betaFeedback

WorksPhysics and Astronomy

Oscillating flow around a circular cylindrical post confined between two parallel plates

Abstract

This work is motivated by the interest in determining the effect of the micro-anatomy of the spinal subarachnoid space on the cerebrospinal fluid flow and on the associated transport of solutes. To that aim, we focus on a canonical model problem in which a circular post of radius aa, confined between two parallel plates separated by a distance 2h2h, is subjected to an oscillatory flow. In particular, we are concerned with the steady, viscous, time-averaged flow that persists in the vicinity on the cylinder, when the stroke length of the oscillating flow, U∞/ωU_{\infty}/ω is smaller or comparable to the radius of the post, aa, where U∞U_{\infty} and ωω are the mean velocity amplitude and the corresponding angular frequency, respectively. First, we analyze the asymptotic limit of small values of the stroke length varying the aspect ratio of the post λλ and the Womersley number MM. First-order steady-streaming corrections at leading order have been computed, together with associated Stokes-drift component, the sum of both yielding the mean Lagrangian velocity field. The confinement effect is seen to induce the three-dimensionality of flow. Moreover, in the mid plane the time-averaged steady flow exhibits a recirculating vortex attached to the wall of the post that decreases as MM increases for low values of MM. However, for values of MM larger than a critical one, Mcr(λ)M_{cr}(λ), a second, outer vortex is also formed. The dependence Mcr(λ)M_{cr}(λ) has been quantified. The analysis has been corroborated experimentally for a fixed aspect ratio. Consideration of an anharmonic oscillating flow shows the fort-and-aft symmetry of the steady flow is broken. Finally, the analysis is experimentally extended to consider an array of equally spaced posts.

Introduction

The spinal subarachnoid space (SSAS) forms an annular slender canal filled of cerebrospinal fluid (CSF) bounded externally by the dura membrane and internally by the pia membrane surrounding the spinal cord. Along it, CSF features an oscillatory velocity, mainly induced by the combined contribution of the cardiac and the respiratory cycles Linninger et al. [2016], Kelley et al. [2023]. The anatomy of the SSAS is known to present a complex geometry, including the presence of microanatomical elements Dauleac et al. [2019], namely, trabeculae, denticulate ligaments and nerve roots, across the width of the canal, that is much smaller than its perimeter and which in turn is much smaller than its length. In particular, nerve roots are present in most spinal segments and can be found on both sides of the canal arranged in bundles that emerge from the spinal cord and exit the vertebral column through the intervertebral discs (see Figure 1). These elements have been reported to play an important role in the transport of substances along the spinal canal Stockman [2006, 2007], Tangen et al. [2015], Pahlavia et al. [2014], Reina et al. [2020], Ayansiji et al. [2023], such as drugs injected intrathecally. The present paper, focused on a simplified geometry, is inspired by the study of the effect of such elements on the flow of CSF and the transport of solutes in the SSAS. Thus, taking into account the anatomy of the canal, the description of the effect of nerve roots in the SSAS can be approached, in a first approximation, as that of a flat canal in the presence of obstacles facing the oscillatory CSF flow, see Figure 2. This configuration is used herein to elucidate the influence of these anatomical elements on the CSF flow, and more specifically on the mean, steady Lagrangian motion that is generated, and which is mainly responsible for the transport of solutes Lawrence et al. [2019]. The mean Lagrangian velocity experienced by a fluid particle in the SSAS is the sum of the steady-streaming velocity, determined by time-averaging the Eulerian velocity field Riley [2001], Sánchez et al. [2018], and the Stokes drift Stokes [1847], a purely kinematic effect associated with the spatial nonuniformity of the pulsatile flow.

Representation of the spinal canal with indication of the nerve roots across it

Figure 1. Representation of the spinal canal with indication of the nerve roots across it.

Figure 2. Schematic configuration of the flow.

The phenomenon of steady or acoustic streaming, documented by Rayleigh [1884] in the last quarter of the 19th century, has been extensively studied since then and it is considered a classical fluid mechanics problem. It is known that an oscillating fluid stream, characterized by a velocity U∞cos⁡(ωt′)U_\infty\cos(\omega t'), interacting with a stationary solid body, leads to a time-averaged steady-streaming motion [12]. The resulting solution depends on the velocity amplitude, U∞U_\infty, the object size, aa, the oscillating angular frequency, ω\omega, and the fluid’s kinematic viscosity, ν\nu. These parameters can be combined to give two dimensionless, controlling parameters, i.e the dimensionless stroke length ε=U∞/ωa\varepsilon= U_\infty/\omega a, and the Womersley number M=(a2ω/ν)1/2M = (a^2\omega/\nu)^{1/2}, related to the Reynolds number by Re=U∞a/ν=εM2Re = U_\infty a/\nu= \varepsilon M^2. For ε≪1\varepsilon\ll1, the problem permits a theoretical description, where the velocity components are expressed as an asymptotic expansion involving powers of ε\varepsilon. The leading-order terms, satisfying convection-free linear equations, manifest as harmonic functions with zero time-averaged values, while the first-order corrections introduce a non-zero steady-streaming component [12].

In the case of two-dimensional oscillating flow around a circular cylinder of radius, aa, much smaller than its length, Holtsmark et al. [1954] derived an analytical description of the Eulerian velocity for ε≪1\varepsilon\ll1, reformulated in primitive variables in Appendix § 8. This analysis provided expressions for both the dominant harmonic velocity and the subsequent first-order velocity corrections, later refined by Chong et al. [ [17]. In the distinguished regime of M∼1M \sim1, the magnitude of the resulting steady-streaming velocity is of the order of that of the Stokes drift Raney et al. [ [18], what requires to take into account both phenomena to fully describe the mean, steady Lagrangian flow. In this case, the symmetry of the problem yields identical recirculatory patterns in all four quadrants. More precisely, when MM is smaller than a critical value, McrM_{cr}, each quadrant displays a single vortex directed towards the cylinder along the oscillation axis. However, for larger values of MM an additional outer vortex is formed, a feature also confirmed by accompanying experiments [16]. As MM increases, the outer vortex strengthens while the inner one diminishes, being ultimately confined to a thin near-surface Stokes layer when M≫1M \gg1. Similar flow characteristics have been observed when the flow is confined between two coaxial cylinders, with the inner one oscillating and the outer one fixed [19].

In the early seventies, Tatsuno [ [20] investigated experimentally the secondary flow induced by an oscillating circular cylinder confined radially for values of M∼1M \sim1 and finite dimensionless stroke lengths ε\varepsilon, finding a behaviour similar to that of Skavlem and Tjotta [ [19] but different values of the critical parameters. While the oscillating flow for ε≪1\varepsilon\ll1 remains periodic and symmetric about the oscillation axis, the solution encountered when ε\varepsilon takes values that are not sufficiently small is known to be more complicated. The periodic viscous flow becomes unstable to axially periodic vortices above a critical value of ε\varepsilon that depends on MM leading to an asymmetrical flow featuring vortex shedding. This symmetry breaking is apparent in the experiments of Tatsuno and Bearman [Tatsuno and Bearman, [21], who identified eight different regimes of flow and reported a three-dimensional instability along the axis of the cylinder for most of the regimes with turbulent motion arising as the Reynolds number Re=εM2Re = \varepsilon M^2 exceeded a critical value. Furthermore, Justesen [Justesen, [22] carried out a numerical study of planar oscillating flow around a circular cylinder in the range of high Reynolds number and finite ε\varepsilon where, depending on the parameters values, several flow regimes that included vortex shedding were documented.

Regarding three-dimensional streaming flows in confined geometries, flow visualization has shown that, in the distinguished limit M∼1M \sim1 with solid boundaries, depending on the boundary layer thickness, the confined case remains very similar to the two-dimensional one towards the central plane [23, 24]. In the same regime, measurements of the three-dimensional velocity field show apparent three-dimensional effects which cannot be ignored, except in the midplane where symmetry enforces no axial motion, i.e. null velocity in the cylinder’s axis direction [25]. Also for M∼1M \sim1 and streaming Reynolds numbers, Res=ε2M2≪1Re_s = \varepsilon^2 M^2 \ll1, when the boundary surface is a bubble excited acoustically confined between two parallel walls, experiments reveal that there are large axial displacements near the bubble [26], which was also analytically reported later [27]. Similarly, three-dimensional streaming flows involving multi-curvature bodies have been also studied [28], as well as configurations where a series of cylinders are placed transverse to Coenen [ [29] or aligned with Alaminos-Quesada et al. [ [30] the oscillating flow, among others.

The objective of the present work is to delve into the complexities of an oscillating, viscous and incompressible flow around a circular cylindrical post confined between two parallel flat plates, as depicted in Figure 2. As mentioned above, the work is motivated by the analysis of the effect on the CSF flow in the spinal canal of existing microanatomical features that bridge irregularly the SSAS, see Figure 1. In this case, taking into account that the radii of the nerve roots range between a∼0.25−2a \sim0.25 - 2 mm Liu et al. [ [31], and the width of the canal is about h∼1−2.5h \sim1 - 2.5 mm Coenen et al. [ [32], Sass et al. [ [33], their aspect ratio, defined by λ=h/a\lambda= h/a, is about λ∼1−10\lambda\sim1 - 10 and the Womersley number M∼1M \sim1. To the best of our knowledge, situations where the obstacle is confined between parallel plates, with ε≪1\varepsilon\ll1, M∼1M \sim1 and the aspect ratio λ\lambda is varied have not been explored.

The paper is organized as follows. In § 2, an standard asymptotic analysis of steady-streaming flows for small stroke lengths, ε≪1\varepsilon\ll1, has been carried out together with the computation of the velocity correction or Stokes-drift. The numerical results are presented in § 3. In particular, the influence of the aspect ratio of the post on the streaming flow and the evaluation of the Lagrangian mean velocity field are reported in this Section. § 4 is devoted to describe the complementary experiments performed, the results obtained being reported in § 5. Finally, concluding remarks, limitations and future work are presented in § 6.

Formulation

A fluid of density ρ\rho and kinematic viscosity ν\nu fills the gap between two infinite parallel plates separated by a distance 2h2h (see Figure 2). Application of a uniform harmonic pressure gradient of angular frequency ω\omega in a given direction xx parallel to the plates results in a unidirectional motion with harmonic velocity v′\mathbf{v}' given by

v′uc=Re⁡{ieiωt′cosh⁡[(1+i2)hν/ω]−cosh⁡[(1+i2)z′ν/ω]cosh⁡[(1+i2)hν/ω]−1}ex(1)\frac{\mathbf{v}'}{u_c} = \operatorname{Re}\left\{ \mathrm{i}e^{\mathrm{i}\omega t'} \frac{\cosh\left[\left(\frac{1+\mathrm{i}}{2}\right)\frac{h}{\sqrt{\nu/\omega}}\right]-\cosh\left[\left(\frac{1+\mathrm{i}}{2}\right)\frac{z'}{\sqrt{\nu/\omega}}\right]}{\cosh\left[\left(\frac{1+\mathrm{i}}{2}\right)\frac{h}{\sqrt{\nu/\omega}}\right]-1} \right\}\mathbf{e}_x \tag*{(1)}

where t′t' and z′z' are the dimensional time and distance to the central plane, respectively, ex\mathbf{e}_x is the unit vector parallel to the pressure gradient, and ucu_c is the peak velocity amplitude, occurring at z′=0z' = 0. The exact solution (1) is altered by the presence of obstacles. We consider in particular the case of a circular cylinder perpendicular to the plates whose radius aa is comparable to the inter-plate semidistance hh, so that λ=h/a∼1\lambda= h/a \sim1. The description assumes frequencies ω\omega comparable to the inverse of the viscous time (a2/ν)−1(a^2/\nu)^{-1} and stroke lengths uc/ωu_c/\omega small compared with the cylinder radius aa, as measured by the parameters

M=(a2ων)1/2∼1andε=uc/ωa≪1.(2)M = \left(\frac{a^2\omega}{\nu}\right)^{1/2} \sim1 \quad\text{and} \quad\varepsilon= \frac{u_c/\omega}{a} \ll1. \tag*{(2)}

Using ω\omega, aa, ucu_c and ρucωa\rho u_c\omega a as scales for the time tt, cartesian coordinates (x,y,z)(x,y,z), velocity v=(u,v,w)\mathbf{v}=(u,v,w), and spatial pressure difference pp reduces the problem to that of integrating

∇⋅v=0,(3)\nabla\cdot\mathbf{v} = 0, \tag*{(3)}
∂v∂t+εv⋅∇v=−∇p+1M2∇2v,(4)\frac{\partial\mathbf{v}}{\partial t}+\varepsilon\mathbf{v}\cdot\nabla\mathbf{v}=-\nabla p+\frac{1}{M^2}\nabla^2\mathbf{v}, \tag*{(4)}

for x2+y2>1x^2+y^2>1 and −λ≤z≤λ-\lambda\le z \le\lambda subject to the nonslip boundary conditions

v=0{at x2+y2=1for −λ≤z≤λat z=±λfor x2+y2>1(5)\mathbf{v}=0\begin{cases} \text{at } x^2+y^2=1 & \text{for } -\lambda\le z \le\lambda\\ \text{at } z=\pm\lambda& \text{for } x^2+y^2>1 \end{cases} \tag*{(5)}

and the far-field condition

v=(Re⁡[ieitU0(z)],0,0) as x2+y2⟶∞ for −λ≤z≤+λ,(6)\mathbf{v}=\left(\operatorname{Re}\left[\mathrm{i}e^{\mathrm{i}t}U_0(z)\right],0,0\right) \text{ as } x^2+y^2\longrightarrow\infty\text{ for } -\lambda\le z \le+\lambda, \tag*{(6)}

where

U0(z)=cosh⁡((1+i2)Mλ)−cosh⁡((1+i2)Mz)cosh⁡((1+i2)Mλ)−1,(7)U_0(z)=\frac{\cosh\left(\left(\frac{1+\mathrm{i}}{2}\right)M\lambda\right)-\cosh\left(\left(\frac{1+\mathrm{i}}{2}\right)Mz\right)}{\cosh\left(\left(\frac{1+\mathrm{i}}{2}\right)M\lambda\right)-1}, \tag*{(7)}

as follows from Eq. (1). In the limit ε≪1\varepsilon\ll1, the dependent variables can be expressed as expansions in powers of ε\varepsilon as,

v=v0+εv1+⋯andp=p0+εp1+⋯(8)\mathbf{v}=\mathbf{v}_0+\varepsilon\mathbf{v}_1+\cdots\quad\text{and}\quad p=p_0+\varepsilon p_1+\cdots \tag*{(8)}

As can be anticipated from the two dimensional configuration, at leading order in the limit ε≪1\varepsilon\ll1 the resulting motion is determined by a linear problem involving the balance of the local acceleration with the pressure and viscous forces. Convective acceleration will be seen to introduce a small relative correction of order ε\varepsilon, including a steady-streaming component, which is to be determined here.

Leading-order flow

The leading-order harmonic solution can be expressed in the form v0=Re⁡[ieitV]\mathbf{v}_0=\operatorname{Re}[\mathrm{i}e^{\mathrm{i}t}\mathbf{V}] and p0=Re⁡[eitP]p_0=\operatorname{Re}[e^{\mathrm{i}t}P], where V(x,y,z)=(U,V,W)\mathbf{V}(x,y,z)=(U,V,W) and P(x,y,z)P(x,y,z), by integration of

∇⋅V=0,(9)\nabla\cdot\mathbf{V}=0, \tag*{(9)}
−V=−∇P+iM2∇2V,(10)-\mathbf{V}=-\nabla P+\frac{\mathrm{i}}{M^2}\nabla^2\mathbf{V}, \tag*{(10)}

subject to the nonslip boundary conditions

V=0{at x2+y2=1 for −λ≤z≤λ,at z=±λ for x2+y2>1(11)\mathbf{V}=0 \begin{cases} \text{at }x^2+y^2=1\text{ for }-\lambda\le z\le\lambda,\\ \text{at }z=\pm\lambda\text{ for }x^2+y^2>1 \end{cases} \tag*{(11)}

and the far-field condition

V=[U0(z),0,0]as x2+y2→∞for −λ≤z≤+λ.(12)\mathbf{V}=[U_0(z),0,0]\quad\text{as }x^2+y^2\to\infty\quad\text{for }-\lambda\le z\le+\lambda. \tag*{(12)}

Steady streaming

Collecting terms of order ε\varepsilon, the following order becomes

∇⋅v1=0,(13)\nabla\cdot\mathbf{v}_1=0, \tag*{(13)}
∂v1∂t+v0⋅∇v0=−∇p1+1M2∇2v1,(14)\frac{\partial\mathbf{v}_1}{\partial t}+\mathbf{v}_0\cdot\nabla\mathbf{v}_0=-\nabla p_1+\frac{1}{M^2}\nabla^2\mathbf{v}_1, \tag*{(14)}

with boundary conditions

v1=0{at x2+y2=1 for −λ≤z≤λ,at z=±λ for x2+y2>1,as x2+y2→∞ for −λ≤z≤λ(15)\mathbf{v}_1=0 \begin{cases} \text{at }x^2+y^2=1\text{ for }-\lambda\le z\le\lambda,\\ \text{at }z=\pm\lambda\text{ for }x^2+y^2>1,\\ \text{as }x^2+y^2\to\infty\text{ for }-\lambda\le z\le\lambda \end{cases} \tag*{(15)}

The first-order corrections v1\mathbf{v}_1 and p1p_1 are 2π2\pi-periodic functions of time. As a result of the nonlinear interactions associated with the convective terms, their time-averaged values vSS=⟨v1⟩\mathbf{v}_{SS}=\langle\mathbf{v}_1\rangle and pSS=⟨p1⟩p_{SS}=\langle p_1\rangle are nonzero, with ⟨⋅⟩=∫02π⋅ dt\langle\cdot\rangle=\int_0^{2\pi}\cdot\,\mathrm{d}t denoting the time-average operator. Taking the time average of (13)–(15) and using the identity ⟨v0⋅∇v0⟩=12Re⁡[V⋅∇V∗]\langle\mathbf{v}_0\cdot\nabla\mathbf{v}_0\rangle=\frac{1}{2}\operatorname{Re}[\mathbf{V}\cdot\nabla\mathbf{V}^{*}], where ∗* denotes complex conjugates, yields the steady-streaming problem

∇⋅vSS=0,(16)\nabla\cdot\mathbf{v}_{SS}=0, \tag*{(16)}
12Re⁡[V⋅∇V∗]=−∇pSS+1M2∇2vSS,(17)\frac{1}{2}\operatorname{Re}[\mathbf{V}\cdot\nabla\mathbf{V}^{*}]=-\nabla p_{SS}+\frac{1}{M^2}\nabla^2\mathbf{v}_{SS}, \tag*{(17)}

with boundary conditions

vSS=0{at x2+y2=1 for −λ≤z≤λ,at z=±λ for x2+y2>1,as x2+y2→∞ for −λ≤z≤λ(18)\mathbf{v}_{SS}=0 \begin{cases} \text{at }x^2+y^2=1\text{ for }-\lambda\le z\le\lambda,\\ \text{at }z=\pm\lambda\text{ for }x^2+y^2>1,\\ \text{as }x^2+y^2\to\infty\text{ for }-\lambda\le z\le\lambda \end{cases} \tag*{(18)}

The steady-streaming velocity vSS=⟨v1⟩\mathbf{v}_{SS}=\langle\mathbf{v}_1\rangle provides the leading-order description for the mean Eulerian velocity ⟨v⟩=εvSS\langle\mathbf{v}\rangle=\varepsilon\mathbf{v}_{SS} in the asymptotic limit ε≪1\varepsilon\ll1.

Lagrangian mean motion

When addressing the oscillating flow around a cylinder, the Lagrangian mean motion of the fluid particles comes partly from the contribution of the Eulerian mean motion (vSS=⟨v1⟩\mathbf{v}_{\mathrm{SS}} = \langle\mathbf{v}_1\rangle) and partly from that of the so-called Stokes drift, arising in non-uniform oscillating flows. Consequently, streamlines visualized in experiments employing dyed fluid do not coincide in general with those determined from the steady-streaming velocity. The velocity of the Lagrangian mean motion is,

vL=vSS+12Im⁡(V⋅∇V∗)(19)\mathbf{v}_{\mathrm{L}} = \mathbf{v}_{\mathrm{SS}} + \frac{1}{2}\operatorname{Im}(\mathbf{V}\cdot\nabla\mathbf{V}^{*}) \tag*{(19)}

where 12Im⁡(V⋅∇V∗)\frac{1}{2}\operatorname{Im}(\mathbf{V}\cdot\nabla\mathbf{V}^{*}) corresponds to the Stokes drift contribution. Of particular interest is the role played by the complex function 12V⋅∇V∗\frac{1}{2}\mathbf{V}\cdot\nabla\mathbf{V}^{*} in determining the Lagrangian mean motion. In the steady streaming formulation, as discussed in (16) and (17), the forcing term is the real part of the complex function, while its imaginary part gives the Stokes-drift velocity. Notably, in the scenario of large MM, viscous forces predominates within a thin Stokes layer. Beyond this layer, the flow is potential, characterized by the function V\mathbf{V} being real. Consequently, one can anticipate a that the contribution of the Stokes drift is negligible in comparison with the steady-streaming one for large values of MM.

Numerical results

For λ∼1\lambda\sim1, no analytic solution is available, and the problem formulated in § 2 must be solved numerically. To that aim, the Eqs. (9), (10) subject to (11), (12) and Eqs. (16), (17) subject to (18) were written in weak form and implemented in the finite element solver COMSOL Multiphysics® v5.6 using the Weak Form PDE toolbox. In the regime herein investigated, i.e. M∼1−10M \sim1 - 10 and ε≪1\varepsilon\ll1, the flow can be anticipated to be symmetric respect to the planes y=0y = 0 and z=0z = 0, and, thus, the computational domain for y≥1y \ge1 and z≥1z \ge1 has been considered imposing symmetry conditions in the aforementioned planes. The domain was discretized using a structured mesh with hexahedral elements. Second-order Lagrange elements were used for the pressure and the velocity. Mesh elements were compressed towards the solid and symmetry boundaryes to assure enough spatial resolution. Both the domain extension and the mesh resolution were varied to assure the domain-size and grid-resolution independence of the results. In the final configuration, the typical element size ratio with respect to the radius, aa, was ranged from 10−310^{-3} at the surfaces to 10−110^{-1} near the far-field boundary, assuring the accuracy of the solution with reasonable computational times. The domain extension in the xx- and yy-directions had a length of twenty-five times the radius of the cylinder, i.e. 25a25a.

Steady streaming

There are different methods to represent and visualize a three-dimensional flow field. Some of these methods can be found in the literature (see, for instance Jeong and Hussain [ [34] and Katsanoulis et al. [ [35]). In the cases of jet or wake-like flows, the QQ-criterion or the λ2\lambda_2-criterion are widely accepted for the representation of the flow through vortical structures. When the exact analytical solution can be found, streamsurfaces are often preferred to characterize the flow. In the particular case at hand, the absence of the normal component of the velocity field in the symmetry, z=0z = 0 and y=0y = 0, and antisymmetry planes, x=0x = 0, and the fact that the symmetry of the flow is conserved when ε≪1\varepsilon\ll1, allows the visualization of the flow topology by means of representation of the steady streaming streamfunction contours in these planes. Only one component of the vorticity field is not zero in each coordinate plane, so the corresponding streamfunction comes from ∇2ψSSi=ΩSSi\nabla^2\psi_{SS}^{i} = \Omega_{SS}^{i} for i=1(z=0),2(y=0),3(x=0)i = 1(z = 0), 2(y = 0), 3(x = 0) with homogeneous Dirichlet boundary conditions in the far-field, in the cylinder and in the walls.

Representative results are shown in Figure 3 for M=1M = 1, 12 and 20, and for λ=0.5\lambda= 0.5, 1 and 5. Because of the flow conditions and the geometrical configuration, the flow is symmetric respect to the plane x=0x = 0. Therefore, only half of the domain (x≥0x \ge0) is shown. The streaming structure arising for finite values of the dimensionless aspect ratio λ\lambda in the central plane z=0z = 0 is qualitatively similar to that of a single cylinder (λ≫1\lambda\gg1). In this plane, the mean secondary Eulerian flow displays a vortex in each quadrant for the lower value of the Womersley number M=1M = 1 (plotted in the left column), being the core of the vortex located along the π/4\pi/4 ray and closer to the cylinder as λ\lambda decreases. This vortex is also known to progressively approach the post wall as MM increases, as can be seen in Figure 3 for M=1M = 1 (left column), M=12M = 12 (central column) and M=20M = 20 (right column). These particular values of MM have been selected to show the transition in the topology of the flow for the three values of λ\lambda, where a second vortex with opposite circulation begins to form outside when MM exceeds a critical value, McrM_{cr}. Thus, the critical value of the Womersley number,

Streamlines in the symmetry and anti-symmetry planes for three values of $M$

Figure 3. Streamlines in the symmetry (z=0z = 0, y=0y = 0) and anti-symmetry (x=0x = 0) planes corresponding to the steady-streaming velocity, vSS\mathbf{v}_{\mathrm{SS}}, for λ=0.5\lambda= 0.5 (a), λ=1\lambda= 1 (b) and λ=5\lambda= 5 (c) for three different values of MM.

McrM_{cr}, represents a threshold for which a change in the topology of the flow is observed.

As seen in Figure 3, the critical value of the Womersley number depends on the aspect ratio λ\lambda. For λ=0.5\lambda= 0.5 (Figure 3a), it can be seen that the transition occurs between M=12M = 12 and M=20M = 20, while for λ=1\lambda= 1 and λ=5\lambda= 5 the transition happens for M<12M < 12 (Figure 3b and Figure 3c). More details will be given below. In the planes y=0y = 0 and x=0x = 0, a similar behaviour can be observed. As it happens in the plane z=0z = 0, the vortex in each quadrant of these planes progressively approaches to the post as MM increases. The core of these vortices is located at z=λ/2z = \lambda/2 for low values of MM and it is situated at lager values of zz, closer to the confining walls, when MM increases. That indicates that the flow tends to be two-dimensional near the central plane (z=0z = 0) for large values of MM and for large values of λ\lambda. When the Womersley number, MM, is increased sufficiently, a Stokes layer appears close to the solid boundaries and in the limit of ε≪1\varepsilon\ll1, M≫1M \gg1, Res≳O(1)Re_s \gtrsim O(1), the flow outside this boundary layer becomes potential.

As previously indicated, the value of McrM_{cr} for which a change in flow topology occurs depends on λ\lambda. For the case of an infinite cylinder, λ≫1\lambda\gg1, the critical value of the Womersley number can be determined from the exact solution [16] (Mcr≈6.08M_{cr} \approx6.08) as the value of MM for which the streamfunction ψSS\psi_{SS} vanishes in the far field. Our numerical simulations indicate that McrM_{cr} varies with λ\lambda, increasing as λ\lambda decreases. This dependency is shown in Figure 4, together with accompanying contours of the steady streaming streaming functions in the central plane z=0z = 0 for different pairs of MM and λ\lambda, shown in the insets. The value of McrM_{cr} is found to vary significantly for low values of λ\lambda. These higher values of McrM_{cr} for lower values of λ\lambda are attributable to the effect of confinement, which also produces a drastic reduction in the magnitude of the streaming motion. In contrast, for sufficiently large values of λ\lambda, the value of McrM_{cr} tends to that of the infinite cylinder (unconfined case). In this regard, as it can be seen for the cases of λ=1,5,∞\lambda= 1, 5, \infty with M=15M = 15 in the three upper insets, the core of the outer vortex moves away from the post as the aspect ratio increases. Identical behaviour is observed in the rest of the cases. Figure 4 also shows by a green point the value of McrM_{cr} corresponding to the experimental configuration reported in § 4.

Figure 4. Critical value of the Womersley number, McrM_{cr}, as a function of λ\lambda. Insets show streamlines in the central plane, z=0z = 0, for different values of MM and λ\lambda. The green point indicates the value of McrM_{cr} corresponding to the experimental configuration reported in § 4.

Figure 6. (a) Schematic view of the experimental facility, consisting of the in vitro model, the pumping system (programmable peristaltic pump and opened-air container) and the image acquisition setup (YAG laser and CCD camera, both synchronized with a PC). (b) In vitro model with the corresponding dimensions. (c) Flow rate waveforms used in the experiments.

Evaluation of the Lagrangian mean velocity

Figure 5 shows streamlines and color contours of vorticity Ω\Omega corresponding to three different velocity fields for λ=1\lambda= 1 and two Womersley numbers, i.e. M=0.5M = 0.5 and M=12M = 12. The first column shows the steady-streaming velocity, vSS\mathbf{v}_{SS}, the second one the Stokes-drift or velocity correction, vSD\mathbf{v}_{SD}, and the last one the steady mean Lagrangian velocity, given by the sum of the two of them, vL=vSS+vSD\mathbf{v}_{L} = \mathbf{v}_{SS} + \mathbf{v}_{SD}. Figure 5(a) displays streamlines and vorticity contours of the velocity fields in the symmetry planes z=0z = 0 and y=0y = 0 for M=0.5M = 0.5, while in Figure 5(b) the same results are presented for M=12M = 12.

Figure 5. Streamlines and colour contours of vorticity Ω\Omega corresponding to the steady-streaming velocity, vSS\mathbf{v}_{SS}, Stokes-drift velocity, vSD\mathbf{v}_{SD}, and steady mean Lagrangian velocity, vL=vSS+vSD\mathbf{v}_{L} = \mathbf{v}_{SS} + \mathbf{v}_{SD} for λ=1\lambda= 1 and M=0.5M = 0.5 (a) and M=12M = 12 (b). For each value of MM, streamlines are represented using a constant spacing δψ=0.01\delta\psi= 0.01 (M=0.5M = 0.5) and δψ=0.03\delta\psi= 0.03 (M=12M = 12), with the corresponding vorticity levels indicated in the colour bar on the right.

Regarding the streaming structure in the plane z=0z = 0 for M=0.5M = 0.5, the flow displays one vortex in each quadrant, with the clockwise circulation (negative vorticity) exhibited by the vortex in the first quadrant corresponding to fluid approaching the cylinder along the oscillation axis y=0y = 0. In the plane y=0y = 0, the flow, that is much weaker than in the middle symmetry plane z=0z = 0, displays also one vortex in the semi infinite space of x>1x > 1, y=0y = 0 with identical behaviour in the region of x<−1x < -1, y=0y = 0. The Stokes-drift results for M=0.5M = 0.5 display a primary clockwise-rotating vortex occupying most of the quadrant, along with a much weaker counter-rotating vortex of negligibly small circulation near the oscillation axis y=0y = 0. For this value of MM, this primary vortex is stronger than the corresponding steady-streaming vortex. The mean Lagrangian velocity field is largely determined by its Stokes-drift component, as reflected in the shape of the corresponding Lagrangian vortex in the right panel of Figure 5(a).

As can be seen for M=12M = 12 in Figure 5(b), in the first quadrant of the center plane z=0z = 0 the flow displays a second vortex with opposite circulation since MM exceeds the corresponding critical value McrM_{cr} for λ=1\lambda= 1 (Mcr≈10.2M_{cr} \approx10.2), which in the case of Stokes-drift prevails over the inner vortex, similar to what happens in an array of cylinders Alaminos-Quesada et al. [2023a]. The hegemony of the steady-streaming over the Stokes-drift in the mean Lagrangian motion becomes apparent in this range of the Womersley number. As inferred from Figure 5, the Stokes drift prevails for sufficiently small values of the Womersley number, whereas in the opposite limit the Stokes-drift motion fades away, as observed by comparing the steady-streaming and the mean Lagrangian motion contours, that are almost identical. Same dominance of the steady-streaming component happens in the plane y=0y = 0, where no second vortex appears for this value of the Womersley number.

Experiments

The complementary experiments carried out allowed us to analyse the long-time-scaled flow patterns of the motion generated for different flow conditions, varying the frequency, the stroke volume of the oscillating flow as well as the waveform. Two types of experiments were performed: the first one focused on the flow field around a single post and the second one on the flow around an array of 5 posts separated a distance ℓ=d/a\ell= d/a, where aa is the radius of the posts and dd the semi-distance between the axes of contiguous posts, being both types of experiments conducted in an adaptable experimental facility using distilled water of density ρ=998.2 kg/m3\rho= 998.2\ \mathrm{kg}/\mathrm{m}^{3} and kinematic viscosity ν=10−6 m2/s\nu= 10^{-6}\ \mathrm{m}^{2}/\mathrm{s} as working fluid. Figure 6(a) shows a sketch of the complete experimental facility, with the test section illuminated with a laser sheet, which includes the programmable pump used to establish the oscillating flow and the acquisition system. As seen in Fig. 6(b), the model consisted of an acrylic channel of total volume V=20×10×2 cm3V = 20 \times10 \times2\ \mathrm{cm}^{3}, where L=20 cmL = 20\ \mathrm{cm}, 2w=10 cm2w = 10\ \mathrm{cm} and 2h=2 cm2h = 2\ \mathrm{cm} are the length, width and height of the channel, respectively. The bottom of the channel was covered with black adhesive vinyl, which acted as a background for the experiments and at the same time prevented laser reflections. A vertical cylindrical post of radius a=5 mma = 5\ \mathrm{mm} and length 2h=2 cm2h = 2\ \mathrm{cm} was perpendicularly placed at the center of the canal, being the corresponding aspect ratio λ=h/a=2\lambda= h/a = 2. Such radius was sufficiently small to avoid any influence from the lateral confining walls on the fluid motion near the post, being the distance between the axis of the post and the walls equal to w=10aw = 10a. This model configuration was also extended to consider a linear array of 5 posts, which were arranged aligned with the fluid motion. These elements were separated a distance 2d=20 mm2d = 20\ \mathrm{mm}, being the dimensionless inter-post semi-distance ℓ=d/a=2\ell= d/a = 2. The in vitro model was connected to a programmable peristaltic pump on one side, which generated the oscillatory flow with a prescribed waveform, and an open-air container on the other side, to allow the displacement of the stroke volume (see Fig. 6a). The pump allowed to vary the relevant parameters of the flow: the stroke volume, ΔV\Delta V, the oscillatory frequency, f=ω/(2π)f = \omega/(2\pi), and the waveform (from a harmonic sinusoidal wave to a MRI-based anharmonic one as displayed in Fig. 6c). The fluid entered and exited the canal through two nozzles placed at the inlet and outlet sections, which also included flow conditioning elements to generate a uniform flow.

The experiments aimed at characterizing the time-averaged flow, induced by the oscillatory stream in the presence of the obstacles. To that end, in order to visualize the flow patterns and to measure the velocity and vorticity fields, particle image techniques were applied, seeding the flow with hollow glass spheres of diameter ϕp≈10 μm\phi_p \approx10\ \mu\mathrm{m}. The experiments consisted of series of long-duration sequences of black-and-white images acquired in phase, at the same frequency as the oscillating flow. The sequences contained between 100 and 200 images depending on the experimental set, which were acquired with a CCD camera of resolution 2048×2048 px2048 \times2048\ \mathrm{px}. The camera was synchronized with a pulsed YAG laser, Fig. 6a), used to generate a horizontal planar green sheet (λlaser=541 nm\lambda_{\mathrm{laser}} = 541\ \mathrm{nm}) of width δh<1 mm\delta h < 1\ \mathrm{mm} at z=0z = 0.

The resulting images were analysed in two different ways to determine the fluid particle trajectories as well as the Lagrangian steady velocity and vorticity fields. To obtain the long-time-scaled Lagrangian trajectories, the images were first enhanced subtracting their background which was calculated using an average filter. Afterwards, the steady fluid flow patterns could be extracted integrating all the snapshots acquired in each experimental set, to yield a cumulative image. It should be emphasized that all the images were taken at the same phase at a sampling frequency, fsf_s, equal to that of the oscillating flow, ff. In addition, the velocity and vorticity fields were also determined using Particle Image Velocimetry (PIV) analysis, applying the MATLAB toolbox PIVlab GUI Thielicke and Stamhuis [2014] to couples of images acquired at an interval of time, Δt=1/f\Delta t = 1/f. To perform the analysis, the images were first pre-processed, masking the cylinder to eliminate solid regions and applying a high-pass filter of window size larger than 150 px to avoid non-uniform background illumination, in combination with an automatic contrast adjustment. The PIV routine was applied to the pre-processed images, using an initial window of 160×160 px160 \times160\ \mathrm{px} with a window reduction up to a minimal size of 32×32 px32 \times32\ \mathrm{px} in 4 steps, applying a 50% window overlapping.

A series of experiments were conducted first to evaluate the effects of the Womersley number, MM, and the dimensionless stroke volume, ε=(U∞/ω)/a\varepsilon= (U_{\infty}/\omega)/a where U∞U_{\infty} is the amplitude of the oscillating flow, on the flow characteristics using a post with a fixed value of the aspect ratio λ=2\lambda= 2, complementing the numerical study reported in § 3. Thus, a value ε=0.2\varepsilon= 0.2 was initially established, assuming that it was small enough to approximate the distinguished limit ε≪1\varepsilon\ll1, and the Womersley number was varied from M=5M = 5 to 11 (experimental sets 1-7 and 13-19 in Table 1). Some of these experiments will be compared with the numerical results reported in § 3. Then, the stroke volume was increased to ε=0.5\varepsilon= 0.5 to evaluate the effect of ε\varepsilon (experimental sets 8-12 and 20-23 in Table 1). The experiments were performed considering two different waveforms in order to investigate the effect of the shape of the flow rate on the induced flow patterns. The first waveform consisted of an harmonic function that provided a periodic flow rate given by Q′(t′)=Qmax⁡cos⁡(ωt′)Q'(t') = Q_{\max}\cos(\omega t'), or similarly, a periodic velocity far from the post u∞′(t′)=U∞cos⁡(ωt′)u'_{\infty}(t') = U_{\infty}\cos(\omega t') with ω=2πf\omega= 2\pi f, u∞′(t′)=Q′(t′)/(4wh)u_{\infty}'(t') = Q'(t')/(4wh) and U∞=Qmax⁡/(4wh)U_{\infty} = Q_{\max}/(4wh) (experimental sets 1-12 in Table 1). For the second waveform, a realistic anharmonic flow rate, obtained by MRI measurements, was used. This waveform, emulating the CSF motion along the spinal canal, was obtained by a modal Fourier analysis, using 8 modes to produce a well fitted signal as described in § 5.3 (experimental sets 13-23 in Table 1). Finally, additional experiments, not included in Table 1, were conducted with a linear array of elements, aligned with the flow as described in § 5.4.

ExperimentWaveformff (Hz)MMU∞U_{\infty} (mm/s)ε\varepsilon
1Sine0.1651.0050.2
2Sine0.2461.5080.2
3Sine0.3272.010.2
4Sine0.4082.510.2
5Sine0.5093.140.2
6Sine0.64104.020.2
7Sine0.78114.900.2
8Sine0.1652.510.5
9Sine0.2463.770.5
10Sine0.3275.030.5
11Sine0.4086.280.5
12Sine0.1697.850.5
13MRI0.1651.0050.2
14MRI0.2561.5080.2
15MRI0.3272.010.2
16MRI0.4082.510.2
17MRI0.5093.140.2
18MRI0.625104.020.2
19MRI0.80114.900.2
20MRI0.1652.510.5
21MRI0.2563.770.5
22MRI0.3275.030.5
23MRI0.4086.280.5

Table 1. Experimental conditions of the different sets of the experiments performed. Here, M=(a2ω/ν)1/2M = (a^{2}\omega/\nu)^{1/2}, f=ω/(2π)f = \omega/(2\pi), U∞U_{\infty} is the peak velocity of the CSF motion for the harmonic waveform, and the harmonic equivalent peak velocity of the CSF motion for the anharmonic one, and ε=(U∞/ω)/a\varepsilon= (U_{\infty}/\omega)/a. ‘Sine’ indicates that the wave form is harmonic and ‘MRI’ that is an anharmonic cardiac signal obtained from Magnetic Resonance Imaging.

Figure 7. (a) Steady pathlines and (b) velocity field and vorticity contours of the flow around a single cylindrical post for ε=0.2\varepsilon= 0.2 and different values of MM. Left side of each panel corresponds to the experimental measurements and right side to the analytical predictions (ε≪1\varepsilon\ll1)

Experimental Results

In this section, the results of the time-averaged Lagrangian flow induced by the interaction of an oscillating stream with a single post, and an array of posts, obtained from the in vitro experiments are described. For the case of a single post, Section 5.1 first discusses the effect of the Womersley number, MM, for small values of the stroke length, ε≃0.2\varepsilon\simeq0.2, to be compared with numerical predictions of Section 3, obtained for the limit ε≪1\varepsilon\ll1. These results are extended to order unity values of ε\varepsilon in Section 5.2, for which the asymptotic description in powers of ε\varepsilon, for ε≪1\varepsilon\ll1, is no longer valid. The analysis in the aforementioned sections is conducted assuming a harmonic sinusoidal wave. In Section 5.3, this study is extended to analyze the time-averaged flow induced by an anharmonic oscillatory wave, similar to that found in realistic CSF flows, as induced by intracranial pressure fluctuations generated by the cardiac cycle. Finally, the time-averaged lagrangian flow originated by the presence of an array of posts aligned with the flow oscillation axis is briefly described in Section 5.4.

Effect of flow frequency, MM

As described in Section 2, for small values of ε≪1\varepsilon\ll1, the motion can be decomposed into the sum of a zero-averaged purely oscillatory velocity field and a first-order correction, whose non-zero average is denoted steady-streaming velocity. The Eulerian velocity field of the two-dimensional flow, λ=∞\lambda= \infty, is described in Appendix A. Such description, although strictly valid for ε≪1\varepsilon\ll1, has been recently reported to provide results that remain reasonably accurate for finite values of ε\varepsilon, close to unity, for a linear array of circular cylinders Alaminos-Quesada et al. [ [30]. Furthermore, the description of the long-time scale motion of a fluid particle includes the additional contribution of the Stokes drift Chong et al. [ [17] that, in combination with the steady streaming, yields the mean Lagrangian velocity field. The Stokes drift has been found to be comparable to the steady streaming for M≤1M \le1, but negligible otherwise Alaminos-Quesada [ [37]. In the case at hand, since the values of MM experimentally considered vary between 5 to 11, it is expected that the time-averaged Eulerian and Lagrangian velocity fields almost coincide in the cases reported in the present section for λ=2\lambda= 2.

In general terms, the results reveal that, as happens in the two-dimensional case, the Lagrangian mean motion in the horizontal symmetry plane z=0z = 0 shows identical recirculation patterns in the four quadrants. Consequently, for the single post case, results in only one quadrant will be reported in the following, see Fig. 7. In this configuration, two different flow topologies can be identified. For values of the Womersley number below a critical one (denoted hereafter as sub-critical values, M<McrM < M_{cr}), a sole vortex forms in each quadrant (left panels in Fig. 7, M=6M = 6), which, in the first quadrant, rotates in clockwise direction, i.e. the flow moves towards the post along the stream’s oscillation axis, y=0y = 0, and away from the post along the antisymmetry axis, x=0x = 0. For values larger than the critical one (or super-critical values), M>McrM > M_{cr}, a secondary external counter-rotating vortex appears, that, differently from what happens in the two-dimensional flow, in this case is closed due to the presence of the confining walls (right panels in Fig. 7, M=10M = 10). The external vortex encloses the internal one close to the wall of the post at distances that decrease as MM increases. As indicated in Section 3, the presence of the plates at the top and bottom boundaries has also an impact on the critical value, McrM_{cr}, at which such flow topology change takes place. In particular, McrM_{cr} increases as λ=h/a\lambda= h/a decreases, being Mcr≃6.08M_{cr} \simeq6.08 in the infinite cylinder case, λ=∞\lambda= \infty, and Mc≃8.5M_c \simeq8.5 in the present configuration, λ=2\lambda= 2 (see Fig. 4), which practically coincides with the value obtained analytically for ε≪1\varepsilon\ll1. In addition, the numerical analysis shows that the time-averaged flow becomes three-dimensional, and it is characterized by the formation of two additional recirculating regions in the planes x=0x = 0 and y=0y = 0, although this issue will not be experimentally explored, due to optical access limitations. To illustrate these observations, Fig. 7 shows the comparison between the time-averaged analytical (ε≪1\varepsilon\ll1) and experimental (ε≃0.2\varepsilon\simeq0.2) results. In particular, Fig. 7(a) shows pathlines of the mean Lagrangian motion obtained experimentally, left or second quadrant, and analytically, right or first quadrant, for values of M=6M = 6, 8 and 10, in columns from left to right, respectively. The first two columns display flow topologies corresponding to M<McrM < M_{cr}, whereas a change in the flow topology becomes apparent for M=10M = 10, where a second external counter-rotating vortex appears, which confines the inner vortex to the high-vorticity Stokes layer close to the post. Figure 7(b) shows the vectors associated to the velocity fields, together with colour contours of vorticity, normalized using the overall maximum analytical and experimental values of velocity and vorticity, respectively. In general terms, very similar and symmetric flow patterns are found for both the theoretical predictions and the measurements, with the analytical or experimental cores of the vortices located, approximately, along the π/4\pi/4 or 3π/43\pi/4 rays, respectively, in all cases, the theory predicting slightly larger radial distances from the post to the vortex core. Finally, it can be also observed that increasing values of MM yield more intense circulations, as inferred from the vorticity contours, while the core of the vortex, for M<McrM < M_{cr}, and that of the internal vortex, for M>McrM > M_{cr}, become closer to the post, as it was previously anticipated.

Effect of ε\varepsilon

The flow rate and stroke volume varies along the spinal canal, affecting the local value of ε\varepsilon, which also varies from order unity values at the cervical region to very small values at the lower regions. The asymptotic description in powers of ε\varepsilon in the limit ε≪1\varepsilon\ll1 ceases to be valid for finite values of the stroke length, ε∼O(1)\varepsilon\sim O(1). To explore this effect, the experimental results for ε∼O(1)\varepsilon\sim O(1) are shown herein for two values of MM, namely M=5<McrM = 5 < M_{cr} (sub-critical regime) and M=9>McrM = 9 > M_{cr} (super-critical regime), which are also compared with their counterparts for ε=0.2\varepsilon= 0.2 (considered here to be a good representation of the distinguished limit ε≪1\varepsilon\ll1). Therefore, Fig. 8 displays pathlines of time-averaged Lagrangian flow, left or second quadrant, and the corresponding velocity and vorticity fields, right or first quadrant, for M=5M = 5, first column, and for M=9M = 9, second column, for ε=0.2\varepsilon= 0.2 in Fig. 8(a), and for ε=0.5\varepsilon= 0.5 in Fig. 8(b). In this figure, the vorticity contours are shown with dimensions, Ω′\Omega' to facilitate the comparison. In both MM cases, similar flow patterns are observed for ε=0.2\varepsilon= 0.2 and ε=0.5\varepsilon= 0.5. Thus, a unique vortex is generated with clockwise rotation in the first quadrant for M=5M = 5, whereas for M=9M = 9, a second external counter-rotating recirculating region appears, enclosing the internal vortex near the post. A close inspection of the results reveals that increasing the stroke length has an impact on the global flow topology and rotation intensity. In that regard, for M=9M = 9, the size of the vortices decrease as ε\varepsilon increases, and their cores move closer to the post. The latter is accompanied with more intense recirculations, as observed from the larger amplitudes of vorticity in the case of ε=0.5\varepsilon= 0.5 compared to those of ε=0.2\varepsilon= 0.2 in Fig. 8. These results agree with those recently reported by Alaminos-Quesada et al. [2023a].

Time-averaged Lagrangian motion in anharmonic oscillating flows: CSF flow

So far the analysis has been focused on the flow around a post subject to a harmonic stream flow of velocity u∞′(t′)=U∞cos⁡ωt′u'_{\infty}(t') = U_{\infty}\cos\omega t'. However, it is common to find oscillatory flows that present anharmonic temporal variations in nature, such as that of the blood in veins and arteries, or the CSF flow in the central nervous system (CNS) Linninger et al. [ [1], Kelley et al. [ [3]. Therefore, in the present section, the study has been experimentally extended to analyze the flow induced by an anharmonic oscillatory flow around a post. Specifically, a wave of temporal dependence equal to that of a CSF flow measured in a human spinal canal Sincomb et al. [2022] has been considered. This velocity waveform can be expressed as a Fourier series u∞′(t′)=∑n=1∞Re⁡[Anei(ωnt′+ϕn)]u'_{\infty}(t') = \sum_{n=1}^{\infty} \operatorname{Re}[A_n e^{i(\omega_n t' + \phi_n)}], where AnA_n, ωn\omega_n and ϕn\phi_n are respectively the amplitude, angular frequency and phase associated to each mode Davidson and Riley [1972], Alaminos-Quesada et al. [2023a,b]. Figure 9(a) shows the flow rate associated to a harmonic waveform while Figure 9(b) represents a period of the flow rate provided by a cardiac anharmonic waveform. In this figure, the positive and negative flow rates are given during systole, of dimensionless duration t1t_1, and diastole, of dimensionless duration t2t_2, respectively with ∫02πQ(t) dt=0\int_{0}^{2\pi} Q(t)\,\mathrm{d}t = 0 or, similarly, ∣∫0t1Q(t) dt∣=∣∫t12πQ(t) dt∣\left|\int_{0}^{t_1} Q(t)\,\mathrm{d}t\right| = \left|\int_{t_1}^{2\pi} Q(t)\,\mathrm{d}t\right|, where t=2πt′/Tt = 2\pi t'/T represents the dimensionless time, and T=1/f=2π/ωT = 1/f = 2\pi/\omega is the period of the waveform.

Figure 9. (a) Harmonic flow waveform (t1=t2=πt_1 = t_2 = \pi). (b) Realistic flow waveform, obtained through MRI measurements and set as Fourier expansion (t1≠t2t_1 \ne t_2). Here, t=2πt′/Tt = 2\pi t'/T represents the dimensionless time, with T=1/f=2π/ωT = 1/f = 2\pi/\omega as the period of the waveform.

Unlike the results reported in the previous sections for the harmonic waveform where the secondary steady flow patterns were doubly symmetric, the flow induced by an anharmonic wave results in a steady motion which is only symmetric with respect to the axis of the flow oscillation y=0y = 0. Consequently, in the following, the first two quadrants, y≥0y \ge0, of the plane z=0z = 0 will be shown to describe the flow. Figure 10(a) shows pathlines of the time-averaged Lagrangian flow for the harmonic wave, with their counterparts for the anharmonic wave shown in Figure 10(b), both for ε=0.2\varepsilon= 0.2 and M=6M = 6, 8 and 10, respectively, where the first mode of the Fourier series, ω1\omega_1, was used to define MM in the anharmonic case and U∞U_{\infty} has been defined as that of the equivalent harmonic waveform signal with the same period and stroke volume. As previously shown, the harmonic wave induces steady flow patterns that are symmetric in each quadrant. However, this symmetry is broken in the anharmonic case, as a result of the inter-modes interaction Alaminos-Quesada et al. [2023a,b]. For the lowest value of MM (M=6M = 6), the lack of fore-and-aft symmetry is already observable, where the vortices appear distorted, with their cores near the post, as in the harmonic case. Despite this fact, differences can be observed in the size of the vortices, with the vortex located in the positive midplane (x>0x > 0) being smaller than its complementary one in the negative one (x<0x < 0). The loss of symmetry is also apparent for larger values of MM, e.g. M=8M = 8, where, as for M=6M = 6, two vortices form close to the post with their cores displaced downwards in the systolic direction (towards the positive xx-coordinate). In addition, the two vortices seem to be surrounded by a region of fundamentally streamwise velocity, which becomes more noticeable for even larger values of MM. It should be noted that, for M=10M = 10, the flow topology has already changed. However, differently from the flow pattern observed when a harmonic waveform is imposed, where a second outer vortex is formed in each quadrant, the steady flow is characterized by the formation of only an outer vortex in the first quadrant (downstream in the systolic direction). Moreover, the two vortices, already present for M<McrM < M_{\mathrm{cr}}, undergo larger distortions and are also displaced downwards in

Steady flow patterns around a cylindrical post

Figure 10. Steady flow patterns around a cylindrical post of aspect ratio λ=2\lambda= 2 for ε=0.2\varepsilon= 0.2 and different values of MM using, a harmonic waveform (a) and a anharmonic one (b).

the systolic direction. These three vortices are surrounded by an almost horizontal external stream, which is in agreement with the results obtained for the infinite cylinder Tatsuno [ [41] or more recently in a linear array of circular cylinders Alaminos-Quesada et al. [ [30].

The effect of considering finite stroke lengths is analyzed in Figure 11, where results for ε=0.5\varepsilon= 0.5 (Figure 11b) are compared with those previously shown for ε=0.2\varepsilon= 0.2 (Figure 11a), for M=6M = 6 and 88. As happened in the harmonic case, the size of the vortices formed is reduced when ε\varepsilon is increased, these being also more distorted and displaced downwards in the systolic direction in this case. Interestingly, as it was previously anticipated, increasing ε\varepsilon has also an impact on the flow topology change, reducing the value of McrM_{cr}. The latter can be observed for both values of MM, for which only two vortices are formed close to the post when ε=0.2\varepsilon= 0.2, one in each quadrant, whereas a third external vortex appears when ε=0.5\varepsilon= 0.5 in the first quadrant already at M=6M = 6.

Steady fluid particles trajectories around a cylindrical post of aspect ratio $\lambda = 2$ for $M = 6$ and $8$, obtained with an anharmonic, subject-specific CSF waveform

Figure 11. Steady fluid particles trajectories around a cylindrical post of aspect ratio λ=2\lambda= 2 for M=6M = 6 and 88, obtained with an anharmonic, subject-specific CSF waveform. (a) ε=0.2\varepsilon= 0.2 and (b) ε=0.5\varepsilon= 0.5.

Time-averaged motion around a linear array of posts

To further investigate the effect of the presence of different elements on the steady flow induced by an oscillatory stream flowing around them, the present section reports experimental results of the time-averaged Lagrangian flow around an array of five circular posts, aligned with the axis of oscillation, which are separated a semi-distance between the axes of contiguous posts dd, with the dimensionless semi-distance ℓ=d/a=2\ell= d/a = 2.

Figure 12 shows time-averaged pathlines in the horizontal symmetry plane z=0z = 0 generated by a harmonic flow for different values of MM, namely M=5M = 5, 6 and 9, and two values of ε\varepsilon, i.e. ε=0.2\varepsilon= 0.2 (Figure 12a) and 0.5 (Figure 12b), for ℓ=2\ell= 2. Qualitatively, the steady flow patterns induced for ℓ=2\ell= 2 are similar to those for ℓ=∞\ell= \infty. Thus, considering the results for ε=0.2\varepsilon= 0.2, for M=5M = 5 and 6, two symmetric, counter rotating vortices are generated, one in each quadrant, with the same direction of rotation as for ℓ=∞\ell= \infty. The presence of contiguous posts renders a confining effect, limiting the shape of each vortex to the semi-distance length in the axial direction. Consequently, deformed vortices are generated, whose cores are displaced towards the antisymmetry axis (x=0x = 0), in agreement with Alaminos-Quesada et al. [2023a]. This effect becomes more apparent for the lowest studied value of the Womersley number, M=5M = 5, since the vortices reduce their size and move closer to the posts when MM increases. The presence of neighbouring posts also has an impact on McrM_{cr}, delaying the flow transition featuring the apparition of external counter rotating vortices in each quadrant to larger values of MM. The latter is noticeable for M=9M = 9, where only two internal vortices are formed, one in each quadrant, whereas its counterpart for ℓ=∞\ell= \infty displays two vortices in each quadrant (Mcr≃8.5M_{cr} \simeq8.5 for ℓ=∞\ell= \infty) as can be seen in Figure 8. Figure 12(b) represents the results obtained for ε=0.5\varepsilon= 0.5 for the same three values of MM, also imposing a harmonic velocity. From an inspection of Figure 12, it cannot be concluded that the size of the vortices substantially change with ε\varepsilon (at least in the range of values of ε\varepsilon at which we have been able to perform the experiments). However, an increase of ε\varepsilon decreases the value of McrM_{cr}, as it was commented in previous sections. This can be observed in the case of M=9M = 9, for which a pair of inner and outer counter rotating vortices are formed in each quadrant when ε=0.5\varepsilon= 0.5, whereas only an inner vortex is seen when ε=0.2\varepsilon= 0.2. The linear array configuration is further explored in Figure 13, where time-averaged pathlines induced by an anharmonic fluid stream are shown for three values of MM, two of them in the sub-critical regime (M=6M = 6 and 9), and the other one in the super-critical regime (M=11M = 11), all of them at z=0z = 0 and for ε=0.2\varepsilon= 0.2. These kinds of waveforms have been found to cause the loss of flow symmetry in all quadrants, thus, this flow is only symmetric respect to the axis y=0y = 0. As a consequence, for M<McrM < M_{cr}, two different and asymmetric inner vortices are induced, with their cores unevenly displaced in the axial direction, which decrease in shape, become closer to the posts and increase in intensity when MM increases. For M>McrM > M_{cr}, a second external vortex appears in the first quadrant, similar to what happened for ℓ→∞\ell\to\infty, but in the case of ℓ=2\ell= 2, the vortices are deformed and axially enclosed within the semi-distance length. Moreover, in all cases the vortices are also surrounded by a region of streamwise velocity.

Steady flow patterns and corresponding dimensional velocity and vorticity fields for $M < M_{\mathrm{cr}}$ and $M > M_{\mathrm{cr}}$

Figure 8. Steady flow patterns (left side of each panel) and corresponding dimensional velocity and vorticity fields (right side of each panel) of the flow around a cylindrical element for M<McrM < M_{\mathrm{cr}} (left column, M=5M = 5) and M>McrM > M_{\mathrm{cr}} (right column, M=9M = 9). Here, a) ε=0.2\varepsilon= 0.2 is considered to satisfy the theoretical approximation (ε≪1\varepsilon\ll1) and b) ε=0.5\varepsilon= 0.5 assumes to be of order-unity.

Steady flow patterns around a linear array of cylindrical posts separated a distance $\ell = 2$ at different values of $M$ induced by a harmonic waveform

Figure 12. Steady flow patterns around a linear array of cylindrical posts separated a distance ℓ=2\ell= 2 at different values of MM (columns) induced by a harmonic waveform for (a) ε=0.2\varepsilon= 0.2 and (b) ε=0.5\varepsilon= 0.5.

Figure 13. Steady flow patterns around a linear array of cylindrical posts separated a distance ℓ=2\ell= 2 for different values of MM, induced by an anharmonic, patient-specific, waveform with ε=0.2\varepsilon= 0.2.

Conclusion

In this work, the problem of oscillatory streams around a cylindrical post, confined between two parallel plates, has been addressed. Eulerian and time-averaged Lagrangian flow fields have been described in detail for λ=O(1)\lambda= O(1) and stroke volumes ε<1\varepsilon< 1. The problem has been initially tackled analytically and numerically for ε≪1\varepsilon\ll1, varying the Womersly number, MM, and the post aspect ratio, λ\lambda. The numerical results indicate that, when the oscillating flow is harmonic, for a given value of λ\lambda, at low values of MM the flow field is sub-critical, showing a recirculating vortex anchored to the post symmetric with respect to the xx and yy axes. The size of the vortex decreases as λ\lambda decreases due to the confining effect of the upper and lower walls, at the same time that it narrows and approaches the surface of the post as MM increases. However for values of MM larger than a critical one, Mcr(λ)M_{\mathrm{cr}}(\lambda), that depends on λ\lambda, the flow becomes super-critical, with the formation of a second, outer vortex also symmetric, whose size decreases as λ\lambda decreases. The evolution of the critical Womersley number with the post aspect ratio is shown Figure 4, with examples of the flow topology displayed for different values of λ\lambda and MM. It can be clearly observed that transition between sub-critical and super-critical regimes is delayed when λ\lambda decreases, i.e., when the distance between the upper and lower walls is decreased, and the magnitude of the streaming motion is reduced. Regarding the Lagrangian mean motion, given by the sum of the steady-streaming and the Stokes drift, it can be concluded that the Stokes drift prevails over the steady-streaming for sufficiently small values of the Womersley number, whereas in the opposite limit the Stokes-drift motion fades away. Nevertheless, for the range of interest of MM in the present work, it has been observed that the steady-streaming dominates over the Stokes-drift.

The numerical results have been complemented by an experimental study, which compares the effect of MM in the Lagrangian steady flow patterns at z=0z = 0 and for λ=2\lambda= 2, showing a good agreement. This experimental work has been extended to situations beyond the theory, where ε\varepsilon is not longer much smaller than one, and cases close to reality, comparing the flow patterns obtained with harmonics and anharmonic waveforms. Furthermore, the mean Lagrangian flow generated by the presence of an array of posts immersed in an oscillatory stream has also been experimentally characterized, emulating the presence of nerve roots in the spinal subarachnoid space. Results obtained for a harmonic flow show time-averaged flow patterns that are identical in all quadrants, in each of which a recirculation cell forms near the post, for M<McrM < M_{\mathrm{cr}}, whose size reduces as MM increases. When M>McrM > M_{\mathrm{cr}}, the flow topology changes, and a second, external vortex is formed. The time-averaged flow induced by an anharmonic waveform is found to lose the symmetry respect to the transversal axis, and two vortices of different size are formed at each side of the post in the sub-critical regime (M<McrM < M_{\mathrm{cr}}). However, in the super-critical regime (M>McrM > M_{\mathrm{cr}}) only an outer cell generates in the quadrant downstream, in the systolic direction. Regardless the waveform, the flow topology transition has been found to also depend on the stroke length ε\varepsilon, taking place at values of McrM_{\mathrm{cr}} that decrease as ε\varepsilon increases. Finally, considering a linear array of posts, the induced flow patterns are similar to those observed with a single post configuration when the stream is harmonic. However, significantly different flow topologies are found for anharmonic waveforms. These results aim to provide a better understanding of the characterization of the influence of microanatomy on CSF flow and lay the foundation for future research, where not only the fluid flow but also the solute transport should be taken into account.

Acknowledgments

This work was supported by the coordinated project, PID2020-115961RB-C31, PID2020-115961RB-C32 and PID2020-115961RA-C33, financed by MICIU/AEI/ 10.13039/501100011033. F. Moral-Pulido wants to thank the Spanish Ministry of Universities for the financial support provided by the Fellowship FPU18/05694. Fruitful discussions with Prof. Antonio L. Sanchez are greatly acknowledged.

Revisting two-dimensional streaming flow over a circular cylinder

Using the same scales as the three dimensional problem for the time tt, cartesian coordinates (x,y)(x,y), velocity v=(u,v)\mathbf{v}=(u,v), and spatial pressure difference pp reduces the problem to that of integrating

∇⋅v=0,(20)\nabla\cdot\mathbf{v}=0, \tag*{(20)}
∂v∂t+εv⋅∇v=−∇p+1M2∇2v,(21)\frac{\partial\mathbf{v}}{\partial t}+\varepsilon\mathbf{v}\cdot\nabla\mathbf{v}=-\nabla p+\frac{1}{M^{2}}\nabla^{2}\mathbf{v}, \tag*{(21)}

for x2+y2>1x^{2}+y^{2}>1, subject to the nonslip boundary conditions

v=0atx2+y2=1(22)\mathbf{v}=0\quad\text{at}\quad x^{2}+y^{2}=1 \tag*{(22)}

and the far-field condition

v=(cos⁡t,0)asx2+y2→∞.(23)\mathbf{v}=(\cos t,0)\quad\text{as}\quad x^{2}+y^{2}\to\infty. \tag*{(23)}

Eulerian velocity for ε≪1\varepsilon\ll1

Following standard practice, we describe the flow by introducing expansions for the different flow variables in powers of ε\varepsilon, i.e.

v=v0+εv1+⋯(24)\mathbf{v}=\mathbf{v}_{0}+\varepsilon\mathbf{v}_{1}+\cdots \tag*{(24)}

and p=p0+εp1+⋯p=p_{0}+\varepsilon p_{1}+\cdots. The leading-order solution satisfies the linear equation

∇⋅v0=0,(25)\nabla\cdot\mathbf{v}_{0}=0, \tag*{(25)}
∂v0∂t=−∇p0+1M2∇2v0,(26)\frac{\partial\mathbf{v}_{0}}{\partial t}=-\nabla p_{0}+\frac{1}{M^{2}}\nabla^{2}\mathbf{v}_{0}, \tag*{(26)}

subject to boundary conditions stated in (22) and (23). The problem can be alternatively written using v0=Re⁡(eitV0)\mathbf{v}_{0}=\operatorname{Re}(e^{it}\mathbf{V}_{0}) with V0(x,y)=(U0,V0)\mathbf{V}_{0}(x,y)=(U_{0},V_{0}) representing a complex velocity. The integration can make use of the stream function Ψ0(x,y)\Psi_{0}(x,y) defined such that

U0=∂Ψ0∂yandV0=−∂Ψ0∂x(27)U_{0}=\frac{\partial\Psi_{0}}{\partial y}\quad\text{and}\quad V_{0}=-\frac{\partial\Psi_{0}}{\partial x} \tag*{(27)}

and

Ω0=−∇2Ψ0,(28)\Omega_{0}=-\nabla^{2}\Psi_{0}, \tag*{(28)}

where the complex function Ω0(x,y)\Omega_{0}(x,y), related to the vorticity by

ω0=∂v0∂x−∂u0∂y=Re⁡(eitΩ0),(29)\omega_{0}=\frac{\partial v_{0}}{\partial x}-\frac{\partial u_{0}}{\partial y}=\operatorname{Re}(e^{it}\Omega_{0}), \tag*{(29)}

satisfies

iM2Ω0=∇2Ω0,(30)iM^{2}\Omega_{0}=\nabla^{2}\Omega_{0}, \tag*{(30)}

as follows from (26). Equations (28) and (30) must be integrated subject to

Ψ0=(x,y)⋅∇Ψ0=0atx2+y2=1,(31)\Psi_{0}=(x,y)\cdot\nabla\Psi_{0}=0\quad\text{at}\quad x^{2}+y^{2}=1, \tag*{(31)}
∂Ψ0∂y−1=∂Ψ0∂x=0asx2+y2→∞.(32)\frac{\partial\Psi_{0}}{\partial y}-1=\frac{\partial\Psi_{0}}{\partial x}=0\quad\text{as}\quad x^{2}+y^{2}\to\infty. \tag*{(32)}

For a general value of MM, the resulting velocity V0(x,y)=(U0,V0)\mathbf{V}_{0}(x,y)=(U_{0},V_{0}) has real and imaginary parts. Note, however, that in the limit of steady creeping flow M≪1M\ll1 the solution is real everywhere, while in the inviscid limit M≫1M\gg1 the solution contains an imaginary part only in the thin Stokes layer of thickness 1/M1/M that develops on the cylinder surface, outside of which Ω0=0\Omega_{0}=0 and Im⁡(V0=0)\operatorname{Im}(\mathbf{V}_{0}=0).

Formulating the problem in cylindrical polar coordinates instead of Cartesians and separating variables, i.e. Ω0(r,θ)=ΩR(r)ΩΘ(θ)\Omega_0(r,\theta)=\Omega_R(r)\Omega_\Theta(\theta), we find the following solution of the equation (30)

Ω0(r,θ)=π2ΩH1(1)(αMr)sin⁡θ,(33)\Omega_0(r,\theta)=\frac{\pi}{2}\Omega H_1^{(1)}(\alpha Mr)\sin\theta, \tag*{(33)}

where Ω\Omega is a constant to be determined by the boundary conditions, H1(1)H_1^{(1)} is the Hankel function of first kind and order one and α=i\alpha=\sqrt{i}.

Using the fact that the equation (28) is linear, it can be applied the superposition principle.

∇2Ψ0=−Ω0:{∇2Ψ^0=0with b.c. (31)–(32)∇2Ψ~0=−Ω0with homogeneous b.c.(34)\nabla^2\Psi_0=-\Omega_0:\quad \begin{cases} \nabla^2\hat{\Psi}_0=0 & \text{with b.c. (31)--(32)}\\ \nabla^2\tilde{\Psi}_0=-\Omega_0 & \text{with homogeneous b.c.} \end{cases} \tag*{(34)}

The solution of ∇2Ψ^0=0\nabla^2\hat{\Psi}_0=0 with the boundary conditions (31) and (32) can be expressed as

Ψ^0=A(r+Br)sin⁡θ,(35)\hat{\Psi}_0=A\left(r+\frac{B}{r}\right)\sin\theta, \tag*{(35)}

that has the same form as potential solution for the flow. The value of the constants of integration AA and BB will be determined below.

The equation ∇2Ψ~0=−Ω0\nabla^2\tilde{\Psi}_0=-\Omega_0 is solved using the Frobenius method, yielding:

Ψ~0=1αMH1(1)(αMr)H0(1)(αM)sin⁡θ(36)\tilde{\Psi}_0=\frac{1}{\alpha M}\frac{H_1^{(1)}(\alpha Mr)}{H_0^{(1)}(\alpha M)}\sin\theta \tag*{(36)}

The leading order stream function will be:

Ψ0=Ψ^0+Ψ~0.(37)\Psi_0=\hat{\Psi}_0+\tilde{\Psi}_0. \tag*{(37)}

In this case, the stream function will be:

Ψ0=[1αMH1(1)(αMr)H0(1)(αM)−12(r+1rH2(1)(αM)H0(1)(αM))]sin⁡θ(38)\Psi_0=\left[\frac{1}{\alpha M}\frac{H_1^{(1)}(\alpha Mr)}{H_0^{(1)}(\alpha M)}-\frac{1}{2}\left(r+\frac{1}{r}\frac{H_2^{(1)}(\alpha M)}{H_0^{(1)}(\alpha M)}\right)\right]\sin\theta \tag*{(38)}

To calculate the constant Ω\Omega it must be imposed the no slip boundary condition for the azimuthal velocity (vθ0=−∂Ψ0∂r=0v_{\theta0}=-\frac{\partial\Psi_0}{\partial r}=0 at r=1r=1). The value of the constant Ω\Omega in the equation (33) is:

Ω=H2(1)(αM)H0(1)(αM).(39)\Omega=\frac{H_2^{(1)}(\alpha M)}{H_0^{(1)}(\alpha M)}. \tag*{(39)}

As follows from collecting terms of order ε\varepsilon, the problem at the following order becomes

∇⋅v1=0,(40)\nabla\cdot\mathbf{v}_1=0, \tag*{(40)}
∂v1∂t+v0⋅∇v0=−∇p1+1M2∇2v1,(41)\frac{\partial\mathbf{v}_1}{\partial t}+\mathbf{v}_0\cdot\nabla\mathbf{v}_0=-\nabla p_1+\frac{1}{M^2}\nabla^2\mathbf{v}_1, \tag*{(41)}

with boundary conditions

v1=0{at x2+y2=1as x2+y2→∞(42)\mathbf{v}_1=0\quad \begin{cases} \text{at }x^2+y^2=1\\ \text{as }x^2+y^2\to\infty \end{cases} \tag*{(42)}

The first-order corrections v1\mathbf{v}_1 and p1p_1 are 2π2\pi-periodic functions of time. As a result of the nonlinear interactions associated with the convective terms, their time-averaged values ⟨v1⟩\langle\mathbf{v}_1\rangle and ⟨p1⟩\langle p_1\rangle are nonzero, with ⟨⋅⟩=∫02π⋅ dt\langle\cdot\rangle=\int_0^{2\pi}\cdot\,\mathrm{d}t denoting the time-average operator. Taking the time average of (40)--(42) and using the identity ⟨v0⋅∇v0⟩=12Re⁡(V0⋅∇V0∗)\langle\mathbf{v}_0\cdot\nabla\mathbf{v}_0\rangle=\frac{1}{2}\operatorname{Re}(\mathbf{V}_0\cdot\nabla\mathbf{V}_0^*), where the asterisk ∗* denotes complex conjugates, provides the steady-streaming problem

∇⋅⟨v1⟩=0,(43)\nabla\cdot\langle\mathbf{v}_1\rangle=0, \tag*{(43)}
12Re⁡(V0⋅∇V0∗)=−∇⟨p1⟩+1M2∇2⟨v1⟩,(44)\frac{1}{2}\operatorname{Re}\left(\mathbf{V}_0\cdot\nabla\mathbf{V}_0^*\right)=-\nabla\langle p_1\rangle+\frac{1}{M^2}\nabla^2\langle\mathbf{v}_1\rangle, \tag*{(44)}

with boundary conditions

⟨v1⟩={0at x2+y2=1,0as x2+y2→∞(45)\langle\mathbf{v}_1\rangle= \begin{cases} 0 & \text{at } x^2+y^2=1,\\ 0 & \text{as } x^2+y^2\to\infty \end{cases} \tag*{(45)}

Introducing the time-averaged stream function ⟨ψ1⟩\langle\psi_1\rangle, with

⟨u1⟩=∂⟨ψ1⟩∂yand⟨v1⟩=−∂⟨ψ1⟩∂x(46)\langle u_1\rangle=\frac{\partial\langle\psi_1\rangle}{\partial y} \quad\text{and}\quad \langle v_1\rangle=-\frac{\partial\langle\psi_1\rangle}{\partial x} \tag*{(46)}

and associated vorticity ⟨ω1⟩=−∇2⟨ψ1⟩\langle\omega_1\rangle=-\nabla^2\langle\psi_1\rangle, reduces (44) to

−M22Re⁡(∂Ψ0∂y∂Ω0∗∂x−∂Ψ0∂x∂Ω0∗∂y)=∇2∇2⟨ψ1⟩(47)-\frac{M^2}{2}\operatorname{Re}\left(\frac{\partial\Psi_0}{\partial y}\frac{\partial\Omega_0^*}{\partial x}-\frac{\partial\Psi_0}{\partial x}\frac{\partial\Omega_0^*}{\partial y}\right)=\nabla^2\nabla^2\langle\psi_1\rangle \tag*{(47)}

involving the complex functions Ψ0\Psi_0 and Ω0∗=−∇2Ψ0∗\Omega_0^*=-\nabla^2\Psi_0^*. The resulting recirculating cells can be obtained by integrating (47) subject to the boundary conditions:

⟨ψ1⟩=(x,y)⋅∇⟨ψ1⟩=0at x2+y2=1,(48)\langle\psi_1\rangle=(x,y)\cdot\nabla\langle\psi_1\rangle=0\quad\text{at }x^2+y^2=1, \tag*{(48)}
⟨ψ1⟩=0as x2+y2→∞.(49)\langle\psi_1\rangle=0\quad\text{as }x^2+y^2\to\infty. \tag*{(49)}

In cylindrical polar coordinates:

−M22Re⁡[1r(∂Ψ0∂θ∂Ω0∗∂r−∂Ψ0∂r∂Ω0∗∂θ)]=∇2∇2⟨ψ1⟩(50)-\frac{M^2}{2}\operatorname{Re}\left[\frac{1}{r}\left(\frac{\partial\Psi_0}{\partial\theta}\frac{\partial\Omega_0^*}{\partial r}-\frac{\partial\Psi_0}{\partial r}\frac{\partial\Omega_0^*}{\partial\theta}\right)\right]=\nabla^2\nabla^2\langle\psi_1\rangle \tag*{(50)}

The equation can be solved using the method of separation of variables, supposing ⟨ψ1⟩=ψ1R(r)sin⁡2θ\langle\psi_1\rangle=\psi_{1R}(r)\sin2\theta, and the method of variation of parameters to do:

ψ1R(r)=r4(148∫1r1r′Φ(r′) dr′+C1)−r2(116∫1rr′Φ(r′) dr′+C2)+(116∫1rr′3Φ(r′) dr′+C3)+1r2(−148∫1rr′5Φ(r′) dr′+C4),(51)\psi_{1R}(r)=r^4\left(\frac{1}{48}\int_1^r\frac{1}{r'}\Phi(r')\,\mathrm{d}r'+C_1\right)-r^2\left(\frac{1}{16}\int_1^r r'\Phi(r')\,\mathrm{d}r'+C_2\right)+\left(\frac{1}{16}\int_1^r r'^3\Phi(r')\,\mathrm{d}r'+C_3\right)+\frac{1}{r^2}\left(-\frac{1}{48}\int_1^r r'^5\Phi(r')\,\mathrm{d}r'+C_4\right), \tag*{(51)}

where

Φ(r)=iM2Im⁡[H2(1)(αMr)H0(1)(αM)+2∥H0(1)(αM)∥2(1r2H2(1)(αM)H0(1)(αr)+H2(1)(αMr)H0(1)(αM))],(52)\Phi(r)=iM^2\operatorname{Im}\left[\frac{H_2^{(1)}(\alpha Mr)}{H_0^{(1)}(\alpha M)}+\frac{2}{\left\lVert H_0^{(1)}(\alpha M)\right\rVert^2}\left(\frac{1}{r^2}H_2^{(1)}(\alpha M)H_0^{(1)}(\alpha r)+H_2^{(1)}(\alpha Mr)H_0^{(1)}(\alpha M)\right)\right], \tag*{(52)}

and

∥H0(1)(αM)∥2=Re⁡(H0(1)(αM))2+Im⁡(H0(1)(αM))2.\left\lVert H_0^{(1)}(\alpha M)\right\rVert^2=\operatorname{Re}\left(H_0^{(1)}(\alpha M)\right)^2+\operatorname{Im}\left(H_0^{(1)}(\alpha M)\right)^2.

The constants

C1=−148∫1∞1rΦ(r) dr,C2=116∫1∞rΦ(r) dr,C3=116∫1∞1rΦ(r) dr−18∫1∞rΦ(r) dr,C4=−124∫1∞1rΦ(r) dr+116∫1∞rΦ(r) dr,(53)\begin{aligned} C_1&=-\frac{1}{48}\int_1^\infty\frac{1}{r}\Phi(r)\,\mathrm{d}r,\quad C_2=\frac{1}{16}\int_1^\infty r\Phi(r)\,\mathrm{d}r,\\ C_3&=\frac{1}{16}\int_1^\infty\frac{1}{r}\Phi(r)\,\mathrm{d}r-\frac{1}{8}\int_1^\infty r\Phi(r)\,\mathrm{d}r,\\ C_4&=-\frac{1}{24}\int_1^\infty\frac{1}{r}\Phi(r)\,\mathrm{d}r+\frac{1}{16}\int_1^\infty r\Phi(r)\,\mathrm{d}r, \tag*{(53)} \end{aligned}

are determined by imposing the boundary conditions 48,49. The Lagrangian mean motion can be defined as:

vL=⟨v1⟩+12Im⁡(V0⋅∇V0∗)(54)\mathbf{v}_{L}=\langle\mathbf{v}_1\rangle+\frac{1}{2}\operatorname{Im}\left(\mathbf{V}_0\cdot\nabla\mathbf{V}_0^*\right) \tag*{(54)}

where 12Im⁡(V0⋅∇V0∗)\frac{1}{2}\operatorname{Im}(\mathbf{V}_0\cdot\nabla\mathbf{V}_0^*) is the Stokes drift or correction. In the present case, the Streamfunction of the Stokes drift can be calculated as:

ψSD=12Im⁡[H2(1)(αMr)H0(1)(αM)+1r2H2(1)(αM)H0(1)(αM)(H0(1)∗(αMr)H0(1)∗(αM)−1)+H0(1)(αMr)H2(1)∗(αMr)H0(1)(αM)H0(1)∗(αM)]sin⁡2θ.(55)\psi_{SD}=\frac{1}{2}\operatorname{Im}\left[\frac{H_2^{(1)}(\alpha Mr)}{H_0^{(1)}(\alpha M)}+\frac{1}{r^2}\frac{H_2^{(1)}(\alpha M)}{H_0^{(1)}(\alpha M)}\left(\frac{H_0^{(1)*}(\alpha Mr)}{H_0^{(1)*}(\alpha M)}-1\right)+\frac{H_0^{(1)}(\alpha Mr)H_2^{(1)*}(\alpha Mr)}{H_0^{(1)}(\alpha M)H_0^{(1)*}(\alpha M)}\right]\sin2\theta. \tag*{(55)}

Declaration of interest.

The authors report no conflict of interest.

References

  1. [1]A.A. Linninger, K. Tangen, C.Y. Hsu, and D. Frim. Cerebrospinal fluid mechanics and its coupling to cerebrovascular dynamic Annu. Rev. Fluid Mech., 48(1): 219—257, 2016.DOI
  2. [3]D.H. Kelley and J.H. Thomas. Cerebrospinal fluid flow. Annu. Rev. Fluid Mech., 55:2237—26, 2023.DOI
  3. [12]N. Riley. Steady streaming. Ann. Rev. Fluid Mech., 33(1):43—65, 2001.
  4. [16]J. Holtsmark, I. Johnsen, T. Sikkel­and, and S. Skavlem. Boundary layer flow near a cylindrical obstacle in an oscillating, incompressible fluid. J. Acoust. Soc. Am., 26(1):26—39, 1954.
  5. [17]K. Chong, S.D Kelly, S. Smith, and J.D. Eldredge. Inertial particle trapping in viscous streaming. Phys. Fluids, 25(3):033602, 2013.DOI
  6. [18]W.P. Raney, J.C. Corelli, and P.J. Westervelt. Acoustical streaming in the vicinity of a cylinder. J. Acoust. Soc. Am., 26(6):1006—1014, 1954.DOI
  7. [19]S. Skavlem and S. Tj̈ötta. Steady rotational flow of an incompressible, viscous fluid enclosed between two coaxial cylinders. J. Acoust. Soc. Am., 27(1):26—33, 1955.DOI
  8. [20]M. Tatsuno. Circulatory streaming around an oscillating circular cylinder at low Reynolds numbers. J. Phys. Soc. Japan, 35(2):915—920, 1973.DOI
  9. [21]M. Tatsuno and P.W. Bearman. A visual study of the flow around an oscillating circular cylinder at low keulegan—carpenter numbers and low stokes numbers. J. Fluid Mech., 211:157—182, 1990.DOI
  10. [22]P. Justesen. A numerical study of oscillating flow around a circular cylinder. J. Fluid Mech., 222:157—196, 1991.DOI
  11. [23]B. R. Lutz, J. Chen, and D. T. Schwartz. Microscopic steady streaming eddies created around short cylinders in a channel:619 Flow visualization and stokes layer scaling. Phys. Fluids, 17(2):0023601, 2005.
  12. [24]B. R. Lutz, J. Chen, and D. T. Schwartz. Hydrodynamic tweezers: 1. noncontact trapping of single cells using steady streaming microeddies. Anal. Chem., 78(15):5429—5435, 2006.pubmed.ncbi.nlm.nih.gov/16878879
  13. [25]W. Tien, D. Dabiri, V.H. Lieu, and D.T. Schwartz. Volumetric velocity measurements of an acoustic streaming microeddy array using color-coded three-dimensional micro particle tracking velocimetry. 10TH Int. Symp PIV – PIV13, Delft, 2013, july2–4.
  14. [26]A. Marin, M. Rossi, B. Rallabandi, C. Wang, Hilgenfeldt S., and C.J. Kähler. Three-dimensional phenomena in microbubble acoustic streaming. Phys. Rev. Applied, 3(041001):1—5, 2015.arxiv.org/abs/1503.04138
  15. [27]B. Rallabandi, A. Marin, M. Rossi, C.J. Kähler, and S. Hilgenfeldt. Three-dimensional streaming flow in confined geometries. J. Fluid Mech., 777:408—429, 2015.DOI
  16. [28]F. K. Chan, Y. Bhosale, T. Parthasarathy, and M. Gazzola. Three-dimensional geometry and topology effects in viscous streaming. J. Fluid Mech., 933:A53, 2022.DOI
  17. [29]W. Coenen. Steady streaming around a cylinder pair. Proc. R. Soc. A: Math. Phys. Eng. Sci., 472(2195), 2016.DOI
  18. [30]J. Alaminos-Quesada, J. J. Lawrence, W. Coenen, and A. L. Sánchez. Oscillating viscous flow past a streamwise linear array of circular cylinders. J. Fluid Mech., 959(A39), 2023.
  19. [31]Y. Liu, X. Zhou, J. Ma, Y. Ge, and X. T. Cao. The diameters and number of nerve fibers in spinal nerve roots. J. Spinal Cord Med., 38(4): 532—537, 2015.DOI
  20. [32]W. Coenen, C. Gutierrez-Montes, S. Sincomb, E. Criado-Hidalgo, K. Wei, K. King, V. Haughton, C. Martínez-Bazán, A.L. Sánchez, and J.C. Lasheras. Subject-specific studies of csf bulk flow patterns in the spinal canal: implications for the dispersion of solute particles in intrathecal drug delivery. AJNR Am. J. Neuroradiol., 40(7): 1242—1249, 2019.DOI
  21. [33]L.R. Sass, M. Khani, G. Natividad, R.S. Tubbs, O. Baledent, and B.A. Martin. A 3D subject-specific model of the spinal subarachnoid space with anatomically realistic ventral and dorsal spinal cord nerve rootlets. Fluids and Barriers of the CNS,, 14(1): 36, 2017.
  22. [34]J. Jeong and F. Hussain. On the identification of a vortex. J. Fluid Mech., 285: 69—94, 1995.
  23. [35]S. Katsanoulis, F. Kogelbauer, R. Kaundinya, J. Ault, and G. Haller. On the identification of a vortex. J. Fluid Mech., 954:A28, 2023.
  24. [37]J. Alaminos-Quesada. Limit of the two-dimensional linear potential theories on the propulsion of a flapping airfoil in forward flight in terms of the reynolds and strouhal number. Phys. Rev. Fluids, 6:123101, 2021.
  25. [41]M. Tatsuno. Secondary flow induced by a circular cylinder performing unharmonic oscillations. J. Phys. Soc. Japan, 50(1):330–337, 1981.DOI

Paper details

vlnc:W_vmd8pcuy
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