driven phenomena in the presence of flow, Chem. Rev. 122, 6986 (2022).DOI
  4. [4]J. T. Ault and S. Shin, Physicochemical hydrodynamics of particle diffusiophoresis driven by chemical gradients, Annu. Rev. Fluid Mech. 57, 227 (2025).DOI
  5. [5]N. Shi, A. Fattah, and T. M. Squires, Generalized, conceptually unified theory of linear phoretic drift and osmotic slip, Phys. Rev. Fluids 10, 043701 (2025).DOI
  6. [6]B. V. Derjaguin, S. S. Dukhin, and A. A. Korotkova, Diffusiophoresis in electrolyte solutions and its role in the mechanism of film formation from rubber latexes by the method of ionic deposition, Kolloidn. Zh. 23, 53 (1961).
  7. [7]B. V. Derjaguin, S. S. Dukhin, and A. A. Korotkova, Diffusiophoresis in electrolyte solutions and its role in the Mechanism of the formation of films from caoutchouc latexes by the ionic deposition method, Prog. Surf. Sci. 43, 153 (1993).DOI
  8. [8]C. Lee, C. Cottin-Bizonne, A.-L. Biance, P. Joseph, L. Bocquet, and C. Ybert, Osmotic flow through fully permeable nanochannels, Phys. Rev. Lett. 112, 244501 (2014).pubmed.ncbi.nlm.nih.gov/24996091
  9. [9]H. Liu and A. A. Pahlavan, Diffusioosmotic reversal of colloidal focusing direction in a microfluidic T-junction, Phys. Rev. Lett. 134, 098201 (2025).DOI
  10. [10]H. A. Stone, A. D. Stroock, and A. Adjari, Engineering flows in small devices: microfluidics toward a lab-on-a-chip, Annu. Rev. Fluid Mech. 36, 381 (2004).DOI
  11. [11]T. M. Squires and S. R. Quake, Microfluidics: Fluid physics at the nanoliter scale, Rev. Mod. Phys. 77, 977 (2005).DOI
  12. [12]F. Paratore, V. Bacheva, M. Bercovici, and G. V. Kaigala, Reconfigurable microfluidics, Nat. Rev. Chem. 6, 70 (2022).DOI
  13. [13]J. T. Ault, S. Shin, and H. A. Stone, Characterization of surface–solute interactions by diffuioosmosis, Soft Matt. 15, 1582 (2019).DOI
  14. [14]J. Teng, B. Rallabandi, and J. T. Ault, Diffusioosmotic dispersion of solute in a long narrow channel, J. Fluid Mech. 977, A5 (2023).DOI
  15. [15]J. T. Ault, P. B. Warren, S. Shin, and H. A. Stone, Diffusiophoresis in one-dimensional solute gradients, Soft Matt. 13, 9015 (2017).DOI
  16. [16]A. Gupta, S. Shim, and H. A. Stone, Diffusiophoresis: from dilute to concentrated electrolytes, Soft Matt. 16, 6975 (2020).DOI
  17. [17]N. Singh, G. T. Vladisavljević, F. Nadal, C. Cottin-Bizonne, C. Pirat, and G. Bolognesi, Reversible trapping of colloids in microgrooved channels via diffusiophoresis under steady-state solute gradients, Phys. Rev. Lett. 125, 248002 (2020).DOI
  18. [18]N. Singh, G. T. Vladisavljević, F. Nadal, C. Cottin-Bizonne, C. Pirat, and G. Bolognesi, Enhanced accumulation of colloidal particles in microgrooved channels via diffusiophoresis and steady-state electrolyte flows, Langmuir 38, 14053 (2022).DOI
  19. [19]R. E. Migacz, M. Castleberry, and J. T. Ault, Enhanced diffusiophoresis in dead-end pores with time-dependent boundary solute concentration, Phys. Rev. Fluids 9, 044203 (2024).
  20. [20]A. Kar, T. Chiang, I. Ortiz Rivera, A. Sen, and D. Velegol, Enhanced transport into and out of dead-end pores, ACS Nano 9, 746 (2015).DOI
  21. [21]S. Shin, J. T. Ault, P. B. Warren, and H. A. Stone, Accumulation of colloidal particles in flow junctions induced by fluid flow and diffusiophoresis, Phys. Rev. X 7, 041038 (2017).DOI
  22. [22]S. Shin, E. Um, S. Babass, J. T. Ault, M. Rahimi, P. B. Warren, and H. A. Stone, Size-dependent control of colloid transport via solute gradients in dead-end channels, Proc. Natl Acad. Sci. USA 113, 257 (2016).DOI
  23. [23]J. T. Ault, S. Shin, and H. A. Stone, Diffusiophoresis in narrow channel flows, J. Fluid Mech. 854, 420 (2018).DOI
  24. [24]D. Ha, S. Seo, K. Lee, and T. Kim, Dynamic transport control of colloidal particles by repeatable active switching of solute gradients, ACS Nano 13, 12939 (2019).DOI
  25. [25]B. M. Alessio, S. Shim, E. Mintah, A. Gupta, and H. A. Stone, Diffusiophoresis and diffusioosmosis in tandem: Two-dimensional particle motion in the presence of multiple electrolytes, Phys. Rev. Fluids 6, 054201 (2021).DOI
  26. [26]B. M. Alessio, S. Shim, A. Gupta, and H. A. Stone, Diffusioosmosis-driven dispersion of colloids: a Taylor dispersion analysis with experimental validation, J. Fluid Mech. 942, A23 (2022).DOI
  27. [27]B. Akdeniz, J. A. Wood, and R. G. H. Lammertink, Diffusiophoresis and diffusio-osmosis into a dead-end channel: Role of the concentration-dependent ζeta potential, Langmuir 39, 2322 (2023).DOI
  28. [28]G. M. Whitesides, The origins and the future of microfluidics, Nature 442, 368 (2006).pubmed.ncbi.nlm.nih.gov/16871203
  29. [29]I. C. Christov, Soft hydraulics: from Newtonian to complex fluid flows through compliant conduits, J. Phys.: Condens. Matter 38, 063001 (2022).
  30. [30]N. Kashaninejad and N.-T. Nguyen, Microfluidic solutions for biofluids handling in on-skin wearable systems, Lab Chip 23, 913 (2023).DOI
  31. [31]J. B. Grotberg and O. E. Jensen, Biofluid mechanics in flexible tubes, Annu. Rev. Fluid Mech. 36, 121 (2004).
  32. [32]S. N. Bhatia and D. E. Ingber, Microfluidic organs-on-chips, Nat. Biotechnol. 32, 760 (2014).DOI
  33. [33]T. T. Al-Housseiny and H. A. Stone, Controlling viscous fingering in tapered Hele-Shaw cells, Phys. Fluids 25 (2013).DOI
  34. [34]J. V. Fontana, C. Cuttle, D. Pihler-Puzović, A. L. Hazel, and A. Juel, Peeling fingers in an elastic Hele-Shaw channel, J. Fluid Mech. 985, A1 (2024).arxiv.org/abs/2310.12940
  35. [35]T. Gervais, J. El-Ali, A. Günther, and K. F. Jensen, Flow-induced deformation of shallow microfluidic channels, Lab Chip 6, 500 (2006).pubmed.ncbi.nlm.nih.gov/16572212
  36. [36]D. Dendukuri, S. S. Gu, D. C. Pregibon, T. A. Hatton, and P. S. Doyle, Stop-flow lithography in a microfluidic device, Lab Chip 7, 818 (2007).pubmed.ncbi.nlm.nih.gov/17593999
  37. [37]B. S. Hardy, K. Uechi, J. Zhen, and H. P. Kavehpour, The deformation of flexible PDMS microchannels under a pressure driven flow, Lab Chip 9, 935 (2009).DOI
  38. [38]I. C. Christov, V. Cognet, T. C. Shidhore, and H. A. Stone, Flow rate–pressure drop relation for deformable shallow microfluidic channels, J. Fluid Mech. 841, 267 (2018).
  39. [39]R. F. Shepherd, F. Ilievski, W. Choi, S. A. Morin, A. A. Stokes, A. D. Mazzeo, X. Chen, M. Wang, and G. M. Whitesides, Multigait soft robot, Proc. Natl Acad. of Sci. USA 108, 20400 (2011).DOI
  40. [40]S. B. Elbaz and A. D. Gat, Dynamics of viscous liquid within a closed elastic cylinder subject to external forces with application to soft robotics, J. Fluid Mech. 758, 221 (2014).DOI
  41. [41]D. Rus and M. T. Tolley, Design, fabrication and control of soft robots, Nature 521, 467 (2015).
  42. [42]P. Polygerinos, N. Correll, S. A. Morin, B. Mosadegh, C. D. Onal, K. Petersen, M. Cianchetti, M. T. Tolley, and R. F. Shepherd, Soft robotics: Review of fluid-driven intrinsically soft devices; manufacturing, sensing, control, and applications in human-robot interaction, Adv. Eng. Mater. 19, 1700016 (2017).DOI
  43. [43]Y. Matia, T. Elimelech, and A. D. Gat, Leveraging internal viscous flow to extend the capabilities of beam-shaped soft robotic actuators, Soft Robotics 4, 126 (2017).DOI
  44. [44]L. Salem, B. Gamus, Y. Or, and A. D. Gat, Leveraging viscous peeling to create and activate soft actuators and microfluidic devices, Soft Robotics 7, 76 (2020).DOI
  45. [45]K. Park, A. Tixier, M. Paludan, E. Østergaard, M. Zwieniecki, and K. H. Jensen, Fluid-structure interactions enable passive flow control in real and biomimetic plants, Phys. Rev. Fluids 6, 123102 (2021).DOI
  46. [46]E. Boyko, H. A. Stone, and I. C. Christov, Flow rate-pressure drop relation for deformable channels via fluidic and elastic reciprocal theorems, Phys. Rev. Fluids 7, L092201 (2022).DOI
  47. [47]A. Martínez-Calvo, A. Sevilla, G. G. Peng, and H. A. Stone, Start-up flow in shallow deformable microchannels, J. Fluid Mech. 885, A25 (2020).
  48. [48]G. Guyard, F. Restagno, and J. D. McGraw, Elastohydrodynamic relaxation of soft and deformable microchannels, Phys. Rev. Lett. 129, 204501 (2022).DOI
  49. [49]M. D. Diviano, M. V. Paludan, A. H. Christensen, E. V. Østergaard, and K. H. Jensen, Smoothing oscillatory peristaltic pump flow with bioinspired passive components, Phys. Rev. Appl. 18, 064013 (2022).
  50. [50]U. Mukherjee, J. Chakraborty, and S. Chakraborty, Relaxation characteristics of a compliant microfluidic channel under electroosmotic flow, Soft Matt. 9, 1562 (2013).DOI
  51. [51]J. M. de Rutte, K. G. H. Janssen, N. R. Tas, J. C. T. Eijkel, and S. Pennathur, Numerical investigation of micro-and nanochannel deformation due to discontinuous electroosmotic flow, Microfluid. Nanofluid. 20, 150 (2016).DOI
  52. [52]S. Rubin, A. Tulchinsky, A. D. Gat, and M. Bercovici, Elastic deformations driven by non-uniform lubrication flows, J. Fluid Mech. 812, 841 (2017).DOI
  53. [53]E. Boyko, R. Eshel, K. Gommed, A. D. Gat, and M. Bercovici, Elastohydrodynamics of a pre-stretched finite elastic sheet lubricated by a thin viscous film with application to microfluidic soft actuators, J. Fluid Mech. 862, 732 (2019).DOI
  54. [54]B. Rallabandi, Fluid-elastic interactions near contact at low Reynolds number, Annu. Rev. Fluid Mech. 56, 491 (2024).DOI
  55. [55]M. Heil and A. L. Hazel, Fluid-structure interaction in internal physiological flows, Annu. Rev. Fluid Mech. 43, 141 (2011).DOI
  56. [56]A. Juel, D. Pihler-Puzović, and M. Heil, Instabilities in blistering, Annu. Rev. Fluid Mech. 50, 691 (2018).DOI
  57. [57]M. Heil, Stokes flow in collapsible tubes: computation and experiment, J. Fluid Mech. 353, 285 (1997).DOI
  58. [58]M. Gomez, D. E. Moulton, and D. Vella, Passive control of viscous flow via elastic snap-through, Phys. Rev. Lett. 119, 144502 (2017).DOI
  59. [59]O. Kodio, I. M. Griffiths, and D. Vella, Lubricated wrinkles: Imposed constraints affect the dynamics of wrinkle coarsening, Phys. Rev. Fluids 2, 014202 (2017).DOI
  60. [60]O. Peretz, A. K. Mishra, R. F. Shepherd, and A. D. Gat, Underactuated fluidic control of a continuous multistable membrane, Proc. Natl Acad. Sci. USA 117, 5217 (2020).DOI
  61. [61]F. Box, G. G. Peng, D. Pihler-Puzović, and A. Juel, Flow-induced choking of a compliant Hele-Shaw cell, Proc. Natl Acad. Sci. USA 117, 30228 (2020).DOI
  62. [62]E. Boyko, R. Eshel, A. D. Gat, and M. Bercovici, Nonuniform electro-osmotic flow drives fluid-structure instability, Phys. Rev. Lett. 124, 024501 (2020).DOI
  63. [63]E. Boyko, D. Ilssar, M. Bercovici, and A. D. Gat, Interfacial instability of thin films in soft microfluidic configurations actuated by electro-osmotic flow, Phys. Rev. Fluids 5, 104201 (2020).
  64. [64]L. G. Leal, Advanced Transport Phenomena: Fluid Mechanics and Convective Transport Processes (Cambridge University Press, 2007).
  65. [65]R. E. Migacz and J. T. Ault, Diffusiophoresis in a Taylor-dispersing solute, Phys. Rev. Fluids 7, 034202 (2022).
  66. [66]S. Lee, J. Lee, and J. T. Ault, The role of variable zeta potential on diffusiophoretic and diffusioosmotic transport, Colloids Surf. A 659, 130775 (2023).DOI
  67. [67]G. M. Rebeiz, RF MEMS: Theory, Design, and Technology (John Wiley & Sons, 2004).
  68. [68]K. Singh, J. R. Lister, and D. Vella, A fluid-mechanical model of elastocapillary coalescence, J. Fluid Mech. 745, 621 (2014).DOI
  69. [69]P. Vanýsek, Ionic conductivity and diffusion at infinite dilution, in CRC Handbook of Chemistry and Physics (CRC Press, 1993) pp. 5–92.
  70. [70]M. Gomez, D. E. Moulton, and D. Vella, Delayed pull-in transitions in overdamped MEMS devices, J. Micromech. Microeng. 28, 015006 (2018).DOI

Paper details

vlnc:W_erb8anre
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