betaFeedback

WorksPhysics and Astronomy

Near-wall sedimentation of short cylinders at low Reynolds numbers: Translational and angular velocity models and experimental validation

Abstract

The sedimentation of short cylinders near rigid walls involves coupled translational and rotational motion, for which readily applicable kinematic descriptions remain limited. In this study, theoretical expressions for the translational and angular velocities were derived by combining slender-body theory, semi-empirical finite-aspect-ratio corrections, and hydrodynamic descriptions of fluid inertia and wall effects. The resulting expressions were assessed through time-resolved sedimentation experiments with rigid cylinders of mean aspect ratios χˉ≈3.8\barχ\approx3.8–5.85.8 in a rectangular duct at particle Reynolds numbers of ReP≈O(10−1)Re_{\mathrm P}\approx O(10^{-1}). The experiments revealed distinct near-wall trajectories, including non-contact glancing motion and contact-induced flipping. For glancing trajectories, the predicted translational and angular velocities reproduced the principal features of the measured kinematics, with good qualitative and, over substantial parts of the trajectories, quantitative agreement. The comparisons also identified different ranges of applicability for the investigated wall corrections and showed that wall-induced effects dominated the angular velocity under the conditions studied. The resulting, experimentally validated expressions provide readily applicable estimates of the instantaneous translational and angular velocities of short cylinders near rigid walls and a reference for investigating additional influences on their sedimentation dynamics, including lower Reynolds numbers and particle or wall elasticity.

Introduction

The sedimentation of cylindrical and rod-shaped particles is relevant to a wide range of environmental, biological, and industrial processes. Examples include the transport and deposition of natural and synthetic fibres, fibrous microplastics, and rod-shaped biological particles. In applications involving particle transport, the settling velocity is frequently a quantity of primary practical interest because it influences particle residence times, transport distances, and deposition. Accordingly, recent studies continue to seek predictive descriptions of the settling velocities of non-spherical particles. Hamidi et al. [1], for example, experimentally investigated the settling of straight and bent rods in a quiescent fluid and developed a semi-empirical description of their terminal settling velocities, highlighting the importance of particle morphology for transport predictions.

In other applications, the motion of a particle relative to a nearby surface is equally important. For instance, the approach of rod-shaped biological particles toward a solid boundary raises questions about their orientation, lateral motion, and rotational behaviour before surface contact and possible attachment. Understanding these near-wall kinematics is therefore relevant to the physical processes preceding biofilm formation, although attachment itself also depends on biological and physicochemical interactions. Unlike a spherical particle, a cylinder exhibits direction-dependent hydrodynamic resistance. Its translational motion depends on its orientation relative to the direction of motion, while nearby boundaries can additionally couple translation and rotation. Consequently, describing the motion of a freely sedimenting cylinder near a wall requires consideration of the simultaneous evolution of its orientation, wall distance, translational velocity, and angular velocity.

The hydrodynamics of cylindrical particles in unbounded fluids have been investigated using analytical, numerical, and experimental approaches. Classical slender-body theory provides descriptions of the translational and rotational resistance of elongated particles in the creeping-flow limit [2]. For short cylinders, however, finite-length effects, including the influence of the cylinder ends, become increasingly important. Kharrouba et al. [3] investigated the hydrodynamic loads on finite-length cylinders over a range of aspect ratios, inclinations, and Reynolds numbers. Pierson et al. [4] subsequently examined the resistance torque on a cylinder rotating about an axis perpendicular to its symmetry axis and derived higher-order slender-body approximations. These studies provide descriptions of hydrodynamic resistance relevant to the short cylinders considered here.

Even in an unbounded fluid, weak fluid inertia can modify the sedimentation of an inclined cylinder by generating a hydrodynamic orienting torque. Khayat and Cox [5] derived inertial corrections to the hydrodynamic force and torque on slender bodies. Cabrera-Booman et al. [6] experimentally investigated freely settling cylinders and compared their measured motion with theoretical descriptions of fluid-inertia effects. These studies demonstrate the importance of orientation and weak inertia in cylinder sedimentation, even without a nearby boundary.

A rigid wall introduces further complexity because the hydrodynamic resistance and induced rotation depend on the instantaneous particle–wall configuration. De Mestre and Russel [7] developed a slender-body description of the hydrodynamic force and torque on a cylinder translating near a plane wall. Russel et al. [8] subsequently combined theoretical analysis and experiments to investigate slender rods sedimenting near a vertical wall. Their results showed how wall-induced rotation can change the lateral motion of an inclined rod. and produce distinct near-wall trajectories, including turns away from the wall without contact. More recently, Koens and Montenegro-Johnson [9] developed an asymptotic description of the local drag on a slender rod parallel to a plane wall. Chun and Jung [10] numerically investigated rod-like particles in confined shear flows and demonstrated that wall proximity and the imposed flow configuration can substantially modify their rotational motion, including tumbling-associated pole-vaulting behaviour. Although their imposed-shear configuration differs from freely sedimenting cylinders in an otherwise quiescent fluid, these investigations illustrate the sensitivity of rod kinematics to the surrounding boundaries and flow conditions.

The near-wall sedimentation of a short cylinder is particularly challenging because neither its orientation nor its distance from the wall generally remains constant. Both quantities evolve during the motion and influence the instantaneous translational and angular velocities. Furthermore, while recent experimental work has sought practically usable descriptions of terminal settling velocities [1], the evaluation of such descriptions at a steady state does not establish their applicability throughout a time-dependent near-wall trajectory. A comparison between measured instantaneous velocities and quasi-steady model velocities along the observed trajectory can therefore provide information about the applicability and limitations of the underlying hydrodynamic approximations.

The choice of particle Reynolds number is also motivated by previous experiments on wall-bounded sphere sedimentation [11]. In that system, rigid spheres sedimenting at ReP≈O(10−1)Re_P \approx O(10^{-1}) exhibited post-acceleration near-wall motion consistent with classical wall-lift descriptions, whereas experiments at ReP≈O(10−2)Re_P \approx O(10^{-2}) revealed persistently unsteady sedimentation despite the small particle Reynolds numbers. These observations do not establish a universal transition Reynolds number or imply that cylinders exhibit the same behaviour. They do, however, motivate the investigation of short, rigid cylinders at ReP≈O(10−1)Re_P \approx O(10^{-1}) as a reference for subsequent studies at smaller Reynolds numbers.

The present study experimentally investigates the sedimentation of short, rigid cylinders with mean aspect ratios of approximately χˉ=3.8\bar{\chi} = 3.8–5.85.8 in a rectangular duct filled with a highly viscous, initially quiescent liquid. The cylinders are released with negligible initial translational velocity, either approximately axially aligned near the centre of the duct or at different initial inclinations and distances from a lateral rigid wall. Time-resolved image measurements are used to determine their complete trajectories, including the evolution of particle orientation, wall distance, translational velocity, and angular velocity. The centre-duct experiments first provide a reference for assessing predicted sedimentation velocities. The near-wall experiments then resolve the coupled translational and angular kinematics, including glancing trajectories and contact-induced flipping behaviour.

To describe the measured kinematics, existing expressions for the hydrodynamic resistance of finite-length cylinders and for wall-induced motion are combined into models that provide translational and angular velocities as functions of the instantaneous particle configuration. The theoretical velocities are evaluated using the experimentally measured orientation and wall distance and compared with the corresponding measured velocities throughout the evaluated trajectories. The objective is therefore to assess the applicability of quasi-steady velocity descriptions to the experimentally observed, time-dependent near-wall motion, rather than to predict complete trajectories from prescribed initial conditions. Expressing the models directly in terms of kinematic quantities provides a comparatively accessible description for applications in which particle velocities, orientations, and surface-approach behaviour are of primary interest.

The remainder of this paper is organized as follows. Section 2 introduces the theoretical descriptions used to estimate the translational and angular velocities of freely sedimenting cylinders, including the effects of finite aspect ratio, weak fluid inertia, and confinement by a rigid wall. Section 3 describes the experimental setup, the cylindrical model particles, and the image-based evaluation of their motion. Section 4 presents and discusses the experimental results. First, the sedimentation of approximately axially aligned cylinders near the centre of the rectangular duct is examined. The subsequent analysis focuses on the near-wall translational and angular kinematics of initially inclined cylinders, comparing the instantaneous measured velocities with the corresponding model estimates along the experimentally observed trajectories. Finally, Section 5 summarizes the main findings, discusses the applicability and limitations of the kinematic descriptions, and outlines directions for further investigation.

Theory

Translational velocity

A rigid circular cylinder of radius rr, total length 2l2l, volume VP=2πr2lV_{\mathrm{P}} = 2\pi r^{2}l, and density ρP\rho_{\mathrm{P}} sediments with velocity magnitude U=∣U∣U = \lvert\mathbf{U}\rvert in an otherwise quiescent fluid of density ρ\rho and dynamic viscosity η\eta. The density ratio approaches unity from above, γ=ρP/ρ→1+\gamma= \rho_{\mathrm{P}}/\rho\to1^{+}, such that the density difference between the particle and the surrounding fluid is small. The cylinder aspect ratio is defined as χ=l/r\chi= l/r.

The cylinder is released from rest at t=0t = 0, while the surrounding fluid is initially quiescent, ∣uf∣=0\lvert\mathbf{u}_{\mathrm{f}}\rvert= 0. The translational particle Reynolds number used throughout most of the present formulation is defined as ReP=ρUl/ηRe_{\mathrm{P}} = \rho U l/\eta.

The spatial and angular quantities used throughout the following are defined in Figure 1. In particular, the orientation angle θ\theta denotes the angle between the cylinder axis and the direction of gravity, whereas the trajectory angle ϕ\phi denotes the angle between the instantaneous translational velocity vector and the direction of gravity. In the wall-bounded configuration, dd denotes the distance between the cylinder center and the plane wall.

Definitions of the spatial and angular quantities and schematic force decompositions

Figure 1. Definitions of the spatial and angular quantities and schematic force decompositions. Left: oblique sedimentation of a cylinder in an unbounded fluid in the creeping-flow regime. Right: translation and rotation of a cylinder near a plane wall at finite particle Reynolds number.

For axial sedimentation, θ=0∘\theta= 0^{\circ}, the steady hydrodynamic drag is balanced by the net gravitational force acting on the cylinder, yielding

(ρP−ρ)VPg=FD,ubθ=0∘, ReP≈0.(1)(\rho_{\mathrm{P}}-\rho)V_{\mathrm{P}}g = F_{\mathrm{D,ub}}^{\theta=0^{\circ},\,Re_{\mathrm{P}\approx0}.} \tag*{(1)}

For transverse sedimentation, θ=90∘\theta= 90^{\circ}, the corresponding force balance becomes

(ρP−ρ)VPg=FD,ubθ=90∘, ReP≈0.(2)(\rho_{\mathrm{P}}-\rho)V_{\mathrm{P}}g = F_{\mathrm{D,ub}}^{\theta=90^{\circ},\,Re_{\mathrm{P}\approx0}.} \tag*{(2)}

Kharrouba et al. [3] developed semi-empirical expressions for the drag force on cylinders with aspect ratios in the range 2≤χ≤102 \le\chi\le10. Their formulation is based on the slender-body approximations of Batchelor [2], Cox [12], and Keller and Rubinow [13]. Higher-order corrections and an empirical finite-length contribution obtained from numerical simulations were incorporated to account for finite-aspect-ratio effects, which become increasingly important for χ≲10\chi\lesssim10.

The axial drag in an unbounded fluid in the creeping-flow limit is approximated as

FD,ubθ=0∘, Re⁡P≈0=4πηlU[ax(1)ln⁡(2χ)+ax(2)[ln⁡(2χ)]2+ax(3)[ln⁡(2χ)]3+ax(4)[ln⁡(2χ)]4−2.4χ2/3(χ−12)1.75](3)F_{\mathrm{D,ub}}^{\theta=0^\circ,\,\operatorname{Re}_{\mathrm{P}}\approx0}=4\pi\eta l U\left[\frac{a_x^{(1)}}{\ln(2\chi)}+\frac{a_x^{(2)}}{[\ln(2\chi)]^2}+\frac{a_x^{(3)}}{[\ln(2\chi)]^3}+\frac{a_x^{(4)}}{[\ln(2\chi)]^4}-\frac{2.4}{\chi^{2/3}\left(\chi-\frac{1}{2}\right)^{1.75}}\right] \tag*{(3)}

where ax(1)=1a_x^{(1)}=1, ax(2)≈0.80685a_x^{(2)}\approx0.80685, ax(3)≈0.82854a_x^{(3)}\approx0.82854, and ax(4)≈1.45243a_x^{(4)}\approx1.45243.

The corresponding transverse drag is

FD,ubθ=90∘, Re⁡P≈0=8πηlU[ay(1)ln⁡(2χ)+ay(2)[ln⁡(2χ)]2+(1−e−c1χ2)(ay(3)[ln⁡(2χ)]3+ay(4)[ln⁡(2χ)]4)].(4)F_{\mathrm{D,ub}}^{\theta=90^\circ,\ \operatorname{Re}_{P}\approx0}=8\pi\eta l U\left[\frac{a_y^{(1)}}{\ln(2\chi)}+\frac{a_y^{(2)}}{[\ln(2\chi)]^2}+\left(1-e^{-c_1\chi^2}\right)\left(\frac{a_y^{(3)}}{[\ln(2\chi)]^3}+\frac{a_y^{(4)}}{[\ln(2\chi)]^4}\right)\right]. \tag*{(4)}

with ay(1)=1a_y^{(1)}=1, ay(2)≈−0.19315a_y^{(2)}\approx-0.19315, ay(3)≈0.21484a_y^{(3)}\approx0.21484, ay(4)≈0.38735a_y^{(4)}\approx0.38735, c1=0.01c_1=0.01, and c2=2.5c_2=2.5. Based on the theory of Khayat and Cox [5], Kharrouba et al. [3] further proposed semi-empirical expressions that account for weak fluid-inertia effects. In the regime 2Re⁡P=O(1)2\operatorname{Re}_{P}=O(1), the inertia-corrected axial drag is

FD,ubθ=0∘, 2Re⁡P=O(1)=4πηlU[ax(1)ln⁡(2χ)+ax(2)+f∥[ln⁡(2χ)]2+ax(3)+f3f∥[ln⁡(2χ)]3+ax(4)+f4f∥[ln⁡(2χ)]4−2.4χ2/3(χ−12)1.75].(5)F_{\mathrm{D,ub}}^{\theta=0^\circ,\ 2\operatorname{Re}_{P}=O(1)}=4\pi\eta l U\left[\frac{a_x^{(1)}}{\ln(2\chi)}+\frac{a_x^{(2)}+f_{\parallel}}{[\ln(2\chi)]^2}+\frac{a_x^{(3)}+f_3f_{\parallel}}{[\ln(2\chi)]^3}+\frac{a_x^{(4)}+f_4f_{\parallel}}{[\ln(2\chi)]^4}-\frac{2.4}{\chi^{2/3}\left(\chi-\frac{1}{2}\right)^{1.75}}\right]. \tag*{(5)}

where f∥≈Re⁡P/4f_{\parallel}\approx\operatorname{Re}_{P}/4, f3=(2Re⁡P)0.07χ0.5f_3=(2\operatorname{Re}_{P})^{0.07}\chi^{0.5}, and f4=(2Re⁡P)0.03χ0.9f_4=(2\operatorname{Re}_{P})^{0.03}\chi^{0.9}.

For transverse motion, only the second-order term is corrected, yielding

FD,ubθ=90∘, 2Re⁡P=O(1)=8πηlU[ay(1)ln⁡(2χ)+ay(2)+f⊥[ln⁡(2χ)]2+(1−e−c1χ2)(ay(3)[ln⁡(2χ)]3+ay(4)[ln⁡(2χ)]4)].(6)F_{\mathrm{D,ub}}^{\theta=90^\circ,\ 2\operatorname{Re}_{P}=O(1)}=8\pi\eta l U\left[\frac{a_y^{(1)}}{\ln(2\chi)}+\frac{a_y^{(2)}+f_{\perp}}{[\ln(2\chi)]^2}+\left(1-e^{-c_1\chi^2}\right)\left(\frac{a_y^{(3)}}{[\ln(2\chi)]^3}+\frac{a_y^{(4)}}{[\ln(2\chi)]^4}\right)\right]. \tag*{(6)}

In the limit 2Re⁡P→02\operatorname{Re}_{P}\to0, f⊥→Re⁡P/2f_{\perp}\to\operatorname{Re}_{P}/2. For larger Reynolds numbers, the complete correlation for f⊥f_{\perp} is given in Eq. (C3) of Kharrouba et al. [3].

At this point, the Reynolds-number notation requires some clarification. The present work primarily uses the half-length-based particle Reynolds number Re⁡P=ρUl/η\operatorname{Re}_{P}=\rho U l/\eta, whereas Kharrouba et al. also express their finite-inertia correlations in terms of a Reynolds number based on the cylinder diameter 2r2r. The two definitions are related by Re⁡=2ρUr/η=2Re⁡Pr/l=2Re⁡P/χ\operatorname{Re}=2\rho U r/\eta=2\operatorname{Re}_{P}r/l=2\operatorname{Re}_{P}/\chi. The factor 2Re⁡P2\operatorname{Re}_{P} appearing explicitly in Eqs. (5) and (6) is retained in the form used by Kharrouba et al., whereas the diameter-based Reynolds number Re⁡\operatorname{Re} is employed below where their correlations are stated in this form. This distinction is maintained throughout to avoid mixing the different characteristic length scales.

For an inclined cylinder in the Stokes regime, the hydrodynamic force can be decomposed into components parallel and perpendicular to the cylinder axis. For a torque-free cylinder sedimenting obliquely in an unbounded fluid, these components are

F∥,ubRe⁡P=0(χ,θ)=FD,ubθ=0∘, Re⁡P≈0cos⁡θ(7)F_{\parallel,\mathrm{ub}}^{\operatorname{Re}_{P}=0}(\chi,\theta)=F_{\mathrm{D,ub}}^{\theta=0^\circ,\ \operatorname{Re}_{P}\approx0}\cos\theta \tag*{(7)}

and

F∥,ubReP=0(χ,θ)=FD,ubθ=90∘, ReP≈0sin⁡θ.(8)F_{\parallel,\mathrm{ub}}^{Re_P=0}(\chi,\theta)=F_{\mathrm{D,ub}}^{\theta=90^\circ,\,Re_P\approx0}\sin\theta. \tag*{(8)}

Even in the creeping-flow limit, an inclined cylinder generally does not sediment exactly parallel to gravity because its hydrodynamic resistance is anisotropic. Consequently, the translational velocity is not, in general, collinear with the gravitational force. What vanishes in this limit is the inertia-induced torque: in an unbounded Stokes flow, the orientation remains constant in time, while the anisotropic translational resistance allows an oblique settling trajectory. The left-hand sketch in Figure 1 therefore represents a constant-orientation creeping-flow configuration rather than a general absence of transverse particle motion.

At finite Reynolds number and particularly in the presence of a wall, fluid inertia gives rise to hydrodynamic torque and modifies the translational motion. A freely sedimenting cylinder therefore does not generally retain a constant orientation angle. Nevertheless, for small Reynolds numbers and moderate orientation angles, the corresponding projection approach remains useful when the inertia-corrected drag is employed.

When lateral motion occurs, the translational velocity is characterized by the trajectory angle ϕ=arctan⁡(UP,y/UP,x)\phi=\arctan(U_{\mathrm{P},y}/U_{\mathrm{P},x}). Accordingly, U=UP,x2+UP,y2=UP,x/cos⁡ϕ=UP,xsec⁡ϕU=\sqrt{U_{\mathrm{P},x}^{2}+U_{\mathrm{P},y}^{2}}=U_{\mathrm{P},x}/\cos\phi=U_{\mathrm{P},x}\sec\phi.

Similarly, the magnitude of the total hydrodynamic drag follows from its components parallel and perpendicular to the cylinder axis, FD=F∥2(θ)+F⊥2(θ)F_{\mathrm{D}}=\sqrt{F_{\parallel}^{2}(\theta)+F_{\perp}^{2}(\theta)}. For small orientation and trajectory angles, θ,ϕ→0∘\theta,\phi\to0^\circ, the streamwise component of the total drag can be approximated by the axial drag of an aligned cylinder, FDcos⁡ϕ≈F∥θ=0∘F_{\mathrm{D}}\cos\phi\approx F_{\parallel}^{\theta=0^\circ}. The total drag is therefore approximated as

FD(χ,θ,ϕ)≈F∥θ=0∘, 2ReP=O(1)cos⁡ϕ=F∥θ=0∘, 2ReP=O(1)sec⁡ϕ.(9)\begin{aligned} F_{\mathrm{D}}(\chi,\theta,\phi)&\approx\frac{F_{\parallel}^{\theta=0^\circ,\,2Re_P=O(1)}}{\cos\phi}\\ &=F_{\parallel}^{\theta=0^\circ,\,2Re_P=O(1)}\sec\phi. \tag*{(9)} \end{aligned}

Brenner [14] developed a general theory according to which the leading-order wall correction for a particle of arbitrary cross-section moving along a principal axis of hydrodynamic resistance can be related to its drag in an unbounded fluid. The theory requires the corresponding wall correction for a sphere to be known and assumes that the characteristic particle dimensions are small compared with the distance from the boundary.

For a particle with characteristic length l∗l^* located at a characteristic distance d∗d^* from a boundary, Brenner’s wall correction may be expressed as

FD,wb∗FD,ub∗=11−C1FD,ub∗(l∗d∗)+O[(l∗d∗)3].(10)\frac{F_{\mathrm{D,wb}}^*}{F_{\mathrm{D,ub}}^*}=\frac{1}{1-C_1F_{\mathrm{D,ub}}^*\left(\frac{l^*}{d^*}\right)+O\left[\left(\frac{l^*}{d^*}\right)^3\right]} . \tag*{(10)}

Here, FD,wb∗=FD,wb/(2πηl∗Uth,wb)F_{\mathrm{D,wb}}^*=F_{\mathrm{D,wb}}/(2\pi\eta l^*U_{\mathrm{th,wb}}) and FD,ub∗=FD,ub/(2πηl∗Uth,ub)F_{\mathrm{D,ub}}^*=F_{\mathrm{D,ub}}/(2\pi\eta l^*U_{\mathrm{th,ub}}) denote the dimensionless wall-bounded and unbounded drag coefficients, respectively.

Ui et al. [15] investigated rods of different aspect ratios sedimenting along the centerline of circular tubes of diameter DD. By varying the tube diameter, measuring the steady sedimentation velocity, and determining the dimensionless drag from the force balance, they extrapolated the unbounded-fluid drag to l/(D/2)=0l/(D/2)=0. They found that the linear relation

FD,wb∗FD,ub∗=1+C1FD,ub∗(l∗d∗)(11)\frac{F_{\mathrm{D,wb}}^{*}}{F_{\mathrm{D,ub}}^{*}}=1+C_{1}F_{\mathrm{D,ub}}^{*}\left(\frac{l^{*}}{d^{*}}\right) \tag*{(11)}

provided a more suitable description of their measurements than the corresponding expansion of Brenner’s relation. The third-order correction is omitted in Eq. (11), since the extrapolation by Ui et al. was restricted to the linear regime.

Happel and Bart [16] derived the wall-bounded drag on a sphere sedimenting along the centerline of a channel with a square cross-section. For a square duct, the corresponding wall coefficient is C1=1.903266C_{1}=1.903266, and the characteristic wall distance is the duct half-width W/2W/2.

Using the first-order Brenner correction, hereafter denoted model (1), the steady velocity of an axially sedimenting cylinder along the centerline of a square duct is approximated by

Uth,wbθ=0∘, Re⁡P≈0∣□(1)≈(ρP−ρ)r2g2ηFD,ub∗, θ=0∘, Re⁡P≈0×[1−1.903266FD,ub∗, θ=0∘, Re⁡P≈0(2lW)].(12)\left.U_{\mathrm{th,wb}}^{\theta=0^\circ,\,\operatorname{Re}_{P}\approx0}\right|_{\square}^{(1)} \approx\frac{(\rho_{P}-\rho)r^{2}g}{2\eta F_{\mathrm{D,ub}}^{*,\,\theta=0^\circ,\,\operatorname{Re}_{P}\approx0}} \times\left[1-1.903266F_{\mathrm{D,ub}}^{*,\,\theta=0^\circ,\,\operatorname{Re}_{P}\approx0}\left(\frac{2l}{W}\right)\right]. \tag*{(12)}

Using the first-order relation proposed by Ui et al., hereafter denoted model (2), the corresponding velocity becomes

Uth,wbθ=0∘, Re⁡P≈0∣□(2)≈(ρP−ρ)r2g2ηFD,ub∗, θ=0∘, Re⁡P≈0×[1+1.903266FD,ub∗, θ=0∘, Re⁡P≈0(2lW)]−1.(13)\left.U_{\mathrm{th,wb}}^{\theta=0^\circ,\,\operatorname{Re}_{P}\approx0}\right|_{\square}^{(2)} \approx\frac{(\rho_{P}-\rho)r^{2}g}{2\eta F_{\mathrm{D,ub}}^{*,\,\theta=0^\circ,\,\operatorname{Re}_{P}\approx0}} \times\left[1+1.903266F_{\mathrm{D,ub}}^{*,\,\theta=0^\circ,\,\operatorname{Re}_{P}\approx0}\left(\frac{2l}{W}\right)\right]^{-1}. \tag*{(13)}

In our previous study [11], we showed that the drag on a rigid sphere sedimenting asymmetrically in a rectangular duct at Re⁡P=O(10−1)\operatorname{Re}_{P}=O(10^{-1}) is governed predominantly by the nearest plane wall, whereas the contributions of the remaining walls are comparatively small. In the near-wall region, at d/R≈2d/R\approx2, the sphere velocity was accurately approximated using Faxén’s wall correction [17].

Following Brenner’s approach, the wall-induced drag correction for an axially sedimenting cylinder near a plane wall can analogously be constructed using the Faxén expansion. The resulting wall-bounded velocity at a cylinder–wall distance dd is

Uth,wbθ=0∘, ReP≈0∣∥(1)≈(ρP−ρ)r2g2ηFD,ub∗,θ=0∘, ReP≈0×[1−916FD,ub∗,θ=0∘, ReP≈0(ld)+18FD,ub∗,θ=0∘, ReP≈0(ld)3].(14)\left.U_{\mathrm{th,wb}}^{\theta=0^\circ,\ \mathrm{Re}_P\approx0}\right|_{\parallel}^{(1)} \approx\frac{(\rho_P-\rho)r^2g}{2\eta F_{\mathrm{D,ub}}^{*,\theta=0^\circ,\ \mathrm{Re}_P\approx0}} \times\left[1-\frac{9}{16}F_{\mathrm{D,ub}}^{*,\theta=0^\circ,\ \mathrm{Re}_P\approx0}\left(\frac{l}{d}\right) +\frac{1}{8}F_{\mathrm{D,ub}}^{*,\theta=0^\circ,\ \mathrm{Re}_P\approx0}\left(\frac{l}{d}\right)^3\right]. \tag*{(14)}

Equation (14) is based on Brenner’s formulation and includes the first- and third-order wall corrections. Based on the relation proposed by Ui et al., and retaining only the first-order Faxén correction, model (2) becomes

Uth,wbθ=0∘, ReP≈0∣∥(2)≈(ρP−ρ)r2g2ηFD,ub∗,θ=0∘, ReP≈0×[1+916FD,ub∗,θ=0∘, ReP≈0(ld)]−1.(15)\left.U_{\mathrm{th,wb}}^{\theta=0^\circ,\ \mathrm{Re}_P\approx0}\right|_{\parallel}^{(2)} \approx\frac{(\rho_P-\rho)r^2g}{2\eta F_{\mathrm{D,ub}}^{*,\theta=0^\circ,\ \mathrm{Re}_P\approx0}} \times\left[1+\frac{9}{16}F_{\mathrm{D,ub}}^{*,\theta=0^\circ,\ \mathrm{Re}_P\approx0}\left(\frac{l}{d}\right)\right]^{-1}. \tag*{(15)}

For weak but finite inertia, the wall correction must additionally account for the inertia-dependent modification of the disturbance flow [11, 18]. Based on model (2), the wall-bounded velocity in the small-angle regime is approximated as

Uth,wbReP≤1(ϕ)∣∥(2)≈(ρP−ρ)r2g2ηFD,ub∗,θ=0∘, 2ReP=O(1)sec⁡ϕ×[1+38ReP,th(ϕ)+916FD,ub∗,θ=0∘, 2ReP=O(1)sec⁡ϕ(ld)K(ϕ)]−1.(16)\left.U_{\mathrm{th,wb}}^{\mathrm{Re}_P\leq1}(\phi)\right|_{\parallel}^{(2)} \approx\frac{(\rho_P-\rho)r^2g}{2\eta F_{\mathrm{D,ub}}^{*,\theta=0^\circ,\ 2\mathrm{Re}_P=\mathcal{O}(1)}\sec\phi} \times\left[1+\frac{3}{8}\mathrm{Re}_{P,\mathrm{th}}(\phi) +\frac{9}{16}F_{\mathrm{D,ub}}^{*,\theta=0^\circ,\ 2\mathrm{Re}_P=\mathcal{O}(1)}\sec\phi\left(\frac{l}{d}\right)\mathcal{K}(\phi)\right]^{-1}. \tag*{(16)}

The corresponding theoretical particle Reynolds number is

ReP,th(ϕ)=(ρP−ρ)ρlr2gcos⁡ϕ2ηFD,ub∗,θ=0∘, 2ReP=O(1).(17)\mathrm{Re}_{P,\mathrm{th}}(\phi)=\frac{(\rho_P-\rho)\rho l r^2g\cos\phi}{2\eta F_{\mathrm{D,ub}}^{*,\theta=0^\circ,\ 2\mathrm{Re}_P=\mathcal{O}(1)}}. \tag*{(17)}

The inertia-dependent wall function is defined as

K(ϕ)=1−43[ReP,th(ϕ)2(l/d)]+2316[ReP,th(ϕ)2(l/d)]3.(18)\mathcal{K}(\phi)=1-\frac{4}{3}\left[\frac{\mathrm{Re}_{P,\mathrm{th}}(\phi)}{2(l/d)}\right]+\frac{23}{16}\left[\frac{\mathrm{Re}_{P,\mathrm{th}}(\phi)}{2(l/d)}\right]^3. \tag*{(18)}

As shown below, Brenner’s model (1) provides a more accurate description of the cylinder velocity at larger wall distances, d/l>2d/l > 2. Including the third-order wall correction yields

Uth,wbRe≤1(ϕ)∣∥(1)≈(ρP−ρ) r2g2ηFD,ub∗,θ=0,2ReP=O(1)sec⁡ϕ[1−38ReP,th(ϕ)−916FD,ub∗,θ=0,2ReP=O(1)sec⁡ϕ(ld)K(ϕ)+18FD,ub∗,θ=0,2ReP=O(1)sec⁡ϕ(ld)3].(19)\left.U_{\mathrm{th,wb}}^{\mathit{Re}\leq1}(\phi)\right|_{\parallel}^{(1)} \approx\frac{(\rho_{\mathrm{P}}-\rho)\,r^{2}g}{2\eta F_{\mathrm{D,ub}}^{*,\theta=0,2\mathit{Re}_{\mathrm{P}}=O(1)}\sec\phi} \left[1-\frac{3}{8}\mathit{Re}_{\mathrm{P,th}}(\phi)-\frac{9}{16}F_{\mathrm{D,ub}}^{*,\theta=0,2\mathit{Re}_{\mathrm{P}}=O(1)}\sec\phi\left(\frac{l}{d}\right)\mathcal{K}(\phi)+\frac{1}{8}F_{\mathrm{D,ub}}^{*,\theta=0,2\mathit{Re}_{\mathrm{P}}=O(1)}\sec\phi\left(\frac{l}{d}\right)^{3}\right]. \tag*{(19)}

Equations (16) and (19) provide approximations for the wall-bounded translational velocity at weak but finite inertia in the regime of small to moderate orientation and trajectory angles. They are based on the small-angle approximation introduced in Eq. (9), while additionally accounting for the inertia-dependent modification of the wall correction.

To extend the velocity model to the broader range of orientations encountered experimentally, the two limiting cylinder configurations are considered separately. For the wall-parallel configuration, θ=0∘\theta=0^\circ, the wall-bounded velocity Uth,wbθ=0∘,Re≤1(χ,d/l)U_{\mathrm{th,wb}}^{\theta=0^\circ,\mathit{Re}\leq1}(\chi,d/l) is calculated using the inertia-corrected axial drag coefficient given by Eq. (5). For the transverse configuration, θ=90∘\theta=90^\circ, the corresponding velocity Uth,wbθ=90∘,Re≤1(χ,d/l)U_{\mathrm{th,wb}}^{\theta=90^\circ,\mathit{Re}\leq1}(\chi,d/l) is calculated analogously using the inertia-corrected transverse drag coefficient given by Eq. (6). In both limiting configurations, the corresponding inertia-corrected wall factor is applied.

The velocity at an arbitrary orientation is subsequently approximated by projecting these two limiting wall-bounded velocities onto the instantaneous cylinder orientation. This construction constitutes a kinematic interpolation between the two limiting resistance configurations rather than an exact wall-bounded resistance relation for an arbitrarily oriented cylinder. It preserves the respective limiting velocities at θ=0∘\theta=0^\circ and θ=90∘\theta=90^\circ, while providing a continuous approximation for intermediate orientations.

The contribution associated with the wall-parallel limiting configuration is approximated as

U∥,th,wb(θ)=Uth,wbθ=0∘,Re≤1(χ,dl)cos⁡θ,(20)U_{\parallel,\mathrm{th,wb}}(\theta)=U_{\mathrm{th,wb}}^{\theta=0^\circ,\mathit{Re}\leq1}\left(\chi,\frac{d}{l}\right)\cos\theta, \tag*{(20)}

whereas the contribution associated with the transverse limiting configuration is

U⊥,th,wb(θ)=Uth,wbθ=90∘,Re≤1(χ,dl)sin⁡θ.(21)U_{\perp,\mathrm{th,wb}}(\theta)=U_{\mathrm{th,wb}}^{\theta=90^\circ,\mathit{Re}\leq1}\left(\chi,\frac{d}{l}\right)\sin\theta. \tag*{(21)}

The magnitude of the theoretical wall-bounded translational velocity is then obtained from the vector sum of both contributions,

Uth,wb(θ)={[Uth,wbθ=0∘, Re⁡≲1(χ,dl)cos⁡θ]2+[Uth,wbθ=90∘, Re⁡≲1(χ,dl)sin⁡θ]2}1/2.(22)U_{\mathrm{th,wb}}(\theta)=\left\{\left[U_{\mathrm{th,wb}}^{\theta=0^\circ,\ \operatorname{Re}\lesssim1}\left(\chi,\frac{d}{l}\right)\cos\theta\right]^2+\left[U_{\mathrm{th,wb}}^{\theta=90^\circ,\ \operatorname{Re}\lesssim1}\left(\chi,\frac{d}{l}\right)\sin\theta\right]^2\right\}^{1/2}. \tag*{(22)}

A direct insertion of the projected transverse force, F⊥(χ,θ)=F⊥θ=90∘sin⁡θF_{\perp}(\chi,\theta)=F_{\perp}^{\theta=90^\circ}\sin\theta, into the wall-bounded force balance is not used. Solving such a force balance for the velocity introduces a factor proportional to 1/sin⁡θ1/\sin\theta, which becomes singular as θ→0∘\theta\to0^\circ, although the physical translational velocity remains finite. The interpolation defined by Eqs. (20)–(22) avoids this artificial singularity and continuously recovers the corresponding limiting velocities for θ→0∘\theta\to0^\circ and θ→90∘\theta\to90^\circ.

Angular velocity

The theoretical angular velocity of a cylinder rotating within the xx-yy plane follows from the torque balance about an axis through the center of mass perpendicular to the plane of motion, i.e. the zz-axis in Figure 1. The angular velocity is defined as ω=dθ/dt\omega=\mathrm{d}\theta/\mathrm{d}t. The rotational equation of motion reads

IPω˙=TH+TI+TW,(23)I_{\mathrm{P}}\dot{\omega}=T_{\mathrm{H}}+T_{\mathrm{I}}+T_{\mathrm{W}}, \tag*{(23)}

where IPI_{\mathrm{P}} is the moment of inertia of the cylinder, THT_{\mathrm{H}} is the hydrodynamic resistance torque associated with the rotation of the cylinder, TIT_{\mathrm{I}} denotes the inertia-induced torque resulting from its translational motion in an unbounded fluid, and TWT_{\mathrm{W}} represents the additional torque arising from the hydrodynamic interaction with the wall.

Analogous to the hydrodynamic drag in the translational force balance, the hydrodynamic resistance torque opposes the rotational motion and can be written as

TH=−CHω,(24)T_{\mathrm{H}}=-C_{\mathrm{H}}\omega, \tag*{(24)}

where CHC_{\mathrm{H}} denotes the rotational resistance coefficient. Under the quasi-steady assumption, rotational particle inertia is neglected, IPω˙≈0I_{\mathrm{P}}\dot{\omega}\approx0, such that the angular velocity follows from

ω=TI+TWCH.(25)\omega=\frac{T_{\mathrm{I}}+T_{\mathrm{W}}}{C_{\mathrm{H}}}. \tag*{(25)}

Rotational resistance

For a slender cylinder rotating about an axis perpendicular to its symmetry axis in an unbounded fluid, the leading-order rotational resistance follows from slender-body theory. Using the present definition of the cylinder aspect ratio, χ=l/r\chi=l/r, the corresponding slenderness parameter is

ϵ=1ln⁡(2χ).(26)\epsilon=\frac{1}{\ln(2\chi)}. \tag*{(26)}

The leading-order rotational resistance coefficient can then be written as [2]

CH(1)=8π3ηl3ϵ.(27)C_{\mathrm{H}}^{(1)} = \frac{8\pi}{3}\eta l^{3}\epsilon. \tag*{(27)}

Equation (27) represents the leading-order slender-body resistance and is recovered in the limit of large aspect ratio.

For cylinders of finite aspect ratio, higher-order contributions to the rotational resistance become relevant. Pierson et al. [4] extended the slender-body expansion of the rotation-induced hydrodynamic torque to fourth order in the small parameter ϵ\epsilon. Their fourth-order expression includes an empirical aspect-ratio-dependent prefactor to improve the approximation for cylinders with moderate aspect ratios.

Using the present definition of the total cylinder length 2l2l, the rotational resistance is written as

CH=8π3ηl3C(χ),(28)C_{\mathrm{H}} = \frac{8\pi}{3}\eta l^{3}\mathcal{C}(\chi), \tag*{(28)}

where the fourth-order approximation reads

C(χ)=ϵ+ϵ2(116−ln⁡2)+ϵ3[16136−π212−113ln⁡2+(ln⁡2)2]+ϵ4(1−1(2χ)1.2)5[−54ζ(3)+103372−ln⁡32+112ln⁡22−16112ln⁡2−π2(1124−18ln⁡4)].(29)\begin{aligned} \mathcal{C}(\chi) ={}& \epsilon+ \epsilon^{2}\left(\frac{11}{6} - \ln2\right) \\ &+ \epsilon^{3}\left[\frac{161}{36} - \frac{\pi^{2}}{12} - \frac{11}{3}\ln2 + (\ln2)^{2}\right] \\ &+ \epsilon^{4}\left(1-\frac{1}{(2\chi)^{1.2}}\right)^{5} \left[-\frac{5}{4}\zeta(3)+\frac{1033}{72}-\ln^{3}2+\frac{11}{2}\ln^{2}2 \right.\\ &\left.\qquad\qquad\qquad\qquad-\frac{161}{12}\ln2-\pi^{2}\left(\frac{11}{24}-\frac{1}{8}\ln4\right)\right]. \tag*{(29)} \end{aligned}

The first term in Eq. (29) recovers the leading-order result in Eq. (27), whereas the additional terms account for higher-order contributions to the rotational resistance associated with a finite aspect ratio.

Inertia-induced rotation in an unbounded fluid

In addition to the rotational resistance torque, translational fluid inertia generates an orienting torque on an inclined cylinder even in an unbounded fluid. Khayat and Cox [5] derived this contribution for slender rods at low Reynolds numbers. Higher-order expressions applicable to cylinders of moderate aspect ratio and finite Reynolds number were subsequently developed by Kharrouba et al. [3].

For the present definitions of the orientation angle θ\theta and trajectory angle ϕ\phi, the instantaneous incidence angle between the cylinder axis and its translational velocity is defined as

α=θ−ϕ.(30)\alpha= \theta- \phi. \tag*{(30)}

Using the correlation of Kharrouba et al. [3], and expressing their total cylinder length as 2l2l, the inertia-induced torque becomes

TI,K=−5π12ηUl2Re1+χRe1.1×[χ(ln⁡χ)2+13.5−30Re1/2χ(ln⁡χ)3e−0.7Re]sin⁡(2α),(31)T_{\mathrm{I,K}}=-\frac{5\pi}{12}\eta U l^{2}\frac{Re}{\sqrt{1+\chi Re^{1.1}}}\times\left[\frac{\chi}{(\ln\chi)^{2}}+\frac{13.5-30Re^{1/2}}{\chi(\ln\chi)^{3}}e^{-0.7Re}\right]\sin(2\alpha), \tag*{(31)}

where U=UP,x2+UP,y2U=\sqrt{U_{\mathrm{P,x}}^{2}+U_{\mathrm{P,y}}^{2}} is the instantaneous translational velocity magnitude and ReRe is the diameter-based Reynolds number defined in § 2.1.

Combining the inertia-induced torque in Eq. (31) with the third-order rotational resistance in Eq. (28) yields the theoretical angular velocity of the translating cylinder in an unbounded fluid,

ωth,ubKP=TI,KCH.(32)\omega_{\mathrm{th,ub}}^{\mathrm{KP}}=\frac{T_{\mathrm{I,K}}}{C_{\mathrm{H}}}. \tag*{(32)}

The superscript KP denotes the combination of the inertia-induced torque correlation of Kharrouba et al. [3] with the rotational resistance of Pierson et al. [4].

Wall-induced rotation

The presence of a plane wall introduces an additional translation–rotation coupling. Russel et al. [8] investigated this interaction for a slender cylinder sedimenting close to a vertical plane wall using slender-body theory together with the corresponding hydrodynamic image system. Their resulting analytical expression for the angular velocity depends on both the cylinder–wall distance and the cylinder orientation.

In the present model, only the limiting solution for a cylinder oriented parallel to the wall, θ=0∘\theta=0^\circ, is retained to describe the wall-induced contribution. The general expression presented by Russel et al. [8] in their Eq. (8) contains a term denoted by FF, whose relation to the quantities FiF_i appearing elsewhere in their derivation is not unambiguous. Although this may indicate a typographical inconsistency in the published expression, this interpretation cannot be verified conclusively. The published general expression is therefore not modified here.

Furthermore, the cylinders considered in the present study have moderate aspect ratios and are consequently far from the asymptotically slender limit underlying the analytical treatment of Russel et al. [8]. For this reason, the θ=0∘\theta=0^\circ solution is used here as the leading-order representation of the wall-distance dependence of the translation–rotation coupling. This choice should not be interpreted as implying that the wall-induced torque is generally independent of the cylinder orientation. Rather, the additional orientation-dependent terms of the complete slender-body expression are not retained in the present model.

In the notation of Russel et al. [8], the dimensionless center–wall distance is X=d/lX=d/l. For θ=0∘\theta=0^\circ, the corresponding wall function is

B(X,0)=4X(X2+1−X)−2XX2+1,X=dl.(33)B(X,0)=4X\left(\sqrt{X^{2}+1}-X\right)-\frac{2X}{\sqrt{X^{2}+1}},\qquad X=\frac{d}{l}. \tag*{(33)}

Hereafter, the wall distance is expressed directly as d/ld/l, consistently with § 2.1.

Using the sign convention for θ\theta adopted in the present study, the leading-order wall-induced angular velocity obtained from Russel et al. [8] becomes

ωwall,R=−3F032πηl2B(dl,0),(34)\omega_{\mathrm{wall,R}}=-\frac{3F_0}{32\pi\eta l^2}B\left(\frac{d}{l},0\right), \tag*{(34)}

where

F0=(ρP−ρ)VPg(35)F_0=(\rho_{\mathrm{P}}-\rho)V_{\mathrm{P}}g \tag*{(35)}

is the net gravitational force acting on the cylinder. The subscript wall, R denotes the wall-induced contribution obtained from the theory of Russel et al. [8].

Equation (34) is based on a leading-order slender-body approximation. Russel et al. [8] themselves noted that higher-order corrections to the resistance coefficients can constitute an important source of error in calculated trajectories. This becomes particularly relevant for the present cylinders, whose aspect ratios are considerably smaller than those associated with the asymptotically slender limit.

The relation between the leading-order Russel wall contribution and the finite-aspect-ratio rotational resistance can be made explicit by rewriting Eq. (34) in terms of a torque. Multiplication by the leading-order rotational resistance coefficient in Eq. (27) gives

TW,R(1)=CH(1)ωwall,R=−F0l4ϵB(dl,0).(36)T_{\mathrm{W,R}}^{(1)}=C_{\mathrm{H}}^{(1)}\omega_{\mathrm{wall,R}} =-\frac{F_0l}{4}\epsilon B\left(\frac{d}{l},0\right). \tag*{(36)}

Equation (36) thus represents the leading-order wall-induced torque underlying the angular-velocity expression of Russel et al. [8].

To account for the finite aspect ratio of the cylinders investigated here, the leading-order wall-coupling term in Eq. (36) is retained, while the leading-order rotational resistance CH(1)C_{\mathrm{H}}^{(1)} is replaced by the third-order rotational resistance CHC_{\mathrm{H}} of Pierson et al. [4]. The resulting wall-induced angular velocity is

ωth,wbRe≈0=TW,R(1)CH=ωwall,RϵC(χ).(37)\omega_{\mathrm{th,wb}}^{\mathit{Re}\approx0}=\frac{T_{\mathrm{W,R}}^{(1)}}{C_{\mathrm{H}}}=\omega_{\mathrm{wall,R}}\frac{\epsilon}{C(\chi)}. \tag*{(37)}

Equation (37) is therefore interpreted as a finite-aspect-ratio-corrected leading-order wall contribution. It does not constitute a complete finite-aspect-ratio solution of the wall-bounded hydrodynamic problem. Instead, it assumes that the dominant finite-χ\chi correction can be represented through the rotational resistance of the cylinder, while the leading-order wall-coupling function of Russel et al. [8] is retained. Higher-order wall corrections and a possible wall-induced modification of the rotational resistance itself are consequently not included.

Combined angular-velocity model

At weak but finite inertia, the theoretical angular velocity is obtained by combining the inertia-induced torque in the unbounded fluid with the wall-induced torque. The quasi-steady torque balance becomes

TI,K+TW,R(1)−CHω=0.(38)T_{\mathrm{I,K}}+T_{\mathrm{W,R}}^{(1)}-C_{\mathrm{H}}\omega=0. \tag*{(38)}

Accordingly,

ωth,wbRe≤1=TI,K+TW,R(1)CH.(39)\omega_{\mathrm{th,wb}}^{\mathrm{Re}\leq1}=\frac{T_{\mathrm{I,K}}+T_{\mathrm{W,R}}^{(1)}}{C_{\mathrm{H}}}. \tag*{(39)}

Using Eqs. (32) and (37), the combined model can equivalently be written in the form used for comparison with the experimental results,

ωth,wbRe≤1=ωth,wbRe≈0+ωth,ubKP.(40)\omega_{\mathrm{th,wb}}^{\mathrm{Re}\leq1}=\omega_{\mathrm{th,wb}}^{\mathrm{Re}\approx0}+\omega_{\mathrm{th,ub}}^{\mathrm{KP}}. \tag*{(40)}

The first contribution, ωth,wbRe≈0\omega_{\mathrm{th,wb}}^{\mathrm{Re}\approx0}, represents the wall-induced rotation in the creeping-flow limit, including the finite-aspect-ratio correction to the rotational resistance according to Eq. (37). The second contribution, ωth,ubKP\omega_{\mathrm{th,ub}}^{\mathrm{KP}}, represents the additional inertia-induced rotation predicted for a translating cylinder in an unbounded fluid according to Eq. (32). No empirical fitting parameter is introduced in either contribution.

Methods

Experimental setup and definitions

Sedimentation experiments with cylindrical model particles were performed in the same rectangular duct previously used for experiments with spherical model particles [11, 19]. Figure 2 shows a photograph of the experimental setup together with top- and side-view schematics illustrating the container geometry, the particle position, the coordinate system, and the evaluated measurement region.

Experimental setup and measurement geometry

Figure 2. Experimental setup and measurement geometry. Left: photograph of the experimental setup, comprising a glass container filled with silicone oil of nominal kinematic viscosity 1000 cSt, a collimated LED light panel with its power supply, a vacuum pipette connected to a vacuum/pressure pump, and a DSLR camera. Upper right: top view of the square container cross-section with width W=140W = 140 mm and depth D=140D = 140 mm. The cylinder is positioned in the mid-plane between the front and rear walls, d2=D/2d_2 = D/2, while dd denotes the distance from the nearest lateral wall. Lower right: side view corresponding to the imaging (xy)(x y)-plane. The container height is 500 mm, and the liquid filling height is approximately 450 mm. The vertical coordinate is referenced to the initial particle-center position, x=0x = 0, located approximately 15 mm below the free surface. The hatched region indicates the evaluated measurement region, extending from x=0x = 0 to approximately 140 mm above the container bottom.

A glass container with internal dimensions of 140 mm×140 mm×500 mm140\ \mathrm{mm} \times140\ \mathrm{mm} \times500\ \mathrm{mm} (W×D×HW \times D \times H) was used for all experiments. The container was filled with silicone oil to a height of approximately 450 mm. The silicone oil had a nominal kinematic viscosity of 1000 cSt. Its dynamic viscosity was measured using a rotational viscometer (IKA® Rotavisc lo-vi), yielding η=0.9797±0.012 Pa s\eta= 0.9797 \pm0.012\ \mathrm{Pa\,s}. The liquid density at room temperature was determined using a hydrostatic weighing system consisting of a Sartorius® ENTRIS BCE 224i-1s analytical balance equipped with a Sartorius® YDK03 density determination kit. The measured density was ρ=971.28±0.134 kg m−3\rho= 971.28 \pm0.134\ \mathrm{kg\,m^{-3}}.

The selected liquid properties enabled sedimentation experiments in the creeping-flow regime, ReP≲0.1Re_{\mathrm{P}} \lesssim0.1. Consequently, centimetre-sized model particles with only a small density difference relative to the surrounding fluid experienced flow conditions comparable to those of micrometre-sized particles sedimenting in aqueous liquids.

Before release, each rod was held by a pipette under weak vacuum and immersed in the liquid. The orientation of the pipette allowed the cylinder to be positioned at different preset angles prior to release. However, this preset angle did not correspond directly to the experimentally determined initial orientation angle θini\theta_{\mathrm{ini}}. Upon release from rest, the cylinders exhibited a pronounced initial reorientation. As liquid penetrated the gap between the cylinder and the pipette tip during detachment, the cylinder experienced a torque towards smaller orientation angles. Consequently, the first measurable orientation angle θini\theta_{\mathrm{ini}}, defined at the instant when the cylinder had completely detached from the pipette, was substantially smaller than the preset angle. For example, the maximum experimentally observed initial orientation of θini≈49∘\theta_{\mathrm{ini}} \approx49^\circ required a preset angle of approximately 65∘65^\circ. Larger preset angles could not be achieved with the present release mechanism because the pipette holder could not be positioned at a more horizontal orientation.

As illustrated in the top view of Figure 2, the rod was positioned in the mid-plane between the front and rear walls, d2=D/2d_2 = D/2, while the distance dd from the nearest lateral wall was varied. After positioning, the vacuum was released and the cylindrical model particle was allowed to sediment under gravity. The cylinder was thus released with negligible initial translational velocity, UP(t=0)≈0U_{\mathrm{P}}(t = 0) \approx0, into an initially quiescent fluid, uf(t=0)=(0,0,0)T\mathbf{u}_{\mathrm{f}}(t = 0) = (0, 0, 0)^{\mathrm{T}}.

The instant of complete detachment from the pipette defines the beginning of the evaluated particle trajectory. The particle-center position at this instant is used as the origin of the vertical coordinate, x=0x = 0, as indicated by the horizontal line located approximately 15 mm below the free surface in the side view of Figure 2.

The sedimentation experiments were recorded using a DSLR camera (Nikon® D7200 equipped with a Sigma® 50 mm ff/1.4 lens). Videos were recorded at a resolution of 1920 × 1080 px and a frame rate of 25 fps.

During the experiments, the container was illuminated from the side opposite the camera using a collimated LED light panel. This backlighting arrangement produced a sharp particle silhouette on the camera sensor. To minimise image distortion, for example due to lens aberrations, the camera was aligned such that the particle trajectory remained close to the centre of the recorded image.

Before each experiment, a high-resolution reference image with a resolution of 3368 × 6000 px was recorded. The cylindrical tube of the vacuum pipette, whose outer diameter was known, served as a reference length for determining the pixel-to-millimetre conversion. Using this calibration, the actual length and diameter of the individual cylinder used in each experiment were determined from the corresponding dimensions measured in the reference image. The spatial and angular quantities used for the evaluation are defined in Figure 1 and § 2.

The focal plane of the camera coincided with the xyxy-plane of the experimental setup. When the cylinder axis remained within the focal plane and the cylinder rotated about the zz-axis through the orientation angle θ\theta, the centre of mass coincided with the centroid of the projected particle area ApxA_{\mathrm{px}}. For a homogeneous cylinder, the centroid is located at the axial half-length ll and at the radial midpoint rr.

Only cylinders rotating within the imaging plane were considered in the present study. Consequently, only rotation about the zz-axis was evaluated. The particle trajectories, orientation angles, and translational velocities were determined using a custom image-processing tool developed in MATLAB®.

Model-particle fabrication

The cylindrical model particles were fabricated from epoxy resin. Weicon® MS 1000 casting resin was prepared according to the manufacturer’s specifications and coloured black by adding 1 wt-% iron oxide powder.

The liquid epoxy resin mixture was drawn into 1 mL syringes (Ominfix®-F LuerSolo, 0.01–1 mL). Each syringe barrel was completely filled with the resin mixture before being placed vertically with the needle adapter pointing upwards. The vertical orientation allowed air trapped in the barrel, as well as air bubbles remaining from the mixing process, to rise and escape through the needle adapter. The syringes were fixed in a suitable holder and cured in an oven according to the manufacturer’s specifications.

After curing, the needle adapter was removed using a tube cutter, thereby opening the syringe barrel completely. The plunger was removed, and the space between the cured polymer and the syringe flange was filled with distilled water. Subsequently, the plunger was reinserted and advanced, forcing water between the polymer cylinder and the syringe wall. This released the cured cylinder from the syringe wall and allowed it to be pressed out through the opened barrel.

The resulting cylinders had a diameter of approximately 2r≈4.72r \approx4.7 mm. The long cylinders were cut into shorter rods with total lengths of approximately 2l≈182l \approx18 mm, 2l≈242l \approx24 mm, and 2l≈282l \approx28 mm, corresponding to aspect ratios of xˉ≈3.8\bar{x} \approx3.8, xˉ≈5.2\bar{x} \approx5.2, and xˉ≈5.8\bar{x} \approx5.8, respectively. After cutting, both end faces of each rod were carefully flattened using fine-grit sandpaper to obtain flat surfaces perpendicular to the cylinder axis. This finishing step ensured well-defined cylinder lengths and planar end faces, thereby minimising geometric deviations between individual model particles.

The cured epoxy resin had a measured solid density of ρs,Epoxy=1159.8±1.9 kg m−3\rho_{\mathrm{s,Epoxy}} = 1159.8 \pm1.9\ \mathrm{kg\ m^{-3}} and behaved as a rigid material. According to the manufacturer’s specifications, the cured epoxy resin had a Young’s modulus of E=2.9 GPaE = 2.9\ \mathrm{GPa}. The solid density was determined by hydrostatic weighing using a Sartorius® ENTRIS BCE 224i-1s analytical balance equipped with a Sartorius® YDK03 density determination kit. The balance had a readability of 0.1 mg, and all density measurements were performed in distilled water.

Results and Discussion

The results are presented in three steps. First, the sedimentation of axially aligned cylinders in the center of the rectangular duct is examined to establish the reference behavior in the absence of a dominant interaction with a single lateral wall. The reproducibility of the sedimentation kinematics and the influence of the cylinder aspect ratio are assessed, and the measured steady velocities are compared with the theoretical predictions introduced in § 2.1. Subsequently, the translational kinematics of initially inclined cylinders sedimenting at different distances from a plane wall are investigated. Particular attention is paid to the coupled evolution of the cylinder orientation, wall distance, and translational velocity and to the applicability of the wall-corrected velocity models. Finally, the corresponding angular kinematics are analyzed and compared with the wall-induced and inertia-corrected angular-velocity models developed in § 2.2. Additional results for cylinders with the smaller aspect ratio χ≈3.8\chi\approx3.8 are provided in Appendix § 6.

Throughout this section, individual experimental runs are shown rather than data averaged over repeated measurements. This distinction is important because the sedimentation dynamics considered here are sensitive to the instantaneous particle configuration. Small and experimentally unavoidable variations in the initial conditions, including the initial orientation θini\theta_{\mathrm{ini}}, the initial wall distance dini/ld_{\mathrm{ini}}/l, and small differences in particle properties between individual specimens, can lead to different subsequent evolutions of the particle orientation, wall distance, and translational and angular velocities. These differences therefore do not necessarily represent random experimental scatter that can be removed by ensemble averaging. In particular, averaging absolute velocities or combining trajectories obtained at slightly different initial conditions would destroy the direct correspondence between θ\theta, d/ld/l, UPU_{\mathrm{P}}, and ω\omega along an individual trajectory and may consequently obscure systematic features of the particle dynamics. Repeated experiments are instead used to assess the reproducibility of the observed behavior, while the individual trajectories and their associated kinematic quantities are retained for the analysis. Numerical scatter arising from the evaluation of the recorded trajectories, particularly from temporal differentiation, was minimized by defining the evaluation time step prior to the analysis and applying it consistently to the experimental data.

Axially aligned sedimentation of short cylinders in the center of a rectangular duct

The sedimentation behavior in the center of the rectangular duct was first examined for axially aligned cylinders with mean aspect ratios of χˉ≈5.2\bar{\chi} \approx5.2 and χˉ≈5.8\bar{\chi} \approx5.8. This configuration represents the symmetric case, in which the four surrounding container walls are approximately equidistant from the particle and therefore exert comparable, yet still measurable, influences on its sedimentation kinematics. It provides a reference for the subsequent experiments in which the symmetry is broken by positioning the cylinder closer to one of the lateral walls.

Figure 3 shows the sedimentation velocity UP,xU_{\mathrm{P},x} as a function of the dimensionless sedimentation distance x/lx/l for cylinders with χˉ≈5.2\bar{\chi} \approx5.2 and χˉ≈5.8\bar{\chi} \approx5.8, released with negligible initial translational velocity in an approximately axially aligned orientation, θini≈0∘\theta_{\mathrm{ini}} \approx0^\circ, in the center of the rectangular duct.

The velocity curves in Figure 3 show that the axially aligned cylinders initially accelerate before approaching an approximately steady sedimentation velocity. The corresponding spatial and angular trajectories (not shown) exhibit similarly stable behavior when the cylinders are released close to the center of the duct with θini≈0∘\theta_{\mathrm{ini}} \approx0^\circ: both the orientation angle θ\theta and the dimensionless wall distance d/ld/l remain approximately constant throughout the evaluated sedimentation path. Small deviations from this symmetric initial configuration, either due to a non-zero initial orientation or a slightly off-center release position, result in gradual changes in the orientation angle. Depending on the orientation of the cylinder relative to the nearest wall, θ\theta may either increase or decrease during sedimentation.

Sedimentation velocity plotted against dimensionless sedimentation distance

Figure 3. Sedimentation velocity UP,xU_{\mathrm{P},x} plotted against the dimensionless sedimentation distance x/lx/l for cylinders (E≈2900E \approx2900 MPa; ρP≈1159.8\rho_{\mathrm{P}} \approx1159.8 kg m−3^{-3}) belonging to two groups with mean aspect ratios χˉ≈5.2\bar{\chi} \approx5.2 and χˉ≈5.8\bar{\chi} \approx5.8, released with negligible initial translational velocity in an approximately axially aligned orientation (θini≈0∘\theta_{\mathrm{ini}} \approx0^\circ) at Re⁡P≈0.1\operatorname{Re}_{\mathrm{P}} \approx0.1 in the center of the duct. Inset: Mean steady sedimentation velocity and corresponding standard deviation for each experiment plotted against the individually measured aspect ratio χ\chi. Colors and symbols correspond to those used for the respective experiments in the main panel. The experimental results are compared with the theoretical sedimentation velocities calculated using UiU_i’s model (2), Eq. (13), with the creeping-flow axial drag of Eq. (3) (light-green plus symbols) and with the inertia-corrected axial drag of Eq. (5) (dark-green cross symbols).

In addition, a focusing behavior is consistently observed across the experiments. Cylinders released slightly off-center tend to migrate toward an equilibrium position within the duct as sedimentation proceeds. This tendency is accompanied by an apparent dependence on cylinder size: the longer cylinders investigated here show a stronger tendency to recover axially aligned orientation and a centered position, whereas the shorter cylinders exhibit larger deviations in orientation and lateral position in response to small differences in their initial conditions.

The inset of Figure 3 shows the mean steady sedimentation velocity obtained from each experiment in the main panel plotted against the individually measured aspect ratio χ\chi, which were χ≈5.1;5.3;5.7;6.0\chi\approx5.1; 5.3; 5.7; 6.0. The experimental values are compared with the theoretical velocities calculated using Ui’s model (2), Eq. (13), with either the creeping-flow axial drag of Eq. (3) or the inertia-corrected axial drag of Eq. (5). Overall, the measured velocities are in good agreement with the predictions based on Ui’s model. For the cylinders with the larger aspect ratio, the experimental values are more closely described when the inertia correction is included, whereas the difference between the two theoretical predictions is less pronounced for the cylinders with the smaller aspect ratio. This trend is consistent with the slightly larger particle Reynolds numbers of the longer cylinders, with the mean RePRe_{\mathrm{P}} increasing from approximately 0.08 for the χˉ≈5.2\bar{\chi} \approx5.2 group to approximately 0.09 for the χˉ≈5.8\bar{\chi} \approx5.8 group.

For comparison, the theoretical velocities were also evaluated using Brenner’s model (1), Eq. (12), which consistently yielded lower velocities than those measured experimentally and is therefore not included in the inset. The origin of this systematic underprediction cannot be unambiguously identified from the present experiments. Differences in the higher-order treatment of the wall correction may contribute to the discrepancy; however, a quantitative assessment of the individual contributions is beyond the scope of the present study.

Near-wall translational kinematics of initially inclined cylinders

After establishing the sedimentation behavior in the symmetric configuration at the center of the duct, the translational kinematics were investigated for initially inclined cylinders with a mean aspect ratio of χˉ≈5.8\bar{\chi} \approx5.8 released at different distances from a lateral wall. The resulting asymmetric wall interaction leads to a coupled evolution of the orientation angle θ\theta, the dimensionless wall distance d/ld/l, and the total translational velocity magnitude ∣U⃗∣|\vec{U}| in the xyxy-plane. Figure 4 summarizes this evolution for the investigated initial conditions and compares the measured velocities with the wall-corrected theoretical models introduced in § 2.1. The individual aspect ratios and initial conditions corresponding to the different experiments are specified in Figure 5.

Translational kinematics of cylinders during near-wall sedimentation

Figure 4. Translational kinematics of cylinders with a mean aspect ratio of χˉ≈5.8\bar{\chi} \approx5.8 during near-wall sedimentation and comparison with the theoretical models. The cylinders were released at different initial wall distances and with different initial orientation angles. (a) Dimensionless wall distance d/ld/l plotted against the orientation angle θ\theta. The combined spatial and angular trajectories reveal glancing and contact-induced flipping behavior depending on the initial conditions. (b) Experimental translational velocity magnitude ∣U⃗∣\lvert\vec{U}\rvert (hollow diamonds) and dimensionless wall distance d/ld/l (filled diamonds) plotted against θ\theta for a cylinder released at a larger initial wall distance. The experimental velocity is compared with the theoretical velocity from Eq. (22), using the inertia-corrected axial and transverse drag expressions of Eqs. (5) and (6), in combination with Brenner’s model [1] (dashed red line) and Ui’s model [2] (solid red line). The green-gray lines show the small-angle approximations based on Ui’s model [2], Eq. (16), and Brenner’s model [1], Eq. (19), respectively. The red lines represent the model using a consistent normalization of both drag components, whereas the cyan line shows the corresponding result when the transverse drag component is normalized by 2πηlUth,ub2\pi\eta lU_{\mathrm{th,ub}} instead of 4πηlUth,ub4\pi\eta lU_{\mathrm{th,ub}}. (c) Total translational velocity magnitude ∣U⃗∣\lvert\vec{U}\rvert (hollow symbols) and dimensionless wall distance d/ld/l (filled symbols) plotted against θ\theta for cylinders released in the near-wall region (dini/l≲2.5d_{\mathrm{ini}}/l \lesssim2.5). The experimental velocities are compared with Eq. (22), using Eqs. (5) and (6) in combination with Ui’s model [2] (solid lines) and Brenner’s model [1] (dashed lines). The individual aspect ratios and initial wall distances and orientation angles corresponding to the different symbols are given in Figure 5.

Angular kinematics of cylinders with a mean aspect ratio of $\bar{\chi} \approx 5.8$ during near-wall sedimentation and comparison with the theoretical models

Figure 5. Angular kinematics of cylinders with a mean aspect ratio of χˉ≈5.8\bar{\chi} \approx5.8 during near-wall sedimentation and comparison with the theoretical models. Angular velocity ω\omega plotted against the orientation angle θ\theta for the individual experiments shown in Figure 4. Hollow symbols represent the experimentally determined angular velocities. The black lines show the finite-aspect-ratio-corrected leading-order wall contribution based on Russel et al., Eq. (37), whereas the cyan lines show the combined angular-velocity model, Eq. (40), including the inertia-induced contribution from Eq. (32). The individually measured aspect ratios and initial conditions are: (a) χ≈6.0\chi\approx6.0, θini≈33.0∘\theta_{\mathrm{ini}} \approx33.0^{\circ}, dini/l≈3.00d_{\mathrm{ini}}/l \approx3.00; (b) χ≈6.3\chi\approx6.3, θini≈32.4∘\theta_{\mathrm{ini}} \approx32.4^{\circ}, dini/l≈2.91d_{\mathrm{ini}}/l \approx2.91; (c) χ≈5.7\chi\approx5.7, θini≈18.4∘\theta_{\mathrm{ini}} \approx18.4^{\circ}, dini/l≈1.59d_{\mathrm{ini}}/l \approx1.59; (d) χ≈6.0\chi\approx6.0, θini≈19.8∘\theta_{\mathrm{ini}} \approx19.8^{\circ}, dini/l≈1.51d_{\mathrm{ini}}/l \approx1.51; (e) χ≈5.9\chi\approx5.9, θini≈32.4∘\theta_{\mathrm{ini}} \approx32.4^{\circ}, dini/l≈2.39d_{\mathrm{ini}}/l \approx2.39; (f) χ≈6.3\chi\approx6.3, θini≈31.4∘\theta_{\mathrm{ini}} \approx31.4^{\circ}, dini/l≈2.34d_{\mathrm{ini}}/l \approx2.34; (g) χ≈5.8\chi\approx5.8, θini≈31.3∘\theta_{\mathrm{ini}} \approx31.3^{\circ}, dini/l≈1.67d_{\mathrm{ini}}/l \approx1.67; (h) χ≈6.0\chi\approx6.0, θini≈31.6∘\theta_{\mathrm{ini}} \approx31.6^{\circ}, dini/l≈1.59d_{\mathrm{ini}}/l \approx1.59; and (i) χ≈5.7\chi\approx5.7, θini≈32.0∘\theta_{\mathrm{ini}} \approx32.0^{\circ}, dini/l≈1.61d_{\mathrm{ini}}/l \approx1.61.

As shown in Figure 4(a), the near-wall trajectories depend strongly on the initial wall distance and orientation. Cylinders released at larger wall distances and moderate initial inclinations approach the wall while simultaneously rotating toward the axially aligned orientation, i.e., toward θ=0∘\theta= 0^\circ. For several trajectories, a minimum wall distance is reached without contact, after which the cylinder moves away from the wall while the orientation angle changes sign. This behavior qualitatively corresponds to the glancing turn described by Russel et al. [8], for which a sedimenting rod approaches a vertical wall, rotates through the vertical orientation, and subsequently moves away with the same end leading.

A distinctly different behavior is observed for cylinders released at comparable initial wall distances but substantially larger initial orientation angles. The three corresponding trajectories in Figure 4(a) were obtained for θini≈46.5∘\theta_{\mathrm{ini}} \approx46.5^\circ, 47.1∘47.1^\circ, and 48.6∘48.6^\circ, at dini/l≈2.50d_{\mathrm{ini}}/l \approx2.50, 2.27, and 2.42, respectively. In contrast to the glancing trajectories obtained at similar initial wall distances but smaller θini\theta_{\mathrm{ini}}, these cylinders continue to approach the wall until physical contact occurs. The subsequent interaction with the wall induces a pronounced rotation through the

horizontally aligned orientation and a change in the sign of θ\theta. This motion is referred to here as contact-induced flipping behavior. The comparison demonstrates that the occurrence of glancing or flipping cannot be determined from the initial wall distance alone and, for otherwise comparable wall distances, is strongly dependent on the initial orientation.

The contact-induced flipping observed here should be distinguished from the reversing turn described by Russel et al. [8]. In their analysis of very slender rods, reversing denotes a non-contact motion at large initial inclination, in which the rod changes its direction of rotation and subsequently moves away from the wall with the opposite end leading. Russel et al. additionally identified an intermediate range of initial orientations for which the rod approaches sufficiently close to the wall that collision occurs. The present flipping trajectories are therefore qualitatively more closely related to this collision regime than to the classical non-contact reversing turn. A direct comparison of the corresponding transition angles is, however, not appropriate because the cylinders investigated here have substantially smaller aspect ratios than the slender rods considered by Russel et al.

The experiment shown in Figure 4(b) exhibits the glancing turn farthest from the wall, reaching θ=0∘\theta= 0^\circ at the largest wall distance among the investigated trajectories. Figure 4(b) compares the measured total translational velocity magnitude with the predictions obtained using the different wall corrections. At larger wall distances, particularly for d/l≳2d/l \gtrsim2, Brenner’s model (1, dashed lines) provides a closer description of the experimental velocities than Ui’s model (2, solid lines). As the cylinder approaches the wall, however, the relative agreement changes, and Ui’s model provides a closer description of the measured velocities. Thus, neither wall correction provides the closest agreement over the entire glancing trajectory; their relative performance depends on the instantaneous wall distance.

The corresponding small-angle approximations reproduce the measured velocities well only in the immediate vicinity of the axially aligned orientation, approximately within ∣θ∣≲5∘|\theta| \lesssim5^\circ. Outside this restricted angular range, the predictions increasingly deviate from the experimental data, consistent with the small-angle assumptions underlying Eqs. (16) and (19).

The comparison in Figure 4(b) also illustrates the influence of the normalization of the transverse drag coefficient on the predicted velocities. With the consistent normalization described in § 2.1, in which the transverse drag coefficient is normalized by 4πηUth,ub4\pi\eta U_{\mathrm{th,ub}}, the predictions based on both wall corrections exhibit larger deviations from the experimental data at large wall distances and large orientation angles. For comparison, the cyan curve shows the result obtained when the transverse drag coefficient is instead normalized by 2πηUth,ub2\pi\eta U_{\mathrm{th,ub}}, while retaining the corresponding velocity components in Eq. (22). This alternative normalization leads to a closer agreement with the measurements in these regions. However, it is not consistent with the normalization of the translational drag coefficients adopted in § 2.1 and therefore does not provide a physically justified alternative to the model based on the consistent normalization. The comparison nevertheless highlights the limitations of the present wall-corrected models at large wall distances and orientation angles.

Figure 4(c) focuses on the glancing trajectories with initial wall distances dini/l≲2.5d_{\mathrm{ini}}/l \lesssim2.5. A characteristic feature of these experiments is the pronounced deceleration as the cylinders enter the immediate vicinity of the wall, where d/l<1d/l < 1. The measured translational velocities decrease sharply with decreasing wall distance, resulting in strongly unsteady kinematics during this part of the trajectory. Despite this pronounced unsteadiness, the measured velocities are well described by the theoretical predictions based on Ui’s wall-distance model (2, solid lines), combined with the inertia-corrected axial and transverse drag expressions. The agreement remains good even for d/l<1d/l < 1, where the strongest deceleration is observed. In contrast, the prediction based on Brenner’s model (1, dashed lines) diverges as the wall distance decreases below d/l=1d/l = 1, yielding unphysical velocity values at the smallest wall distances considered.

The quantitative comparison with the theoretical models is restricted to trajectories exhibiting glancing behavior. For the contact-induced flipping trajectories, substantial deviations between the measured and predicted velocities already develop before physical contact with the wall. This is consistent with the construction of the translational model in Eq. (22), which interpolates between the limiting wall-bounded velocities for axial and transverse cylinder orientations. At intermediate orientations around ∣θ∣≈45∘|\theta| \approx45^\circ, the axial and transverse contributions enter the interpolation with comparable weights, whereas neither limiting configuration provides a close representation of the hydrodynamics of an inclined cylinder. Consequently, deviations associated with both limiting approximations can contribute simultaneously, making this angular range particularly sensitive to the underlying model assumptions. Once physical contact with the wall occurs, the cylinder motion is additionally governed by contact mechanics and the associated forces and torques, which are not included in the present hydrodynamic model. The contact-induced flipping trajectories are therefore used to characterize the observed motion but are excluded from the quantitative comparison with the theoretical velocity models.

Corresponding experiments were also performed with shorter cylinders having a mean aspect ratio of χˉ≈3.8\bar{\chi} \approx3.8, and the results are presented in Figure A1. The principal trends observed for the χˉ≈5.8\bar{\chi} \approx5.8 cylinders are also reproduced at the smaller aspect ratio. In particular, Brenner’s model (1) provides a closer description at larger wall distances, whereas Ui’s model (2) better captures the translational velocities as the minimum wall distance reached during glancing decreases. The pronounced near-wall deceleration is likewise observed for the shorter cylinders. However, the trajectory behavior becomes less distinct at the smaller aspect ratio. In particular, the distinction between glancing and contact-induced flipping is less pronounced, with both types of motion occurring for comparable combinations of initial wall distance and orientation. This increased sensitivity and variability at smaller χ\chi may be associated with a stronger influence of finite-aspect-ratio effects, including the hydrodynamic influence of the cylinder ends.

Translational and angular kinematics of shorter cylinders during near-wall sedimentation

Figure A1. Translational and angular kinematics of shorter cylinders with a mean aspect ratio of χˉ≈3.8\bar{\chi} \approx3.8 during near-wall sedimentation and comparison with the theoretical models. (a) Dimensionless wall distance d/ld/l plotted against the orientation angle θ\theta for cylinders released at different initial wall distances and orientation angles. The combined spatial and angular trajectories reveal glancing and contact-induced flipping behavior depending on the initial conditions. (b–d) Selected experiments exhibiting glancing behavior for three different initial conditions. The left panels show the total translational velocity magnitude in the xyxy-plane, ∣U∣|\mathbf{U}| (hollow symbols), and the dimensionless wall distance d/ld/l (filled symbols) plotted against θ\theta, together with the theoretical translational velocities based on Eq. (22) with Brenner´s model (dashed lines) and Ui´s model solid lines. The right panels show the experimentally determined angular velocity ω\omega (hollow symbols) plotted against θ\theta. The black lines represent the finite-aspect-ratio-corrected leading-order wall contribution based on Russel et al., Eq. (37), whereas the cyan lines represent the combined angular-velocity model, Eq. (40), including the inertia-induced contribution from Eq. (32), as described in Figure 5. The individually measured aspect ratios and initial conditions of the selected experiments are: (b) χ≈3.8\chi\approx3.8, θini≈22.7∘\theta_{\mathrm{ini}} \approx22.7^\circ, dini/l≈2.90d_{\mathrm{ini}}/l \approx2.90; (c) χ≈3.9\chi\approx3.9, θini≈22.7∘\theta_{\mathrm{ini}} \approx22.7^\circ, dini/l≈2.16d_{\mathrm{ini}}/l \approx2.16; and (d) χ≈3.7\chi\approx3.7, θini≈32.7∘\theta_{\mathrm{ini}} \approx32.7^\circ, dini/l≈3.55d_{\mathrm{ini}}/l \approx3.55.

Near-wall angular kinematics of initially inclined cylinders

Following the analysis of the translational kinematics, the angular motion is considered for the same near-wall sedimentation experiments shown in Figure 4. The analysis focuses on the glancing trajectories, for which the continuous evolution of the angular velocity can be compared with the theoretical models introduced in § 2.2 without the additional influence of physical wall contact. Figure 5 shows the experimentally determined angular velocity ω\omega as a function of the orientation angle θ\theta for cylinders with a mean aspect ratio of χˉ≈5.8\bar{\chi} \approx5.8. The individual panels correspond to the same experiments presented in Figure 4, with equivalent colors and symbols to facilitate direct comparison with the corresponding d/ld/l-θ\theta trajectories. The experimental angular velocities are compared with the finite-aspect-ratio-corrected leading-order wall contribution based on Russel et al., Eq. (37), and with the combined angular-velocity model, Eq. (40), which additionally includes the inertia-induced contribution from Eq. (32).

Although the angular velocity is plotted as a function of θ\theta in Figure 5, the resulting curves should not be interpreted as a dependence on the orientation angle alone. The wall-induced contribution in Eq. (37) is, to leading order, governed by the instantaneous dimensionless wall distance d/ld/l, which itself evolves continuously with θ\theta along each glancing trajectory. Consequently, the apparent variation of ω\omega with θ\theta reflects primarily the coupled evolution of the cylinder orientation and wall distance, θ(t)\theta(t) and d(t)/ld(t)/l, rather than a purely angular dependence of the wall-induced rotation. The corresponding d/ld/l-θ\theta trajectories are shown in Figure 4 and should therefore be considered together with the angular-velocity curves in Figure 5.

Before discussing the comparison with the theoretical models, two features of Figure 5 should be noted. First, the panels are arranged according to the wall distance reached at the axially aligned orientation. The dimensionless wall distance at θ=0∘\theta= 0^\circ progressively decreases from Figure 5(a) to (i). Thus, panel (a), corresponding to the experiment shown in Figure 4(b), represents the glancing trajectory passing through the axially aligned orientation farthest from the wall, whereas the subsequent panels represent trajectories reaching progressively smaller wall distances. This arrangement facilitates a direct assessment of the increasing influence of the wall on the angular kinematics.

Second, several experimental curves exhibit increased scatter in the measured angular velocity in the immediate vicinity of θ=0∘\theta= 0^\circ. This scatter is primarily an artifact of the image evaluation and numerical differentiation rather than an indication of rapid physical fluctuations in the cylinder rotation. The orientation angle is determined frame by frame from the segmented cylinder geometry and is therefore subject to small variations associated with the finite image resolution and segmentation. These comparatively small uncertainties in θ\theta are amplified when calculating the angular velocity from its temporal variation and become particularly apparent as the signed orientation passes through θ=0∘\theta= 0^\circ. Accordingly, the individual fluctuations in ω\omega close to the axially aligned orientation should not be interpreted as physical unsteadiness of the rotational motion.

A common feature of the measured angular-velocity curves is the pronounced maximum in ∣ω∣|\omega| at the beginning of the evaluated trajectories, corresponding to the largest positive orientation angles on the right-hand side of the panels. As sedimentation proceeds, the magnitude of the angular velocity initially decreases rapidly. This initial rotational transient coincides with the translational acceleration phase following release. As discussed in § 3, the measured initial orientation θini\theta_{\mathrm{ini}} corresponds to the first orientation that can be reliably determined after complete detachment from the vacuum pipette and therefore differs from the preset release angle. During detachment, the cylinder ends remain in close proximity to the pipette tips, while liquid enters the developing gaps between the cylinder and the tips. The resulting hydrodynamic interaction, together with the release process itself, can therefore influence the initial translational and rotational motion.

Consequently, the pronounced initial variation in ω\omega cannot be attributed unambiguously to an intrinsic transient hydrodynamic response of the freely sedimenting cylinder. Such a contribution may in principle be present, analogous to the transient inertial wall attraction previously observed for spherical particles released from rest [11], but cannot be separated from release-induced effects with the present experimental configuration. The initial angular-velocity peak and its subsequent rapid decrease are therefore reported here as part of the measured kinematics but are not interpreted quantitatively in terms of the theoretical near-wall rotation models.

A second common feature of the experiments is the good qualitative agreement between the measured angular velocities and the theoretical predictions over the entire range of investigated wall distances. In particular, the theoretical curves reproduce the characteristic evolution of ω\omega along the glancing trajectories, supporting the interpretation that the angular kinematics are governed primarily by the changing wall distance. For the moderate wall distances represented by Figure 5(b)–(Figure 5g), the agreement is also quantitatively good over substantial parts of the trajectories. In this range, the combined model, Eq. (40), differs only slightly from the leading-order wall contribution, Eq. (37), indicating that the additional inertia-induced contribution from Eq. (32) remains comparatively small.

A systematic difference between the two theoretical predictions nevertheless becomes apparent when the approach toward and the motion away from the wall are considered separately. During the approach toward the wall, i.e., before the cylinder passes through the axially aligned orientation at θ=0∘\theta= 0^\circ, inclusion of the inertia-induced contribution generally shifts the prediction slightly away from the experimental data, such that Eq. (37) provides the closer description. After passage through θ=0∘\theta= 0^\circ, when the cylinder moves away from the wall, the combined model generally provides a slightly closer description of the measured angular velocity. Thus, although the wall-induced contribution dominates the predicted rotational motion, the influence of the additional inertia-induced contribution depends on the respective part of the glancing trajectory.

More pronounced quantitative deviations are observed at the largest wall distance in Figure 5(a) and, in particular, for trajectories entering the immediate wall region with d/l<1d/l < 1. Nevertheless, even in the latter regime, the theoretical curves reproduce the characteristic evolution of the measured angular velocity remarkably well. This qualitative agreement indicates that the leading-order wall-distance dependence retained in Eq. (37) captures an essential contribution to the rotational dynamics even in the immediate wall region. The remaining quantitative deviations are likely associated with several limitations of the theoretical description.

First, Eq. (37) retains only the leading-order wall-distance dependence based on the axially aligned configuration of Russel et al. and does not explicitly include orientation-dependent contributions to the wall-induced rotation. Its variation along a glancing trajectory therefore results primarily from the changing instantaneous wall distance. At finite inclination, particularly in the immediate vicinity of the wall, the neglected orientation-dependent terms may become quantitatively important and could explain why the model captures the characteristic evolution of ω\omega while deviations remain in its absolute magnitude.

Second, higher-order contributions to the wall interaction are likewise not represented by the leading-order description. Such contributions associated with the wall-reflected disturbance flow are expected to become increasingly relevant as the wall distance decreases. The good qualitative agreement observed even for d/l<1d/l < 1 therefore indicates that the leading wall-distance dependence captures the dominant trend of the rotational motion, while neglected higher-order wall contributions may account for part of the remaining quantitative discrepancy.

Third, the strongest near-wall deviations occur in a region in which both the translational and rotational kinematics vary substantially along the trajectory. The theoretical wall contribution represents an instantaneous, quasi-steady description of the hydrodynamic interaction and therefore assumes that the surrounding disturbance flow adjusts sufficiently rapidly to the changing particle configuration. At finite Rep\mathrm{Re}_p, a delayed adjustment of the wall-reflected disturbance flow may contribute to the observed deviations when the particle-wall configuration changes rapidly. Such transient effects provide a plausible additional explanation for the quantitative differences in the immediate wall region, although their contribution cannot be isolated from the present experiments.

The quantitative discrepancy at the comparatively large wall distance in Figure 5(a) cannot readily be attributed to the same mechanisms that become increasingly important in the immediate wall region. In this case, the wall-induced rotational contribution is substantially weaker, and higher-order near-wall corrections would not be expected to provide the primary explanation. Deviations associated with the finite-aspect-ratio approximation, the approximate representation of the weak wall interaction, and experimental uncertainties may therefore become comparatively more apparent. The present data do not permit the individual contributions to this far-field discrepancy to be identified unambiguously.

The corresponding angular-velocity measurements for the shorter cylinders with χˉ≈3.8\bar{\chi} \approx3.8 are shown in Figure A1. The characteristic evolution observed for the longer cylinders is reproduced, while the quantitative agreement with the theoretical predictions is generally improved. In particular, the model closely describes the measured angular velocity for the glancing trajectory shown in Figure A1(d). The same trend was found for trajectories reaching even smaller wall distances, including d/l<1d/l < 1. Although the quantitative deviations increase in the immediate wall region compared with those observed at intermediate wall distances, they remain smaller than for the cylinders with χˉ≈5.8\bar{\chi} \approx5.8 under comparable near-wall conditions. Thus, the increasing deviations at d/l<1d/l < 1 are observed at both aspect ratios, but their magnitude depends on the cylinder aspect ratio. The improved agreement for the shorter cylinders is also consistent with the aspect-ratio dependence of the rotational-resistance expansion reported by Pierson et al. [4], although the respective contributions of the resistance approximation and the approximations inherent in the wall model cannot be separated from the present experiments.

Overall, the angular-velocity measurements demonstrate that the near-wall rotation of the cylinders is governed predominantly by the wall-induced contribution, with the instantaneous wall distance providing the principal dependence along the glancing trajectories. The leading-order wall model captures the characteristic evolution of the angular velocity over the entire investigated range of wall distances and aspect ratios, including trajectories entering d/l<1d/l < 1, although its quantitative accuracy depends on both wall distance and aspect ratio. The additional inertia-induced contribution remains comparatively small under the present conditions and modifies the agreement differently during the approach toward and departure from the wall. Together with the translational results of § 4.2, these observations demonstrate that the near-wall sedimentation of finite-aspect-ratio cylinders is characterized by a strongly coupled evolution of wall distance, orientation, translational velocity, and angular velocity, while comparatively simple wall-corrected models already capture the principal features of the experimentally observed kinematics.

Summary and Conclusions

The present study investigated the sedimentation of short, rigid cylinders in a rectangular duct filled with a highly viscous, initially quiescent liquid at particle Reynolds numbers of Rep≈O(10−1)Re_p \approx O(10^{-1}). Cylinders with mean aspect ratios of approximately χˉ=3.8\bar{\chi} = 3.8–5.85.8 were released from rest, either approximately axially aligned near the centre of the duct or at different initial inclinations and distances from a lateral rigid wall. Time-resolved image measurements were used to determine their trajectories, orientations, translational velocities, and angular velocities. The measured kinematics were compared with theoretical descriptions accounting for finite-cylinder resistance, weak fluid inertia, and wall-induced hydrodynamic effects. The model velocities were evaluated using the instantaneously measured particle orientation and wall distance along the experimental trajectories; complete trajectories were not predicted from prescribed initial conditions.

For approximately axially aligned cylinders sedimenting near the centre of the rectangular duct, the experiments provided a reference configuration in which the influence of the surrounding walls could be assessed without the pronounced changes in orientation and lateral position observed in the near-wall experiments. The measured sedimentation velocities demonstrated that confinement remains relevant even when the cylinder is positioned approximately symmetrically between the duct walls. The comparison with the theoretical descriptions showed that accounting for weak fluid inertia improves the representation of the measured velocities and that the wall-correction approach based on Ui et al. [15] provides closer agreement than the alternative based on Brenner [14] for the investigated centre-duct configurations. These results establish the importance of considering both finite-cylinder resistance and confinement when estimating the sedimentation velocity of short cylinders, even in a nominally symmetric configuration.

The experiments with initially inclined cylinders near a lateral wall revealed a coupling between translation and rotation. The evolving orientation and wall distance affected both the sedimentation velocity and the lateral motion, while wall-induced rotation altered the subsequent particle–wall configuration. Two distinct types of near-wall trajectories were observed. During glancing motion, the cylinder rotated through the axially aligned orientation in the vicinity of the wall and subsequently moved away without changing which end was leading. In contrast, contact-induced flipping involved physical contact with the wall and a subsequent rotation of the cylinder through a nearly horizontal orientation. This distinction is important because the non-contact reversing trajectories reported for very slender rods in earlier studies should not be identified with the contact-induced flipping observed here. The present velocity models do not include contact forces; consequently, their quantitative assessment was restricted to the glancing trajectories.

The translational measurements revealed different ranges of applicability for the investigated wall-correction models. At larger wall distances, the Brenner-based description generally provided a closer representation of the measured sedimentation velocities. At intermediate and small wall distances, the Ui-based description agreed more closely with the measurements, whereas the Brenner-based description failed in the close-wall regime. The translational velocity description incorporating both axial and transverse components reproduced the principal variations in the measured sedimentation velocities as the cylinders’ orientations and wall distances evolved, although deviations remained in parts of the trajectories.

The angular measurements showed that the near-wall rotation of the investigated cylinders was governed predominantly by the wall-induced contribution represented by the leading-order term retained from the theory of Russel et al. [8]. The inertia-induced contribution was comparatively small in the investigated Reynolds-number range. Including it did not uniformly improve agreement with the experiments: it tended to increase deviations during the approach toward the wall but improved the representation of parts of the subsequent departure. The angular-velocity model incorporating wall-induced and inertia-induced contributions reproduced the principal features of the measured rotational kinematics, including the change in angular velocity as the cylinder passed through the near-wall region. Quantitative agreement was generally closer at intermediate wall distances, whereas larger deviations occurred in some trajectories farther from the wall and in the immediate near-wall region. The latter discrepancies indicate that the leading-order wall description is insufficient to represent every stage of the motion, particularly when the cylinder approaches the boundary closely.

The experiments with the shorter cylinders, χˉ≈3.8\bar{\chi} \approx3.8, were used to assess the applicability of the kinematic descriptions beyond the aspect ratios considered in the main near-wall experiments. The translational kinematics exhibited comparable trends, indicating that the principal features of the wall-induced changes in sedimentation velocity and lateral motion were also observed for the shorter cylinders. For the angular kinematics, the model showed even closer quantitative agreement with the measurements than for the longer cylinders. This finding supports the use of the higher-order slender-body approximation for rotational resistance in the present model: accounting for finite aspect ratio is important when extending a description originally developed for slender rods to short cylinders. However, the comparison alone does not establish that the higher-order approximation is the sole reason for the improved agreement, since the wall-induced contribution and the experimentally realized trajectories also affect the predicted angular velocity.

Taken together, the results show that the investigated velocity descriptions can reproduce important features of the instantaneous translational and angular kinematics of short cylinders sedimenting near a rigid wall, provided that their ranges of applicability are respected. Evaluating the models at the measured orientation and wall distance allows their predictions to be compared throughout an evolving experimental trajectory. Agreement with the measured instantaneous velocities does not, however, establish that the models can independently predict complete trajectories from initial conditions. The experiments also identify limitations associated with close wall proximity, larger inclinations, and contact-induced motion, for which additional hydrodynamic contributions or contact mechanics would need to be considered.

The present experiments provide a reference for investigating the near-wall sedimentation of short cylinders at lower particle Reynolds numbers and with additional influencing factors such as elasticity. Earlier experiments reported in the author’s dissertation [20] revealed markedly different kinematics at ReP≈O(10−2)Re_{\mathrm{P}} \approx O(10^{-2}), including flipping of initially axially aligned cylinders and substantial variability between repeated releases. Additional, as yet unpublished experiments with initially inclined cylinders have also revealed flipping without contact with the wall. This non-contact behavior contrasts with the contact-induced flipping observed in the present study and demonstrates that the range of near-wall motions extends beyond the trajectories examined here. However, the greater variability observed at lower Reynolds numbers, together with the limited number of experiments available so far, precludes quantitative conclusions about the conditions governing these different motions. Further systematic measurements are needed to establish their dependence on Reynolds number, aspect ratio, initial orientation, and wall distance, and to determine whether the instantaneous velocity descriptions assessed in the present study remain applicable under these conditions.

Additional results

Figure A1 shows the experimentally determined dimensionless wall distance d/ld/l, total translational velocity magnitude ∣U⃗∣\lvert\vec{U}\rvert and the angular velocity ω\omega as a function of the orientation angle θ\theta for short cylinders with a mean aspect ratio of χˉ≈3.8\bar{\chi} \approx3.8.

References

References

  1. [1]Hamidi, A., Daramsing, D., Gordon, M.D., Hanson, R.E.: Low reynolds number settling of bent rods in quiescent fluid. Fluids 11(3), 72 (2026) https://doi.org/10.3390/fluids11030072
  2. [2]Batchelor, G.K.: Slender-body theory for particles of arbitrary cross-section in stokes flow. Journal of Fluid Mechanics 44(3), 419–440 (1970) https://doi.org/10.1017/S002211207000191X
  3. [3]Kharrouba, M., Pierson, J.-L., Magnaudet, J.: Flow structure and loads over inclined cylindrical rodlike particles and fibers. Physical Review Fluids 6(4) (2021) https://doi.org/10.1103/PhysRevFluids.6.044308
  4. [4]Pierson, J.-L., Kharrouba, M., Magnaudet, J.: Hydrodynamic torque on a slender cylinder rotating perpendicularly to its symmetry axis. Physical Review Fluids 6(9) (2021) https://doi.org/10.1103/PhysRevFluids.6.094303
  5. [5]Khayat, R.E., Cox, R.G.: Inertia effects on the motion of long slender bodies. Journal of Fluid Mechanics 209, 435–462 (1989) https://doi.org/10.1017/S0022112089003174
  6. [6]Cabrera, F., Sheikh, M.Z., Mehlig, B., Plihon, N., Bourgoin, M., Pumir, A., Naso, A.: Experimental validation of fluid inertia models for a cylinder settling in a quiescent flow. Physical Review Fluids 7(2) (2022) https://doi.org/10.1103/PhysRevFluids.7.024301
  7. [7]Mestre, N.J., Russel, W.B.: Low-reynolds-number translation of a slender cylinder near a plane wall. J Eng Math 9, 81–91 (1975) https://doi.org/10.1007/BF01535390
  8. [8]Russel, W.B., Hinch E.J., Leal L.G. and Tieffenbruck G.: Rods falling near a vertical wall. Journal of Fluid Mechanics 83, 273–287 (1977) https://doi.org/10.1017/S00221120770001190
  9. [9]Koens, L., Montenegro-Johnson, T.D.: Local drag of a slender rod parallel to a plane wall in a viscous fluid. Physical Review Fluids 6(6) (2021) https://doi.org/10.1103/PhysRevFluids.6.064101
  10. [10]Chun, B., Jung, H.W.: Near-wall dynamics of non-Brownian rod-like particle suspensions in confined shear flows. Journal of Fluid Mechanics 1009, 51 (2025) https://doi.org/10.1017/jfm.2025.225
  11. [11]Noichl, I., Schönecker, C.: Inertial forces and elastohydrodynamic interaction of spherical particles in wall-bounded sedimentation experiments at low $$Re_P$$. The European Physical Journal E 49(7), 51 (2026) https://doi.org/10.1140/epje/s10189-026-00594-8
  12. [12]Cox, R.G.: The motion of long slender bodies in a viscous fluid part 1. general theory. Journal of Fluid Mechanics 44(4), 791–810 (1970) https://doi.org/10.1017/S0022112070000215X
  13. [13]Joseph B. Keller and Sol I. Rubinow: Slender-body theory for slow viscous flow. Journal of Fluid Mechanics 75(4), 705–714 (1976) https://doi.org/10.1017/S0022112076000475
  14. [14]Brenner, H.: Effect of finite boundaries on the stokes resistance of an arbitrary particle. Journal of Fluid Mechanics 12(1), 35–48 (1962) https://doi.org/10.1017/S0022112062000026
  15. [15]Ui, T.J., Hussey, R.G., Roger, R.P.: Stokes drag on a cylinder in axial motion. The Physics of Fluids 27(4), 787–795 (1984) https://doi.org/10.1063/1.864706
  16. [16]Happel, J., Bart, E.: The settling of a sphere along the axis of a long square duct at low reynolds’ number (29), 241–258 (1974)DOI
  17. [17]Faxén, H.: Der widerstand gegen die bewegung einer starren kugel in einer zähen flüssigkeit, die zwischen zwei parallelen ebenen wänden eingeschlossen ist. Annalen der Physik (373), 89–119 (1922)DOI
  18. [18]Tashibana, M., Kitasho, K.: The motion of a falling sphere in a viscous fluid and the effects of side walls. Memoirs of the Faculty of Engineering Fukui University 24(2) (1976)
  19. [19]Noichl, I., Schönecker, C.: Dynamics of elastic, nonheavy spheres sedimenting in a rectangular duct. Soft Matter 18(12), 2462–2472 (2022) https://doi.org/10.1039/D1SM01789F
  20. [20]Noichl, I.: Elastohydrodynamic interaction of flowing particles with surfaces. PhD thesis, Rheinland-Pfälzische Technische Universität Kaiserslautern-Landau (2026). https://doi.org/10.26204/KLUEDO/9525

Paper details

vlnc:W_pj9p5djs
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

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