1:
2: % version of 04/14/04
3:
4: % Choose pra, prb, prc, prd, pre, prl, prstab, or rmp for journal
5: % Add 'draft' option to mark overfull boxes with black boxes
6: % Add 'showpacs' option to make PACS codes appear
7: % Add 'showkeys' option to make keywords appear
8: %\documentclass[aps,prl,preprint,showpacs,groupedaddress]{revtex4}
9: %\documentclass[aps,prb,preprint,showpacs,superscriptaddress]{revtex4}
10: \documentclass[aps,prl,twocolumn,showpacs,groupedaddress]{revtex4}
11:
12: \bibliographystyle{apsrev}
13:
14: \usepackage{graphicx}
15: \usepackage{dcolumn}
16: \usepackage{bm}
17:
18: \begin{document}
19:
20: %\preprint{}
21:
22: %Title of paper
23: \title{
24: Fluid Vesicles with Viscous Membranes in Shear Flow}
25:
26: \author{Hiroshi Noguchi}
27: %\email[]{Electronic mail: hi.noguchi@fz-juelich.de}
28: \author{Gerhard Gompper}
29: %\homepage[]{Your web page}
30: %\altaffiliation{}
31: \affiliation{
32: Institut f\"ur Festk\"orperforschung, Forschungszentrum J\"ulich,
33: 52425 J\"ulich, Germany}
34:
35: %\date{\today}
36:
37: \begin{abstract}
38: The effect of membrane viscosity on the dynamics of vesicles in
39: shear flow is studied. We present a new simulation technique, which combines
40: three-dimensional multi-particle
41: collision dynamics for the solvent with a dynamically-triangulated
42: membrane model. Vesicles are found to transit
43: from steady tank-treading to unsteady tumbling motion with increasing
44: membrane viscosity. Depending on the reduced volume and membrane viscosity,
45: shear can induce both discocyte-to-prolate and prolate-to-discocyte
46: transformations.
47: This dynamical behavior can be understood from a simplified model.
48: \end{abstract}
49:
50: \pacs{87.16.Dg, 47.55.-t, 87.17.Aa}
51: %\keywords{}
52:
53: \maketitle
54:
55: The dynamical behavior of vesicles --- closed lipid membranes in aqueous
56: solution --- under shear flow
57: is an important subject not only of fundamental research but also in
58: medical applications. For example, in microcirculation,
59: the deformation of red blood cells reduces the flow resistance of microvessels.
60: In diseases such as diabetes mellitus and sickle cell anemia,
61: red blood cells have reduced deformability and often block microvascular flow.
62: Although red blood cells do not have a nucleus and other intracellular
63: organelles,
64: they are of more complex than simple lipid vesicles, since their plasma membrane
65: has an attached spectrin network, which modifies its elastic and rheological
66: properties.
67: The deformability of cells and vesicles is determined by their shape,
68: the viscosity of the internal fluid, and the viscoelasticity of the
69: membrane \cite{chie87}.
70:
71:
72: The dynamical behavior of vesicles in shear flow has been studied experimentally
73: \cite{chie87,fisc78}, theoretically \cite{kell82,tran84}, and numerically
74: \cite{krau96,pozr03,beau04}.
75: The vesicle shape is determined by the competition of the curvature elasticity
76: of the membrane, the constraints of constant volume $V$ and constant
77: surface area $S$, and the external hydrodynamic forces.
78: One of the difficulties in theoretical studies of the hydrodynamic effects
79: on the vesicle dynamics is the boundary condition for the embedding
80: fluid on the vesicle surface, which changes its shape dynamically.
81: In some previous studies, a fluid vesicle was therefore modeled as an
82: ellipsoid with fixed shape \cite{kell82}. More recently, the
83: time-evolution of the shape was studied numerically using a boundary
84: integral method in
85: three spatial dimensions \cite{krau96} or an advected-field
86: method in two spatial dimensions \cite{beau04}.
87: The red blood cell membrane has also been modeled as an elastic capsule of
88: discocyte shape \cite{pozr03}.
89:
90: Two types of dynamics have been found in these studies, a steady state
91: with a tank-treading
92: motion of the membrane and a finite inclination angle with the flow
93: direction, and an unsteady state with a tumbling (flipping) motion.
94: A transition
95: from tank-treading to tumbling motion with an increasing viscosity of the
96: internal fluid has been predicted
97: for fluid vesicles with fixed ellipsoidal shape in three dimensions
98: \cite{kell82}, and with the advected-field method in two dimensions
99: \cite{beau04}.
100: When the shape is relaxed dynamically in three dimensions, all discocyte
101: vesicles were
102: surprisingly found to transform into prolates in shear flow,
103: even for the smallest shear rates studied ~\cite{krau96}.
104:
105: In this letter, we focus on the effect of the membrane viscosity on the
106: dynamics of vesicles in shear flow. This is an important question, because
107: the membrane of red blood cells, for example, becomes more viscous on aging
108: \cite{tran84,nash83} or in diabetes mellitus \cite{tsuk01}. Experiments
109: indicate that the energy dissipation in the membrane is larger than that
110: inside a red blood cell \cite{tran84}. Furthermore, it has been shown recently
111: that vesicles can not only be made from lipid bilayers, but also from bilayers
112: of block copolymers \cite{disc99}. These ``polymersomes'' have been shown
113: to have a membrane viscosity which is several orders of magnitude larger than
114: for liposomes \cite{dimo02}.
115:
116: Several mesoscopic simulation techniques for fluid flow have been developed
117: in recent years. We present here the first simulation studies for a
118: combination of mesoscopic model for
119: the solvent and a coarse-grained, dynamically-triangulated surface model
120: for the membrane.
121: This approach has four main advantages: (i) The membrane is described
122: explicitly, so that their properties like the viscosity can be varied
123: easily; (ii) thermal fluctuation of both the solvent and the membrane are
124: fully and consistently taken into account; (iii) the method can easily
125: be generalized to more complex flow geometries; and (iv) no numerical
126: instabilities can occur.
127:
128: We employ a particle-based hydrodynamics method \cite{male99,male00b,ihle01,kiku03,kiku03a,lamu01,ripo04} to simulate the solvent,
129: which is called multi-particle collision dynamics (MPCD)
130: \cite{lamu01,ripo04} or
131: stochastic rotation dynamics \cite{ihle01,kiku03}.
132: This method was applied, for example, to flow around a solid object
133: \cite{lamu01} and to polymer dynamics \cite{male00b,ripo04}.
134: The fluids in the interior and exterior of the vesicle
135: are taken to be the same, in particular to have the same
136: viscosity $\eta_0$.
137:
138: As the MPCD model is described in detail in
139: Refs.~\cite{male99,male00b,ihle01,kiku03}, we can be very brief in
140: explaining the mesoscopic simulation technique.
141: The solvent is described by $N_{\text s}$ point-like particles of mass
142: $m_{\text s}$ moving in a rectangular box of size $L_x\times L_y \times L_z$.
143: The algorithm consists of alternating streaming and collision steps.
144: In the streaming step, the particles move ballistically and the position of
145: each particle ${\bf r}_{i}$ is updated according to
146: \begin{equation}
147: {\bf r}_{i}(t+h)= {\bf r}_{i}(t) + {\bf v}_{i}(t) h,
148: \end{equation}
149: where ${\bf v}_{i}$ is the velocity of particle $i$ and $h$
150: is the time interval between collisions. In the collision step, the particles
151: are sorted into cubic cells of lattice constant $a$.
152: The collision step consists of a stochastic rotation of the relative velocities
153: of each particle in a cell,
154: \begin{equation}
155: {\bf v}_{i}(t)= {\bf v}_{\text {cm}}(t) +
156: {\bf \Omega}(\varphi)({\bf v}_{i}(t)-{\bf v}_{\text {cm}}(t)),
157: \end{equation}
158: where ${\bf v}_{\text {cm}}$ is the velocity of the center of mass of all
159: particles in the cell. The matrix ${\bf \Omega}(\varphi)$ rotates velocities by a
160: fixed angle $\varphi$ around an axis, which is chosen randomly for each cell.
161: In our simulation, the angle $\varphi=\pi/2$ is employed. We apply a
162: random-shift procedure \cite{ihle01} before each collision step to ensure
163: Galilean invariance.
164:
165: For the membrane, we employ a dynamically-triangulated surface model
166: \cite{gg:gomp97f}, in which the membrane is described by
167: $N_{\text {mb}}$ vertices which are connected by tethers to form a
168: triangular network. The vertices have excluded volume and mass
169: $m_{\text {mb}}$. The shapes and fluctuations of the membrane are
170: controlled by curvature elasticity with the energy \cite{helf73}.
171: \begin{equation}
172: H_{\text {cv}} = \frac{\kappa}{2} \int (C_1+C_2 )^2 dS,
173: \end{equation}
174: where $\kappa$ is the bending modulus, and $C_1$ and $C_2$ are the principal
175: curvatures at each point of the membrane. The curvature energy is discretized
176: as described in Ref.~\cite{gomp96}. To model the fluidity of the
177: membrane, tethers can be flipped between the two possible diagonals of two
178: adjacent triangles. These bond flips provide also a convenient way to vary
179: the membrane viscosity $\eta_{\text {mb}}$, because it increases
180: with decreasing bond-flip rate. In contrast to previous studies of
181: dynamically triangulated surfaces, which were all done by Monte Carlo
182: simulations, we introduce a smooth bond-interaction potential, which
183: makes the model amenable for molecular dynamics simulations.
184:
185: The solvent particles interact with the membrane in two ways. First, the
186: membrane vertices are included in the MPCD collision procedure \cite{male00b}.
187: Second, the solvent particles are scattered elastically or via bounce-back
188: from membrane triangles. We use here the procedure suggested in
189: Ref.~\cite{kiku03a} for a spherical particle.
190:
191:
192: \begin{figure}
193: \includegraphics{fig1.eps}
194: \caption{ \label{fig:ang}
195: (Color online)
196: Dependence of the average inclination angle $\langle\theta\rangle$
197: ($-\pi/2\le\theta<\pi/2$)
198: on the membrane viscosity $\eta_{\text {mb}}^*$ for reduced shear rate
199: $\dot\gamma^*=0.92$ and various reduced volumes $V^*$.
200: The error bars are estimated from three independent runs.
201: Squares and circles represent discocyte and prolate vesicles at
202: $V^*=0.59$, respectively. Triangles and diamonds represent prolate vesicles
203: at $V^*=0.78$ and $V^*=0.91$, where the prolate is the only stable shape.
204: The solid lines and broken line are calculated by K-S theory with
205: prolate ($V^*=0.59, 0.78$, and $0.91$) and
206: oblate ellipsoids ($V^*=0.59$), respectively.
207: }
208: \end{figure}
209:
210:
211: To induce a shear flow, we employed Lee-Edwards boundary condition
212: \cite{kiku03,alle87}, which gives a linear flow profile
213: $(v_x,v_y,v_z)=(\dot\gamma z,0, 0)$ in the MPCD fluid.
214: The particle density was set to $\rho=10m_{\text {s}}/a^3$
215: ($N_{\text s}=450000$, $L_x=50a$, $L_y=L_z=30a$).
216: In experimental conditions of red blood cells and liposomes,
217: the Reynolds number Re$=\dot\gamma\rho R_{\text 0}^2/\eta_{\text {0}}$ is
218: very small (Re $\sim 10^{-3}$), where $R_{\text 0}=\sqrt{S/4\pi}$ is
219: the effective vesicle radius. Therefore,
220: we chose a short mean free path $h\sqrt{k_{\text B}T/m_{\text {s}}}=0.025 a$,
221: where $k_{\text B}$ is the Boltzmann constant and $T$ is the temperature
222: \cite{ripo04}.
223: Then the viscosity of solvent fluid is
224: $\eta_{\text {0}}= 20.1\sqrt{m_{\text {s}}k_{\text B}T}/a^2$ \cite{kiku03}.
225: We used $\kappa=20k_{\text B}T$, $N_{\text {mb}}=500$, and
226: $m_{\text {mb}}=10m_{\text {s}}$.
227: The volume $V$ and surface area $S=405 a^2$ of a vesicle are kept constant
228: to about 1\% accuracy.
229: With these parameters, we obtain a Reynolds number Re$\simeq 0.1$.
230: The results are conveniently expressed in terms of dimensionless
231: variables:
232: the reduced volume $V^*=V/(4\pi R_{\text 0}^3/3)$, the intrinsic time scale
233: $\tau=\eta_{\text {0}}R_{\text 0}^3/\kappa$,
234: the reduced shear rate $\dot\gamma^*=\dot\gamma\tau$, and the relative membrane
235: viscosity $\eta_{\text {mb}}^*=\eta_{\text {mb}}/\eta_{\text 0}R_{\text 0}$.
236: Details of the numerical scheme will be published elsewhere \cite{nogu04}.
237:
238: At $\eta_{\text {mb}}^*=0$, a vesicle exhibits tank-treading motion for
239: all simulated reduced volumes in the range $0.59\le V^*\le 0.97$.
240: We calculated the average inclination angles $\langle\theta\rangle$, and found
241: them to agree very well with those obtained by the boundary integral method
242: \cite{krau96}.
243:
244: With increasing membrane viscosity $\eta_{\text {mb}}^*$, the inclination
245: angle $\theta$ decreases, as shown in Fig.~\ref{fig:ang}.
246: The qualitative features of the simulation data are reproduced very well
247: by the theory of Keller and Skalak
248: (K-S) \cite{kell82,tran84}. Note that there are no adjustable parameters.
249: Due to the approximations in the K-S theory, an agreement on a
250: quantitative level cannot be expected: (i) an ellipsoidal shape is
251: assumed, which is only mimics the real shapes of vesicles,
252: (ii) the flow on the surface of the
253: droplet is not locally area conserving, as it must be for a incompressible
254: membrane, and (iii) thermal fluctuations are ignored in the theory, but
255: are present in the simulations.
256: In the K-S theory, the vesicle transits from tank-treading to tumbling
257: motion when the angle $\theta$ reaches $0$. In contrast, we observe
258: tumbling intermittently to occur already for nonzero $\langle\theta\rangle$,
259: since our simulation includes thermal fluctuation.
260: For example, the vesicle with $V^*=0.78$ starts tumbling at
261: $\eta_{\text {mb}}^*=1.22$.
262: This intermittent tumbling smoothes out the decrease in $\langle\theta\rangle$
263: around the transition point, compare Fig.~\ref{fig:ang}.
264:
265: \begin{figure}
266: \includegraphics{fig2.eps}
267: \caption{ \label{fig:fe}
268: Free-energy profile $F(\alpha)$ of the asphericity $\alpha$ for $V^* = 0.59$
269: in the absence of shear
270: flow. The sliced snapshots of stable (discocyte) and metastable (prolate and
271: stomatcyte) shapes are also shown. }
272: \end{figure}
273:
274: We now focus on the case $V^*=0.59$. At this reduced volume, the oblate
275: discocyte shape is stable and the prolate and
276: stomatocyte shapes are metastable in the absence of shear flow.
277: Fig.~\ref{fig:fe} shows the free energy $F$ as a function of
278: the asphericity $\alpha$, calculated
279: with a version of the generalized-ensemble Monte Carlo method \cite{berg03}.
280: The asphericity
281: $\alpha = [({\lambda_1}-{\lambda_2})^2+ ({\lambda_2}-{\lambda_3})^2+
282: ({\lambda_3}-{\lambda_1})^2]/[2 R_{\text g}^4]$,
283: with the eigenvalues ${\lambda_1} \leq {\lambda_2} \leq {\lambda_3}$ of the
284: moment-of-inertia tensor and the squared
285: radius of gyration $R_{\text g}^2=\lambda_1+\lambda_2+\lambda_3$, is a
286: convenient measure to distinguish oblate and prolate shapes, where
287: $\alpha=0$ for spheres, $\alpha=1$ for thin rods, and $\alpha=0.25$ for
288: thin discs \cite{rudn86}. The shear flow changes this stability.
289: For membrane viscosity $\eta_{\text {mb}}^*=0$ and shear rates
290: $\dot\gamma^*\ge 1.66$, the discocyte state is found to be
291: destabilized and to transform into a prolate,
292: in agreement with the results of Ref.~\cite{krau96}. However, for
293: a smaller shear rate of $\dot\gamma^*=0.92$, the discocyte vesicle
294: retains its shape. Speculations about shear to be a singular
295: perturbation \cite{krau96} can therefore be ruled out.
296:
297: \begin{figure}
298: \includegraphics{fig3.eps}
299: \caption{\label{fig:d1}
300: (Color online)
301: Time development of (a) asphericity $\alpha$ and (b) inclination angle
302: $\theta$, for $V^*=0.59$, $\eta_{\text {mb}}^*=1.62$ and $\dot\gamma^*=1.84$.
303: The broken lines are obtained from Eqs.~(\ref{eq:al}) and (\ref{eq:thet})
304: with $\zeta_{\alpha}=100$, $A=15$, and $B(\alpha)=1.1-0.18\alpha$.
305: }
306: \end{figure}
307:
308: The inclination angle $\theta$ of prolates decreases faster than that
309: of discocytes
310: with increasing $\eta_{\text {mb}}^*$, see Fig.~\ref{fig:ang}.
311: At a large membrane viscosity of $\eta_{\text {mb}}^*=1.62$, the
312: prolate enters the tumbling phase,
313: while the discocyte remains in the tank-treading phase.
314: The reason is that the discocyte has a flat dimple region and is
315: less affected by the membrane viscosity than the prolate.
316: Remarkably, for small shear rates $\dot\gamma^*\le 1.84$,
317: the (metastable) prolate starts tumbling, but after a $\pi$ or $2\pi$
318: rotation, transforms into a tank-treading discocyte, see Fig.~\ref{fig:d1}.
319: Only for larger shear rates of $\dot\gamma^*\ge 2.76$, the tumbling
320: continues --- accompanied
321: by shape oscillations between prolate and discocyte (Fig.~\ref{fig:d2}).
322: At intermediate membranes viscosities, $\eta_{\text {mb}}^*=0.49$ and
323: $\eta_{\text {mb}}^*=0.87$, and shear rate $\dot\gamma^*=0.92$, the
324: prolate transforms into a discocyte
325: after tank-treading motion for a time of
326: $(70\pm40)\tau$ or $(40\pm20)\tau$ by thermal fluctuation, respectively.
327: For a larger shear rate of $\dot\gamma^*=1.84$, the tumbling continues
328: intermittently.
329:
330:
331: \begin{figure}
332: \includegraphics{fig4.eps}
333: \caption{\label{fig:d2}
334: (Color online)
335: Time development of (a) asphericity $\alpha$ and (b) inclination angle
336: $\theta$ at $\dot\gamma^*=2.76$.
337: The other parameters are set to same values in Fig.~\ref{fig:d1}.
338: }
339: \end{figure}
340:
341: K-S theory~\cite{kell82,tran84} explains the
342: $\eta_{\text {mb}}^*$-dependence of the stability of tank-treading
343: (compare Fig.~\ref{fig:ang}), but cannot be applied to describe
344: the dynamics, including morphological changes.
345: We suggest the simple equations
346: \begin{eqnarray}
347: \zeta_{\alpha} \dot{\alpha}=
348: -\kappa^{-1}\partial F/\partial \alpha + A\dot\gamma^* \sin(2\theta)
349: \label{eq:al}\\
350: \dot{\theta}=0.5\dot\gamma^*\{-1+B(\alpha) \cos(2\theta)\} \label{eq:thet}.
351: \end{eqnarray}
352: The force $\partial F/\partial \alpha$ is calculated from the free
353: energy $F(\alpha)$ of Fig.~\ref{fig:fe}.
354: The second term of Eq.~(\ref{eq:al}) is the deformation force due to the
355: shear flow. Its angular dependence can be derived from the shape
356: equations of Ref.~\cite{krau96}, while the amplitude is assumed to
357: be independent of the asphericity $\alpha$ (to leading order).
358: Eq.~(\ref{eq:thet}) is adopted from K-S theory \cite{kell82,tran84}.
359: Here, $B$ is a constant which depends on viscosities and ellipsoid shape.
360: For $B>1$, a steady angle $\theta=0.5\arccos(1/B)$ exists and
361: tank-treading motion occurs, while
362: for $B<1$, there is no stable angle and tumbling motion occurs.
363: In our case, the vesicle shape can be time dependent, so that $B$ is
364: no longer constant. For simplicity, we assume a
365: linear dependence of $B$ on the asphericity, $B(\alpha)=B_0-B_1\alpha$.
366: To obtain tank-treading discocytes and tumbling prolate, we need
367: $B(0.2)>1$ and $B(0.7)<1$, respectively.
368: Then, Eqs.~(\ref{eq:al}) and (\ref{eq:thet}) reproduce the simulated
369: dynamics very well, see Figs.~\ref{fig:d1} and \ref{fig:d2}.
370: The vesicle is found, for example, to relax after some tumbling to a stable,
371: tank-treading discocyte state at $\dot\gamma^*=1.84$, and to
372: relax to a limit-cycle oscillation at $\dot\gamma^*=2.76$.
373:
374: The vesicle deformation due to shear depends on the inclination
375: angle $\theta$. Shear flow increases the elongation of a vesicle for
376: $0<\theta<\pi/2$ (where $\dot\gamma^*\sin(2\theta)>0$),
377: and the angles $\theta$ of tank-treading motion belong to this region.
378: On the other hand, shear flow reduces the elongation for $-\pi/2<\theta<0$
379: (where $\dot\gamma^*\sin(2\theta)<0$) during tumbling.
380: The force in the former case induces the discocyte-to-prolate transformation,
381: in the latter case the prolate-to-discocyte transformation.
382: With increasing membrane viscosity $\eta_{\text {mb}}^*$,
383: the inclination angle $\theta$ of a tank-treading discocyte decreases,
384: and larger shear rates $\dot\gamma^*$ are necessary to generate the required
385: elongational forces to induce a discocyte-to-prolate transition.
386:
387: It is also interesting to compare the effect of membrane viscosity
388: $\eta_{\text {mb}}$ and internal viscosity $\eta_{\text {in}}$.
389: In both cases, an increase of the viscosity
390: induces a decrease of the inclination angle $\theta$ and a transition from
391: tank-treading to tumbling.
392: However, the effect of internal viscosity $\eta_{\text {in}}$ is less
393: dependent on the vesicle morphology.
394: K-S theory \cite{kell82} shows that
395: the tank-treading phase of oblate vesicle is destabilized a little
396: faster than that of prolate
397: with an increase of $\eta_{\text {in}}$ (for $\eta_{\text {mb}}^*=0$);
398: the transition viscosity is $\eta_{\text {in}}/\eta_{\text {0}}= 2.8$
399: and $3.3$ of oblate and prolate ellipsoids at $V^*=0.59$, respectively.
400: Thus, only a sufficiently high membrane viscosity $\eta_{\text {mb}}^*$
401: can induce a prolate-to-discocyte transformation.
402:
403: The membrane viscosity of human red blood cells is estimated from
404: the analysis of the tank-treading motion to be
405: $\eta_{\text {mb}}=10^{-7}$Ns/m \cite{tran84}, while a
406: micropipette recovery-time technique gives
407: $\eta_{\text {mb}}=10^{-6}$Ns/m \cite{nash83}.
408: When the viscosity of the external fluid is set to the same value of
409: the intracellular fluid,
410: $\eta_{\text 0}=10^{-2}$Pa~s, and $R_{\text 0}=3.3\mu$m,
411: the relative membrane viscosity is then found to be in the range
412: $\eta_{\text {mb}}^*=1...10$.
413: Thus, the effect of this membrane viscosity is sufficiently large (compare
414: Fig.~\ref{fig:ang}) to strongly affect the dynamics of erythrocytes.
415: The viscoelasticity of membrane can be changed by varying the chemical
416: composition of the solvent \cite{chie87}.
417: It is difficult to separate the effects of viscosity and elasticity, however.
418: On the other hand, the membrane viscosity of polymersome can be changed
419: by varying the polymer chain length. Thus,
420: polymersomes seem to be very well suited to study the effect of
421: membrane viscosity experimentally.
422:
423: To summarize, we have applied the MPCD method to a vesicle with viscous
424: membrane under a simple shear flow.
425: The membrane viscosity qualitatively changes the vesicle dynamics.
426: The shear induces a discocyte-to-prolate or prolate-to-discocyte
427: transformation at low or high membrane viscosity, respectively.
428: The deformations of other vesicles shapes, such as stomatocyte and budded
429: vesicles, will be interesting subjects in the further studies.
430:
431:
432: \begin{acknowledgments}
433: We would like to thank N.~Kikuchi (Oxford), A.~Lamura (Bari), and R.G. Winkler
434: for helpful discussions.
435: HN's stay at FZJ was supported by the Postdoctoral Fellowships for Research
436: Abroad of the Japan Society for the Promotion of Science (JSPS).
437: \end{acknowledgments}
438:
439:
440: \begin{thebibliography}{100}
441:
442: \bibitem{chie87}
443: S. Chien, Ann.\ Rev.\ Physiol. {\bf 49}, 177 (1987).
444:
445: \bibitem{fisc78}
446: T.~M. Fischer, M. St{\"o}hr-Liesen, and H. Schmid-Sch{\"o}nbein, Science {\bf
447: 202}, 894 (1978).
448:
449: \bibitem{kell82}
450: S.~R. Keller and R. Skalak, J.\ Fluid\ Mech. {\bf 120}, 27 (1982).
451:
452: \bibitem{tran84}
453: R. Tran-Son-Tay, S.~P. Sutera, and P.~R. Rao, Biophys.\ J. {\bf 46}, 65
454: (1984).
455:
456: \bibitem{krau96}
457: M. Kraus, W. Wintz, U. Seifert, and R. Lipowsky, Phys. Rev. Lett. {\bf 77},
458: 3685 (1996).
459:
460: \bibitem{beau04}
461: J. Beaucourt {\it et~al.}, Phys.\ Rev.\ E {\bf 69}, 011906 (2004).
462:
463: \bibitem{pozr03}
464: C. Pozrikidis, Annals\ Biomed.\ Eng. {\bf 31}, 1194 (2003).
465:
466: \bibitem{nash83}
467: G.~B. Nash and H.~J. Meiselman, Biophys.\ J. {\bf 43}, 63 (1983).
468:
469: \bibitem{tsuk01}
470: K. Tsukada, E. Sekizuka, C. Oshio, and H. Minamitani, Microvasc.\ Res. {\bf
471: 61}, 231 (2001).
472:
473: \bibitem{disc99}
474: B.~M. Discher {\it et~al.}, Science {\bf 284}, 1143 (1999).
475:
476: \bibitem{dimo02}
477: R. Dimova {\it et~al.}, Eur. Phys. J. E {\bf 7}, 241 (2002).
478:
479: \bibitem{male99}
480: A. Malevanets and R. Kapral, J. Chem. Phys. {\bf 110}, 8605 (1999).
481:
482: \bibitem{male00b}
483: A. Malevanets and J.~M. Yeomans, Europhys. Lett. {\bf 52}, 231 (2000).
484:
485: \bibitem{ihle01}
486: T. Ihle and D.~M. Kroll, Phys. Rev. E {\bf 63}, 020201(R) (2001).
487:
488: \bibitem{kiku03}
489: N. Kikuchi, C.~M. Pooley, J.~F. Ryder, and J.~M. Yeomans, J.\ Chem.\ Phys. {\bf
490: 119}, 6388 (2003).
491:
492: \bibitem{kiku03a}
493: N. Kikuchi, Ph.~D. thesis, University of Oxford 2003.
494:
495: \bibitem{lamu01}
496: A. Lamura, G. Gompper, T. Ihle, and D.~M. Kroll, Eruophys.\ Lett. {\bf 56},
497: 319 (2001).
498:
499: \bibitem{ripo04}
500: M. Ripoll, K. Mussawisade, R.~G. Winkler, and G. Gompper, preprint (2004).
501:
502: \bibitem{gg:gomp97f}
503: G. Gompper and D.~M. Kroll, J. Phys. Condens. Matter {\bf 9}, 8795 (1997).
504:
505: \bibitem{helf73}
506: W. Helfrich, Z. Naturforsch. {\bf 28c}, 693 (1973).
507:
508: \bibitem{gomp96}
509: G. Gompper and D.~M. Kroll, J.\ Phys.\ I\ (France) {\bf 6}, 1305 (1996).
510:
511: \bibitem{alle87}
512: M.~P. Allen and D.~J. Tildesley, {\it Computer simulation of
513: liquids } (Clarendon Press, Oxford, 1987).
514:
515: \bibitem{nogu04}
516: H. Noguchi and G. Gompper, (in preparation).
517:
518: \bibitem{berg03}
519: B.~A. Berg, H. Noguchi, and Y. Okamoto, Phys.\ Rev.\ E {\bf 68}, 036126
520: (2003).
521:
522: \bibitem{rudn86}
523: J. Rudnick and G. Gaspari, J. Phys. A {\bf 19}, L191 (1986).
524:
525: \end{thebibliography}
526:
527:
528:
529: \end{document}
530:
531:
532: