betaFeedback

WorksPhysics and Astronomy

Enucleated incompressible red blood cells in shear flow: theoretical analysis of shape instabilities

Abstract

Red blood cells (RBCs) are essential for oxygen transport, and their remarkable ability to undergo significant deformations during flow is a crucial feature for their physiological function. At intermediate shear rates typical of the microcirculation, RBCs can adopt complex, multi-lobed shapes, signifying a dynamic instability. Here we adopt a perturbative theoretical framework of a quasi-spherical RBC under external shear flow to study such shape instabilities. To better capture RBC maturation and enucleation, we first extend the framework to explicitly account for different excess areas between the stress-free and current membrane shapes. We revisit the reduced equations of motion obtained for an ellipsoidally-shaped RBC, and demonstrate the effect of different excess areas and initial orientation on the dynamical trajectories. Then, we introduce additional spatial modes and show that an emerging instability critically depends on the RBC's shear and bending moduli, the internal to external viscosity ratio, and the excess area, mainly through the RBC's membrane tension. We also study the instability-induced saturation of the membrane tension, and the resulting excess area redistribution at long times. The theoretical framework and the emerging picture of the different instabilities provide insights into the emergence of stomatocyte and trilobe shapes exhibited by RBCs under external flow.

Introduction

Red blood cells (RBCs) are the primary cellular constituents of blood and are fundamental to the maintenance of healthy circulation. Due to their abundance and deformability, RBCs play a central role in the rheological properties of blood, enabling efficient flow and oxygen transport through the body’s vasculature [1, 2, 3]. RBCs mature within the bone marrow, undergoing a critical process of enucleation where the cell nucleus is expelled [4, 5, 6]. Once released into the blood stream, RBCs typically adopt a biconcave discocyte shape and form a suspension of deformable particles [7]. This suspension allows proper transport of oxygen and nutrients and is vital for physiological function [8, 9].

The dynamics of individual RBCs have long been a subject of study not only for their physiological relevance but also for their rich dynamics, serving as a model system for soft matter under flow [9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23]. When subjected to shear flows, RBCs display a variety of behaviors: RBCs can exhibit rigid-body-like tumbling [9, 10], where the entire cell rotates; tank treading, where the membrane continuously circulates around a fixed interior [11, 12]; or swinging, characterized by oscillations around a steady orientation [13]. These diverse dynamics motivated the development of several theoretical frameworks. Since RBCs largely maintain their overall biconcave discocyte shape during such simple motions, the models treated RBCs as rigid or nearly rigid objects [14, 15, 16]. These rigid-body theories, including those based on the classical Jeffery orbits, successfully predicted the transitions between tumbling, tank-treading, and swinging dynamics [10, 13, 17, 18, 23].

Recently, experiments reported a wealth of complex RBC shapes emerging at intermediate shear rates that cannot be explained by rigid-body models. Specifically, unusual morphologies such as the symmetry-broken stomatocytes and various multi-lobed shapes were observed under external shear rates, typically ranging from 100 to 1000 s−1^{-1} [3, 12]. Although several numerical studies explored the deformation of RBCs under flow [24, 25], the underlying physical mechanisms driving these shape changes remain debated. Specifically, as multi-lobed shapes were observed in both 3D RBC or 2D vesicle simulations, it is unclear what roles are played by elasticity, bending, membrane tension, excess area, and the internal and external viscosities, on initiating shape instabilities, and how they affect the emerging morphologies [3, 12, 24, 25].

Here, we theoretically explore the emergence of different RBC shapes under shear flow. To this end, we revisit a theoretical framework that fully accounts for membrane mechanics and deformations, and describes the RBC’s spatio-temporal shape dynamics, enabling an analytical investigation of RBC stability in the quasi-spherical shape limit [26, 27, 28, 29, 30, 31]: Vlahovska et. al. [26] presented the linearization of the full, non-linear framework using a perturbative expansion, derived the equations of motion for the leading-order components, analyzed a simplified case of an ellipsoidal RBC and demonstrated dynamics of different, well-known scenarios (e.g., capsules and vesicles). By incorporating different excess areas for the stress-free and current shapes — inspired by RBC’s enucleation during maturation — and by including additional spatial modes, we probe the emerging RBC dynamics, instabilities, and the emerging shape configurations at long times, for a wide range of material parameters. This allows us to study the interplay between membrane tension, elasticity, and bending resistance. As membrane tension is the only mechanism that couples the spatial modes at leading order in our framework, the present study examines the role of membrane tension as the key driver of RBC shape instabilities.

We begin by stating the governing equations, dimensionless numbers, and parameterizations of the problem in Sec. II. We revisit the equations of motion of the reduced system, considering only the ellipsoidal deformation modes with different stress-free and current excess areas, and demonstrate the resulting dynamics in Sec. III. In Sec. IV, we introduce additional spatial modes to develop a linear instability criterion, whose validity and resulting phase diagram are verified numerically. Then, in § 1.4, we analyze the long-time behavior, providing an analytical approximation for the excess area distribution after an instability occurs. We conclude by discussing the emerging physical picture and its implications in Sec. VI.

Mathematical framework

We treat the RBC as a closed, fluid-filled membrane suspended in an externally driven ambient fluid [26, 27, 28, 29, 30, 31]. Both internal and external fluids are assumed to be incompressible,

∇⋅vex/in=0(1)\nabla\cdot\mathbf{v}^{\mathrm{ex/in}} = 0 \tag*{(1)}

where vex/in\mathbf{v}^{\mathrm{ex/in}} denote the external and internal velocity fields respectively.

The momentum balance equation is greatly simplified, as the RBC’s characteristic size leads to a low Reynolds number, and we neglect inertial contributions. The fluids are assumed to be Newtonian, resulting in Stokes equations:

∇⋅Tex/in=−∇pex/in+ηex/in∇2vex/in=0(2)\nabla\cdot\mathbf{T}^{\mathrm{ex/in}} = -\nabla p^{\mathrm{ex/in}} + \eta^{\mathrm{ex/in}}\nabla^{2}\mathbf{v}^{\mathrm{ex/in}} = 0 \tag*{(2)}

where Tex/in\mathbf{T}^{\mathrm{ex/in}} are the stress tensors, ηex/in\eta^{\mathrm{ex/in}} are the external and internal viscosities, and pex/inp^{\mathrm{ex/in}} are the external and internal pressures, respectively.

The presence of the membrane imposes two essential sets of boundary conditions at the interface. First, the fluid velocities vex/in\mathbf{v}^{\mathrm{ex/in}} evaluated on the membrane boundary ∙∣m\bullet|_{\mathrm{m}} are assumed to be continuous across the membrane, and match the membrane velocity vm\mathbf{v}^{\mathrm{m}}. Additionally, the forces exerted by the fluids balance with the membrane force density τm\boldsymbol{\tau}^{\mathrm{m}}. These two conditions are cast as

vex∣m=vin∣m=vm,n⋅(Tex−Tin)∣m=τm(3)\left.\mathbf{v}^{\mathrm{ex}}\right|_{\mathrm{m}}=\left.\mathbf{v}^{\mathrm{in}}\right|_{\mathrm{m}}=\mathbf{v}^{\mathrm{m}}, \qquad\left.\mathbf{n}\cdot\left(\mathbf{T}^{\mathrm{ex}}-\mathbf{T}^{\mathrm{in}}\right)\right|_{\mathrm{m}}=\boldsymbol{\tau}^{\mathrm{m}} \tag*{(3)}

where n\mathbf{n} is the outward unit normal to the membrane.

To proceed, we need to specify the contributions to the RBC membrane force density τm\boldsymbol{\tau}^{\mathrm{m}}. The RBC’s membrane is a composite structure, consisting of a lipid bilayer connected to a spectrin-based cytoskeletal network [32, 33]. Consequentially, the resulting membrane force density is decomposed into three components: elastic, bending, and membrane tension:

τm=τμ+τκ+τs(4)\boldsymbol{\tau}^{\mathrm{m}}=\boldsymbol{\tau}^{\mu}+\boldsymbol{\tau}^{\kappa}+\boldsymbol{\tau}^{s} \tag*{(4)}

Here we omit the potential contributions that arise due to membrane viscosity (these have been shown to reduce deformations and cause damped oscillatory deformation of the RBC [34]).

The elastic response of the cytoskeletal network is captured using the Skalak constitutive model [35]. Denoting by x\mathbf{x} the current configuration, X\mathbf{X} the stress-free configuration, and d=x−X\mathbf{d}=\mathbf{x}-\mathbf{X} the displacement, the elastic force density can be approximated for small deformations as a two-dimensional equivalent to Hooke’s law [26, 36, 37]

τμ=2(KA−μ)(∇s⋅d)Hn−(KA−μ)∇s∇s⋅d−μ∇s⋅[∇sd⋅Is+Is⋅(∇sd)T](5)\boldsymbol{\tau}^{\mu}=2(K_{A}-\mu)(\nabla_{s}\cdot\mathbf{d})H\mathbf{n}-(K_{A}-\mu)\nabla_{s}\nabla_{s}\cdot\mathbf{d}-\mu\nabla_{s}\cdot\left[\nabla_{s}\mathbf{d}\cdot\mathbf{I}_{s}+\mathbf{I}_{s}\cdot(\nabla_{s}\mathbf{d})^{T}\right] \tag*{(5)}

Here KAK_{A} is the area modulus and μ\mu is the planar shear modulus, H=12∇s⋅nH=\frac{1}{2}\nabla_{s}\cdot\mathbf{n} is the mean curvature, Is=I−nn\mathbf{I}_{s}=\mathbf{I}-\mathbf{n}\mathbf{n} is a projection operator onto the membrane’s surface, and ∇s=Is⋅∇\nabla_{s}=\mathbf{I}_{s}\cdot\nabla is the surface gradient [27, 36].

The resistance of the lipid bilayer to bending is obtained from the Helfrich energy [38, 39], yielding the bending force density τκ\boldsymbol{\tau}^{\kappa} as

τκ=−κ(4H3−4KH+2∇s2H)n(6)\boldsymbol{\tau}^{\kappa}=-\kappa\left(4H^{3}-4KH+2\nabla_{s}^{2}H\right)\mathbf{n} \tag*{(6)}

where κ\kappa is the bending modulus, and KK is the Gaussian curvature.

Finally, the membrane tension contribution is given by [26, 40]

τs=2ΣHn−∇sΣ(7)\boldsymbol{\tau}^{s}=2\Sigma H\mathbf{n}-\nabla_{s}\Sigma \tag*{(7)}

Here the membrane tension Σ\Sigma acts as a Lagrange multiplier enforcing constant local membrane area, and is not a material parameter. This implies Σ\Sigma varies spatially and temporally, and is determined as part of the solution.

A key aspect of our approach is the explicit consideration of the change in RBC geometry due to enucleation. Before the enucleation, the RBC is approximately an ellipsoid-shaped membrane of surface area AA enclosing a volume V0V_0. We define the typical stress-free radius R0R_0 of a sphere of a similar volume, i.e., R0≡(3V04π)1/3R_0 \equiv\left(\frac{3V_0}{4\pi}\right)^{1/3}, and the initial excess area of the ellipsoid shape with respect to the sphere as Δ0≡AR02−4π\Delta_0 \equiv\frac{A}{R_0^2}-4\pi. Then, the RBC undergoes enucleation, maturing into the discocyte shape. The resulting shape is a membrane of the same surface area AA, but enclosing a volume V<V0V<V_0, with an effective radius R=(3V4π)1/3R=\left(\frac{3V}{4\pi}\right)^{1/3}, and excess area of Δ≡AR2−4π\Delta\equiv\frac{A}{R^2}-4\pi; for typical RBC Δ≃4\Delta\simeq4 [26, 41, 42]. As V<V0V<V_0, R<R0R<R_0, and Δ>Δ0\Delta>\Delta_0, the excess area is increased at the cost of locked internal stresses in the RBC membrane. The effect of reduced volume is schematically shown in Figure 1. As we shall see, this increase in excess area alters the RBC response to external flow, and allows instabilities to emerge.

Figure 1. A sketch of the enucleation of RBCs. (a) Before the enucleation the stress-free RBC takes an ellipsoidal shape with total surface area AA, volume V0V_0, and an effective radius R0≡(3V0/4π)1/3R_0 \equiv(3V_0/4\pi)^{1/3}, corresponding to excess area Δ0\Delta_0 (here Δ0=0.01\Delta_0 = 0.01). The stress-free surface is parametrized using ρ0(θ,φ,t)\rho_0(\theta,\varphi,t), where θ\theta is the polar angle, and φ\varphi is the azimuthal angle, and tt is the dimensionless time. (b) After enucleation the RBC acquires a new, discocyte shape with volume V<V0V<V_0 (depicted by the dashed circle), and an effective radius R=(3V/4π)1/3R=(3V/4\pi)^{1/3}, while the total surface area AA is conserved. This current shape is characterized by Δ\Delta (here Δ=1\Delta=1), parametrized by ρ(θ,φ,t)\rho(\theta,\varphi,t), with the outward unit normal to the membrane n\boldsymbol{n}, and is axisymmetric around the zz-axis. Colors heuristically indicate distance from the center of the stress-free and enucleated RBC respectively.

Under an imposed shear flow v∞=γ˙yx^\boldsymbol{v}^{\infty}=\dot{\gamma}y\hat{\boldsymbol{x}} (here x^\hat{\boldsymbol{x}} is a unit vector in the xx-direction, and γ˙\dot{\gamma} is the shear rate), the above description leads to several dimensionless numbers. The internal-to-external viscosity ratio λ≡ηin/ηex\lambda\equiv\eta^{\mathrm{in}}/\eta^{\mathrm{ex}} plays a crucial role in RBC dynamics [8, 13, 42, 43, 44]. For a typical RBC of radius RR, comparing the bending modulus κ\kappa, and the (2D) shear modulus μ\mu, to the typical stresses imposed by the external drive, we define the bending number B\mathcal{B} and the capillary number C\mathcal{C} as

B=ηexγ˙R3κ,C=ηexγ˙Rμ.(8)\mathcal{B}=\frac{\eta^{\mathrm{ex}}\dot{\gamma}R^3}{\kappa}, \qquad\mathcal{C}=\frac{\eta^{\mathrm{ex}}\dot{\gamma}R}{\mu}. \tag*{(8)}

In what follows, we treat the RBC as an area-incompressible membrane, implying ∇s⋅d=0\nabla_s\cdot\boldsymbol{d}=0, rendering the area modulus KAK_A irrelevant.

To analytically explore the membrane dynamics, we employ a perturbative framework, expanding the RBC shape in spherical harmonics [26, 27, 28, 29, 30, 31]. The current RBC membrane ρ(θ,φ,t)\rho(\theta,\varphi,t), and the stress-free shape ρ0(θ,φ,t)\rho_0(\theta,\varphi,t), depicted in Figure 1, are expanded in spherical harmonics Yl,mY_{l,m}'s, as [26]

ρ(θ,φ,t)=1+∑l=2∞∑m=−llYl,m(θ,φ)fl,m(t),(9)\rho(\theta,\varphi,t)=1+\sum_{l=2}^{\infty}\sum_{m=-l}^{l}Y_{l,m}(\theta,\varphi)f_{l,m}(t), \tag*{(9)}
ρ0(θ,φ,t)=1+∑l=2∞∑m=−llYl,m(θ,φ)gl,m(t).(10)\rho_0(\theta,\varphi,t)=1+\sum_{l=2}^{\infty}\sum_{m=-l}^{l}Y_{l,m}(\theta,\varphi)g_{l,m}(t). \tag*{(10)}

Here we used RR and R0R_0, respectively, to rescale ρ\rho and ρ0\rho_0 into this dimensionless form. We assume that, while the RBC has undergone enucleation Δ0<Δ<4π\Delta_0<\Delta<4\pi, the radii ratio is RR0=Δ0+4πΔ+4π≃1\frac{R}{R_0}=\sqrt{\frac{\Delta_0+4\pi}{\Delta+4\pi}}\simeq1. The amplitudes fl,m(t)f_{l,m}(t) and gl,m(t)g_{l,m}(t) describe the current and stress-free shape dynamics (see Appendix § 1.7 for details about the spherical harmonics Yl,mY_{l,m}'s). The summation in Eq. (9) starts at l=2l=2 as we are interested in the RBC’s deformations — the l=0l=0 modes correspond to volumetric modes (which are not allowed as the fluids are incompressible), and l=1l=1 modes correspond to center of mass motion.

The current and stress-free excess areas Δ\Delta and Δ0\Delta_0 are approximated as

Δ=∑l=2∞∑m=−llalfl,m(t)fl,m∗(t)+O(f3),Δ0=∑l=2∞∑m=−llalgl,m(t)gl,m∗(t)+O(g3),(11)\Delta=\sum_{l=2}^{\infty}\sum_{m=-l}^{l}a_l f_{l,m}(t)f_{l,m}^{*}(t)+O(f^3),\qquad\Delta_0=\sum_{l=2}^{\infty}\sum_{m=-l}^{l}a_l g_{l,m}(t)g_{l,m}^{*}(t)+O(g^3), \tag*{(11)}

where al=(l+2)(l−1)/2a_l=(l+2)(l-1)/2, ∗\ast indicates the complex conjugate, and we emphasize that while fl,m(t)f_{l,m}(t) and gl,m(t)g_{l,m}(t) evolve in time, Δ\Delta and Δ0\Delta_0 remain constants. However, in the remainder of the paper, we suppressed the explicit time-dependence from the fl,mf_{l,m}’s and gl,mg_{l,m}’s for readability. These relations explicitly link the dynamic amplitudes fl,mf_{l,m}’s and gl,mg_{l,m}’s to their respective excess areas Δ\Delta and Δ0\Delta_0. Finally, the membrane tension is also expanded in a similar manner as

Σ(θ,φ,t)=σ(t)+∑l=2∞∑m=−llYl,m(θ,φ)σl,m(t),(12)\Sigma(\theta,\varphi,t)=\sigma(t)+\sum_{l=2}^{\infty}\sum_{m=-l}^{l}Y_{l,m}(\theta,\varphi)\sigma_{l,m}(t), \tag*{(12)}

where σ(t)\sigma(t) is the spatially homogeneous part of the membrane tension and σl,m(t)\sigma_{l,m}(t)’s are the time-dependent spatial projections of the membrane tension on the respective spherical harmonics Yl,mY_{l,m}’s.

The equations of motion are obtained by substituting the perturbative expansions of ρ\rho, ρ0\rho_0, and Σ\Sigma, together with the expressions for Δ\Delta and Δ0\Delta_0, Eqs. (9)–(12), into the fluid dynamics Eqs. (1)–(3) and the membrane constitutive relations of τμ\tau^\mu, τκ\tau^\kappa, and τs\tau^s, Eqs. (4)–(7), and collecting leading-order terms [26]. This results in a set of coupled equations for the spherical harmonic amplitudes fl,mf_{l,m}’s [26]:

∂tfl,m=[iωm+Υl(t)]fl,m+Sl,m(t),(13)\partial_t f_{l,m}=[i\omega m+\Upsilon_l(t)]f_{l,m}+S_{l,m}(t), \tag*{(13)}

where the equations are cast in dimensionless time tt (using the shear rate γ˙−1\dot{\gamma}^{-1} as the time scale), ω=12\omega=\frac{1}{2} is the rotational frequency, and Υl(t)\Upsilon_l(t) and Sl,m(t)S_{l,m}(t) include the forces originating from both the fluid flow and membrane response as

Υl(t)=B−1[Γl1+σ(t)Γl2]+C−1Γl3,(14)\Upsilon_l(t)=\mathcal{B}^{-1}\left[\Gamma_l^1+\sigma(t)\Gamma_l^2\right]+\mathcal{C}^{-1}\Gamma_l^3, \tag*{(14)}
Sl,m(t)=Cl,m−C−1Γl3gl,m.(15)S_{l,m}(t)=C_{l,m}-\mathcal{C}^{-1}\Gamma_l^3g_{l,m}. \tag*{(15)}

Here Γl1,2,3\Gamma_l^{1,2,3} are rational functions solely dependent on ll and λ\lambda, σ(t)\sigma(t) is the spatially homogeneous part of the membrane tension Σ\Sigma [see Eq. (11)], and Cl,mC_{l,m}’s arise from the viscous forces (more information is provided in Appendix § 1.8). The dynamics of the stress-free shape amplitudes gl,mg_{l,m}’s are assumed to simply rotate with the external flow [26, 36], dynamically evolving according to

∂tgl,m=iωmgl,m.(16)\partial_tg_{l,m}=i\omega m g_{l,m}. \tag*{(16)}

The system of equations Eqs. (13)–(16) is closed by specifying the dynamics of the spatially homogeneous part of the membrane tension σ(t)\sigma(t). As mentioned, Σ\Sigma acts as a Lagrange multiplier and dynamically evolves to enforce a constant excess area Δ\Delta. By enforcing ∂tΔ=0\partial_t\Delta=0, σ(t)\sigma(t) is expressed as [26]

σ(t)=−B∑l=2∞∑m=−llΓl2alfl,mfl,m∗[∑l=2∞∑m=−llal(B−1Γl1fl,m+C−1Γl3gl,m+Sl,m)fl,m∗].(17)\sigma(t)=-\frac{\mathcal{B}}{\sum_{l=2}^{\infty}\sum_{m=-l}^{l}\Gamma_l^2a_lf_{l,m}f_{l,m}^{*}}\left[\sum_{l=2}^{\infty}\sum_{m=-l}^{l}a_l\left(\mathcal{B}^{-1}\Gamma_l^1f_{l,m}+\mathcal{C}^{-1}\Gamma_l^3g_{l,m}+S_{l,m}\right)f_{l,m}^{*}\right]. \tag*{(17)}

This coupled system, Eqs. (13)–(17), allows us to explore how the dimensionless numbers B\mathcal{B}, C\mathcal{C} and λ\lambda, as well as the stress-free and enucleated excess areas, Δ0\Delta_0 and Δ\Delta, influence the system’s stability, and eventually its long-time behaviors. While the perturbative approach assumes Δ,Δ0≪1\Delta,\Delta_0\ll1, we demonstrate the effects below for modest Δ\Delta values — doing so allows us to emphasize the emerging physics, and to adopt values closer to previous experiments and simulations.

Unperturbed dynamics

Interestingly, the system of Eqs. (13)–(17) contains a single essential nonlinear contribution — the coupling of different fl,mf_{l,m}’s through the membrane tension σ(t)\sigma(t). Before analyzing the system’s dynamics for multiple ll values, it is instructive to examine the dynamics in the l=2l=2 subspace only. This reduction, treating the RBC as an ellipsoid, was shown to capture fundamental RBC dynamic regimes such as tumbling, tank-treading, and swinging or vacillating-breathing [26]. We demonstrate that the emerging dynamics are determined not only by the membrane elasticity and external flow rate, but also by the reference excess area Δ0\Delta_0.

We simplify the initial and stress-free structures by assuming they are axially symmetric, concentrating all the excess area in the (l,m)=(2,0)(l,m)=(2,0) components of the RBC’s stress-free and enucleated structures. To examine the effects of initial orientation, we reorient this initial structure by applying the Wigner D matrix [45] (a representation of the 3D rotation operator), Dm,0(l)(β)D^{(l)}_{m,0}(\beta) where β\beta is the rotation angle (see Appendix § 1.7 for details). That is, we set f2,m(0)=−Δ2a2Dm,0(2)(β)f_{2,m}(0)=-\sqrt{\frac{\Delta}{2a_2}}D^{(2)}_{m,0}(\beta), and g2,m(0)=−Δ02a2Dm,0(2)(β)g_{2,m}(0)=-\sqrt{\frac{\Delta_0}{2a_2}}D^{(2)}_{m,0}(\beta). As D2,0(2)=1232sin⁡2βD^{(2)}_{2,0}=\frac{1}{2}\sqrt{\frac{3}{2}}\sin^2\beta, the resulting initial conditions are symmetric for f2,±2(0)f_{2,\pm2}(0) and g2,±2(0)g_{2,\pm2}(0). In what follows, we focus on two extreme cases: β=0\beta=0 corresponding to the RBC’s symmetry axis aligned perpendicular to the flow plane, which is an out-of-plane orientation (OP), and β=π/2\beta=\pi/2 corresponding to an in-plane (IP) symmetry axis, as shown in the inset of Figure 2(b). For these two values the m=±1m=\pm1 components vanish.

Figure 2. Reduced system dynamics with different excess areas. The dynamics of rr in (a) and Ψ\Psi in (b) for C=0.04\mathcal{C}=0.04, λ=5\lambda=5, and Δ0=0.01\Delta_0=0.01, for β=0.001\beta=0.001 and β=π/2\beta=\pi/2, and for different Δ\Delta values Δ=0.01,1\Delta=0.01,1 (corresponding to S≃2S\simeq2, and Λ≃0.19,1.9\Lambda\simeq0.19,1.9 respectively). At long times, the β=0.001\beta=0.001 dynamics saturate rr to a fixed value, such that the RBC is elongated and oriented at a fixed angle Ψ\Psi with the external flow. Even though SS is not very large, the dynamics converge to the steady states of Eq. (20), as the OP orientation is rotationally symmetric in the shear plane. The β=π/2\beta=\pi/2 exhibits oscillations — for Δ=0.01\Delta=0.01 the dynamics converge to oscillations around a fixed angle, characteristic of swinging. For Δ=1\Delta=1 the rr oscillations are accompanied by a continuous rotation, visible in the decrease of Ψ\Psi, which is equivalent to an RBC that deforms and tumbles. Here the solution deviates from the steady-state dynamics of Eq. (20), as SS is insufficiently large, and the dynamics are presented over short times. Insets: the OP orientation in blue and the IP orientation in red (here we use Δ=3\Delta=3 to emphasize the differences).

We analytically reduce the equations of motion by parameterizing f2,±2f_{2,\pm2}, f2,0f_{2,0}, and g2,±2g_{2,\pm2}, g2,0g_{2,0} as [26, 28]

f2,±2(t)=Δ2a2r(t)e±2iψ(t),f2,0(t)=−Δa2[1−r(t)2],(18)\begin{aligned} f_{2,\pm2}(t)&=\sqrt{\frac{\Delta}{2a_2}}r(t)e^{\pm2i\psi(t)}, \qquad f_{2,0}(t)=-\sqrt{\frac{\Delta}{a_2}}\left[1-r(t)^2\right], \tag*{(18)} \end{aligned}
g2,±2(t)=Δ02a2r0(t)e±2iϕ(t),g2,0(t)=−Δ0a2[1−r0(t)2],(19)\begin{aligned} g_{2,\pm2}(t)&=\sqrt{\frac{\Delta_0}{2a_2}}r_0(t)e^{\pm2i\phi(t)}, \qquad g_{2,0}(t)=-\sqrt{\frac{\Delta_0}{a_2}}\left[1-r_0(t)^2\right], \tag*{(19)} \end{aligned}

where the variables r(t)r(t) and r0(t)r_0(t) represent, respectively, the deformation amplitudes in the shear plane of the current and stress-free shapes, while the angles ψ(t)\psi(t) and ϕ(t)\phi(t) represent the respective in-plane orientations. The f2,0f_{2,0} and g2,0g_{2,0} components are obtained by the conservation of the excess areas Δ\Delta and Δ0\Delta_0, respectively. Finally, we chose f2,0f_{2,0} and g2,0g_{2,0} with a negative sign consistent with the emerging oblate shape of RBCs.

Substituting these parameterizations into the membrane tension σ(t)\sigma(t) of Eq. (17) yields the reduced membrane tension σR(t)\sigma_R(t) as

σR(t)=−6+30πB3Δrsin⁡Ψ+2B3CΔ0Δ[r0rcos⁡(Φ−Ψ)+(r02−1)(r2−1)−ΔΔ0],(20)\sigma_R(t)=-6+\frac{\sqrt{30}\pi B}{3\sqrt{\Delta}}r\sin\Psi+\frac{2B}{3C}\sqrt{\frac{\Delta_0}{\Delta}}\left[r_0r\cos(\Phi-\Psi)+\sqrt{(r_0^2-1)(r^2-1)}-\sqrt{\frac{\Delta}{\Delta_0}}\right], \tag*{(20)}

where we defined Ψ≡2ψ\Psi\equiv2\psi and Φ≡2ϕ\Phi\equiv2\phi, and suppressed the explicit time dependence from rr, r0r_0, Ψ\Psi and Φ\Phi for readability. Inserting the parameterizations Eq. (16) together with Eq. (20) into the equations of motion Eqs. (13)-(16) yields reduced dynamical equations for rr, Ψ\Psi and Φ\Phi as [26]

r′=Λ−1(1−r2)sin⁡Ψ+(ΛS)−1[r0(1−r2)cos⁡(Φ−Ψ)−r(1−r02)(1−r2)],(21)r'=\Lambda^{-1}(1-r^2)\sin\Psi+(\Lambda S)^{-1}\left[r_0(1-r^2)\cos(\Phi-\Psi)-r\sqrt{(1-r_0^2)(1-r^2)}\right], \tag*{(21)}
Ψ′=−1+Λ−1cos⁡Ψr+(ΛS)−1r0rsin⁡(Φ−Ψ),(22)\Psi'=-1+\Lambda^{-1}\frac{\cos\Psi}{r}+(\Lambda S)^{-1}\frac{r_0}{r}\sin(\Phi-\Psi), \tag*{(22)}
Φ′=−1,(23)\Phi'=-1, \tag*{(23)}

where ′≡ddt\prime\equiv\frac{d}{dt} is the time-derivative operator. The dynamics are governed by two composite dimensionless parameters

Λ=Δ(23λ+32)830π,S=C15π2Δ0,(24)\Lambda=\frac{\sqrt{\Delta}(23\lambda+32)}{8\sqrt{30}\pi}, \qquad S=C\sqrt{\frac{15\pi}{2\Delta_0}}, \tag*{(24)}

and the initial value r0r_0. Clearly, this reduced system depends on the combined effects of the viscosity ratio λ\lambda, the capillary number CC, and the excess areas Δ\Delta and Δ0\Delta_0. The equations above are fully determined once Λ\Lambda, SS and r0r_0 are specified together with the initial condition characterized by β\beta. Interestingly, while the bending number BB appears in the reduced membrane tension Eq. (20), it is absent from the equations of motion of the reduced system, Eq. (18) — this is due to the fact that σR\sigma_R is determined only using the l=2l=2 components. Previous analysis [26] demonstrated Eq. (20) can exhibit various dynamical behaviors, e.g., fluid membrane and capsule dynamics. As recent experimental evidence suggests a shape instability at intermediate and large shear rates, we are interested in the solutions for large CC, and consequently large SS values. In the limit S≫1S\gg1, the leading order steady-state solutions are obtained as [26]

rss−=1,Ψss−=arccos⁡Λ,for Λ<1,(25)\begin{aligned} r_{ss}^{-}&=1, \qquad\Psi_{ss}^{-}=\arccos\Lambda, \qquad\text{for }\Lambda<1, \tag*{(25)} \end{aligned}
rss+=Λ−1,Ψss+=0,for Λ>1,(26)\begin{aligned} r_{ss}^{+}&=\Lambda^{-1}, \qquad\Psi_{ss}^{+}=0, \qquad\text{for }\Lambda>1, \tag*{(26)} \end{aligned}

where we used the notation ∙ss−\bullet_{\mathrm{ss}}^{-} for the Λ<1\Lambda<1 steady-state solutions, and ∙ss+\bullet_{\mathrm{ss}}^{+} for the Λ>1\Lambda>1 ones.

Unlike in previous analysis [26], SS defined in Eq. (24) incorporates the reference shape excess area Δ0\Delta_0, while Λ\Lambda depends on Δ\Delta. SS then becomes the effective elastic constant, quantifying the ability of the shear flow to overcome the energy barrier associated with the tank-treading of the heterogeneous membrane [18, 23]. This modified definition of SS implies that a dynamical transition can occur either by increasing the shear rate through C\mathcal{C}, or by decreasing Δ0\Delta_0. In the latter case, the reference shape gets close to a sphere, reducing the heterogeneity of the reference shape, which facilitates its circulation and favors tank-treading, consistent with previous studies [13, 18, 23, 33]. Clearly, such a dynamical transition is unavailable for vesicles, as their fluidized membrane does not hold memory of a previous state. In our framework, this corresponds to either Δ0→0\Delta_0\to0 or C→∞\mathcal{C}\to\infty (see Appendix C).

We briefly examine the dynamics of Eq. (18) as a function of the initial orientations and the excess areas Δ\Delta and Δ0\Delta_0. The initial orientation is controlled by the angle β\beta and results in different values of the initial deformation amplitude r(0)r(0). We assume that the initial orientation of the current shape and the stress-free shape are the same. Ideally, we would initialize an OP RBC (with β=0\beta=0), and then apply rotations with D±2,0(2)(β)D_{\pm2,0}^{(2)}(\beta) to both f2,mf_{2,m}’s and g2,mg_{2,m}’s to explore the role of initial orientations. Practically, setting the stress-free and initial structures in an OP orientation (β=0\beta=0) results in r=0r=0, which causes a singularity of Eq. (18). To avoid this singularity, we impose a small initial rotation β≃0\beta\simeq0 (rendering the f2,±1f_{2,\pm1} components negligible). To obtain the IP case we apply D±2,0(2)(π/2)D_{\pm2,0}^{(2)}(\pi/2) to f2,mf_{2,m}’s and g2,mg_{2,m}’s (yielding f2,±2=−123Δ2Δ0f_{2,\pm2}=-\frac{1}{2}\sqrt{\frac{3\Delta}{2\Delta_0}}, f2,0=12ΔΔ0f_{2,0}=\frac{1}{2}\sqrt{\frac{\Delta}{\Delta_0}} and similar values for g2,±2g_{2,\pm2} and g2,0g_{2,0} upon replacing Δ\Delta with Δ0\Delta_0, giving r(0)=r0(0)=3/2r(0)=r_0(0)=\sqrt{3}/2, and Ψ(0)=Φ(0)=π\Psi(0)=\Phi(0)=\pi).

We analyze the dynamics for different excess areas Δ\Delta in Figure 2, fixing λ=5\lambda=5, a typical value for physiological conditions [3, 46]. We probed different Δ\Delta values to demonstrate the effect of internal stresses, using Δ=Δ0=0.01\Delta=\Delta_0=0.01 as a reference, non-stressed case. For both the OP β≃0\beta\simeq0 and IP β=π/2\beta=\pi/2 orientations, increasing Δ\Delta lowers the rr value at long times, indicating a less elongated particle in the shear plane. In the OP case of β≃0\beta\simeq0, the emerging RBC dynamics correspond to RBC alignment along the vorticity axis with a fixed deformation and orientation. This motion is actually accompanied by the tank treading of the membrane. This tank-treading motion differs from the classical case [11] associated with an IP orientation. In the present case, the motion can be interpreted as a deformed form of rolling [10]: the RBC’s short axis is aligned with the vorticity axis, the membrane tank-treads around the cell’s short axis, while shear stresses deform the cell within the shear plane [47] and tend to align it with the flow direction (Ψ\Psi decreases), as interpreted from the blue curves in Figure 2(a)-(Figure 2b). In the IP case of β=π/2\beta=\pi/2, the heterogeneity of the reference shape makes the particle shape and angle oscillate over time [13]. Increasing Δ\Delta promotes a transition from that swinging motion to tumbling, visible in Figure 2(b). As changing Δ\Delta amounts to changing Λ\Lambda, similar transitions have been observed for Δ=Δ0\Delta=\Delta_0 upon increasing λ\lambda [15, 26, 30].

The enucleation process allows the RBCs to lock stresses within their membranes, manifested by an excess area Δ\Delta different from their reference value Δ0\Delta_0. The effective elasticity constant SS is then renormalized, taking into account the reference excess area Δ0\Delta_0 instead of Δ\Delta. Additional examples of the dynamics available for the system described above are available in Appendix C. While the reduced system successfully captures various RBC dynamics, it is inherently limited to the ellipsoidal modes, and hence cannot provide insight into the possible reshaping of RBCs. This motivates us to include and analyze the dynamics in the presence of higher-order spherical harmonics. In what follows we demonstrate that enucleation, and the mismatch Δ≠Δ0\Delta\ne\Delta_0, not only modify the unperturbed dynamics, but also allow for shape instabilities to emerge.

Perturbations and instabilities

Understanding the reduced system’s dynamics, we now return to the full perturbative formalism of Eqs. (13)–(17), including contributions from higher-order modes, l>2l>2. In this section, we start by deriving the stability criterion to clarify the role of the different parameters of the problem. We then use the reduced dynamics to obtain quantitative instability predictions, which we then verify numerically by integrating Eqs. (13)–(17).

According to Eqs. (13) and (13), an instability is expected once the real part of the growth rate becomes positive ℜ[Υl]>0\Re[\Upsilon_l]>0 for any mode l>2l>2. Using the expression for Υl\Upsilon_l in Eq. (13a), we can recast this set of instability criteria per ll-mode, as a simplified condition on the instantaneous membrane tension

Π2+σ(t)Π+4α<0 ,(27)\Pi^2+\sigma(t)\Pi+4\alpha<0\,, \tag*{(27)}

where we defined Π≡l(l+1)\Pi\equiv l(l+1) for convenience. The above simplified criterion also introduces the dimensionless ratio between the in-plane shear elasticity to the bending stiffness, α≡B/C=μR2/κ\alpha\equiv B/C=\mu R^2/\kappa, reminiscent of the Föppl–von Kármán number [48, 49], but for composite membranes. An instability occurs once the bending and elastic responses are insufficient to counter the forces induced by the external flow and internal stresses. A similar instability criterion was obtained for vesicles under external drive [50].

When σ(t)\sigma(t) is sufficiently negative, the system is unstable. We treat ll as a continuous variable, resulting in unstable, excited ll-modes existing between l±=12(−1+1−2σ±2σ2−16α)l_\pm=\frac{1}{2}\left(-1+\sqrt{1-2\sigma\pm2\sqrt{\sigma^2-16\alpha}}\right), with a maximum at lc=12(−1+1−2σ)l_c=\frac{1}{2}\left(-1+\sqrt{1-2\sigma}\right). At the critical point, we set the inequality of Eq. (27) to equality, allowing us to isolate σ(t)=−4α+Π2Π\sigma(t)=-\frac{4\alpha+\Pi^2}{\Pi}. The maximal ll value then yields l∗=12(−1+1+8α)l_*=\frac{1}{2}\left(-1+\sqrt{1+8\sqrt{\alpha}}\right), and the corresponding σ∗=−4α\sigma_*=-4\sqrt{\alpha}. Importantly, as physically only integer ll values are allowed, even if the instability criterion is met, we need to ensure it also holds for a subset of integer values within l−≤l≤l+l_-\le l\le l_+.

The theoretical criterion allows us to predict whether a certain parameter combination is stable or not, and also which modes are expected to grow the most. An instability is initiated when σ<σ∗\sigma<\sigma_*, corresponding to the solid curve in Figure 3a. The growth of the excited modes causes an increase in σ\sigma towards σ∗\sigma_* that stabilizes the unstable modes. This increase in σ\sigma pushes the stability curve, from the unstable curve [corresponding to the solid line in Figure 3a], towards the marginally-stable curve [corresponding to the dashed line in Figure 3a], as is shown schematically by the bold arrow in Figure 3a. As the dynamics evolve, some of the modes are stabilized, and eventually the modes that are unstable for the longest times correspond to l∗l_*. Finally, as the instability criterion differs from the growth rates Υl\Upsilon_l by an ll-dependent factor, lcl_c is not necessarily the fastest-growing mode. We schematically visualize lcl_c, l∗l_*, l±l_\pm, the instability criterion σ(t)=σ∗\sigma(t)=\sigma_*, and the rates BΥl\mathcal{B}\Upsilon_l in Figure 3a, showing explicitly a possible deviations of lcl_c from the instantaneous fastest growing mode.

To induce a shape change, the excess area needs to be redistributed between l=2l=2 and other ll’s. We demand that the instability occurs for l≥3l\ge3, which implies that l+>3l_+>3. We simplify this demand by assuming that the unstable range in ll is approximately symmetric, which allows us to check whether l∗≥2.5l_*\ge2.5 instead. As this condition is time independent and depends solely on α\alpha, it implies α≥(35/8)2\alpha\ge(35/8)^2.

Predicting the instability onset criteria analytically is challenging, because the equations of motion are nonlinearly coupled through σ(t)\sigma(t). To advance, we utilize our assumption that the initial RBC shape is an ellipsoid, so the RBC’s excess area is initially concentrated in l=2l=2 (except from potential perturbations), allowing us to approximate σ(t)≃σR(t)\sigma(t)\simeq\sigma_R(t) of Eq. (20). As long as the dynamics are stable, the only significant shape amplitudes remain within the l=2l=2 subspace, and the system evolves according to the reduced description of Eq. (18). For an instability to occur σ(t)≃σR(t)<σ∗\sigma(t)\simeq\sigma_R(t)<\sigma_*. In general, this non-trivial evolution links BB, CC, Δ0\Delta_0, Δ\Delta and λ\lambda, as well as the initial conditions. To gain some intuition for this criterion, we can estimate σ(t)\sigma(t) by evaluating σR(t)\sigma_R(t) at convenient states — the initial conditions or the long-time behaviors for Λ<1\Lambda<1 and Λ>1\Lambda>1 of Eq. (20).

We first use the OP orientation and set r,r0→0r,r_0\to0 in Eq. (20), which render Ψ\Psi and Φ\Phi irrelevant. The initial value of σR(t)\sigma_R(t) is approximated as σRin≃−6+23α(Δ0/Δ−1)\sigma_R^{\mathrm{in}}\simeq-6+\frac{2}{3}\alpha\left(\sqrt{\Delta_0/\Delta}-1\right), and demanding that this evaluates to σ∗\sigma_* yields a critical ratio

Linear instability criterion

(a)

Dynamics of excess area modes for $\beta=0.001$

(b)

Dynamics of excess area modes for $\beta=\pi/2$

(c)

Membrane tension dynamics for the OP case

(d)

Period-averaged membrane tension dynamics

(e)

Figure 3. Linear instability criteria and dynamics. (a) The simplified criterion −Π2−σ(t)Π−4α>0-\Pi^2-\sigma(t)\Pi-4\alpha>0 [note the negative sign compared to Eq. (27)], evaluated for α=25\alpha=25, at the critical stress σ(t)=σ∗≡−4α\sigma(t)=\sigma_\ast\equiv-4\sqrt{\alpha} (denoted by a dashed line), and for σ(t)<σ∗\sigma(t)<\sigma_\ast (denoted by a solid black line). Positive values correspond to unstable modes, while negative values are stable modes. Black points represent l∗l_\ast and lcl_c (connected by a black arrow), as well as the root l+l_+. We also plot in green the corresponding curve for BΥl\mathcal{B}\Upsilon_l for σ(t)<σ∗\sigma(t)<\sigma_\ast, as a function of ll (using the same α\alpha and σ\sigma values, but on a different vertical scale), to demonstrate that while the roots are identical to the simplified expression, lcl_c can differ from the fastest-growing mode (vertical black and green dashed lines serve as a guide to the eye). The bold curved arrow indicates that initially-unstable dynamics cause σ\sigma to increase, and the stability curve dynamically evolves from the solid, unstable curve towards the marginally-stable, dashed curve. (b)-(c) Δl\Delta_l dynamics for C=0.04C=0.04, B=2\mathcal{B}=2, Δ=1\Delta=1, Δ0=0.01\Delta_0=0.01, λ=5\lambda=5 for β=0.001\beta=0.001 in (b) and β=π/2\beta=\pi/2 in (c). In both cases the initial ellipsoid transitions the excess area from Δ2\Delta_2 to Δ3\Delta_3 and Δ4\Delta_4, until l=4l=4 is stabilized, and Δ4\Delta_4 starts decreasing. The emerging dynamics are reminiscent to those of the reduced system — smooth trajectories for β≃0\beta\simeq0 and oscillatory ones for β=π/2\beta=\pi/2 (see Figure 2). (d)-(e) The temporal dynamics of the membrane tension σ(t)\sigma(t) in black, the reduced system’s membrane tension σR(t)\sigma_R(t) in thick gray, and σlc\sigma_l^c evaluated for l=3,4l=3,4. For the OP case in (d), σ(t)\sigma(t) initially follows σR(t)\sigma_R(t) but is below σ3c\sigma_3^c and σ4c\sigma_4^c, which causes both to increase and Δ2\Delta_2 to decrease. Eventually σ>σ4c\sigma>\sigma_4^c, causing Δ4\Delta_4 to decrease. In (e) the period-averaged σ(t)\sigma(t) follows similar dynamics.

α∗init=9[1−(Δ0/Δ)1/4]2.\alpha_\ast^{\mathrm{init}}=\frac{9}{\left[1-\left(\Delta_0/\Delta\right)^{1/4}\right]^2}.

An instability should satisfy both l∗≥2.5l_\ast\geq2.5 [corresponding to α≥(35/8)2\alpha\geq(35/8)^2], and α>α∗init\alpha>\alpha_\ast^{\mathrm{init}}. With Δ0=0.01\Delta_0=0.01 and Δ=1\Delta=1 the two values are similar, and we predict an instability once both of these criteria are met.

In case the underlying reduced dynamics evolves quickly towards their long-time dynamics, we approximate σR(t)\sigma_R(t) in a similar fashion, using the long-time solution for the reduced system of Eq. (18). Using the steady-state approximate solutions Eq. (20) in σR(t)\sigma_R(t), we equate the time-averaged membrane tension ⟨σR⟩\langle\sigma_R\rangle to σ∗\sigma_\ast (as detailed in Appendix D). For the Λ<1\Lambda<1 and Λ>1\Lambda>1 scenarios in the limit of weak elasticity S≫1\mathcal{S}\gg1, we obtain the critical CC values as

C∗−=23(6+BA−)2B,C∗+=19(1−A+)2B,(28)C_\ast^-=\frac{2}{3\left(\sqrt{6}+\sqrt{\mathcal{B}A^-}\right)^2}\mathcal{B},\qquad C_\ast^+=\frac{1}{9}\left(1-\sqrt{A^+}\right)^2\mathcal{B}, \tag*{(28)}

where we defined the auxiliary A−≡1330π(1−Λ2)ΔA^- \equiv\frac{1}{3}\sqrt{\frac{30\pi(1-\Lambda^2)}{\Delta}}, and A+≡Δ0Δ1−Λ2Λ(r02−1)A^+\equiv\sqrt{\frac{\Delta_0}{\Delta}\frac{1-\Lambda^2}{\Lambda}(r_0^2-1)}, respectively. While the Λ>1\Lambda>1 solution depends trivially on B\mathcal{B} (i.e., the instability criterion could be cast in terms of α\alpha), the Λ<1\Lambda<1 case results in a non-trivial critical condition.

To test these predictions, we numerically integrate Eqs. (13)–(17) for l∈[2,10]l\in[2,10], and m∈[−l,l]m\in[-l,l]. We initialize the fl,mf_{l,m}’s and gl,mg_{l,m}’s by rotating a fixed rest structure, as described previously (see also Appendix A). For convenience, we first define the excess area stored in the ll-th subspace as Δl≡al∑m=−llfl,mfl,m∗\Delta_l\equiv a_l\sum_{m=-l}^{l}f_{l,m}f_{l,m}^{\ast}. We choose to start with f2,mf_{2,m}’s and g2,mg_{2,m}’s as used in Sec. III, that is we incorporate Δ2=Δ\Delta_2=\Delta and Δ0\Delta_0 into the axial direction f2,0f_{2,0} and g2,0g_{2,0}, respectively. To test the stability of higher modes, we introduce small initial perturbations of amplitude δ\delta to fl,0f_{l,0} with l>2l>2, such that Δl=δ\Delta_l=\delta for l∈[3,10]l\in[3,10] (here we chose δ=10−3\delta=10^{-3}), while we keep the stress-free shape purely ellipsoidal, setting gl,m=0g_{l,m}=0 for all mm and l>2l>2. By considering such axial perturbations, we assume that the enucleation process preserves the planar symmetry of the RBC but breaks the top-bottom reflection symmetry. To rotate this shape, we apply Dm,0(l)(β)D_{m,0}^{(l)}(\beta) to fl,0f_{l,0} and gl,0g_{l,0} (e.g., set β=10−3\beta=10^{-3} for the OP case, or β=π2\beta=\frac{\pi}{2} for the IP case). We then integrate Eqs. (13)–(17) for t∈[0,T]t\in[0,T] (we take T=500T=500 in what follows).

We show two examples of the dynamics obtained for Δ>Δ0\Delta>\Delta_0 in Figure 3b–(Figure 3e), with similar parameter values as used in Figure 2. Panels (b)-(c) show the dynamics of Δl\Delta_l for l=2l=2–44, for the OP and IP cases respectively. In both cases, Δ2\Delta_2 decreases while Δ3\Delta_3 and Δ4\Delta_4 increase initially. Eventually, Δ4\Delta_4 decreases while Δ3\Delta_3 continues to increase until the system reaches a steady-state or oscillates around such a state. The redistribution of Δ\Delta implies the RBC changes its shape. Panels (d)-(e) show the corresponding membrane tension σ(t)\sigma(t) obtained from the full dynamics in black, as well as the reduced membrane tension σR(t)\sigma_R(t) [obtained from the reduced system, Eq. (20)] in gray. On top of the σ(t)\sigma(t) dynamics, we also plot in dashed lines the instability criterion σlc\sigma_l^c evaluated for different ll values. The initial dynamics of σ(t)\sigma(t) closely follows σR(t)\sigma_R(t) while an instability develops, causing σ(t)\sigma(t) to increase. As σ(t)<σ3c,σ4c\sigma(t)<\sigma_3^c,\sigma_4^c initially, both Δ3\Delta_3 and Δ4\Delta_4 grow. This causes σ(t)\sigma(t) to grow until σ4c<σ(t)\sigma_4^c<\sigma(t), causing Δ4\Delta_4 to decrease while Δ3\Delta_3 keeps increasing (Δ2\Delta_2 decreases to conserve the total excess area). Eventually σ(t)=σ3c\sigma(t)=\sigma_3^c, stabilizing all growth in the system. While the initiation of the instability happens for both OP and IP orientations, the emerging dynamics are rather different; this results from the fact that the instability criterion could be cast in terms of Δl\Delta_l, which is agnostic to the RBC orientation (see Appendix § 1.8).

To validate the predictions of the stability boundary, we focus on the OP configuration, and systematically sample the B−C\mathcal{B}-\mathcal{C} space. For each case, we numerically find which ll mode corresponds to the largest excess area increase, by comparing the initial Δl(0)\Delta_l(0) distribution to the final state Δl(T)\Delta_l(T)’s and finding the ll value corresponding to the largest Δl\Delta_l growth, lΔ=argmaxl[Δl(T)−Δl(0)]l_{\Delta}=\mathop{\rm argmax}_l[\Delta_l(T)-\Delta_l(0)]. We then use lΔl_{\Delta} values to color the emerging (B,C)(\mathcal{B},\mathcal{C}) phase space. We sample such phase spaces for λ=1\lambda=1 and λ=5\lambda=5 for different Δ\Delta values, and plot the results for the OP cases in Figure 4. On top, we overlay the theoretically-predicted α\alpha values that correspond to different l∗l_* values, resulting in the dashed diagonal lines in Figure 4. We also plot the approximate C∗\mathcal{C}_* relevant for the specific cases [Eq. (28)], and α∗init\alpha_*^{\mathrm{init}} when relevant.

Figure 4. Out of plane (OP) stability phase diagrams for different λ\lambda values [panels (a)-(b)], and different Δ\Delta values [panels (b)-(c)]. log⁡B−log⁡C\log\mathcal{B}-\log\mathcal{C} phase diagrams for (a) λ=1\lambda=1, Δ=1\Delta=1 (Λ≃0.7\Lambda\simeq0.7), (b) λ=5\lambda=5, Δ=1\Delta=1 (Λ≃1.9\Lambda\simeq1.9), and (c) λ=5\lambda=5, Δ=0.1\Delta=0.1 (Λ≃0.6\Lambda\simeq0.6) in the OP orientation (Δ0=0.01\Delta_0=0.01 for all cases). Each phase diagram was obtained by sampling (B,C)(\mathcal{B},\mathcal{C}) pairs spanning B∈[exp⁡(−1),exp⁡(6)]\mathcal{B}\in[\exp(-1),\exp(6)], and C∈[exp⁡(−5.5),exp⁡(2)]\mathcal{C}\in[\exp(-5.5),\exp(2)], and simulating each perturbed system for a duration of T=500T=500. Colors indicate the value of lΔl_{\Delta}, signifying the ll value for which the excess area increased the most (and empty regions correspond to stable dynamics with lΔ=2l_{\Delta}=2), as obtained from the numerical simulations (see legend). Solid diagonal black lines show the theoretical predictions for l∗=2.5l_*=2.5 (where we expect l=3l=3 to be excited), and the additional diagonal dashed lines show the theoretically predicted transition from excited ll value to the next one. Red curves show the value of C∗\mathcal{C}_* obtained from the steady-state calculation [Eq. (28), see also Appendix § 1.10]. Finally, the blue diagonal line shows α∗init\alpha_*^{\mathrm{init}} effecting the instability boundary at low B\mathcal{B} and C\mathcal{C} values [αcinit\alpha_c^{\mathrm{init}} is not shown in panels (a) and (b) as it roughly overlaps with the l=3l=3 critical line, as mentioned above]. Increasing the viscosity ratio λ\lambda, and the excess area Δ\Delta [panel (b) contrasted with (a) and (c)] destabilizes RBCs of high C\mathcal{C} (weak elasticity).

The theoretical predictions are verified numerically. Figure 4(a) shows the λ=1\lambda=1, Δ=1\Delta=1 case, corresponding to Λ<1\Lambda<1. Upon introducing the different excess areas we observe an instability in a large region of the (B,C)(\mathcal{B},\mathcal{C}) phase diagram. The critical C∗\mathcal{C}_* value plays a dominant role in stabilizing high C\mathcal{C} cases. This stabilizing effect is diminished as we increase λ\lambda to 5 in panel (b), for which Λ>1\Lambda>1. Finally, to examine the stabilizing effect of α∗init\alpha_*^{\mathrm{init}}, we reduce Δ\Delta to 0.1 (so that α∗init\alpha_*^{\mathrm{init}} deviates from the l∗≥2.5l_*\geq2.5 criterion) in panel (c) — the stabilizing effect of α∗init\alpha_*^{\mathrm{init}} is observed for low B\mathcal{B}-C\mathcal{C} values.

In general, the value of α\alpha corresponds well to the emerging maximally-unstable observed ll, implying that the mechanical ratio α\alpha dictates the emerging excited harmonics. Additionally, comparing the different panels demonstrates clearly that lower viscosity ratios λ\lambda and lower excess areas Δ\Delta both stabilize the dynamics for large C\mathcal{C} values. The stabilizing effects of low λ\lambda values, or lower Δ\Delta values, are in fact expected, as the underlying dynamics is governed by the reduced equations, Eq. (18), which depend on the composite parameter Λ\Lambda. Similar effects occur for the IP case — we discuss and compare those to the OP cases in Appendix § 1.11.

We also probed stability as a function of C\mathcal{C} and λ\lambda (as probed in [24]). We probed the C−λ\mathcal{C}-\lambda phase space for fixed α\alpha values, α=40,100\alpha=40,100 and α=500\alpha=500 for the OP case. As a fixed α\alpha value always induces an instability for the same ll value, we only probed whether the dynamics are stable or not. We also derive an analogue instability criteria λ∗(C)\lambda_*(C) for fixed α\alpha for the Λ<1\Lambda<1 case, as

λ∗=823(30πC2α2/Δ−4(α−3)4Cα−4).(29)\lambda_*=\frac{8}{23}\left(\frac{\sqrt{30\pi C^2\alpha^2/\Delta}-4(\sqrt{\alpha}-3)^4}{C\alpha}-4\right). \tag*{(29)}

In the Λ>1\Lambda>1 case, λ∗\lambda_* is independent of CC (see Appendix § 1.10). We plot the analytical predictions on top of the numerical results for Δ=1\Delta=1 in Figure 5(a) and Δ=0.1\Delta=0.1 in Figure 5(b). As expected, regions of large CC and low λ\lambda values are stable (see Figure 4). This stabilization at high CC and low λ\lambda is reminiscent of the numerically-obtained phase space reported in [24]. However, the current theoretical framework induces instabilities for fixed Λ\Lambda values, which differs from the numerical results of [24], hinting that a different instability mechanism may be relevant for the full numerical simulations.

Figure 5. Out of plane (OP) stability diagrams in log⁡C−λ\log C-\lambda space, obtained for α=40\alpha=40, 100 and 500 for Δ=1\Delta=1 in (a) and for Δ=0.1\Delta=0.1 in (b). Markers represent unstable dynamics for the associated α\alpha value (and empty regions correspond to stable dynamics). We plot the predictions for λ∗(C)\lambda_*(C) of Eq. (29) in both panels for the three α\alpha values (see Appendix § 1.10). Increasing α\alpha pushes the boundary to higher CC values; a similar effect happens for increased Δ\Delta values. Note that in panel (b), α=40\alpha=40 is actually below the α∗init\alpha_*^{\mathrm{init}} criterion, rendering the entire log⁡C−λ\log C-\lambda space stable (we still plot the theoretical prediction λ∗\lambda_*). The log⁡C−λ\log C-\lambda phase diagrams are consistent with the observations from Figure 4 — increasing λ\lambda destabilizes RBCs with large CC.

The observed diagrams imply that RBC become unstable for large α\alpha values, which is an intrinsic structural property. If Λ\Lambda is sufficiently low (either due to a lower viscosity ratio λ\lambda or due to lower excess area Δ\Delta), the structural criterion is further stabilized [as seen in Figure 4(a) and (c)]. Irrespective of the value of Λ\Lambda, the instabilities persist for weak flows associated with low BB and CC values. This implies that even in the presence of weak flows, RBCs can become unstable. This behavior is somewhat counter-intuitive, as in the absence of flow the RBC should be stable (this is an underlying assumption about the initial RBC state). However previous studies of depression of spherical capsules and vesicles have also reported reminiscent phenomena [51, 52]. The derivation of Eqs. (13)–(17) relied on the coupling between the deformations along the RBC membrane, and those normal to it (here depicted as fl,mf_{l,m}’s) through both membrane tension and the surrounding flow [26]. In the absence of flow, the linear elastic deformations induce forces only along the RBC membrane, while bending acts in the normal direction, leaving the membrane tension to balance both. The residual internal forces generated via the enucleation process balance both the forces along the RBC membrane and the normal forces, possibly inducing buckling via local compressive forces. This nontrivial coupling mechanism is inherently nonlinear [51], and is currently missing from the framework — possibly inducing the non-physical instabilities at low CC values.

The analysis above demonstrates that an enucleated RBC can develop substantial negative membrane tension, which drives an instability. Membrane elasticity allows RBCs to hold memory both of their reference shape, here quantified by Δ0\Delta_0, and of their equilibrium shape, via elastic and bending forces that balance the membrane stresses. These two qualities allow RBCs to undergo, and sustain, an instability. A similar instability criterion was obtained for vesicles [50], understandably without the effects of the reference excess area Δ0\Delta_0, and without elasticity. In those cases, a sudden change in the external flow profile yielded negative membrane tension and a wrinkling instability, that was later on stabilized as the vesicle reoriented along the new flow direction [50, 53]. When subjected to an elongation flow, such vesicles could develop pearling instabilities, under sufficiently high strain-rates [54]. The situation here is different — the instability persists, and allows for different modes to grow, resulting in non-trivial shapes, possibly performing oscillatory dynamics. In shear flow, in the absence of elastic forces, and in the absence of a mismatch Δ0≠Δ\Delta_0 \ne\Delta sustaining negative tensions, this type of instability could not be attained.

Post-instability excess area redistribution

Once an instability is initiated, unstable modes grow at the expense of the other modes. This growth causes the membrane tension to increase until, eventually, the system reaches, or oscillates around, a steady state — the excess area must be conserved, so growth cannot persist indefinitely. The membrane tension saturates at ∼σ∗\sim\sigma_*, as shown in Figure 3d-(Figure 3e), implying that all of the modes are now stable. The emerging post-instability shapes potentially correspond to the observed morphologies in experiments and simulations — here we obtain analytical predictions for their spatial characteristics.

Once a steady state is reached, the excess area is mainly distributed between the initial l=2l=2 ellipsoidal modes, and l=l∗l=l_* modes. Denoting the long-time quantities with ⋅~\tilde{\mathbin{\cdot}}, we assume that Δ≃Δ~2+Δ~l∗\Delta\simeq\tilde{\Delta}_2+\tilde{\Delta}_{l_*}, and that σ~(t)≃σ∗\tilde{\sigma}(t)\simeq\sigma_*. Then, we can approximate the values of Δ~2\tilde{\Delta}_2 from Eq. (17) as

Δ~2≃2Γl∗1Γ22Γl∗1−Γl∗1Γ22+αΓ21Γl∗3−αΓl∗2Γ23BS~2mf~2m∗.(30)\tilde{\Delta}_2 \simeq\frac{2\Gamma_{l_*}^{1}}{\Gamma_2^2\Gamma_{l_*}^{1}-\Gamma_{l_*}^{1}\Gamma_2^2+\alpha\Gamma_2^1\Gamma_{l_*}^{3}-\alpha\Gamma_{l_*}^{2}\Gamma_2^3}B\tilde{S}_{2m}\tilde{f}_{2m}^{*}. \tag*{(30)}

Importantly, this expression implies that the steady-state excess area distribution is independent of the initial perturbation amplitudes. We demonstrate this feature in Figure 6(a)-(Figure 6d), showing the Δl\Delta_l and σ(t)\sigma(t) dynamics for several cases of identical parameters, but with different initial perturbation amplitude δ\delta. Panels (a) and (c) show the Δl\Delta_l dynamics (for l=2−4l=2-4) for different orientations β\beta and different initial perturbation amplitudes δ\delta, while panels (b) and (d) show the complementary membrane tension dynamics. Note that while the time to reach the post-instability regime depends on δ\delta, the final excess area distribution and membrane tension values (or their time-averaged equivalents) remain similar.

Post-instability dynamics

Figure 6. Post-instability dynamics. (a) The excess areas Δl\Delta_l for l=2l=2–44 for B=2\mathcal{B}=2, C=0.04\mathcal{C}=0.04, λ=5\lambda=5, Δ=1\Delta=1, and Δ0=0.01\Delta_0=0.01. Solid, dashed, and dotted-dashed lines show the trajectories for Δ2,3,4\Delta_{2,3,4} with different initial perturbation amplitudes of δ=10−3\delta=10^{-3}, 10−1010^{-10}, and different orientations [see legend of (b)]. While the actual trajectories differ, the long-time state is similar. Solid black line shows the theoretical prediction for Δ~2\widetilde{\Delta}_2, in agreement with the steady-state Δ‾2\overline{\Delta}_2 attained by the simulations (shown in green). (b) The membrane tension for similar parameters as in (a), reemphasizing that while the dynamics differ, the steady-state values are approximately independent of δ\delta. (c)–(d) Same as (a)–(b) but for λ=1\lambda=1. Insets: the long-time configurations obtained from the OP orientation β≃0\beta\simeq0 (resulting in a stomatocyte) and the IP orientation β=π/2\beta=\pi/2 (resulting in a trilobe) configurations. Note that the two configurations are plotted from a different angle (see their relative axes). Colors capture the distance from the origin. (e) A comparison between the theoretical Δ~2\widetilde{\Delta}_2 and the values obtained from the simulations Δ‾2\overline{\Delta}_2 for a range of C\mathcal{C} values, and B=1\mathcal{B}=1–1010 for both λ=1\lambda=1 and λ=5\lambda=5. An auxiliary diagonal line is added as a guide to the eye. Inset: the ratio Δ~2/Δ‾2\widetilde{\Delta}_2/\overline{\Delta}_2, as a function of log⁡α\log\alpha, showing deviations as α\alpha decreases towards the instability boundary from above.

The steady-state excess area Eq. (30) requires the full solution, specifically knowing f~2,m\tilde{f}_{2,m} at long times (where, in principle, we need to know g~2,m\tilde{g}_{2,m} as well). To estimate f~2,m\tilde{f}_{2,m}, we use substitutions for f~2,m\tilde{f}_{2,m} similar to those of Eq. (16), but now with Δ~2\tilde{\Delta}_2 being the excess area within the l=2l=2 mode, which is unknown in advance, r~\tilde{r} as the new rr value, and Ψ~\tilde{\Psi} as the new Ψ\Psi value attained post-instability. Explicitly, we take

f~2,±2=Δ~22a2r~e∓iΨ~,f~2,0=−Δ~2a2(1−r~2).(31)\tilde{f}_{2,\pm2}=\sqrt{\frac{\tilde{\Delta}_2}{2a_2}\tilde{r}}e^{\mp i\tilde{\Psi}}, \qquad\tilde{f}_{2,0}=-\sqrt{\frac{\tilde{\Delta}_2}{a_2}(1-\tilde{r}^2)}. \tag*{(31)}

Using the approximate membrane tension σ~(t)≃σ∗\tilde{\sigma}(t)\simeq\sigma_*, the equations of motion for r~\tilde{r} and Ψ~\tilde{\Psi} in the reduced system after an instability are

r~′=Λ~−1sin⁡Ψ~+(Λ~S~)−1r0cos⁡(Φ~−Ψ~)+230πΛ~−1(6BC−1C−9B)Δ~2r~,(32)\tilde{r}'=\tilde{\Lambda}^{-1}\sin\tilde{\Psi}+(\tilde{\Lambda}\tilde{S})^{-1}r_0\cos(\tilde{\Phi}-\tilde{\Psi})+\frac{2}{\sqrt{30\pi}}\tilde{\Lambda}^{-1}\left(\frac{6}{\sqrt{BC}}-\frac{1}{C}-\frac{9}{B}\right)\sqrt{\tilde{\Delta}_2}\tilde{r}, \tag*{(32)}
Ψ~′=−1+Λ~−1cos⁡Ψ~r~+(Λ~S~)−1r0r~sin⁡(Φ~−Ψ~),(33)\tilde{\Psi}'=-1+\tilde{\Lambda}^{-1}\frac{\cos\tilde{\Psi}}{\tilde{r}}+(\tilde{\Lambda}\tilde{S})^{-1}\frac{r_0}{\tilde{r}}\sin(\tilde{\Phi}-\tilde{\Psi}), \tag*{(33)}

where we defined the post-instability Λ~≡Δ~2(23λ+32)830\tilde{\Lambda}\equiv\frac{\sqrt{\tilde{\Delta}_2}(23\lambda+32)}{8\sqrt{30}}, depending on the unknown Δ~2\tilde{\Delta}_2. While some of the terms are reminiscent of the unperturbed case of Eq. (18), most of the terms are modified through Λ~\tilde{\Lambda}. There are additional terms in the r~\tilde{r} equation that introduce an explicit BB-dependence, which is absent from the unperturbed system of Eq. (18).

We now look for the steady state of r~\tilde{r} and Ψ~\tilde{\Psi} for the OP case at the post-instability state described by Eq. (26), and find that

r~ss=[(16(9C−6CB+B)(23λ+32)CB)2+1]−1/2Λ~−1,Ψ~ss=arctan⁡16(9C−6CB+B)(23λ+32)CB.(34)\tilde{r}_{ss}=\left[\left(\frac{16(9C-6\sqrt{CB}+B)}{(23\lambda+32)CB}\right)^2+1\right]^{-1/2}\tilde{\Lambda}^{-1}, \qquad\tilde{\Psi}_{ss}=\arctan\frac{16(9C-6\sqrt{CB}+B)}{(23\lambda+32)CB}. \tag*{(34)}

As f~2,±2∼Δ~2r~∼Δ~2Λ~−1\tilde{f}_{2,\pm2}\sim\sqrt{\tilde{\Delta}_2}\tilde{r}\sim\sqrt{\tilde{\Delta}_2}\tilde{\Lambda}^{-1} and Λ~∼Δ~2\tilde{\Lambda}\sim\sqrt{\tilde{\Delta}_2}, we do not need to know Δ~2\tilde{\Delta}_2 to get the value of f~2,±2\tilde{f}_{2,\pm2}, however the value of f~2,0\tilde{f}_{2,0} still depends on Δ~2\sqrt{\tilde{\Delta}_2}. By substituting the initial values of g2,mg_{2,m}'s, and the new expressions for f~2,m\tilde{f}_{2,m} from Eq. (25)–(27) into Eq. (24), we obtain a quadratic equation for Δ~2\widetilde{\Delta}_2, which we solve numerically, for a range of B\mathcal{B} and C\mathcal{C} parameters.

The predicted Δ~2\widetilde{\Delta}_2 values are shown in thick black lines in Figure 6(a),(c), and σ~(t)=σ∗\widetilde{\sigma}(t)=\sigma_\ast are shown as thick black lines in the corresponding panels (b) and (d). We show the systematic agreement between our theoretical estimate Δ~2\widetilde{\Delta}_2 and the numerically-obtained values denoted here as Δ‾2\overline{\Delta}_2 in Figure 6(e). Deviations are observed as α\alpha approaches the stability boundary from above, as shown in the inset of Figure 6(e).

Finally, as the effective instability occurs for Δl\Delta_l and not for individual fl,mf_{l,m} component (see Appendix § 1.8), the long-time structures inherit their symmetry from their initial state and their orientation with respect to the flow. We present two examples of emerging shapes in the insets of Figure 6(d) — an initially OP RBC evolves into a stomatocyte (left inset), while an IP RBC evolves into a multi-lobed shape (right inset). This emerges due to our assumption about the conservation of axial symmetry during the enucleation process. This is one possible explanation for the emergence of different RBC morphologies under the same external flow profiles, reported in [3].

Discussion

Inspired by recent observations of large deformations in RBCs under shear flow [3, 24, 42], we extended a perturbative framework [26] by first including different excess areas for the stress-free and current shapes, Δ0\Delta_0 and Δ\Delta, respectively. We first observed that in the ellipsoidal approximation, where only l=2l=2 modes are considered, the emerging dynamics is reminiscent of the case with Δ=Δ0\Delta=\Delta_0, with a modification to the governing parameter SS. In the present framework, SS encapsulates not only the elastic properties of the RBC membrane through C\mathcal{C}, but also a memory of its stress-free shape through Δ0\Delta_0.

Then, we incorporated additional spherical harmonics as perturbations, and examined the emerging dynamics. We derived an instability criterion, predicted the most prone spherical harmonics for growth, and approximated the stability boundaries as the RBC’s mechanical properties are changed. In the present framework, the membrane tension is the only nonlinear coupling mechanism driving the instability, consistent with the discussion from Abbasi et. al [25]. However, our stability analysis allows the identification of the role of membrane bending resistance and elasticity: we demonstrated the effects of the bending number B\mathcal{B} on the RBC’s stability (absent from the ellipsoidal equations of motion), and showed that α=B/C\alpha=\mathcal{B}/\mathcal{C}, the ratio of bending and elastic forces, controls the excited harmonics: modes of higher ll values are excited as α\alpha increases. We have also demonstrated that changing the viscosity ratio λ\lambda or the excess area Δ\Delta can change the stability boundaries, mainly through the dynamics of the membrane tension. We also probed the stability in log⁡C−λ\log C-\lambda space, and obtained results reminiscent of those obtained from full numerical solutions [24].

Finally, the simplicity of the model allowed us also to understand the post-instability long-time behaviors. An instability induces a redistribution of the excess areas, until the membrane tension σ(t)\sigma(t) stabilizes at σ∗\sigma_*. Once this state is reached, we assumed only the l=2l=2 and l∗l_* modes contribute to the excess area, and derived the effective reduced equations of motion after an instability. Using the expected steady-state solutions, we found an estimated excess area for the l=2l=2 subspace, which we confirmed numerically. As a result of the initial axi-symmetric perturbation, different three-dimensional shapes emerged under the same flow conditions, as also observed in experiments and simulations [3]. Combining the initial conditions with the value of α\alpha and our stability predictions allowed us to understand the long-time dynamics and to demonstrate that α\alpha controls the emerging shape (e.g., the spherical harmonics that become prominent) while the initial conditions control the emerging dynamics (e.g., whether the long-time trajectories reach a steady state or exhibit oscillations).

Experimentally, RBCs are known to adopt morphologies different from their typical discocyte shape, specifically when subjected to intermediate and large shear rates [3, 42]. The theoretical framework proposed above constitutes a first step in understanding such type of instabilities. For example, experimentally, in steady shear flow, trilobes were observed for γ˙∼1000 s−1\dot{\gamma}\sim1000\ \mathrm{s}^{-1} and ηex≃1 cP\eta^{\mathrm{ex}}\simeq1\ \mathrm{cP}. Taking R≃3 μmR\simeq3\ \mu\mathrm{m}, μ≃2.5 μPa⋅m\mu\simeq2.5\ \mu\mathrm{Pa}\cdot\mathrm{m} and κ≃2.0×10−19 J\kappa\simeq2.0\times10^{-19}\ \mathrm{J} yields B∼102\mathcal{B}\sim10^2 and C∼1C\sim1 (log⁡B≈4.5\log B\approx4.5 and log⁡C≈0\log C\approx0), which are in the unstable region at λ=5\lambda=5 (Figure 4b). However, sustained stomatocyte and trilobe morphologies appeared in the experiments above a critical shear rate [3] — a feature not captured by our theoretical framework, indicating some elements are still missing. The proposed theoretical framework suggests additional dependencies on λ\lambda, α\alpha and Δ\Delta, which were not probed experimentally yet.

To further understand the instability mechanisms, and to test the theory above, additional experiments of diluted RBCs or single RBCs are required. Modifying the shear rate γ˙\dot{\gamma} and the external viscosity ηex\eta^{\mathrm{ex}} should allow probing the significance of the viscosity ratio λ\lambda to the instability (see Figure 5 and the related discussion) — the instabilities are expected to disappear as λ\lambda decreases. Modification to the RBC membrane (e.g., using membrane-hardening techniques [3]) should allow probing the importance of elastic forces — though as mentioned, the instabilities for low CC and low B\mathcal{B} values are unphysical. Probing RBCs with different excess areas (e.g., obtained using osmotic pressure effects), would help elucidate the role of the locked internal stresses and elasticity, where generally the more deflated is the RBC, the more susceptible it becomes to instabilities. Fine control over the initial orientation would also allow clarifying the role of the initial orientation with respect to the external flow. Probing the RBCs surface and whether the deformed shapes are sustained or intermittent could shed light on the underlying differences between vesicles and RBCs.

While the theoretical approach above allowed a thorough analysis of mechanical instabilities, this approach also lacks some aspects that probably play an important role in real RBC dynamics. First, the entire approach is based on perturbative expansion from a spherical shape. Clearly, when an RBC undergoes enucleation, the emerging discocyte is very different from a sphere — this geometric difference may be crucial in the emerging dynamics. Additionally, we have taken moderate Δ\Delta values, which may be insufficiently small, causing the perturbative assumption to break down. In realistic RBCs Δ≲4\Delta\lesssim4 [26, 41, 42], and it remains questionable whether the perturbative theory respects such values.

Second, as the theory is approximately linear, by construction it lacks some possibly important non-linear contributions. Currently, the membrane tension term is the only nonlinear coupling in the emerging equations. Yet, it is clear that the interactions between RBCs and their surrounding flow is an inherently non-linear problem — the flow deforms the RBC membrane that modifies the flow again. Additionally, the RBC membrane itself has a non-linear response (due to its initial biconcave geometry, and its non-linear constitutive response). These effects may be important for the instability onset — for example, nonlinear elastic contributions could excite higher harmonics and induce an instability in regions rendered as stable by the above theory. In the long-time limit, coarsening effects could smooth high ll perturbations towards lower ll values possibly altering the long-time sustained shape.

Finally, the theoretical description above assumes that the stress-free shape does not reorient during its dynamics, and maintains a constant orientation with respect to the flow plane. This limits the current theoretical framework to a fixed angle with respect to the flow. Statistically, RBCs orientation with respect to the external flow is not well controlled, and could induce competition between rotation towards/from the flow plane and an instability growth. For example, in [42] the rate of rotation clearly played a crucial role in inducing instabilities, but those were not sustained (hence termed “intermittent”). Capturing the rotation of the stress-free structure with respect to the ambient flow field could induce new dynamical behaviors and enrich the stability analysis.

Overall, the simplicity of the perturbative framework reported here allowed us to gain many insights into the instability onset and dynamics of RBCs under shear flow. These observations serve as first steps in understanding RBC dynamic instabilities, and could motivate additional numerical and experimental investigations into the instability and post-instability dynamics of RBCs in flow. Such future directions would help understand the emergence of various RBC shapes, elucidate the roles played by the RBC membrane mechanics on the emerging blood rheology, and potentially provide insights into biologically relevant problems, such as the identification of membrane anomalies from the observation of RBC dynamics.

Acknowledgments

A.M. acknowledges support from the James S. McDonnell Foundation Postdoctoral Fellowship Award in Complex Systems. The authors gratefully acknowledge the computing resources provided by GENCI (under grant allocations A0160307194 and A0180307194) and by the ISDM computing center (Montpellier, France). The authors also thank Dr. Manouk Abkarian (CNRS Montpellier), and the two anonymous reviewers, for helpful discussion and comments.

Spherical harmonics and Wigner D matrix

We parameterize the RBC surface using an expansion in the spherical harmonics Yl,mY_{l,m}’s, as defined in [26]

Yl,m(θ,ϕ)=[2l+14π(l−m)!(l+m)!]1/2(−1)mPlm(cos⁡θ)eimϕ,(35)Y_{l,m}(\theta,\phi)=\left[\frac{2l+1}{4\pi}\frac{(l-m)!}{(l+m)!}\right]^{1/2}(-1)^mP_l^m(\cos\theta)e^{im\phi}, \tag*{(35)}

where Plm(cos⁡θ)P_l^m(\cos\theta) are the Legendre polynomials.

As mentioned in the main text, we first initiate an axisymmetric RBC structure, parameterized by Yl0Y_{l0}’s only. This corresponds to a structure oriented out of plane (OP). To rotate this structure, we need to apply the rotation operator R(β1,β2,β3)\mathcal{R}(\beta_1,\beta_2,\beta_3) defined here as

R(β1,β2,β3)=e−iβ1Jze−iβ2Jye−iβ3Jz,(36)\mathcal{R}(\beta_1,\beta_2,\beta_3)=e^{-i\beta_1J_z}e^{-i\beta_2J_y}e^{-i\beta_3J_z}, \tag*{(36)}

where the JJ’s are the generators of the SO(3) Lie algebra, and β\beta’s are the Euler angles (using the zz–yy–zz convention). The Wigner D-matrix is the rotation R\mathcal{R} projected onto the specific subspace of interest, defined here as

Dm′,m(l)(β1,β2,β3)=⟨l,m′∣R(β1,β2,β3)∣l,m⟩.(37)D_{m',m}^{(l)}(\beta_1,\beta_2,\beta_3)=\langle l,m'|\mathcal{R}(\beta_1,\beta_2,\beta_3)|l,m\rangle. \tag*{(37)}

Here we use rotations in the yy direction only, simplifying the dependence of R\mathcal{R} and DD to depend on β2≡β\beta_2\equiv\beta (sometimes referred to as Wigner’s d-matrix).

For example, the entries of Dm,0(2)(β)D_{m,0}^{(2)}(\beta) are obtained [55] as

D0,0(2)(β)=14(3cos⁡2β+1),D±1,0(2)(β)=±32cos⁡βsin⁡β,D±2,0(2)(β)=1232sin⁡2β.D_{0,0}^{(2)}(\beta)=\frac{1}{4}(3\cos2\beta+1),\qquad D_{\pm1,0}^{(2)}(\beta)=\pm\sqrt{\frac{3}{2}}\cos\beta\sin\beta,\qquad D_{\pm2,0}^{(2)}(\beta)=\frac{1}{2}\sqrt{\frac{3}{2}}\sin^2\beta.

The perturbative framework

The theoretical framework considered in the main text results from a systematic perturbation expansion of the full equations of motion [26, 27, 29]. We consider the internal and external fluids as incompressible Newtonian fluids with negligible Reynolds number (mainly due to the spatial scale of the RBC). The external fluid is driven at infinity as v∞=γ˙yx^\mathbf{v}^{\infty}=\dot{\gamma}y\hat{\mathbf{x}}. At the RBC boundary, the fluid velocities identify with the membrane velocity, and the emerging stress difference between the external and internal fluids is balanced by the membrane stresses.

The full details and corresponding equations of motion are provided in [26]. The perturbative expansion eventually yields the equations of motion for fl,mf_{l,m}’s as

∂tfl,m=[iωm+Γl(t)]fl,m+Sl,m(t),(38)\partial_t f_{l,m}=[i\omega_m+\Gamma_l(t)]f_{l,m}+S_{l,m}(t), \tag*{(38)}

where

Γl(t)=B−1[Γ1l+σ(t)Γ2l]+C−1Γ3l,(39)\Gamma_l(t)=\mathcal{B}^{-1}\left[\Gamma_1^l+\sigma(t)\Gamma_2^l\right]+\mathcal{C}^{-1}\Gamma_3^l, \tag*{(39)}
Sl,m(t)=Cl,m−C−1Γ3l,m,(40)S_{l,m}(t)=C_{l,m}-\mathcal{C}^{-1}\Gamma_{3l,m}, \tag*{(40)}

and

d(λ,l)=(2l3+3l2+4)+λ(2l3+3l2−5),(41)d(\lambda,l)=(2l^3+3l^2+4)+\lambda(2l^3+3l^2-5), \tag*{(41)}
Γ1l=−(l+2)(l+1)2l2(l−1)d(λ,l),(42)\Gamma_1^l=-\frac{(l+2)(l+1)^2l^2(l-1)}{d(\lambda,l)}, \tag*{(42)}
Γ2l=−(l+2)(l+1)l(l−1)d(λ,l),(43)\Gamma_2^l=-\frac{(l+2)(l+1)l(l-1)}{d(\lambda,l)}, \tag*{(43)}
Γ3l=−4(l+2)(l−1)d(λ,l).(44)\Gamma_3^l=-\frac{4(l+2)(l-1)}{d(\lambda,l)}. \tag*{(44)}

The external driving contributes

Cl,m=d(λ,l)−1[cl,m,0∞l(l+1)(2l+1)+cl,m,2∞(4l3+6l2−4l−3)].(45)C_{l,m}=d(\lambda,l)^{-1}\left[c^\infty_{l,m,0}\sqrt{l(l+1)(2l+1)}+c^\infty_{l,m,2}(4l^3+6l^2-4l-3)\right]. \tag*{(45)}

In the case of shear, the only components that contribute to cl,m,0∞c^\infty_{l,m,0} and cl,m,2∞c^\infty_{l,m,2} are c2,±2,0∞=∓iπ/5c^\infty_{2,\pm2,0}=\mp i\sqrt{\pi/5} and c2,±2,2∞=∓i2π/15c^\infty_{2,\pm2,2}=\mp i\sqrt{2\pi/15}.

Finally, the membrane tension Σ(t)\Sigma(t) is also expressed using spherical harmonics as

Σ(θ,ϕ,t)=σ(t)+∑σl,m(t)Yl,m(θ,ϕ).(46)\Sigma(\theta,\phi,t)=\sigma(t)+\sum\sigma_{l,m}(t)Y_{l,m}(\theta,\phi). \tag*{(46)}

Solving for the tangential stress balance specifies the values of σl,m\sigma_{l,m} [26], which were already substituted into the normal equations of motion, Eq. (38). The value of the spatially-homogeneous part of the membrane tension σ(t)\sigma(t) is obtained by requiring a constant excess surface area ∂tΔ=0\partial_t\Delta=0, which yields Eq. (17).

The expression for Υl\Upsilon_l, Eq. (13a), clearly demonstrates that the growth rate depends on ll via the Γ\Gamma functions, but not on mm. This implies that within a certain ll subspace, perturbations grow proportionally to their initial value. Under shear, and assuming that the stress-free shape has only ellipsoidal components, Sl,mS_{l,m} vanishes, and the dynamics reduces to ∂tΔl=2ΥlΔl\partial_t\Delta_l=2\Upsilon_l\Delta_l. This implies that for l≥3l\ge3 one can examine the instability through Δl\Delta_l.

Survey of unperturbed dynamics

The reduced system detailed in Sec. III, with Δ=Δ0\Delta= \Delta_0 was thoroughly developed and studied in [26]. Here, we do not intend to perform a thorough investigation of the underlying available dynamics of our modified equations of motion Eq. (18) (such analyses were performed in [26]), but rather to showcase various dynamics emerging from the current formalism, and compare and contrast those with the corresponding vesicle dynamics.

We show in Figure 7 the emerging rr and Ψ\Psi dynamics, in the top and bottom panels respectively for three values of λ\lambda, when the RBC is oriented out-of-plane. Panels (a),(d) show the dynamics for λ=1\lambda= 1, (b),(e) for λ=3\lambda= 3, and (c),(f) for λ=5\lambda= 5. In each panel, we show the case of large C=106C = 10^{6} value, corresponding to large SS value, which is the vesicle limit, low C=1C = 1 value obtained with Δ=Δ0\Delta= \Delta_0, and finally the case of low SS value with Δ≠Δ0\Delta\ne\Delta_0.

Figure 7. Unperturbed dynamics of out-of-plane vesicle-like RBC (in dashed green lines), unstressed RBC (in dotted-dashed light blue lines), and enucleated RBC (in blue lines) under different viscosity ratios λ\lambda. Top panels show the amplitude dynamics r(t)r(t), and the bottom panels show the angular dynamics Ψ(t)\Psi(t). Panels (a) and (d) correspond to λ=1\lambda=1, (b) and (e) to λ=3\lambda=3, and (c) and (f) to λ=5\lambda=5. The dynamics are discussed above, however in general the enucleated RBC exhibits dynamics in-between the vesicle-like RBC and the unstressed RBC.

The dynamics for λ=1\lambda= 1 are almost indistinguishable — all cases converge to tank-treading along Ψ=π/2\Psi= \pi/2, corresponding to ψ=π/4\psi= \pi/4. For higher λ\lambda values, the vesicle-like RBC tumbles and deforms the most [see panels (b)-(c)], while the unstressed RBC converges to a steady-state quickly [at t≃20t \simeq20 in (b), and t≃40t \simeq40 in (c)]. Interestingly, the enucleated RBC exhibits damped oscillations, but the damping occurs on a longer time-scale, showing an intermediate dynamics between the vesicle case and the Δ=Δ0\Delta=\Delta_0 case.

We repeat a similar examination for the in-plane case in Figure 8. Again the dynamics for λ=1\lambda= 1 are similar across the three different cases, with the exception of small angle oscillations for the unstressed RBC. This is the signature of swinging [13] — the elasticity of the membrane modulating its circulation and thus the particle inclination. These oscillations are reduced when having Δ0<Δ\Delta_0 < \Delta, consistently with previous studies [18, 23]. For λ=3\lambda= 3, both the vesicle-like RBC and the unstressed RBC perform amplitude oscillations in r(t)r(t), where again the internally-stressed RBC exhibits a dynamical behavior in-between the vesicle-like RBC and the unstressed RBC. It exhibits decaying amplitude oscillations, where the decay occurs on time-scales longer than those of the unstressed RBC. The vesicle-like RBC tumbles, the unstressed RBC performs swinging, and the stressed RBC starts tumbling, and transitions into a swinging (the averaged angle is non-zero). When λ\lambda is further increased, all three RBCs perform deformed tumbling.

Figure 8. Unperturbed dynamics of in-plane vesicle-like RBC (in dashed green lines), unstressed RBC (in dotted-dashed brown lines), and enucleated RBC (in red lines) under different viscosity ratios λ\lambda. Top panels show the amplitude dynamics r(t)r(t), and the bottom panels show the angular dynamics Ψ(t)\Psi(t). Panels (a) and (d) correspond to λ=1\lambda=1, (b) and (e) to λ=3\lambda=3, and (c) and (f) to λ=5\lambda=5. Inset of panel d: a zoom in showing small angle oscillations for the unstressed RBC. The dynamics are discussed above, however in general the enucleated RBC exhibits dynamics in-between the vesicle-like RBC and the unstressed RBC.

Steady-state solutions and derived instability criteria

The reduced system of Eq. (18) admits several steady-state solutions. Here we focus on the leading-order steady-state solution rss−,Ψss−r_{ss}^{-},\Psi_{ss}^{-} when Λ<1\Lambda<1, and rss+,Ψss+r_{ss}^{+},\Psi_{ss}^{+} for Λ>1\Lambda>1, provided in Eq. (20) (detailed in [26]). To obtain the instability criteria for each of these cases, we use the above solutions in σR(t)\sigma_R(t) and perform time averaging ⟨σR⟩≡12π∫02πσR(t′)dt′\langle\sigma_R\rangle\equiv\frac{1}{2\pi}\int_0^{2\pi}\sigma_R(t')dt', allowing us to eliminate the Φ\Phi dependence in Eq. (20). This yields analytical expressions for the membrane tension for the weak elastic scenarios, as

⟨σR−⟩=−6+30π31−Λ2ΔB−2B3C,for Λ<1,(47)\langle\sigma_R^{-}\rangle=-6+\frac{\sqrt{30\pi}}{3}\sqrt{\frac{1-\Lambda^2}{\Delta}}\mathcal{B}-\frac{2\mathcal{B}}{3\mathcal{C}}, \qquad\text{for }\Lambda<1, \tag*{(47)}
⟨σR+⟩=−6+2B3C(Δ0(1−Λ2)(1−r02)ΔΛ2−1),for Λ>1.(48)\langle\sigma_R^{+}\rangle=-6+\frac{2\mathcal{B}}{3\mathcal{C}}\left(\sqrt{\frac{\Delta_0(1-\Lambda^2)(1-r_0^2)}{\Delta\Lambda^2}}-1\right), \qquad\text{for }\Lambda>1. \tag*{(48)}

Note that ⟨σR−⟩\langle\sigma_R^{-}\rangle depends on B\mathcal{B}, and not only on the ratio B/C\mathcal{B}/\mathcal{C}, hinting that the derived criterion may have a non-trivial dependence on B\mathcal{B}. Additionally, note that ⟨σR+⟩\langle\sigma_R^{+}\rangle depends explicitly on r0r_0, implying the instability criterion depends on the initial orientation. Finally, we demand that ⟨σR±⟩≤σ∗\langle\sigma_R^{\pm}\rangle\leq\sigma_* for the expressions in Eq. (D1), allowing us to obtain the critical C∗\mathcal{C}_* values for instability shown in Eq. (28). Note that to obtain a similar criterion for λ∗(C)\lambda_*(\mathcal{C}) for fixed C\mathcal{C} values, the Λ<1\Lambda<1 case yields a non-trivial dependence due to an explicit dependence on B\mathcal{B}, while the Λ>1\Lambda>1 case depends on C\mathcal{C} only through α\alpha, yielding a C\mathcal{C}-independent λ∗\lambda_*.

In-plane stability

We probed the B\mathcal{B}-C\mathcal{C} phase spaces for the IP cases, in a similar fashion to the analysis presented in Sec. IV. Here, we had to define an average observable as the IP cases yield oscillatory dynamics [e.g., see Figure 3c]. Probing only the last entry of Δl\Delta_l in a given trajectory is prone to error, as it depends critically on where this last point occurred relative to the oscillation cycle. To circumvent this issue, we defined ⟨Δl⟩\langle\Delta_l\rangle as an average over the last 10 sampled entries obtained from the numerical simulations (corresponding to t∈490−500t \in490-500). The emerging phase diagrams, complementary to those presented in Figure 4, are plotted in Figure 9, showing similar results.

Figure 9. IP stability phase diagrams similar to those of Figure 4, for different λ\lambda values [panels (a)-(b)], and different Δ\Delta values [panels (b)-(c)]. Log-log phase diagrams for (a) λ=1\lambda= 1, Δ=1\Delta= 1, (b) λ=5\lambda= 5, Δ=1\Delta= 1 and (c) λ=5\lambda= 5, Δ=0.1\Delta= 0.1 in the IP orientation (Δ0=0.01\Delta_0 = 0.01 for all cases). Each phase diagram was obtained by sampling (B,C)(\mathcal{B},\mathcal{C}) pairs spanning B∈[exp⁡(−1),exp⁡(6)]\mathcal{B} \in[\exp(-1),\exp(6)], and C∈[exp⁡(−5.5),exp⁡(2)]\mathcal{C} \in[\exp(-5.5),\exp(2)], and simulating each perturbed system for a duration of T=500T = 500. Colors indicate the value of lΔl_{\Delta}, indicating the ll value for which the excess area increased the most, as obtained from the numerical simulations (see legend). Solid diagonal line show the theoretically-predicted value of l∗=2.5l_* = 2.5 (where we expect l=3l=3 to be excited), and the additional diagonal dashed lines show the theoretically-predicted transition from excited ll value to the next one. Red curves show the value of C∗\mathcal{C}_* obtained from the steady-state calculation (see Appendix § 1.10). Finally, blue diagonal shows αcinit\alpha_c^{\mathrm{init}} effecting the instability boundary at low B\mathcal{B} and C\mathcal{C} values [αcinit\alpha_c^{\mathrm{init}} is not shown in panels (a) and (b) as it roughly overlaps with the l=3l=3 critical line, as mentioned above]. The emerging picture is similar to Figure 4, where RBCs of high C\mathcal{C}'s are destabilized by high viscosity contrast λ\lambda and large excess area Δ\Delta.

An interesting behavior occurs for Δ=0.1\Delta=0.1, λ=1\lambda=1, corresponding to Figure 4(c) and Figure 9(c). There are regions that are initially unstable but become stable at the steady state of the reduced system (occurring for high B\mathcal{B} and C\mathcal{C} values, in the region above C∗\mathcal{C}_* but below α∗init\alpha_*^{\mathrm{init}}). Then, there are also cases that are initially stable, but are destabilized in the steady state (occurring for low B\mathcal{B} and C\mathcal{C} values, in the region above α∗init\alpha_*^{\mathrm{init}} but below C∗\mathcal{C}_*). We provide two exemplary trajectories from the first scenario in Figure 10 — initially unstable trajectories are stabilized by the approach to the steady-state solution of the reduced system. The second scenario is much harder to observe as the initial stability causes the perturbations to decay rapidly.

Figure 10. Suppressed initial instability. (a) Initial dynamics for Δ=0.1\Delta= 0.1, Δ0=0.01\Delta_0 = 0.01, λ=5\lambda= 5, B=exp⁡(4)\mathcal{B} = \exp(4), C=exp⁡(−2.7)\mathcal{C} = \exp(-2.7) showing an initial growth suppressed by the underlying reduced dynamics. (b) Similar plot for B=exp⁡(4)\mathcal{B} = \exp(4), C=exp⁡(−1)\mathcal{C} = \exp(-1) (all other parameters kept the same).

References

References

  1. [1]S. Chien, Shear dependence of effective cell volume as a determinant of blood viscosity, Science, 168, 977–979 (1970).pubmed.ncbi.nlm.nih.gov/5441028
  2. [2]E.W. Merrill, E.R. Gilliland, T.S. Lee, and E.W. Salzman, Blood rheology: effect of fibrinogen deduced by addition, Circ. Res., 18, 437–446 (1966).pubmed.ncbi.nlm.nih.gov/4952703
  3. [3]L. Lanotte, J. Mauэр, S. Mendez, D.A. Fedosov, J.M. Fromental, V. Claveria, F. Nicoud, G. Gompper, and M. Abkarian, Red cells’ dynamic morphologies govern blood shear thinning under microcirculatory flow conditions, PNAS, 113, 13289–13294 (2016).
  4. [4]G. Keirthivasan, A. Wickrema, and J.D. Crispino, Erythroblast enucleation, Stem Cells Int., 2011, 139851 (2011).
  5. [5]S.J. Lu, Q. Feng, J.S. Park, L. Vida, B.S. Lee, M. Strausbauch, P.J. Wettstein, G.R. Honig, and R. Lanza, Biologic properties and enucleation of red blood cells from human embryonic stem cells, Blood, 112, 4475–4484 (2008).
  6. [6]V. Menon and S. Ghaffari, Erythroid enucleation: a gateway into a “bloody” world, Exp. Hematol., 95, 13–22 (2021).
  7. [7]T.M. Fischer, Shape memory of human red blood cells, Biophys. J., 86, 3304–3313 (2004).
  8. [8]T.W. Secomb, Red blood cell mechanics and capillary blood rheology, Cell Biophys., 18, 231–251 (1991).
  9. [9]H.L. Goldsmith, J. Marlow, and F.C. MacIntosh, Flow behaviour of erythrocytes - I. Rotation and deformation in dilute suspensions, Proc. R. Soc. B, 182, 351–384 (1972).DOI
  10. [10]J. Dupire, M. Socol, and A. Viallat, Full dynamics of a red blood cell in shear flow, PNAS, 109, 20808–20813 (2012).
  11. [11]T.M. Fischer, M. Stöhr-Liesen, and H. Schmid-Schönbein, The red cell as a fluid droplet: tank tread-like motion of the human erythrocyte membrane in shear flow, Science, 202, 894–896 (1978).
  12. [12]T. Fischer and H. Schmid-Schönbein, Tank tread motion of red cell membranes in viscometric flow: behavior of intracellular and extracellular markers (with film), Red Cell Rheology, 347–361, (Springer, 1978).
  13. [13]M. Abkarian, M. Faivre, and A. Viallat, Swinging of red blood cells under shear flow. Phys. Rev. Lett., 98, 188302 (2007).
  14. [14]G.B. Jeffrey, The motion of ellipsoidal particles immersed in a viscous fluid, Proc. R. Soc. A, 102, 161–179 (1922).
  15. [15]S.R. Keller and R. Skalak, Motion of a tank-treading ellipsoidal particle in a shear flow, J. Fluid Mech., 120, 27–47 (1982).
  16. [16]J.M. Skotheim and T.W. Secomb, Red blood cells and other nonspherical capsules in shear flow: Oscillatory dynamics and the tank-treading-to-tumbling transition, Phys. Rev. Lett., 98, 078301 (2007).
  17. [17]M. Abkarian and A. Viallat, Vesicles and red blood cells in shear flow. Soft Matter, 4, 653–657 (2008).
  18. [18]S. Mendez and M. Abkarian, In-plane elasticity controls the full dynamics of red blood cells in shear flow, Phys. Rev. Fluids, 3, 101101 (2018).
  19. [19]M.M. Guest, T.P. Bond, R.G. Cooper, and J.R. Derrick, Red blood cells: change in shape in capillaries, Science, 142, 1319–1321 (1963).
  20. [20]D. Barthes-Biesel, Motion and deformation of elastic capsules and vesicles in flow, Annu. Rev. Fluid Mech, 48, 25–52 (2016).
  21. [21]B. Kaoui, G. Biros, and C. Misbah, Why do red blood cells have asymmetric shapes even in a symmetric flow? Phys. Rev. Lett., 103, 188101 (2009).
  22. [22]J. Dupire, M. Abkarian, and A. Viallat, Chaotic dynamics of red blood cells in a sinusoidal flow. Phys. Rev. Lett., 104, 168101 (2010).
  23. [23]J. Dupire, M. Abkarian, and A. Viallat, A simple model to understand the effect of membrane shear elasticity and stress-free shape on the motion of red blood cells in shear flow, Soft Matter, 11, 8372-8382 (2015).
  24. [24]J. Mauer, S. Mendez, L. Lanotte, F. Nicoud, M. Abkarian, G. Gompper, and D.A. Fedosov, Flow-induced transitions of red blood cell shapes under shear. Phys. Rev. Lett., 121, 181103 (2018).DOI
  25. [25]M. Abbasi, A. Farutin, A. Nait-Ouhra, H. Ez-Zahraouy, A. Benyoussef, and C. Misbah, Dynamics and rheology of a single two-dimensional multilobe vesicle in a confined geometry, Phys. Rev. Fluids, 7, 093603 (2022).DOI
  26. [26]P.M. Vlahovska, Y.N. Young, G. Danker, and C. Misbah, Dynamics of a non-spherical microcapsule with incompressible interface in shear flow, J. Fluid Mech., 678, 221–247 (2011).
  27. [27]P.M. Vlahovska and R.S. Gracia, Dynamics of a viscous vesicle in linear flows, Phys. Rev. E, 75, 016313 (2007).pubmed.ncbi.nlm.nih.gov/17358259
  28. [28]C. Misbah, Vacillating breathing and tumbling of vesicles under shear flow, Phys. Rev. Lett, 96, 028104 (2006).pubmed.ncbi.nlm.nih.gov/16486649
  29. [29]P. Ola, The behavior of closed inextensible membranes in linear and quadratic shear flows, Physica A, 278, 87–106 (2000).
  30. [30]B. Kaoui, A. Farutin, and C. Misbah, Vesicles under simple shear flow: Elucidating the role of relevant control parameters Phys. Rev. E, 80, 061905 (2009).pubmed.ncbi.nlm.nih.gov/20365188
  31. [31]G. Danker, T. Biben, T. Podgorski, C. Verdier, and C. Misbah, Dynamics and rheology of a dilute suspension of vesicles: Higher-order theory Phys. Rev. E, 76, 041905 (2007).DOI
  32. [32]G. Lim H.W., M. Wortis, and R. Mukhopadhyay, Stomatocyte-discocyte-echinocyte sequence of the human red blood cell: Evidence for the bilayer–couple hypothesis from membrane mechanics, PNAS, 99, 16766–16769 (2002).
  33. [33]Z. Peng, A. Masychek, and Q. Zhu, Erythrocyte responses in low-shear-rate flows: effects of non-biconcave stress-free state in the cytoskeleton, J. Fluid Mech., 742, 96–118 (2014).DOI
  34. [34]A. Yazdani and P. Bagchi, Influence of membrane viscosity on capsule dynamics in shear flow, J. Fluid Mech., 718, 569–595 (2013).DOI
  35. [35]R. Skalak, A. Tozeren, R.P. Zarda, and S. Chien, Strain energy function of red blood cell membranes, Biophys. J., 13, 245–264 (1973).pubmed.ncbi.nlm.nih.gov/4697236
  36. [36]D. Barthes-Biesel and J.M. Rallison, The time-dependent deformation of a capsule freely suspended in a linear shear flow, J. Fluid Mech., 113, 251–267 (1981).
  37. [37]D.A. Edwards, H. Brenner, and D.T. Wasan, Interfacial Transport Processes and Rheology, Elsevier (2013).
  38. [38]W. Helfrich, Elastic properties of lipid bilayers: theory and possible experiments, Z. Naturforsch. C, 28, 693–703 (1973).pubmed.ncbi.nlm.nih.gov/4273690
  39. [39]U. Seifert, Configurations of fluid membranes and vesicles, Adv. Phys., 46, 13–137 (1997).DOI
  40. [40]U. Seifert, Fluid membranes in hydrodynamic flow fields: Formalism and an application to fluctuating quasisperical vesicles in shear flow, Eur. Phys. J. B, 8, 405–415 (1999).DOI
  41. [41]E. Evans and Y.C. Fung, Improved measurements of the erythrocyte geometry, Microvasc. Res., 4, 335–347 (1972).pubmed.ncbi.nlm.nih.gov/4635577
  42. [42]M. Levant and V. Steinberg, Intermediate regime and a phase diagram of red blood cell dynamics in a linear flow, Phys. Rev. E, 94, 062412 (2016).DOI
  43. [43]T.M. Fischer and R. Korzeniewski, Threshold shear stress for the transition between tumbling and tank-treading of red blood cells in shear flow: dependence on the viscosity of the suspending medium, J. Fluid Mech., 736, 351–365 (2013).
  44. [44]C. Pafferott, G.B. Nash, and H.J. Meiselman, Red blood cell deformation in shear flow. Effects of internal and external phase viscosity and of in vivo aging, Biophys. J., 47, 695–704 (1985).
  45. [45]M.E. Rose, Elementary theory of angular momentum, Courier Corporation (1995).
  46. [46]R. Wells and H. Schmid-Schönbein, Red cell deformation and fluidity of concentrated cell suspensions, J. Appl. Physiol., 27, 213–217 (1969).pubmed.ncbi.nlm.nih.gov/5796309
  47. [47]W. Yao, Z. Wen, Z. Yan, D. Sun, W. Ka, L. Xie, and S. Chien, Low viscosity ektacytometry and its validation tested by flow chamber, J. Biomech., 34, 1501–1509 (2001).pubmed.ncbi.nlm.nih.gov/11672725
  48. [48]D.A. Fedosov, H. Noguchi, and G. Gompper, Multiscale modeling of blood flow: from single cells to blood rheology, Biomech. Model. Mechanobiol., 13, 239–258 (2014).DOI
  49. [49]B. Audoly and Y. Pomeau, Elasticity and Geometry: From hair curls to the nonlinear response of shells, Oxford University Press (2010).
  50. [50]K.S. Turitsyn and S.S. Vergeles, Wrinkling of vesicles during transient dynamics in elongational flow, Phys. Rev. Lett., 100, 028103 (2008).pubmed.ncbi.nlm.nih.gov/18232932
  51. [51]C. Quilliet, Depressions at the surface of an elastic spherical shell submitted to external pressure, Phys. Rev. E, 74, 046608 (2006).pubmed.ncbi.nlm.nih.gov/17155193
  52. [52]F. Quemeneur, C. Quilliet, M. Faivre, A. Viallat, and B. Pépin-Donat, Gel phase vesicles buckle into specific shapes, Phys. Rev. Lett., 108, 108303 (2012).pubmed.ncbi.nlm.nih.gov/22463462
  53. [53]V. Kantsler, E. Segre, and V. Steinberg, Vesicle dynamics in time-dependent elongation flow: Wrinkling instability, Phys. Rev. Lett., 99, 178102 (2007).pubmed.ncbi.nlm.nih.gov/17995373
  54. [54]V. Kantsler, E. Segre, and V. Steinberg, Critical dynamics of vesicle stretching transition in elongational flow, Phys. Rev. Lett., 101, 048101 (2008).pubmed.ncbi.nlm.nih.gov/18764366
  55. [55]Inc. Wolfram Research, Mathematica, version 14.1, Champaign, IL, 2024.

Paper details

vlnc:W_axj7adk9
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

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

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents