206
DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION by Pratik Pranay A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Chemical Engineering) at the UNIVERSITY OF WISCONSIN – MADISON 2011

DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

DYNAMICS OF SUSPENSIONS OF ELASTIC

CAPSULES IN POLYMER SOLUTION

by

Pratik Pranay

A dissertation submitted in partial

fulfillment of the requirements for the degree of

Doctor of Philosophy

(Chemical Engineering)

at the

UNIVERSITY OF WISCONSIN – MADISON

2011

Page 2: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)
Page 3: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

i

Acknowledgments

I express my deepest gratitude towards Prof. Michael D. Graham, for constantly

guiding me in my area of research and exposing me to various projects throughout

my PhD work. Without his constant support and motivation it would not have been

possible to work on such diversified and challenging aspects of research. I will always

remember the insightful discussions with him on simple to complicated problems that

we undertook in my PhD work and hope to carry the same basic understanding and

way of undertaking new problems in future.

I thank Profs. Rawlings, Klingenberg, Yin and Chesler for taking time to be on

my thesis committee.

It has been a great experience working with many former and current members

of the Graham group. I thank Samartha G. Anekal, Patrick T. Underhill, Juan P.

Hernandez-Ortiz, Wei Li, Mauricio Lopez, Hongbo Ma and Aslin Izmitli for providing

assistance during my first few years. I have enjoyed interacting with Li Xi who has

been my office-mate for the past four years especially in giving me a good company

at late night hours. Pieter J. A. Janssen has been a good friend and I will miss our

thoughtful discussions and jokes on every topic. I thank Amit Kumar for helping me

in simulations and providing answers to all my questions. I have enjoyed working with

Page 4: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

ii

Kushal Sinha both inside and outside the group. Yu Zhang has been a good friend

throughout my PhD life. In the relatively short time I had with the new members,

Rafael Henrıquez-Rivera, Timothy Feyereisen, Shifan Mao, Prof. Shinji Tamano and

Friedemann Hahn have been a good company.

During my stay in the Madison, I have made many good friends, who made my

times in Madison enjoyable. I have been fortunate to have met Deepa Sanwal and I

thank her for her company, love and friendship over the last couple of years. I treasure

my friendship with Rohit Malshe and will miss the unforgettable times spent together.

Vikram Adhikarla and Maneesh Mishra have been good firnds in Madison and I wish

the best for them in future. I will miss running with Profs. Klingenberg and Root.

Manan Chopra, Santosh Reddy, Maneesh Rathi and Murali Rajamani have been

trustworthy friends and have provided me with timely advice and help.

Above all, I am thankful to my parents whose motivation and inspiration have

helped me grow at all stages of my life.

Research projects presented in this dissertation was supported by the National

Science Foundation.

Page 5: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

iii

Abstract

Many serious medical conditions, from hemorrhage to coronary artery disease to di-

abetes, are associated with disruptions in blood flow. It has been observed that

potential beneficial effects on hemodynamics arise from addition of low concentra-

tions of high molecular weight long-chain polymer molecules known as drag-reducing

additives (DRAs) to blood. These are so named because of their drag reducing ef-

fects on the turbulent flows, but since blood flow in small vessels is not turbulent, the

effect of these polymers on hemodynamics must have a separate origin that are not

understood. The present work represents an initial step towards gaining fundamental

understanding of these observed effects of DRAs on blood flow, and more broadly,

shedding light on the dynamics of complex multiphase fluids.

The dynamics and pair collisions of fluid-filled elastic capsules during Couette flow

in Newtonian fluids and dilute solutions of high molecular weight (drag-reducing)

polymers are investigated via direct simulation. Capsule membranes are modeled

using either a neo-Hookean constitutive model or a model introduced by Skalak et

al. (R. Skalak, A. Tozeren, R. P. Zarda, and S. Chien, “Strain energy function of red

blood-cell membranes”, Biophysical Journal 13, 245–280 (1973)), which includes an

energy penalty for area changes. This model was developed to capture the elastic

Page 6: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

iv

properties of red blood cells. Polymer molecules are modeled as bead-spring trimers

with finitely extensible non-linearly elastic (FENE) springs; parameters were chosen

to loosely approximate 4000 kD poly(ethylene oxide). Simulations are performed

with a novel Stokes flow formulation of the Immersed Boundary Method (IBM) for

the capsules, combined with Brownian dynamics for the polymer molecules. Results

for isolated capsules in shear indicate that at the very low concentrations considered

here, polymers have little effect on the capsule shape. In the case of pair collisions, the

effect of polymer is strongly dependent on the elastic properties of the capsules’ mem-

branes. For neo-Hookean capsules or for Skalak capsules with only a small penalty for

area change, the net displacement in the gradient direction after collision is virtually

unaffected by polymer. For Skalak capsules with a large penalty for area change,

polymers substantially decrease the net displacement when compared to the Newto-

nian case and the effect is enhanced on increasing the polymer concentration. The

differences between the polymer effects in the various cases is associated with the

extensional flow generated in the region between the capsules as they leave the colli-

sion. The extension rate is highest when there is a strong resistance to a change in

membrane area, and is substantially decreased in the presence of polymer.

Suspensions of fluid-filled elastic capsules in a Couette flow in Newtonian fluids and

dilute solutions of high molecular weight (drag-reducing) polymers are investigated.

A simple theoretical model is presented to describe the cross-stream migration of

deformable capsules in suspensions which comes from a balance of shear-induced dif-

fusion and wall-induced migration due to capsule deformability. The model provides

an explicit theoretical prediction of the dependence of capsule-depleted layer thick-

ness on the capillary number. A computational approach is then used to examine

Page 7: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

v

the motion of elastic capsules in polymer solutions. Capsule membranes are modeled

using a neo-Hookean constitutive model and polymer molecules are modeled as bead-

spring chains with FENE springs; parameters were chosen to loosely approximate

4000 kD poly(ethylene oxide). Results for an isolated capsule near a wall indicate

that the wall-induced migration depends strongly on the capillary number, expressing

capsule deformability. Numerical simulation of suspensions of capsules in Newtonian

fluid illustrates the inhomogeneous distribution of capsules at steady-state with the

formation of capsule-depleted layer near the walls. The thickness of this layer is found

to be strongly dependent on the capillary number. The shear-induced diffusivity, on

the other hand, show a weak dependence on capillary number. These results indicate

that the mechanism of wall-induced migration is the primary source for determining

the capsule-depleted layer thickness of capsules in suspensions. Numerical simulation

results show that both the wall-induced migration and the shear-induced diffusive

motion of the capsules are suppressed under the influence of polymer. Results on sus-

pensions of capsules illustrate that the net effect of polymers is to reduce the thickness

of the capsule-depleted layer resulting in a redistribution of capsules at steady state.

The results are in qualitative agreement to the experimental observations.

Page 8: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

vi

Contents

Acknowledgments i

Abstract iii

List of Figures x

List of Tables xxii

1 Introduction 1

2 Background 6

2.1 Dynamics of blood flow in the microcirculation . . . . . . . . . . . . . 6

2.2 Drag reducing additives in blood flow . . . . . . . . . . . . . . . . . . 9

2.3 Simulations of blood flow . . . . . . . . . . . . . . . . . . . . . . . . . 12

3 Models and discretization methods 17

3.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3.2 Capsule Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3.3 Polymer Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.4 Fluid velocity calculation . . . . . . . . . . . . . . . . . . . . . . . . . 28

Page 9: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

vii

3.5 Volume correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

4 Single capsule dynamics 38

4.1 Method and code validation . . . . . . . . . . . . . . . . . . . . . . . 38

4.2 Single capsule in shear . . . . . . . . . . . . . . . . . . . . . . . . . . 42

4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

5 Pair collisions in shear 46

5.1 Newtonian fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

5.2 Polymer solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

5.3 Effect of Ca . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

5.4 Mechanism of polymer effects . . . . . . . . . . . . . . . . . . . . . . 60

5.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

6 Suspensions of capsules in Couette flow 63

6.1 Theory for suspension of capsules in shear flow . . . . . . . . . . . . 63

6.1.1 Wall-induced migration of a single capsule . . . . . . . . . . . 63

6.1.2 Shear-induced diffusion . . . . . . . . . . . . . . . . . . . . . . 65

6.1.3 Model for steady-state distribution . . . . . . . . . . . . . . . 66

6.1.4 Steady-state capsule-depleted layer near a single wall . . . . . 68

6.2 Migration of a single capsule in a Couette Flow . . . . . . . . . . . . 69

6.2.1 Newtonian fluid . . . . . . . . . . . . . . . . . . . . . . . . . . 69

6.2.2 Validation of dipole approximation . . . . . . . . . . . . . . . 72

6.2.3 Polymer fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

6.3 Suspensions of capsules in a Couette flow . . . . . . . . . . . . . . . . 84

6.3.1 Newtonian solution . . . . . . . . . . . . . . . . . . . . . . . . 84

Page 10: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

viii

6.3.2 Polymer solution . . . . . . . . . . . . . . . . . . . . . . . . . 91

6.3.3 Capsule-depleted layer . . . . . . . . . . . . . . . . . . . . . . 95

6.3.4 Diffusion at steady state . . . . . . . . . . . . . . . . . . . . . 98

6.4 Comparison with the theoretical model . . . . . . . . . . . . . . . . . 102

6.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

7 Conclusion 107

8 Current work: Suspensions of capsules in a pressure-driven flow 110

8.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

8.2 Migration of an isolated capsule in a pressure-driven Flow . . . . . . 112

8.3 Suspensions of capsules in a pressure-driven Flow . . . . . . . . . . . 116

8.4 Discussion and future work . . . . . . . . . . . . . . . . . . . . . . . . 122

9 Future work 124

9.1 Suspensions of Red Blood Cells . . . . . . . . . . . . . . . . . . . . . 124

9.2 Drug delivery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

9.3 Leucocyte margination under the influence of polymers . . . . . . . . 130

A Single Polymer Dynamics 132

B Comparison of GGEM/IBM with other methods 137

B.1 Overview of Conventional IBM Method . . . . . . . . . . . . . . . . . 137

B.2 IBM and GGEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

B.3 BIM and GGEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

C Implementation Issues-Slit 149

Page 11: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

ix

C.1 Different flow profiles: . . . . . . . . . . . . . . . . . . . . . . . . . . 150

C.2 Assignment: putting ρg on the mesh - collocation . . . . . . . . . . . 152

C.3 Solution approach - FFT/finite differences . . . . . . . . . . . . . . . 153

C.3.1 Treatment of periodic boundary condition: . . . . . . . . . . . 154

C.3.2 Proper treatment of k = 0: . . . . . . . . . . . . . . . . . . . . 156

C.3.3 Finite Difference: . . . . . . . . . . . . . . . . . . . . . . . . . 160

C.4 Interpolation: getting ug at the particle positions . . . . . . . . . . . 165

C.4.1 Lagrange interpolation of degree 2 : . . . . . . . . . . . . . . . 166

C.4.2 Interpolation of global velocity: . . . . . . . . . . . . . . . . . 166

Page 12: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

x

List of Figures

1.1 Fractional survival vs. time of rats subjected to hemorrhagic shock and

not resuscitated (CON), injected with normal saline (NS), or injected

with the same volume of saline containing 10 ppm of drag reducing

polymer (DRP)1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 Schematic of suspensions of fluid-filled elastic capsules in shear flow of

a very dilute polymer solution. . . . . . . . . . . . . . . . . . . . . . . 5

2.1 Schematic of blood flow in the microcirculation, illustrating phenomena

of migration, margination and plasma-skimming. . . . . . . . . . . . . 7

2.2 Experimental data (symbols) on the thickness of cell-free layer as a

function of flow rate for a suspensions of RBCs in Newtonian (Control)

and drag-reducing polymer (DRP) solutions from Kameneva et al.2. . 11

Page 13: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

xi

4.1 (a) Deformation parameter as a function of time for single non-prestressed

Neo-Hookean (NH) capsules in shear flow in a cubic box of size 12.5a.

Symbols are simulations from Lacet al.3. Lines are results from present

work. (b) Steady state deformation parameter as a function of Ca

for single non-prestressed Neo-Hookean (NH) capsules in shear flow

in a cubic box of size 12.5a. For comparison, results of Doddi and

Bagchi4(DB), Lac et al.3(LB), Ramanujan and Pozrikidis5(RP) are

also shown. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

(a) D vs. t∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

(b) steady-state D vs. Ca . . . . . . . . . . . . . . . . . . . . . . . 39

4.2 Xi Dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4.3 Low Ca Limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

4.4 (a) Steady state values of deformation parameter at different Ca for

single non-prestressed NH and SK (C = 10) capsules (shown by bold

lines) and for preinflated NH and SK capsules (shown by dotted lines)

in shear flow. (b) Deformation parameter for preinflated NH and SK

(C = 10) capsules under shear flow in a cubic box of size 12.5a. Images

correspond to capsule shapes taken at t∗ = 10.. . . . . . . . . . . . . . 43

(a) Effect of preinflation on steady state deformation parameter . . 43

(b) Deformation parameter for preinflated NH and SK capsules . . 43

4.5 Difference in (a) deformation parameter and (b) inclination angle for

preinflated capsules in Newtonian fluid and a polymer fluid (β = 0.997)

under shear flow in a cubic box of size 12.5a. Subscript N and P

correspond to Newtonian and polymer case respectively. . . . . . . . 44

Page 14: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

xii

(a) DN −DP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

(b) θN − θP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

5.1 Schematic of pair collisions of fluid-filled elastic capsules in shear flow

of a very dilute polymer solution. . . . . . . . . . . . . . . . . . . . . 47

5.2 Pair collisions of preinflated NH capsules in a Newtonian fluid . . . . 48

(a) verification with Lac et al.6 . . . . . . . . . . . . . . . . . . . . 48

(b) effect of screening parameter α . . . . . . . . . . . . . . . . . . 48

5.3 Relative trajectories for pair collisions of preinflated NH capsules in a

Newtonian fluid at different Ca. Images correspond to snapshots taken

at collision (∆x = 0). . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5.4 Relative trajectories for pair collisions of preinflated NH and SK cap-

sules (C = 10) in a Newtonian fluid at different Ca. Images correspond

to snapshots taken at collision (∆x = 0). . . . . . . . . . . . . . . . . 50

5.5 Deformation parameter as a function of relative separation in the stream-

wise direction (∆x) for pair collisions of pre-inflated NH (Ca = 0.142)

and SK (Ca = 0.60, C = 10) capsules in a Newtonian fluid. . . . . . . 51

5.6 Paircollision Polymer Effect . . . . . . . . . . . . . . . . . . . . . . . 53

5.7 Pair collision with different initial Separation . . . . . . . . . . . . . . 54

5.8 Variation of relative trajectories with non-dimensional area dilation

modulus, C, for SK (Ca = 0.60) capsules in Newtonian (thick lines)

and polymeric (thin lines, β = 0.997) fluid. . . . . . . . . . . . . . . . 55

Page 15: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

xiii

5.9 Relative trajectories as a function of (1− β) for SK (Ca = 0.60, C =

10) capsules under shear flow in a Newtonian (thick lines) and poly-

meric (thin lines) fluid. Inset shows the difference in ∆y of SK capsules

at ∆x = 8a in Newtonian and polymeric fluid, (∆yN − ∆yP), plotted

against (1− β). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

5.10 Variation of (a) maximum displacement (∆ymax −∆y0) and (b) net final

displacement (∆yfinal −∆y0) with Ca for NH and SK (C = 10) capsules

in Newtonian (thick lines) and polymeric (thin lines, β = 0.997) fluid.

Note: ∆yfinal is calculated at ∆x = 8a. The figure in the inset of (a)

shows the schematic of these displacements. . . . . . . . . . . . . . . 58

(a) maximum displacement (∆ymax −∆y0) vs. Ca . . . . . . . . . . 58

(b) net final displacement (∆yfinal −∆y0) vs. Ca . . . . . . . . . . . 58

5.11 Largest eigenvalue λmax of the deformation rate at the origin as a func-

tion of ∆x for pair collisions under various conditions. Thick lines

are from Newtonian simulations; thin lines are from simulations with

polymers, β = 0.997. . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

6.1 Schematic of suspensions of fluid-filled elastic capsules in Couette flow. 64

6.2 (a) Schematic of migration of an isolated capsule away from the nearest

wall in a Couette flow. (b) Schematic showing pair collision of the

capsules. The lines shows the trajectory of a capsule undergoing a

collision process in the shear (x− y) plane. . . . . . . . . . . . . . . . 67

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

Page 16: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

xiv

6.3 Migration of a capsule in a Newtonian fluid in a Couette flow. (a)

Trajectory of the center of mass of a capsule y as a function of time t∗

in the wall-normal direction. The walls are at y = 0 and y = By = 10a.

(b) Capsule deformation D as a function of center of mass of a capsule

y. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

6.4 (a) Difference between first (N1) and second (N2) normal stress differ-

ences as a function of y. (b) N1 − N2 evaluated at y = 2.5a (quarter

channel height) as a function of Ca. The symbols represent the simu-

lation results and the dashed line represents the exponential fit. . . . 71

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

6.5 Validation of a point dipole approximation. (a) Trajectory of the center

of mass of an isolated capsule y at Ca = 0.30 as a function of time t∗

in the wall-normal direction for different values of initial condition y0.

The walls are at y = 0 and y = By = 10a. Symbols are simulation

results and lines are the fits using eq. 6.13.(b) Trajectory of a capsule

as a function time for different values of Ca. . . . . . . . . . . . . . . 73

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

Page 17: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

xv

6.6 (a) Migration velocity umig as a function of center of mass of a capsule

y. (d) Comparison of the numerical value of the slope k obtained from

simulations (by fitting eq. 6.13) and the theoretical value obtained

using eq. 6.12, at different values of Ca. . . . . . . . . . . . . . . . . . 74

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

6.7 Migration of a capsule in a Newtonian (solid lines) and polymer (dashed

lines, β = 0.994,Wi = 20) solutions. (a) Trajectory of the center of

mass of a capsule y as a function of time t∗ in the wall-normal direction.

(b) Steady state capsule deformation D as a function of Ca. . . . . . 77

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

6.8 (a) N1 −N2 evaluated at y = 2.5a as a function of Ca. (b) Migration

velocity umig evaluated at y = 2.5a (quarter channel height) as a func-

tion of N1 −N2. The symbols represent the simulation results and the

dashed line represents the linear fit. . . . . . . . . . . . . . . . . . . . 78

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

6.9 Effect of (a) polymer concentration expressed as 1 − β at fixed Wi

(= 20) and (b) Wi at fixed β (= 0.994) on the trajectory of an isolated

capsule (Ca = 0.30) in the wall-normal direction of in a Couette flow.

Symbols are simulation results and lines are the fits. . . . . . . . . . . 81

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

Page 18: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

xvi

6.10 Migration velocity umig evaluated at y = 2.5a as a function of (c) 1−β

at fixed Wi (= 20) and (d) Wi at fixed β (= 0.994) for an isolated

capsule (Ca = 0.30). . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

6.11 (a) Snapshots of the suspensions of capsules (Ca = 0.60,φ of 0.10) in

a Newtonian fluid at t∗ = 1 (left) and t∗ = 300 (right) in a Newtonian

fluid in a cubic box of size 10a . (b) Trajectories of the center of mass

of capsules (Ca = 0.60, φ = 0.10) in the wall-normal direction as a

function of time. The walls are at y = 0 and y = By = 10a. . . . . . . 85

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

6.12 (a) Average distance from the centerline < |y−ycenter| > of suspensions

(φ = 0.10) of capsules in a Newtonian fluid in a a cubic box of size

10a as a function of time t∗. (b) Steady state distribution of capsules

(φ = 0.10) as a function of y. The walls are at y = 0, 10a and y = 5a

is the channel centerline. The walls are at y = 0, 10a and y = 5a is the

channel centerline. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

Page 19: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

xvii

6.13 (a) Average distance from the centerline < |y−ycenter| > of suspensions

of capsules (φ = 0.10, Ca = 0.60) in a Newtonian fluid in a cubic box

of size 16a as a function of time t∗. (b) Steady state distribution of

capsules (φ = 0.10, Ca = 0.60) as a function of y. The walls are at

y = 0, 16a and y = 8a is the channel centerline. . . . . . . . . . . . . 89

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

6.14 The effect of volume fraction φ on the steady state distribution of

capsules at Ca = 0.60 as a function of y. The walls are at y = 0, 10a

and y = 5a is the channel centerline. . . . . . . . . . . . . . . . . . . 90

6.15 (a) Snapshots of the suspensions of capsules (Ca = 0.30) at t∗ = 10

in a polymer (β = 0.994,Wi = 20) solutions in a cubic box of size

10a. Polymer molecules are shown as thin black lines (b) Average

distance from the centerline < |y−ycenter| > of suspensions of capsules in

Newtonian (solid line) and polymer(dashed lines, β = 0.994,Wi = 20)

solutions as a function of time t∗. . . . . . . . . . . . . . . . . . . . . 92

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

6.16 (a) Steady state distribution of capsules as a function of y in Newtonian

(solid line) and polymer (dashed line, β = 0.994,Wi = 20) solutions

in a cubic box of size 10a. (b) Steady state distribution of capsules in

the “bulk” (2.5a ≤ y ≤ 7.5a) region as a function of Ca in Newtonian

(solid line) and polymer(dashed line, β = 0.994,Wi = 20) solutions. . 93

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

Page 20: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

xviii

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

6.17 Steady state distribution of capsules (Ca = 0.60) in Newtonian and

polymer (Wi = 20) solutions with different values of β. . . . . . . . . 95

6.18 (a) Dependence of capsule-depleted layer thickness on Ca for suspen-

sions of capsules in Newtonian and polymer (Wi = 20) solutions with

different values of β in a cubic box of size 10a. Symbols are the sim-

ulation results and lines are the fits. The standard deviation is based

on results from different initial configurations. (b) Experimental data

(symbols) on the thickness of cell-free layer as a function of flow rate

for a suspensions of RBCs from Kameneva et al.7. . . . . . . . . . . . 96

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

6.19 (a) Mean squared displacement of suspensions of capsules in the wall-

normal direction at steady state in Newtonian and polymer ( β =

0.994,Wi = 20) solutions as a function as a function of time t∗ and the

corresponding short-time diffusivities (b) in the wall-normal direction

as a function of Ca. The standard deviation (error bar) is based on

results from different initial configurations. . . . . . . . . . . . . . . . 99

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

6.20 Short-time diffusivities in the wall-normal direction as a function of y in

Newtonian (solid line) and polymer(dashed line, β = 0.994,Wi = 20)

solutions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

Page 21: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

xix

6.21 Comparison of the thickness of the capsule free layer for suspensions

(φ = 0.10) of capsules in polymer ( β = 0.994,Wi = 20) solution

obtained from simulations and predicted from theory (eq. 6.16) as a

function of Ca. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

8.1 Schematic of suspensions of fluid-filled elastic capsules in a pressure-

driven flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

8.2 Migration of NH capsules in a Newtonian fluid in a pressure-driven

flow. (a) Trajectory of the center of mass of a capsule y as a function

of time t∗ in the wall-normal direction. The walls are at y = 0 and

y = By = 10a. (b) Capsule deformation D as a function of center of

mass of a capsule y. . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

8.3 (a) Difference between first (N1) and second (N2) normal stress differ-

ences as a function of y. (b) The trajectory of an isolated capsule in a

Newtonian fluid for different channel heights. Solid lines are simulation

results for channel of height By = 10a and dashed are for By = 40a. . 114

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

8.4 (a) Average distance from the centerline < |y−ycenter| > of suspensions

(φ = 0.10) of NH capsules in a Newtonian fluid in a a cubic box of size

10a as a function of time t∗. (b) Average velocity of the center of mass

of capsules in flow (x)direction as a function of time t∗. The dashed line

represents the average velocity of the undisturbed flow (〈U〉/U = 2/3) 117

Page 22: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

xx

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

8.5 (a) Steady-state distribution of the velocity of capsule’s center (sym-

bols) in flow direction (x) as a function of y in a Newtonian fluid in

a cubic box of size 10a. The dashed lines are the quadratic fits to

the symbols. The solid line (black) represents parabolic profile of the

undisturbed velocity. (b) Steady state distribution of capsules as a

function of y. The walls are at y = 0 and y = 10a. . . . . . . . . . . . 119

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

8.6 Average distance from the centerline < |y − ycenter| > of suspensions

(φ = 0.10) of NH (Ca = 0.142) capsules in a Newtonian fluid for

different channel heights. . . . . . . . . . . . . . . . . . . . . . . . . . 121

9.1 A schematic showing suspensions of RBCs in the microcirculation. . . 125

9.2 Examples of recently sythesized types of particles with potential for

drug delivery. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

9.3 A schematic showing hypothesized distributions of drug delivery par-

ticles in the microcirculation. . . . . . . . . . . . . . . . . . . . . . . 129

Page 23: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

xxi

A.1 (a) Average steady state RMS end-to-end distance < R0 > and (b)

average steady state polymer stress < τ p > as a function of Wi for

a single polymer molecule in an unbounded shear flow. x, y and z

represents flow, gradient and neutral directions respectively. “HI” de-

note simulations including hydrodynamic interactions in the Brownian

term. “FD” represents simulations neglecting hydrodynamic interac-

tions in the Browninan term. . . . . . . . . . . . . . . . . . . . . . . . 135

(a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

(b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

B.1 Schematic of different length scales . . . . . . . . . . . . . . . . . . . 141

(a) length scales associated with conventional IBM method . . . . . 141

(b) length scales associated with IBM/GGEM . . . . . . . . . . . . 141

Page 24: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

xxii

List of Tables

Page 25: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

1

Chapter 1

Introduction

Many serious medical conditions contribute to or arise from disruptions in normal

blood flow. Examples include: hemorrhage (simple replacement with fluids other

than whole blood leads to short-term improvement but often to severe long-term

consequences whose origin is not well-understood8), sepsis (systemic infection arising

from severe injury or surgical complications9), blood hyperviscosity syndromes (which

accompany many disorders and injuries, including severe burns10), vascular diseases

such as atherosclerosis (which lead to poor circulation and potentially to catastrophic

events such as heart attacks or strokes), and diabetes (one of whose long-term com-

plications is chronic peripheral vascular disease, which results in poor wound healing

in the extremities and may lead to gangrene). Various approaches exist for treatment

of these disorders. For example, many drugs used in prevention and treatment of

atherosclerosis impede formation of plaques in arteries or formation of blood clots,

but these drugs do not directly affect blood flow per se. Vasodilators and vasocon-

strictors control blood flow by inducing blood vessels to change size; vasodilators

Page 26: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

2

in particular are commonly used in the treatment of heart disease. However, these

agents only affect the blood flow in large vessels (> 100µm diameter), and they have

no known effect on blood flow distribution at the capillary level (∼ 5µm diameter)

where exchange of nutrients between tissue and blood takes place. There are other

approaches that directly modify the behavior of blood flow, especially at the micro-

circulatory level of vessels having size below 100µm. For example, the addition of a

“plasma expander” for treatment of severe hemorrhage – generally a dextran solution

– to the blood is standard8. This has the effect of increasing blood volume, decreasing

the red blood cell (RBC) concentration and thus the blood viscosity and, because of

the presence of the dextran, approximately maintaining the proper osmotic pressure

in the blood. While having significant short-term benefit, this treatment has long-

term side effects, which can include severe and frequently lethal edema in the lungs8.

For chronic disorders such as peripheral vascular disease, drugs such as pentoxifylline

(Trental R©) are commonly used. Pentoxifylline does not affect plasma viscosity or

RBC concentration, but significantly increases the deformability of the RBC, result-

ing in a decreased blood viscosity and probably also improving the trafficking of the

RBCs in the microcirculation10.

Recently, studies in animals have shown that a radically different approach to mod-

ification of hemodynamics has promising beneficial effects11,12,13,14,15,16,17,18,19,20,21,22? 23,24,2,7,1,25:

In this approach, fluid containing dissolved long-chain water-soluble polymers is added

to blood; the resulting polymer concentration in the blood is as as low as several

ppm. For example, a study on dogs26 reports, upon addition of the polymer ad-

ditive to blood, a 27% increase in cardiac output despite a significant decrease in

arterial blood pressure - from 130/80 to 110/80 mmHg, and pulse, from 180 min−1 to

Page 27: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

3

Figure 1.1: Fractional survival vs. time of rats subjected to hemorrhagic shock andnot resuscitated (CON), injected with normal saline (NS), or injected with the samevolume of saline containing 10 ppm of drag reducing polymer (DRP)1.

120 min−1, corresponding to a reduction in overall hemodynamic resistance of about

40%. Two further recent studies with dogs found that that these polymer additives

decreased pressure drop across a coronary stenosis and improved oxygen transport

to heart tissues downstream of the stenosis27,28. Another study examined the effect

of these polymer additives on small-volume fluid resuscitation of rats subjected to

massive hemorrhage7,1. As shown in Fig. 1.1, the survival rate for rats resuscitated

with polymer solution was dramatically higher – more than double – than for rats

treated with normal saline. Additionally, measurements of whole-body O2 consump-

tion and CO2 production showed rates about 50% higher for the polymer solution

treated animals compared to the saline treated ones. There are a number of points

that should be noted while considering the origin of the observed effects of polymers.

Page 28: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

4

For a given chemical composition of the polymer, but varying molecular weight, the

physiological effects were absent at low molecular weight but significant at high molec-

ular weight. Additionally, chemically different polymers (e.g. poly(ethylene oxide),

polyacrylamide, certain polysaccharides) yielded very similar results as along as the

molecular weight was sufficiently high (typically > 106 Dalton). Furthermore, the

hemodynamic effects of these additives are virtually immediate, indicating that al-

though the additives are ultimately cleared from the blood by the immune system,

the immune response occurs on a time scale much slower than the hemodynamic one.

These observations indicate that the origin of the physiological effects is physical

rather than chemical.

The polymers used in all these studies are members of a class of molecules known

as “drag reducing additives” (DRAs). When added to simple single-phase fluids

such as water at ppm levels, DRAs significantly reduce drag in turbulent flows29,30.

Flow in small blood vessels is not turbulent, so turbulent drag reduction per se is

not the explanation for the hemorrhage recovery and other results described above.

Nevertheless, during flow, the collisions of the RBCs lead to fluctuations in fluid

velocity that are somewhat analogous to the fluctuations present in turbulent flow.

The effect of these fluctuations on the dynamics of the individual polymer molecules

and the feedback from the polymer dynamics back to the dynamics of the RBCs are

completely unknown.

The aim of the present work is to take first steps toward understanding the influ-

ence of drag reducing polymers on blood flow. We will introduce a theoretical, mod-

eling and computational approach to study the dynamics of suspensions of red blood

cells (RBCs) in a polymer solution. RBCs are modelled as fluid-filled elastic cap-

Page 29: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

5

Wall

beads experience Stokes drag tomotion relative relative to

fluid

nodes move with the fluidvelocity

γ

By

x

y

z

2a

Wall

Figure 1.2: Schematic of suspensions of fluid-filled elastic capsules in shear flow of avery dilute polymer solution.

sules (liquid drops enclosed by a solid elastic membrane) and polymer molecules are

simulated as bead spring chains with parameters chosen to model 4×106 D polyethy-

lene oxide (PEO). We present a novel, highly efficient formulation of the immersed

boundary method for Stokes flow. Specifically, we will highlight the effect of polymer

molecules on a number of important phenomena exhibited by the blood cells in the

microcirculation, with an ultimate aim to shed lights on the dynamics of complex

multiphase fluids. Fig. 1.2 illustrates the basic situation of interest.

Page 30: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

6

Chapter 2

Background

2.1 Dynamics of blood flow in the microcirculation

The so called microcirculation is defined by vessels having diameter less than 100µm.

A red blood cell has an average diameter of about 8µm. The viscosity of plasma is

around 1.2cP31,32. For the flow in the microcirculation, with shear rates ranging from

100s−1 to 1000s−1, the Reynolds number is much less than unity. Many important

phenomena are observed which are illustrated in Fig. 2.1. The RBCs migrates away

from the wall towards the center of the vessel. This leads to inhomogeneous distribu-

tion of RBCs – having higher concentration at the core of the vessel and a formation

of several microns thick cell-free layer. The thickness of this layer appears to grow

slowly with increasing flow rate33,2 and is assumed to follow an approximate relation-

ship : thickness ∼ (flowrate)n, with n reported to be in the range of 0.3 − 0.533,34.

The origin of this relationship is not known. The migration of RBCs towards the

center of a vessel is assumed to increase the probability of less deformable leucocytes,

Page 31: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

7

red blood cell

marginated leucocytecell−free layer

side branch:plasma−skimminglowers hematocrit

Figure 2.1: Schematic of blood flow in the microcirculation, illustrating phenomenaof migration, margination and plasma-skimming.

such as white blood cells and platelets, to be found near the blood vessel walls, a

phenomenon known as margination. Margination is found to play a key role in the

process of inflammation35,36. Due to the presence of the cell-free layer, fewer blood

cells are drawn from the nearby wall region of large blood vessels in to the smaller

branching capillary and is hypothesized to be the cause for the lower hematocrit in

these capillaries – a phenomenon known as plasma-skimming effect. In suspensions,

the cells collide with each other intermittently. These collisions lead to substantial

velocity fluctuations which drives the diffusive motion of cells and solutes in blood

flow.

A number of mechanisms may contribute to the formation of non-uniform distri-

bution of RBCs. A deformable particle (capsule, drops, vesicle) migrates away from

Page 32: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

8

the wall even at zero Reynolds number37,38,39. Migration arises due to the distur-

bance velocity in the fluid caused by the deformable particle as tries to relax to its

equilibrium position. This disturbance velocity is asymmetric due to the presence

of a nearby wall and tends to push the particle away from the wall. A deformable

particle, to the leading order, can be treated as a point dipole and the wall-induced

migration effect arises due to the disturbance velocity in flow caused by the image of

the point dipole on the other side of the wall. In suspensions, the cells do not migrate

continuously towards the center40.

In suspensions, the cells do not migrate continuously towards the center. The

wall-induced migration is balanced by the diffusive motion of the cells, leading to the

development of cell depleted layer. Although there has been considerable progress

towards understanding the mechanisms of wall-induced migration and shear-induced

diffusion of these deformable particles, the balance of these effects that lead to the

development of cell-depleted layer and concentration distribution remain poorly un-

derstood. Migration is also observed in suspensions of rigid particles but it can occur

only through particle interactions and at high concentrations. The net flux comes

from the contribution of gradient in shear rate and the diffusion due to concentration

gradient41,42. This shear-induced particle drift in this case also leads to the blunt-

ing of the velocity profiles43. For the case of dilute polymers, where the Brownian

diffusion is substantial, the polymer depleted layer comes from the balance between

wall-induced migration and Brownian diffusion44. In suspensions of cells, the Brow-

nian diffusion of the cells are negligible where as the shear-induced diffusion, due to

random collisions, is substantial. For the case of dilute suspensions of deformable

drops, analogous to the suspensions of deformable cells, King and Leighton45 fol-

Page 33: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

9

lowed by Hudson46 performed a theoretical analysis on the spatial distribution of

drops accounting for wall-induced migration and shear-induced diffusion. Their anal-

ysis predicted a drop-free region analogous to the cell-free layer observed in the case of

blood flow. However, no analytical or asymptotic form was obtained for the drop-free

layer. As part of the thesis, we derive a simple variant of the these theories, which

are based on a fundamental understanding of stochastic processes, to describe the

cross-stream migration of deformable particles in suspensions.

2.2 Drag reducing additives in blood flow

As noted above, the in vivo experiments on animals show potential beneficial effects

on hemodynamics arise from addition DRAs to blood. The effects of DRAs on blood

flow have also been observed in in vitro experiments2,7. Of particular interest is the

work done by Kameneva et al.2, who studied the dependence of cell-free layer with

flow rate in their microchannel experiment of blood. They showed that the addition

of small amount of these polymer additives to the suspensions of RBCs resulted in

a redistribution of RBCs with a significant reduction in the thickness of the cell-

free layer. As shown in Fig .2.2, addition of concentration of drag-reducing polymer

as low as 10ppm can cause significant reduction in the thickness of cell-free layer.

Additionally, addition of these polymer additives also reduce the concentration of

leucocytes in the cell-free layer47.

The effects of DRAs on both turbulent flow and blood flow are determined by

their physical nature as long-chain flexible linear polymers rather than their specific

chemical composition. At sufficiently high molecular weight, concentrations of the

Page 34: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

10

order of 10 ppm are sufficient to induce reductions in turbulent drag of 50% or more,

and concentrations at this same level lead to the physiological effects described above.

At these concentrations, the shear viscosity of the fluid (with or without blood cells)

is nearly indistinguishable from that without polymers. In contrast, the extensional

viscosity might be orders of magnitude larger than the polymer contribution to the

shear viscosity and, more importantly, even larger than the viscosity of the solvent.

In particular, the degree of drag reduction in turbulent flow is found to correlate quite

well with the ratio between extensional and shear viscosities, the Trouton ratio. The

origin of the differences between shear and extensional viscosities of DRAs is closely

associated with their long-chain nature48.For example, consider a molecule of 4000

kD poly(ethylene oxide), or PEO. This molecule has a fully extended length L of

about 30 µm. In quiescent solution, however, it exists in a random coil configuration

with an RMS end-to-end distance R0 of only a few hundred nanometers. Since a

red blood cell has dimensions on the order of several microns, we see that there is

no separation of length scales in this situation between the polymers and the cells.

The relaxation time λ, for this molecule, estimated as the time the molecule takes

to diffuse its own radius in water, is of the order of 10 ms. Given a characteristic

strain rate γ for the flow (e.g. an estimate of shear rate or extension rate), one can

define the Weissenberg number Wi. Only when Wi & 1 will polymer chains stretch

significantly in flow. How much the polymer chains stretch depends on the specific

nature of the flow. For an extensional flow with Wi & 1, the Trouton ratio is roughly

proportional to (L/R0)2 , which for 4000 kD PEO is > 104. Thus a solution of this

polymer in which the polymer makes a negligible contribution to the stress in shear

flow can exhibit large stresses in extensional flow.

Page 35: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

11

Flow Rate (ml/min)

Cel

l−F

ree

Laye

r (µm

) ControlDRP, 10ppm

Figure 2.2: Experimental data (symbols) on the thickness of cell-free layer as a func-tion of flow rate for a suspensions of RBCs in Newtonian (Control) and drag-reducingpolymer (DRP) solutions from Kameneva et al.2.

These ideas can be applied to blood flow in the microcirculation. If we estimate

the velocity U of blood in a vessel of radius R = 16 µm to be 12 mm/s (using data

from cat mesentery given by Fung32), then the Reynolds number for blood in this

vessel is much less than unity. On the other hand, using U/R as an estimate of γ,

we find for the 4000 kD PEO described above a Weissenberg number of around 10, is

large enough to significantly stretch polymer chains. It is important to note that only

in the microcirculation are strain rates likely to be large enough to significantly stretch

DRAs. Furthermore, local shear rates in the gap between an RBC and an arteriole

or capillary wall may be very large, and collisions between RBCs during flow lead

to substantial velocity fluctuations and in particular to transient extensional flows

in the neighborhood of colliding cells. By analogy with the role of fluctuations in

Page 36: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

12

turbulence, we hypothesize that these fluctuations significantly affect the stretching

of the polymer molecules and correspondingly their effects on the flow.

2.3 Simulations of blood flow

As noted above, the Reynolds number for the flow in the microcirculation is small,

hence the flow is governed by the Stokes equation. At the length and time scales of

interest, the details of the molecular structure of a cell can be neglected and the entire

cell can be modeled as a fluid-filled elastic capsule, which is a fluid droplet enclosed

by a solid elastic membrane. There are many computational studies on the dynamics

of capsules in flow5,49,50,51,52,53,54,55,3,56,6,57,58,4,37,59. Most of these studies of capsule

dynamics have focused on the behavior of individual capsules in flow; the work of

Lac et al.3 is of particular relevance to the present work. These authors studied

the dynamics of individual capsules in unbounded shear and extensional flow at low

Reynolds number. Fluid viscosities inside and outside the capsule were matched, with

value ηs. Two models of the membrane elasticity were used: the neo-Hookean (NH)

model, which approximates the behavior of a cross-linked rubber, and a model (SK)

proposed by Skalak and coworkers60 to describe the elasticity of a red blood cell. For

the NH model, the surface shear modulus G completely parameterizes the elasticity,

while for the SK model an additional parameter C appears, which represents the

energy penalty for area change. These models will be described in section 3.2 in

further detail. For cell membranes, C � 1. The ratio of viscous to elastic stresses

is measured by a capillary number Ca = ηsγa/G, where a is the equilibrium capsule

radius. Simulations were performed with the boundary element method51 and values

Page 37: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

13

of C between 0.5 and 10 were considered. In an intermediate range of Ca around unity,

the deformations of the capsule were qualitatively similar to those observed for a liquid

drop. Potentially important distinctions arose at lower or higher Ca, particularly in

planar extensional flow. At small Ca, the surface wrinkles due to a buckling instability.

Lac et. al.56 have reported that the instability is due to their numerical method (use

of B-spline polynomials). Li and Sarkar61 also reported that wrinkling is numerical

rather than physical; their simulated wrinkles depend on resolution. This wrinkling

instability can be prevented by slightly inflating the capsule so that there is a small

pre-stress in the membrane even at equilibrium (see Lac and Barthes-Biesel56). At

large Ca, the interface develops regions with very large curvature. In the numerical

method of Lac et al.56, these situations lead to loss of existence of steady solutions. Li

and Sarkar61 also reported stability problems at high Ca. This loss of existence does

not occur in the smoothed spectral boundary element simulations of Dodson and

Dimitrakopoulos58 or in the noval coupling method of Walter et. al.62, indicating

that the loss of existence is numerical rather than physical. At a given value of

Ca, as C increases, the degree of deformation of the membrane decreases, and the

range of Ca where steady solutions can be obtained dramatically increases. Similar

observations are found in simulations of uniaxial extension. In planar extension,

related work by Dodson and Dimitrakopoulos predicts the possibility of multiple

steady state configurations of the capsule58.

The dynamics of pair collisions of capsules in shear has also been studied. Bagchi

et al.63 studied this problem in the context of evaluating models of aggregation of

RBCs. Their simulations were two-dimensional, using a NH model for the mem-

brane, a Lagrangian finite element discretization of the membrane first presented by

Page 38: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

14

Charrier et al.64 and a front-tracking immersed boundary method (IBM) for the fluid

dynamics65,66. They predicted that increasing internal viscosity or membrane rigid-

ity – both of which suppress cell deformation – favor aggregation, in agreement with

physiological observations. Doddi and Bagchi4 performed a related study in three-

dimensions to determine the effect of inertia on pair collisions of capsules. Jadhav et

al.67 performed three-dimensional simulations, also using the NH model and IBM for

the fluid dynamics of collisions during shear, focusing on the shape of the “contact

region” between the colliding capsules: i.e. the region where the capsules are in clos-

est proximity. At low capillary number, the two capsules remain convex throughout

the collision and this region is an elongated disk. At high capillary numbers however,

the capsules each display a dimple in the collision region that arises due to the high

pressure generated by the squeezing flow in that region. Correspondingly, the contact

region becomes a distorted annulus. Lac et al.57,6 examined pair collisions of NH

capsules, focusing on the net displacements due to collisions – i.e. the displacements

that lead to shear-induced diffusion in suspensions. The main result of these studies

was that although capsule shapes in shear appear qualitatively similar to those of

drops, the collision dynamics of capsules and drops can be qualitatively different.

Recent advances in numerical methods have enabled researchers to study large

scale simulations of deformable capsules68,69,70. Of particular interest is the work

done by Freund34, who used an accelerated boundary element formulation to study

the margination of leucocytes in presence of deformable red blood cells in a 2-D mi-

crovessel. Through his model, he was able to predict the blunting of velocity profiles

(plug-flow behavior) and observed the formation of a cell-free layer the thickness of

which increased with the flow rate. Accelerated boundary integral formulations69 are

Page 39: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

15

gaining considerable interest to be implemented to study multiscale simulations of

deformable particle. Doddi and Bagchi70 used an immersed-boundary formulation

to perform three dimensional simulations of multiple deformable cells (capsules and

RBCs) in microchannel. Through their simulations, they showed the blunting of ve-

locity profiles, the cell-free layer dependence on the hematocrit and the non-monotonic

behavior of effective viscosity on vessel diameter (Farhaeus-Lindqvist effect). Other

methods such as lattice Boltzmann method (LBM) have also been used to study the

suspensions of RBCs71,72,73 but the focus of these studies have been on the rheology

rather than the dynamics of suspensions of cells. Recent numerical simulations of

pair collisions of capsules in polymer solutions of our current work (Pranay et al.74)

showed dramatic effect of DRAs on collision dynamics – polymers suppressed the

net displacement of capsules after the collision. The uniaxial extensional flow gener-

ated in the gap of the colliding capsules stretched the polymers significantly and the

stretching worked against the separation of the departing capsules after the collision.

To our knowledge, no numerical study of the dynamics of suspensions of capsules in

polymer solutions has been performed.

The remainder of this work is organized as follows. Chapter 3 describe the models

and discretization methods used for the capsules and polymer molecules. Chapter 4

reports results for single capsules in shear, as background for Chapter 5, which de-

scribes and discusses pair collisions in Newtonian fluids and polymer solutions. Chap-

ter 6 describes and discusses the dynamics of suspensions of capsules in Newtonian

fluids and polymer solutions in a Couette flow. Specifically in Section 6.1 we describe

the theory for the suspensions of capsules as a balance of two competing mechanisms

and derive the closed-form expression for the capsule-free layer and Chapter 6.2 re-

Page 40: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

16

ports results for the migration of a single capsule in a Couette flow in Newtonian

and polymer solutions. Chapter 7 provides a brief summary of the current study.

Chapter 9 shows some preliminary work on pressure-driven flows along with some of

the proposed future work with theory. Appendix A reports the validation for using

an approximate treatment of the Brownian motion of polymer molecules. Appendix

B elucidates the relationship between the present approach to velocity computations

and conventional immersed boundary and boundary integral methods. Appendix C

provides the details of implementation of the stokes flow solver (GGEM) used in the

current study.

Page 41: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

17

Chapter 3

Models and discretization methods

3.1 Overview

We consider the motion of liquid capsules and polymer molecules, immersed in a

surrounding fluid of viscosity ηs, under shear flow with strain rate γ between two

parallel plates as shown in Figure 1.2. The simulation box is periodic in the flow

(x) and vorticity (z) directions. The wall-normal direction is y. The lengths of the

domain in x, y and z are Bx, By and Bz, respectively. Time is non-dimensionalized

with 1/γ as t∗ = γt. Each capsule is comprised of a nominally spherical elastic

membrane of radius a, enclosing a Newtonian liquid with density and viscosity equal

to that of the outside fluid. The membrane has a two-dimensional elastic modulus

G and negligible thickness, and we characterize the capsule dynamics as a function

of capillary number Ca = ηsγa/G. As detailed below, two models for the membrane

elasticity will be studied. The polymer molecules are modeled (as detailed below)

as bead-spring chains, with three beads connected together by finitely extensible

Page 42: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

18

non-linear elastic (FENE) springs. The maximum end-to-end length of the polymer

molecules is sufficiently large to mimic a high-molecular weight (> 106 Da) polymer,

the details of which will be described in section 3.3. The polymer contribution to

the fluid viscosity is denoted ηp. The relevant dimensionless numbers for the polymer

molecules are the Weissenberg number, Wi, the viscosity ratio β = ηs/(ηs + ηp), and

the extensibility parameter, Ex, defined as the steady state Trouton ratio in the limit

of high extension rate. We consider the case of very dilute long-chain polymers, where

1−β � 1 but Ex> 1, so that polymer effects are negligible in pure shear flow but can

be important in flows with a substantial extensional component. Finally, the relative

sizes of polymer chains and capsules is important. The polymer molecules have a

mean equilibrium end-to-end distance of about 450 nm and fully extended length 30

µm, while the equilibrium radius a of the capsules is 3 µm.

3.2 Capsule Model

We use two different models for the capsule membrane. The first is a neo-Hookean

model, which mimics the behavior of rubber-like materials, and the second is a model

of the red blood cell (RBC) membrane that was originally proposed by Skalak et al.60.

The Skalak model can be parametrized to yield a strong resistance to area change

relative to its resistance for shear deformation. The mathematical details of the two

models and their application will be discussed after we introduce some formalism for

describing the kinematics of membrane deformation.

This formalism is mostly clearly presented for deformations in a plane, which for

the moment we take to be the xy plane. Let (x, y) and (X, Y ) denote respectively

Page 43: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

19

the undeformed and deformed coordinates of a material point, with respect to a fixed

set of Cartesian axes. If u and v denote the displacements of the material point in x

and y directions respectively, then

X = x+ u,

Y = y + v.(3.1)

The relation between the position vectors of two points infinitesimally close to each

other, before and after deformation is given by

dX

dY

=

1 + ∂u∂x

∂u∂y

∂v∂x

1 + ∂v∂y

dx

dy

, (3.2)

or compactly as

dX = F · dx. (3.3)

The square of the distance between the two neighboring points after deformation is

given by

dS2 = dX · dX = dx ·G · dx,

G = FT · F,(3.4)

where G is a symmetric, positive definite matrix. The elements of G are given by

the expressions

G11 =

(

1 +∂u

∂x

)2

+

(

∂v

∂x

)2

,

G22 =

(

∂u

∂y

)2

+

(

1 +∂v

∂y

)2

,

G12 = G21 =

(

1 +∂u

∂x

)(

∂u

∂y

)

+

(

1 +∂v

∂y

)(

∂v

∂x

)

.

(3.5)

Page 44: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

20

The principal stretch ratios, λ1 and λ2, are defined as the eigen values of G, and are

given by the expressions,

λ21 =

12

(

G11 +G22 +√

(G11 −G22)2 + 4G2

12

)

,

λ22 =

12

(

G11 +G22 −√

(G11 −G22)2 + 4G2

12

)

.(3.6)

For a thin membrane that displays no resistance to bending, the strain energy

density W of the membrane is a function of λ1 and λ2. Following Barthes-Biesel et

al.55, for a neo-Hookean model the strain energy density function is given by

WNH =G

2

[

I1 − 1 +1

I2 + 1

]

. (3.7)

Here G is the two-dimensional shear modulus for the membrane, having units of force

per unit length. The two invariants, I1 and I2 are given by

I1 = λ21 + λ2

2 − 2, I2 = λ21λ

22 − 1. (3.8)

The Skalak model60 has the strain energy density

WSK =G

4

[(

I21 + 2I1 − 2I2)

+ CI22]

. (3.9)

The Skalak model contains a shear modulus G and an additional parameter C associ-

ated with the energy penalty for area change; the area dilation modulusK is related to

the shear modulus G (using the nomenclature of Skalak et. al.60) as K = 2G(1+2C).

Typically, C � 1 indicating approximate area-incompressibility.It has been shown55

that under a simple uniaxial deformation, results for the Skalak model reach an

Page 45: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

21

asymptotic value for C ≥ 10. We will be considering C = 0, 1 and 10 below, focus-

ing on the case C = 10. We adopt the finite element method developed by Charrier

et al.64 to describe the surface of the deformable particle, and a variant of the im-

mersed boundary method75 to describe the fluid-structure interaction. These choices

result in a discretized description that is first-order accurate in the element size. In

this approach, the membrane is discretized into flat triangular elements, in which the

strain is uniform and which are assumed to remain flat even after deformation. The

element corners, or nodes, are taken to move with the local fluid velocity as required

by the no-slip boundary condition. That is,

dxci

dt= u (xc

i) , (3.10)

where xci is the position of the ith node and u (xc

i) the fluid velocity evaluated at that

node. This expression is integrated with a second-order Adams-Bashforth method.

Results presented here are computed with a time step ∆t of 5·10−3. As is conventional

in immersed boundary descriptions of fluid-structure interaction75, the forces exerted

by the fluid on the membrane and vice versa are taken to be localized at these

nodes. Since the inertia and Brownian fluctuations of the membrane are taken to

be negligible, the total force exerted on any point on the membrane is equal to zero

at any instant. Thus, the elastic membrane forces f ci on each node of the capsule have

to be balanced by the hydrodynamic forces fhi exerted on that node:

f ci + fhi = 0, (3.11)

Page 46: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

22

The reaction force exerted by the node on the fluid is thus given by

fhf (xci) = −fhi = f ci . (3.12)

The force exerted by the nodes on the fluid enters into the computation of the fluid

velocity field, as described in Section 3.4.

The description of the capsule motion is completed by the computation of the

nodal forces f ci . In the Charrier et al.64 approach, these forces are determined by

computing displacements of the vertices of the deformed elements with respect to the

undeformed elements and applying the principle of virtual work. To facilitate the

determination of the principal stretch ratios λ1 and λ2, each element in the deformed

state and the corresponding un-deformed element are transformed to a plane by

rigid body rotations, using a transformation matrix Rα, for each element α. The

deformation of each element is then calculated using the positions of the nodes in the

deformed state relative to their positions in the undeformed state. The deformation

at any point inside the element is calculated by interpolating linearly from the nodes.

The principal stretch ratios can then be calculated from the nodal displacements.

Having chosen a suitable membrane strain energy density function, and having

calculated the nodal displacements, the forces at node i of an element along two per-

pendicular directions in the plane of the element are calculated using the expressions

fLx,i = −Ae

[

∂W∂λ1

∂λ1

∂ui+ ∂W

∂λ2

∂λ2

∂ui

]

,

fLy,i = −Ae

[

∂W∂λ1

∂λ1

∂vi+ ∂W

∂λ2

∂λ2

∂vi

]

.(3.13)

Here, fLx and fL

y are the nodal forces along the two perpendicular directions, x and

Page 47: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

23

y, the “local” or transformed co-ordinate system, and Ae is the undeformed area of

the element. After computing the deformation forces in the transformed co-ordinate

system, the global components of the nodal forces for element can be calculated by

transforming back to the global Cartesian coordinates using the expression fαi =

(Rα)T · fLi . The resulting forces are the nodal forces with respect to a fixed Cartesian

(global) coordinate system. The total elastic force on a capsule node is calculated as

the sum of forces resulting from the deformations of triangular elements surrounding

that node, and is given by

f ci =∑

α

fαi , (3.14)

where the summation is over all triangular elements to which the node belongs.

3.3 Polymer Model

Each polymer molecule in our simulations is described using a coarse-grained model-

ing approach appropriate for highly flexible polymer molecules76,77. In this approach,

each molecule is represented as a string of beads connected by springs. Each bead

represents a large number of polymer segments and has a given Stokes law friction

coefficient ζ ; corresponding drag and Brownian forces are exerted on it by the fluid.

The springs connecting the beads reflect the connectivity of the polymer chain and the

effect of entropy in driving the chains toward an equilibrium random coil conforma-

tion. The simplest nontrivial bead-spring chain model (and the least computationally

expensive) is a dumbbell model with only two beads. This captures the longest time

and length scales of the chain and in many cases is adequate for qualitative prediction

Page 48: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

24

of flow properties of polymer solutions. However, in the situation under consideration

at present, the fully extended length of a polymer chain is substantially larger than

a capsule – if we used a dumbbell model in this case, it would be highly probable

that although the two beads of the dumbbell remain outside the capsules, the spring

connecting the beads would pass right through a capsule. To minimize the possibility

of such an unphysical event, we use a three-bead chain. We now turn to specific

aspects of the polymer model.

The polymer is very loosely modeled after a 4×106 D poly(ethylene oxide) (PEO)

molecule. The number density of molecules is n. We take the contour length of

each molecule, LC , to be 30 µm, with a Kuhn length, l = 1 nm corresponding to

nk = LC/l = 3×104 Kuhn segments per polymer molecule. The spring force exerted

on each bead by its nearest neighbor(s) is given by the so called FENE (Finitely

Extensible Nonlinear Elastic) force law77:

fpi,i±1 = −Hqi,i±1

1−( qi,i±1

L

)2 , (3.15)

where H is the spring constant, qi,i±1 = xpi±1 − xp

i is the vector connecting the two

beads, and L = LC/2 is the contour length of each chain segment. The spring constant

H is calculated using the relation

H =3kBT

R20,s

, (3.16)

where kb is Boltzmann’s constant, T is temperature and R0,s is the average end to end

distance of each chain segment at equilibrium. Thus the total spring force exerted on

Page 49: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

25

each bead is

fpi = fpi,i+1 + fpi,i−1. (3.17)

For beads at either end of the chain, one of these terms will be missing. Assuming

good solvent conditions, which is valid for PEO in water, the average end to end

distance is estimated using:

R0,s ≈(

nk

Nb − 1

)35

l = 320 nm, (3.18)

where Nb is the number of beads in the chain (here it is always three). For a three-

bead chain this expression yields an RMS end-to-end distance R0 =√

2R20,s = 452

nm. The bead radius ab is taken to be 130 nm, and the corresponding Stokes drag

coefficient is

ζ = 6πηsab. (3.19)

For chains with such small numbers of beads, it is reasonable to estimate dynamic

properties with the Rouse model77. The polymer contribution to the viscosity is

estimated from this model to be

ηp =2nkBTζ

3H, (3.20)

and the stress relaxation time as

λ =ηp

nkBT= 6.7ms. (3.21)

Page 50: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

26

The Weissenberg number is given by

Wi = λγ, (3.22)

where γ is the shear rate. In our polymer simulations, Wi = 5. The parameter

β =ηs

ηs + ηp(3.23)

measures the ratio of the viscous stress to the total stress in simple shear. The

polymer concentration is proportional to 1 − β. The extensibility parameter Ex

measures the steady state ratio between polymer stresses and viscous stress in uniaxial

extensional flow in the limit of high extension rate. For a three-bead FENE chain,

with hydrodynamic interactions between beads neglected, it is straightforward to find

a closed form expression for this parameter:

Ex =nζL2

C

12ηs(3.24)

In our simulations with polymers, we use β = 0.9985, 0.997 and 0.994, which

correspond to number density n of 0.0235µm−3, 0.0470µm−3 and 0.0940µm−3, or a

mass fraction of 0.157 ppm, 0.314 ppm and 0.628 ppm respectively. The numbers of

polymer chains in the simulation domain for these three cases are 2300, 4600 and 9200,

respectively. These are very low concentrations, much smaller than typically used in

experiments. These were chosen to keep the cost of the computations reasonable,

while still allowing polymer effects to be substantial. In particular, 1− β � 1 for all

situations studied here, indicating that in shear flow, the stresses due to the polymer

Page 51: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

27

chains will be very small. On the other hand, Ex ranges from 4.3 to 17.2, indicating

that in extensional flow, polymeric extensional stresses can be much larger than the

viscous extensional stresses. So although the polymer concentration is very low, we

are still in the regime of primary interest for drag reducing polymer solutions: the

situation where the shear viscosity of the solution is barely changed by the polymer

additives, but the extensional viscosity is changed substantially.

The equations of motion for each bead of the chain is determined by the balance

of Stokes drag between the bead and the surrounding fluid, the spring forces acting

between neighboring beads, and the Brownian forces exerted by the fluid on the

bead. It is important to emphasize that, in contrast to the situation for nodes on

the surface of the capsules, the beads are not taken to move with the fluid velocity.

In other words, we do not aim to resolve the details of the fluid motion on the scale

of each polymer bead, but rather use a standard coarse-grained description of the

interaction between the polymer and the surrounding fluid. As we see below, this

difference between the treatment of the capsule-fluid and polymer-fluid interaction

leads to some subtleties in the computation of the fluid velocity (see Figure 1.2)

Turning to the Brownian force, the fluctuation-dissipation theorem implies a non-

trivial coupling between the Brownian motion and the configurations of the polymer

molecules and capsules. This coupling is expensive to compute78. In the present sim-

ulations we are interested only in dilute solutions at high Weissenberg number, where

Brownian motion is relatively unimportant (Weissenberg number is proportional to

particle Peclet number). Therefore, we do not compute the full Brownian term, but

rather apply to each bead the Brownian displacement it would exhibit in isolation

in an unbounded domain. Under this approximation, the force balance leads to the

Page 52: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

28

following evolution equation79 for polymer bead positions xpi :

dxpi =

[

u (xpi ) +

1

ζfpi

]

dt+

2kBT

ζdW, (3.25)

where u (xpi ) is the fluid velocity at the position of bead i and dW is a vector of

independent random variables, each chosen from a Gaussian distribution with zero

mean and variance dt. This equation is integrated with the stochastic explicit Euler

method79 . The same time step is used for this equation and the evolution equation

for capsule node positions.

3.4 Fluid velocity calculation

The fluid velocity u (x) is driven by the imposed velocities of the top and bottom

walls and by the forces that the capsules and polymer beads exert on the fluid. The

forces exerted by the polymer beads are localized at the bead positions and the forces

exerted by the capsule are localized at the nodal positions. In both cases we will

treat these localized forces as regularized delta functions as is conventional in polymer

dynamics79 and in immersed boundary methods for fluid-structure interactions65,75,80.

The method now described for determining the interaction between capsules and

fluid motion is a variant of the immersed boundary method that takes advantage

of a recently-developed algorithm78 for efficiently computing Stokes flow driven by

regularized point forces in arbitrary geometries. This approach can also be formulated

starting from the boundary integral equation for a deformable particle in flow; the

relationships between the present method and the conventional immersed boundary

and boundary integral methods are described in the Appendix B.

Page 53: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

29

As mentioned above, we describe the forces exerted by the membranes and poly-

mers as a distribution of regularized point forces: i.e. as a force density distribution

ρ(x) in the fluid given by:

ρ(x) =nodes∑

i=1

δc (x− xci) f

ci (x

ci) +

beads∑

i=1

δp (x− xpi ) f

pi (x

pi ) , (3.26)

with regularized delta functions that have a quasi-Gaussian form:

δc (r) = ξ3cπ3/2 e

(−ξ2cr2) [5

2− ξ2c r

2]

,

δp (r) =ξ3p

π3/2 e(−ξ2pr

2) [52− ξ2pr

2]

.(3.27)

Here f ci and fpi are the forces exerted by nodes (at positions xci) and beads (at positions

xpi ) respectively, and ξc and ξp are the corresponding regularization parameters for

the delta functions; their reciprocal represents the length scale over which the force

is spread.

Consider first ξc, the regularization parameter for points on the capsule surface.

The approach that we are using for the capsule dynamics is essentially a Stokes

flow/Green’s function-based variant of the IBM developed by Peskin65,75,80. In this

method, force distributions at moving interfaces or membranes are discretized as

distributions of regularized point forces, where the length scale for smoothing the

delta function scales as the grid spacing used in the simulation. That is, if hc is a

characteristic node spacing (or element size) on the surface, then we choose

ξc ∼ h−1c (3.28)

This ensures that the force density associated with each node is spread over the

Page 54: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

30

length scale of the associated elements, thereby preventing fluid from penetrating the

membrane surface, and also preventing the unphysically large fluid (and thus node)

velocity that would be present at the node if ξc is made too large (in which case δc(r)

approaches a true delta function). As with the conventional IBM, the simulation

results are insensitive to the choice of the regularization parameter provided its value

lies within a range where ξch = O (1) . Computations to determine the specific value

of ξc are presented in the results section.

For the polymer beads, the choice for ξp is straightforward and scales as the size of

the bead radius, ξp ∼ a−1p . Further, the pair mobility tensor will be positive definite

for ξ−1p ≥ 3ap/

√π. We take this value here.

Having specified the form of the force density associated with the capsule and

polymer molecules, we turn to determining the velocity field driven by that density.

This velocity u is the solution of Stokes’ equations,

−∇P (x) + η∇2u(x) + ρ(x) = 0,

∇ · u(x) = 0(3.29)

subject to periodic boundary conditions in x and z and a no-slip boundary condition

on y = ±By/2:

u(y = ±By/2) = (±γBy/2) ex. (3.30)

Eq. C.1 needs to be solved at each time step with a different force density. To

accomplish this, we use a particle-particle/particle-mesh method that has recently

been developed for Stokes flow problems in nonperiodic geometries, called GGEM

(General Geometry Ewald-Like Method)78,81. We will denote the combination of

Page 55: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

31

GGEM with our immersed boundary treatment of the fluid-surface interaction as

IBM/GGEM. Its specific implementation for the present situation is now described.

In what immediately follows, no distinction is made between nodes on a capsule and

polymer beads. The differences between the two will be described in detail later.

GGEM starts with splitting the regularized force density, Eq. 3.26 into two parts,

ρ(x) = ρl(x) + ρg(x),

ρl(x) =∑

i

[δγi (x− xi)− g (x− xi)] fi,

ρg(x) =∑

i

g (x− xi) fi,

(3.31)

where the sums are over both polymer beads and capsule points, so γi = c or p

depending on whether point i is a capsule node or polymer bead. The “screening

function”, g (x) is a regularized delta function that is used to split the force density

into a “local” and “global” part denoted by the subscripts l and g such that the

velocity field driven by ρl(x) decays very rapidly away from the location of each

point force. This function g (x) takes the same mathematical form as the regularizing

functions and is given by

g (x) =α3

π3/2e(−α2r2)

[

5

2− α2r2

]

, (3.32)

where α is the so-called screening parameter and r = |x|. Formally, this parameter

is arbitrary; it is chosen to yield a computationally efficient algorithm. By linearity,

the flow field is a superposition of flow driven by ρl(x) and ρg(x):

u(x) = ul(x) + ug(x). (3.33)

Page 56: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

32

Somewhat analogously to Ewald summation methods (see e.g. Sierou and Brady82),

the general idea in GGEM is to separately compute the velocity fields due to the

two force densities. The total velocity is then simply the sum of the individual

contributions.

Consider first the local velocity ul(x), which results from the force density ρl(x).

This is computed analytically using solutions to Stokes’ equations in an unbounded

domain – the error incurred by assuming an unbounded domain will be cancelled

out in the global solution described below. One can show that the modified force

distribution, Eq. (B.8) yields a velocity field given by

ul(x) =∑

i

GR,γil (x− xi) · fi, (3.34)

with the local or screened Green’s function GR,γl given by

GR,γl (x) =

1

8πηs

[

δ +xx

r2

]

[

erf (ξγr)

r− erf (αr)

r

]

− 1

8πηs

[

δ − xx

r2

]

[

π1/2e−α2r2 − 2ξγ

π1/2e−ξ2γr

2

]

, (3.35)

where as previously, γ = c or p. Because of the presence of the screening function

in the local force density, this function decays rapidly to zero, as exp(−α2r2). The

calculation of the local velocity filed at a given point x begins by first identifying

point forces located within a sphere of fixed radius 4/α. Any force outside this cut-off

radius is ignored. The choice of α is discussed below, and is constant for a given

simulation. For any point x in the fluid which is not on the membrane surface or a

Page 57: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

33

polymer bead, the local contribution to the velocity is given by

ul (x) =

n∑

i=1

GR,cl (x− xc

i) · f ci +m∑

i=1

GR,pl (x− xp

i ) · fpi . (3.36)

Here n and m respectively are the number of capsule nodal points and polymer beads

which lie within the cut-off sphere with center at x. For a nodal point xcj on the

capsule, the local velocity can be written as

ul

(

xcj

)

=n

i=1,i 6=j

GR,cl (x− xc

i) · f ci +m∑

i=1

GR,pl (x− xp

i ) · fpi + limx→0

GR,cl (x) · f cj

=n

i=1,i 6=j

GR,cl (x− xc

i) · f ci +m∑

i=1

GR,pl (x− xp

i ) · fpi +1

8πηs

[

4ξc√π− 4α√

π

]

· f cj .(3.37)

This approach avoids problems that arise in naive evaluation of Eq. B.15 when

r = 0. For a point xpj corresponding to a polymer bead, the local velocity is given by

ul

(

xpj

)

=

n∑

i=1

GR,cl (x− xc

i) · f ci +m∑

i=1,i 6=j

GR,pl (x− xp

i ) · fpi , (3.38)

The difference between Eqs. 3.37 and 3.38 is the exclusion of the “self-term” in the

local velocity calculation for the polymer beads in the latter. (Recall that in contrast

to the capsule nodes, the polymer beads do not move with the fluid velocity, so the

velocity “seen” by the polymer bead should not include the singular velocity that the

bead itself generates as it moves through the fluid.) Eqs. 3.36, 3.37 and 3.38 give the

local contribution to the velocity at any point in the system.

We now proceed to the calculation of the global contribution to the velocity, ug (x),

which is due to the global part of the force distribution, ρg (x) in Eq. B.8 and the

Page 58: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

34

imposed boundary conditions for velocities on top and bottom walls (Eq. 3.30). If

ρg = 0 then ug = u∞. In GGEM the solution to the Stokes equations with a forcing

function ρg (x) is calculated numerically, requiring that the total velocity u (x) satisfy

appropriate boundary conditions. At points xW on the bounding walls, we therefore

have that

ug (xW) = −ul (xW) + u (xW) , (3.39)

where ul (xW) is the known local velocity field evaluated at the wall and u (xW) is

the boundary condition which comes from imposing velocities on top and bottom

walls (Eq. 3.30). Using the known local velocity field at the wall in the boundary

condition, we cancel out the error introduced by using unbounded-geometry solutions

for the local problem. For periodic boundary conditions (in the streamwise and

vorticity directions in the current system), we discretize using Fourier collocation.

Accordingly, the error in the solution in the periodic directions scales exponentially

with mesh resolution. However, it should be noted that the adoption of the Fourier

methods is merely for convenience and accuracy – in principle could be replaced by

other methods such as finite difference or finite elements. In the wall-normal direction,

we use a second-order finite difference scheme – the error decays quadratically with

mesh spacing in y. We choose the number of mesh points in each directions Mi

according to

Mi =√2αBi ; i = x, y, z, (3.40)

where the Bi’s are the box lengths in the simulation. It is important to note that

in traditional immersed boundary methods, the mesh spacing for the fluid solver is

determined by the regularization parameter for the capsule nodes. In contrast, with

Page 59: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

35

the present method, the mesh spacing is set by the choice of α. See Appendix ?? for

further discussion of this point. Unless otherwise noted, the computations reported

here use αa = 1. All of the small scale motions are part of the local computation,

which is done analytically so there is no discretization error. The global contribution

to the velocity is calculated at the mesh points and interpolated to any point, x, in

the system using quadratic Lagrange polynomials. It is this interpolation step that

ultimately controls the order of accuracy of the solution.

Finally, we note that, as with the local velocity calculation, the global velocity

calculation at a polymer bead needs to exclude the self-interaction term. The global

velocity field associated with this term is the free-space velocity driven by the bead’s

contribution to ρg(x). This velocity field is determined by the free-space regularized

Stokeslet

GR,α∞ (x) =

1

8πηs

[

δ +xx

r2

]

(

erf (αr)

r

)

+1

8πηs

[

δ − xx

r2

]

(

π1/2e−α2r2

)

, (3.41)

which reduces to the Oseen-Burgers tensor as α → ∞. Now the velocity on a polymer

bead can be written as

u(

xpj

)

= ul

(

xpj

)

+ ug

(

xpj

)

− limx→0

GR,α∞ (x) · fpj

= ul

(

xpj

)

+ ug

(

xpj

)

− 1

8πηs

4α√π· fpj . (3.42)

Page 60: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

36

3.5 Volume correction

While the regularization of point forces on the capsule membrane reduces fluid move-

ment across the membrane surface, fluid penetration is not completely eliminated at

the level of discretization used in our simulations. This causes the volume of fluid

enclosed by the capsule to vary, which is undesirable. To prevent this, we add a

constraint to the motion of capsule nodes such that the volume remains constant.

Consider a deformed capsule centered at the origin. The volume V of the capsule is

given by

V =

∫∫∫

V

dV =1

3

∫∫∫

V

(∇ · x) dV, (3.43)

where x is the position vector of a point on the surface of the capsule. Using the

divergence theorem we get

V =1

3

∫∫∫

V

(∇ · x) dV =1

3

∫∫

S

(x · n) dS, (3.44)

where n is the outward unit normal to the surface S. Following Freund34, we seek

displacement corrections z that minimize the function

I = Ψ

1

3

∫∫

S

(x + z) · n dS − V0

+

∫∫

S

(z · z) dS, (3.45)

where V0 is the initial volume of the undeformed capsule and Ψ is a Lagrange multi-

plier. The minimization is achieved by equating the variation of I, δI to zero,

δI = δΨ

1

3

∫∫

S

(z · n) dS + (V − V0)

+

∫∫

S

δz ·(

2z+Ψ

3n

)

dS = 0, (3.46)

Page 61: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

37

Since δΨ and δz are arbitrary, we get

2z+Ψ

3n = 0,

1

3

∫∫

S

(z · n) dS + (V − V0) = 0.(3.47)

Solving the two equations, we get the corrections z as

z = −3V − V0

An, (3.48)

Where A is the area of capsule membrane. This correction to the nodal displacement

is applied at every time step.

Page 62: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

38

Chapter 4

Single capsule dynamics

4.1 Method and code validation

In this section we report validation tests of our methodology and code for behav-

ior of single capsule behavior. For this, we compare results of single capsules in

shear flow with earlier results. In this case the capsule membranes are in a stress-

free state at equilibrium. Fig. 4.1(a) shows plots of Taylor deformation parameter,

D = (Lmax − Lmin) / (Lmax + Lmin), as a function of dimensionless time NH capsules

at various values of Ca, where Lmin and Lmax are the smallest and largest dimensions

of the capsule in the shear plane. The capsule is placed at the center of a simula-

tion box. For this and for all future results, the capsule surface is discretized into

1280 triangular elements, corresponding to 642 nodes. Results from the boundary

integral simulations of Lac et al.3 are also shown; good agreement is found. Figure

4.1(b) shows a comparison of steady state deformation parameter for NH capsules as

a function of Ca with the numerical results of Doddi and Bagchi4, Lac et al.3 and

Page 63: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

39

0 2 4 6 80

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

t∗

D

Ca = 0.075Ca = 0.15Ca = 0.30Ca = 0.45Ca = 0.60

(a) D vs. t∗

0 0.1 0.2 0.3 0.4 0.5 0.60

0.1

0.2

0.3

0.4

0.5

0.6

Ca

D

SimulationDBLBRP

(b) steady-state D vs. Ca

Figure 4.1: (a) Deformation parameter as a function of time for single non-prestressedNeo-Hookean (NH) capsules in shear flow in a cubic box of size 12.5a. Symbols aresimulations from Lacet al.3. Lines are results from present work. (b) Steady statedeformation parameter as a function of Ca for single non-prestressed Neo-Hookean(NH) capsules in shear flow in a cubic box of size 12.5a. For comparison, results ofDoddi and Bagchi4(DB), Lac et al.3(LB), Ramanujan and Pozrikidis5(RP) are alsoshown.

Page 64: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

40

0.4 0.6 0.8 1 1.2 1.4 1.610

−5

10−4

10−3

10−2

10−1

1/(hcξ

c)

|D −

DLB

|/DLB

Figure 4.2: Relative deviation of deformation parameter (D) compared to the resultof Lac et al.3 (DLB) plotted as a function of normalized regularization parameter ξchc

for a single non-prestressed NH capsule in shear flow at Ca = 0.30 in a cubic box ofsize 12.5a.

Ramanujan and Pozrikidis5. Again we observe good agreement with the previous

numerical results with closest agreement to those of Lac et al.3. For the neo-Hookean

capsule at Ca = 0.6, a slight overshoot in D is observed before it reaches its steady

state value. A similar overshoot is seen in the orientations and stress for a suspension

of Brownian rigid rods77 for sufficiently large Weissenberg number (which is analo-

gous to Capillary number in the present problem). This arises because at short times,

the capsule surface tends to move affinely with the flow, with relaxation to the steady

shape and motion following at later times.

Continuing the comparison of our results with those of Lac et al.3, Fig. 4.2 shows

the absolute value of the relative error in the steady-state deformation parameter

(assuming for the moment that the solution of Lac et al.3 is exact) as a function of

the regularization parameter ξc for NH capsules at Ca= 0.30. As expected for an

Page 65: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

41

0 0.01 0.02 0.03 0.04 0.050

0.02

0.04

0.06

0.08

0.1

Ca

D

NH, SimulationSK, SimulationNH, TheorySK, Theory

Figure 4.3: Steady state deformation parameter at low Ca for single non-prestressedNeo-Hookean (NH) and Skalak (SK) capsules (C = 10) in shear flow in a cubic boxof size 25a. Lines are theoretical predictions from Barthes-Biesel55 and symbols arefrom present simulations.

immersed boundary method, the solution is most accurate when this parameter is

neither too small nor too large. For the remainder of the paper, this parameter is set

to (hcξc)−1 = 0.75, where the relative error for this particular case is 3 × 10−4. As

another validation of our methodology, we compare to analytical results of Barthes-

Biesel et al.55 for the dependence of deformation parameter D on capillary number

for capsules in the small capillary number limit in unbounded shear flow, for both the

NH and SK models. This comparison is shown in Fig. 4.3 – agreement is excellent,

especially considering that the simulations were performed in a bounded domain of

size 25a, rather than the unbounded domain considered in the perturbation analysis.

Further validation of the methodology in the case of pair collisions is described in

Section 5.1.

Page 66: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

42

4.2 Single capsule in shear

For single capsules with relaxed membranes at equilibrium, the steady state defor-

mation parameter in Newtonian flow as a function of capillary number is shown for

NH and SK capsules in Fig. 4.4(a). (Unless otherwise noted, all SK results are

with C = 10, representing the case where area changes are strongly penalized.) As

expected, the NH capsule deforms more than the SK capsule for a given capillary

number, because of the presence of substantial energy penalty for area changes in

the SK model. Results for these capsules, however, are complicated by the presence

of wrinkling instabilities at small Ca and large curvatures at higher Ca, as noted in

Section 2.3. The wrinkling instabilities are more severe in the case of pair collisions.

To avoid wrinkling, Lac et al.56,6 suggested preinflating the capsule to ensure that

stresses on the surface remain purely tensile. Therefore, following Lac et al.56, we

also examine D vs Ca for capsules that have been preinflated to have a radius that

is enlarged by a factor of 1.05. The value for preinflation was chosen based on simu-

lations of pair collisions by Lac et al.56,6, who observed that a minimum preinflation

of 5% was required to prevent the appearance of compressive stresses. These results

are also shown in Fig. 4.4(a). Preinflation has a slight negative effect on D for the

NH capsule and a larger negative effect on the SK capsule. From Fig. 4.4(a) it can

be seen that the deformation for the NH capsule is always larger than that for the

SK capsule. For example, the D for the SK capsule at Ca = 0.6 is virtually the same

(0.20) as the D for the NH capsule at Ca = 0.142. We expand on this observation in

Fig. 4.4(b), which shows the time evolution of D for the NH capsule at Ca = 0.142

and the SK (with C = 10) capsule at Ca = 0.60. Not only are the values of D close,

but the transient evolution and the detailed shapes of the capsules (shown in the inset

Page 67: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

43

0 0.1 0.2 0.3 0.4 0.5 0.6 0.70

0.1

0.2

0.3

0.4

0.5

0.6

Ca

DNH non−preinflatedSK non−preinflatedNH preinflatedSK preinflated

(a) Effect of preinflation on steady state deformation parameter

0 2 4 6 8 100

0.1

0.2

0.3

0.4

0.5

t*

D

NH, Ca = 0.142SK, Ca = 0.60NH, Ca = 0.60

(b) Deformation parameter for preinflated NH and SK capsules

Figure 4.4: (a) Steady state values of deformation parameter at different Ca for singlenon-prestressed NH and SK (C = 10) capsules (shown by bold lines) and for prein-flated NH and SK capsules (shown by dotted lines) in shear flow. (b) Deformationparameter for preinflated NH and SK (C = 10) capsules under shear flow in a cubicbox of size 12.5a. Images correspond to capsule shapes taken at t∗ = 10..

Page 68: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

44

0 2 4 6 8 10−4

−3

−2

−1

0

1

2

3

4x 10

−3

t*

DN

− D

PNH, Ca = 0.142SK, Ca = 0.60NH, Ca = 0.60

(a) DN −DP

0 2 4 6 8 10−4

−3

−2

−1

0

1

2

3

4x 10

−3

t*

(θN

− θ

P)/π

NH, Ca = 0.142SK, Ca = 0.60NH, Ca = 0.60

(b) θN − θP

Figure 4.5: Difference in (a) deformation parameter and (b) inclination angle forpreinflated capsules in Newtonian fluid and a polymer fluid (β = 0.997) under shearflow in a cubic box of size 12.5a. Subscript N and P correspond to Newtonian andpolymer case respectively.

Page 69: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

45

of Fig. 4.4(b)) are quite similar. For comparison, results for the NH capsule at Ca

= 0.6 are also shown. For the remainder of this work, preinflated capsules are used,

and when comparisons between NH and SK models are made, we present results at

constant D and at constant Ca.

To study the effect of a very small amount of polymer (β = 0.997) on the defor-

mation of an isolated capsule, the following protocol was used: Initially, the capsules

are held fixed and rigid, while the polymer is subjected to shear flow for a period

of 20 time units. In this time interval, the polymer molecules are stretched to their

steady-state values. After the equilibration of the polymer molecules, the capsules are

allowed to deform. Fig. 4.5, the effect of polymer on the deformation of an isolated

capsule is illustrated, via plots of the difference of D and θ between the Newtonian

(subscript N) and polymer (subscript P) cases. The figure indicates that in the shear-

dominated flow around an isolated capsule, the polymer stretching is not substantial

and the effect of the polymer on the capsule dynamics is completely negligible (the

deviation of D from its Newtonian value is O (10−3) ).

4.3 Discussion

The dynamics of isolated capsules in shear are found to be almost completely un-

affected by the polymer additives at the very dilute concentrations considered here.

This result is a reflection that the flow field in this case is shear-dominated, and thus

does not lead to substantial polymer stretching.

Page 70: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

46

Chapter 5

Pair collisions in shear

5.1 Newtonian fluid

We turn now to pair collisions between capsules in shear flow. Fig. 5.1 shows a

schematic of the simulation box for pair collisions. For the present work we consider

the case where the initial coordinates of the capsule centers are (−4a, 0.25a, 0) and

(4a, −0.25a, 0), with the center of the box located at (0, 0, 0). Thus the initial stream-

wise separation between the capsule centers, ∆x0 is −8a and the initial wall-normal

separation, ∆y0 is −0.5a. The initial displacement in the neutral (z) direction is zero

for all cases studied here. Note that the initial position of the second capsule can

be found from the initial position of the first particle by rotating by π around the

z-axis. A slight complication arises in the simulations because, as is well-known, a

deformable particle in shear flow near a wall will generally migrate away from the

wall37,40,83,84. Migration is minimized by using a large wall-normal domain size. The

effect of migration on capsules was quantified by placing a single capsule with its

Page 71: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

47

x

yFlow

Wall

Wall

∆x

By

Bx

Bz

z 2a

∆y

Figure 5.1: Schematic of pair collisions of fluid-filled elastic capsules in shear flow ofa very dilute polymer solution.

center at one of the initial positions of the pair collisions and then subjecting it to

shear flow for a period of time corresponding to the time of collision. The net verti-

cal distance traveled by the capsule in this time interval is the displacement due to

migration. This displacement was less than 1% of the capsule radius.

As a validation for the pair collision code, Fig. 5.2(a) shows the relative trajecto-

ries (i.e. ∆y vs. ∆x) of colliding NH capsules at Ca = 0.45 in the absence of polymers

in a cubic box of side 25a, along with the boundary integral result (in an unbounded

domain) of Lac et al.6. We observe good agreement. To illustrate the insensitivity of

the results to the choice of α, Fig. 5.2(b) shows the relative trajectories (i.e. ∆y vs.

Page 72: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

48

−8 −6 −4 −2 0 2 4 6 80

0.5

1

1.5

2

∆x/a

∆y/a

simulationLac et. al.

(a) verification with Lac et al.6

−8 −6 −4 −2 0 2 4 6 80

0.5

1

1.5

2

∆x/a

∆y/a

4/(α a) = 34/(α a) = 44/(α a) = 6

(b) effect of screening parameter α

Figure 5.2: Pair collisions of preinflated NH capsules in a Newtonian fluid in a cubicbox of size 25a. Relative separation of the two capsules in y direction ∆y is plottedas a function of relative separation in x direction ∆x: (a) NH capsules at Ca = 0.45.Symbols are simulation result of Lac et al.6 and lines are present simulation; (b) effectof varying screening parameter α on collision dynamics of NH capsules at Ca = 0.30.

Page 73: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

49

−8 −6 −4 −2 0 2 4 6 8 100

0.5

1

1.5

2

∆x/a

∆y/a

NH,Ca = 0.10

NH,Ca = 0.30 NH,Ca = 0.60

NH, Ca = 0.10NH, Ca = 0.30NH, Ca = 0.60

Figure 5.3: Relative trajectories for pair collisions of preinflated NH capsules in aNewtonian fluid at different Ca. Images correspond to snapshots taken at collision(∆x = 0).

∆x) of colliding NH capsules at Ca = 0.30 at different values of α keeping the reso-

lution of the mesh fixed. The figure shows that the results converge for 4/(αa) ≥ 3.

For all remaining simulation results this parameter is fixed at αa = 1. The effect

of dimensionless time step ∆t∗ and capsule mesh size hc on the numerical method

was studied for collision dynamics of NH (Ca = 0.142) and SK (Ca = 0.60, C =

10) capsules. The numerical method was found to converge linearly with time step

(the linear convergence with time step was also obtained by Walter et. al.62 using

Runge-Kutta second order method) and linearly with capsule mesh size as expected

from an immersed boundary method.

As expected from prior studies of pair collisions between particles, the vertical

distance between the particles is larger after the collision than before. The maximum

displacement decreases with increasing Ca. At low Ca, the net final displacement

Page 74: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

50

−8 −6 −4 −2 0 2 4 6 8 100

0.5

1

1.5

2

∆x/a

∆y/a

NH, Ca = 0.142SK, Ca = 0.60NH, Ca = 0.60

Ca = 0.60 (NH)

Ca = 0.142 (NH)

Ca = 0.60 (SK)

Figure 5.4: Relative trajectories for pair collisions of preinflated NH and SK capsules(C = 10) in a Newtonian fluid at different Ca. Images correspond to snapshots takenat collision (∆x = 0).

increases with increasing Ca and at high Ca, the net final displacement decreases

with increasing Ca57. In particular, for Ca ≤ 0.10 for NH capsules, the net final

displacement is found to decrease with Ca. The focus of our study corresponds

to the regime of high Ca (Ca > 0.10) where stiffer capsules results in larger net

final displacements. The effect of Ca on maximum and net final displacements is

discussed in section 5.3. Fig. 5.3 shows the relative trajectories of colliding NH

capsules at various Ca in the absence of polymers, as well as shapes of the capsules

upon “impact”, defined when ∆x = 0. For this and all future results, we take

Bx = 25a, By = 20a, Bz = 10a. For all collisions in a Newtonian fluid, we observed

that the symmetry of the initial capsule configuration was conserved – the entire

collision process appears identical if the system is rotated by π around the z-axis. It

should be noted that with unit viscosity ratio we are always in the tank-treading as

Page 75: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

51

−8 −6 −4 −2 0 2 4 6 8 100

0.05

0.1

0.15

0.2

0.25

0.3

∆x/a

D

NH, Ca = 0.142SK, Ca = 0.60

Figure 5.5: Deformation parameter as a function of relative separation in the stream-wise direction (∆x) for pair collisions of pre-inflated NH (Ca = 0.142) and SK (Ca= 0.60, C = 10) capsules in a Newtonian fluid.

opposed to tumbling regime.

Fig. 5.4 compares relative trajectories of the NH and SK models (with C = 10) in

the absence of polymer. Recalling that the SK model at Ca = 0.60 and the NH model

at Ca = 0.142 have the same value ofD (= 0.20), results at these conditions are shown.

For comparison, the result for the NH capsule at Ca = 0.60 is also shown. Despite

the virtually identical shapes and transient evolution of capsule deformability of the

NH and SK capsules in isolation at D = 0.20 (see Fig. 4.4(b)), there are nontrivial

differences in the collision dynamics: the SK capsule displays a larger displacement

than the NH capsule throughout the collision process, and the final displacement is

larger by about 0.15a. This is a reflection of the fact that the SK capsule is stiffer

than the NH capsule. In Fig. 5.4, we show snapshots of the two cases during collision;

for the SK capsules (middle), the squeeze flow region between the colliding capsules

Page 76: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

52

is smaller than for the NH capsules (left), an indication that the SK capsules are

deforming less in response to the increased pressure found in that region during the

collision.

We further examine the difference between the NH and SK pair collisions by

observing the deformation parameter during the collision. Fig. 5.5 shows plots of

deformation parameter for a collision of NH capsules at Ca = 0.142 and SK capsules

(with C = 10) at Ca = 0.6. Since the Newtonian pair collision is symmetric with

respect to the two capsules involved, the capsules display identical dynamics so only

one curve is shown for each simulation. The deformation parameter D shows larger

variations in deformation for the NH capsule than for the SK capsule, particularly

in the time period when the capsules are approaching one another (∆x < 0). Even

though the two capsules display the same value of D in isolation during shear flow,

their responses clearly differ substantially during collisions. This fact will play an

important role in the effects of polymers on the collision.

5.2 Polymer solution

We turn now to the effect of added polymer on the dynamics of the pair collision

process. We will be exploring the effects of membrane model, initial displacements

(∆x0,∆y0) and polymer concentration 1−β on this process. To begin, Fig. 5.6 shows

the collision trajectories for the conditions of Figure 5.4 (thick lines), along with col-

lision trajectories for capsules in the polymer solution (thin lines) with β = 0.997.

As in the case of single capsules under shear flow in the polymer fluid, the polymer

molecules are first equilibrated for 20 time units during which the capsules remain

Page 77: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

53

−8 −6 −4 −2 0 2 4 6 8 100

0.5

1

1.5

2

∆x/a

∆y/a

NH, Ca = 0.142SK, Ca = 0.60NH, Ca = 0.60

Figure 5.6: Relative trajectories for pair collisions of preinflated NH and SK capsules(C = 10) in a Newtonian fluid (thick lines) and polymeric fluid (thin lines, β = 0.997).

unaltered. After the equilibration, the capsules are allowed to deform and trans-

late according to the flow. Because of the stochastic nature of the simulations with

polymers, several realizations with difference initial polymer positions and different

random number sequences were run. The differences between realizations were neg-

ligible. In particular, the standard deviation of the relative separation of capsules in

y−direction (∆y ) for the realizations was less than 0.4% of capsule radius at the

time of collision (∆x = 0) and less than 3.8% of capsule radius after the collision

process (∆x = 8a). The primary observation is that polymers reduce the total dis-

placement for all three cases, although the difference between the Newtonian and the

polymer fluid is very small for the NH capsules. Indeed for the two NH cases, the

trajectories without and with polymers are very similar until ∆x ≈ 2, by which time

the collision process is almost complete. In contrast, for the SK capsule at Ca = 0.6,

Page 78: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

54

−8 −6 −4 −2 0 2 4 6 8 10

0

0.5

1

1.5

∆x/a

∆y/a

− ∆

y 0/a

NH, Ca = 0.142SK, Ca = 0.60

∆y0/a = 0.75

∆y0/a = 0.50

∆y0/a = 1.00

Figure 5.7: Relative trajectories for pair collisions of NH and SK (C = 10) capsuleswith different initial separations ∆y0 in Newtonian (thick lines) and polymeric (thinlines, β = 0.997) fluids. The initial y-separation is subtracted off to facilitate com-parison of the results. Note: Newtonian and polymeric results for the NH cases aretoo close to each other to be distinguishable on the plot.

the discrepancy is substantial (0.16a) and begins just before ∆x = 0, while the par-

ticle centers are still moving apart vertically (i.e. just before the maximum in ∆y).

Additionally, this collision breaks the rotation symmetry of the Newtonian collisions

– the ultimate magnitudes of the y-positions of the two capsules differ by about 3%.

This symmetry-breaking was not the origin of the change in relative displacement,

however, a simulation with the rotation symmetry enforced yielded a very similar

relative trajectory.

Fig. 5.7 shows the collision trajectories of NH and SK (with C = 10) capsules

with different initial separations in Newtonian (thick lines) and polymeric (thin lines)

solutions with β = 0.997. Unsurprisingly, the collision dynamics are strongly affected

Page 79: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

55

−8 −6 −4 −2 0 2 4 6 80

0.5

1

1.5

2

∆x/a

∆y/a

C = 0C = 1C = 10

Figure 5.8: Variation of relative trajectories with non-dimensional area dilation mod-ulus, C, for SK (Ca = 0.60) capsules in Newtonian (thick lines) and polymeric (thinlines, β = 0.997) fluid.

by initial separation – smaller initial separation leads to higher maximum displace-

ment (around ∆x = 0) and higher net final displacement after the collision. As

observed in the previous results, the displacement of SK capsules is larger than NH

ones throughout the collision process – an illustration of the fact that SK capsules

are stiffer than NH capsules. The effect of polymers on different initial separations is

interesting. The figure shows that effect of polymer (thin lines) depends not only on

the model but also on the proximity of the capsules at the time of collision. The ef-

fect of polymers on NH capsules is not significant for any initial separation, consistent

with the previous results (Fig. 5.6). However the effect on SK capsules is nontrivial –

viscoelastic effects strongly depend on initial separations. As the separation between

the capsules increases the effect of polymer diminishes and the behavior of capsules

is similar to that in the absence of polymers.

Page 80: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

56

−8 −6 −4 −2 0 2 4 6 80

0.5

1

1.5

2

∆x/a

∆y/a

0 0.004 0.0080

0.2

0.4

0.6

0.8

(1 − β)

(∆y N

− ∆

y P)/a

Newtonianβ = 0.9985β = 0.9970β = 0.9940

Figure 5.9: Relative trajectories as a function of (1− β) for SK (Ca = 0.60, C = 10)capsules under shear flow in a Newtonian (thick lines) and polymeric (thin lines)fluid. Inset shows the difference in ∆y of SK capsules at ∆x = 8a in Newtonian andpolymeric fluid, (∆yN −∆yP), plotted against (1− β).

One of the significant differences between NH and SK capsule models is the pres-

ence of an explicit area penalty in the SK model. For all previous results of SK

capsules we set C = 10. Figure 5.8 shows the effect of C on the collision dynamics at

Ca = 0.60. Consider the Newtonian case (thick lines) first. The collision dynamics

are strongly affected by the parameter C. It is observed that the capsules with high

values of C (= 10) have more relative displacement from the initial state than the

capsules with low value of C (= 0, 1). This is consistent with the previous observa-

tions that stiffer capsules have higher relative displacements throughout the collision

process (Fig. 5.3 and Fig. 5.4).

Turning to the effect of polymer on the collision dynamics of SK capsules with

various values of C (thin lines on Fig. 5.8), we see that polymers suppress the net

Page 81: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

57

displacement for C = 10 while there is almost no effect at lower values (C = 0, 1).

The weak effect of polymer on SK capsules at low C values is consistent with the

observations with NH capsules, which also do not have a strong resistance to area

changes. These results indirectly indicate that during pair collisions, the resistance

to area change of the capsules leads to relatively stronger polymer stretching, a point

revisited in section 5.4.

Now we briefly examine the effect of polymer concentration on the collision dy-

namics. Results will be reported in terms of of 1 − β, which is a linear function

with concentration in the dilute regime considered here. Fig. 5.9 shows the effect

of polymer concentration on the collision dynamics of SK capsules (with C = 10)

at Ca = 0.60 with values of concentration that are half and double the value used

above, corresponding to β values of 0.9985 and 0.9940, respectively. (Results for

Newtonian collisions and collisions with β = 0.997 are repeated here for comparison.)

The effect of polymer concentration is substantial. The inset shows the change in net

displacement from the Newtonian case as a function of 1 − β, illustrating that the

effect of polymer concentration on collision dynamics is nonlinear even at very low

concentrations.

5.3 Effect of Ca

We turn now to the effect of Ca on collision dynamics of capsules in Newtonian

and polymeric fluid. Fig. 5.10(a) shows the maximum displacement during collision

(∆ymax −∆y0) as Ca varies for NH and SK (C = 10) capsules in Newtonian (thick

lines) and polymeric (thin lines, β = 0.997) fluid. As Ca increases, the maximum

Page 82: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

58

0 0.5 1 1.50.8

0.9

1

1.1

1.2

1.3

1.4

1.5

Ca

(∆y m

ax −

∆y 0)/

a

NHSK

(∆yfinal

−∆y0)

(∆ymax

−∆y0)

(a) maximum displacement (∆ymax −∆y0) vs. Ca

0 0.5 1 1.50

0.2

0.4

0.6

0.8

1

Ca

(∆y fin

al −

∆y 0)/

a

NHSK, C = 10

(b) net final displacement (∆yfinal −∆y0) vs. Ca

Figure 5.10: Variation of (a) maximum displacement (∆ymax −∆y0) and (b) net finaldisplacement (∆yfinal −∆y0) with Ca for NH and SK (C = 10) capsules in Newtonian(thick lines) and polymeric (thin lines, β = 0.997) fluid. Note: ∆yfinal is calculated at∆x = 8a. The figure in the inset of (a) shows the schematic of these displacements.

Page 83: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

59

displacement decreases monotonically – the capsules become less of an obstruction

to each other with increasing Ca, and accordingly the maximum displacement due to

collision decreases. The effect of polymer is minimal on the maximum displacement

for both NH and SK capsules indicating that the effect of polymer is not significant

when the capsules collide (∆x ≈ 0).

The effect of Ca and polymer addition on the net final displacement of capsules

is more complex. Fig. 5.10(b) shows this quantity as a function of Ca for NH

and SK (C = 10) capsules. At low Ca, where the deformations are small, the net

final displacement decreases as Ca decreases. This is because at low Ca, as Ca

decreases, the capsules approach the rigid particle limit and the net displacement

of rigid particles is zero due to the reversibility of Stokes flow. At high Ca, where the

deformations of the capsules are significant, each capsule becomes more elongated in

x but less so in y as Ca increases. Thus the effective “collision cross section” between

the particles decreases with increasing Ca, and accordingly the net displacement due

to collision decreases. The effect of polymer on net final displacement is substantial,

indicating that polymer effect becomes significant when the capsules begin to leave

each other after the collision – a point which will be revisited in section 5.4. The figure

also shows that the effect of polymer is more sensitive at low Ca regime for both NH

and SK capsules – the effect of polymer is stronger for stiffer capsules. However, the

effect of polymers on SK capsules are larger than that on NH capsules, consistent

with the previous observations.

Page 84: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

60

−8 −6 −4 −2 0 2 4 6 80

0.5

1

1.5

2

2.5

3

3.5

∆x/a

λ max

(0,0

,0)

Thick Lines − NewtonianThin Lines − Polymer

NH, Ca 0.142SK, Ca = 0.60, C = 0SK, Ca = 0.60, C = 10

Figure 5.11: Largest eigenvalue λmax of the deformation rate at the origin as a functionof ∆x for pair collisions under various conditions. Thick lines are from Newtoniansimulations; thin lines are from simulations with polymers, β = 0.997.

5.4 Mechanism of polymer effects

To gain an understanding of mechanism of the polymer effects on the pair collisions,

we examined the spatial dependence of the largest eigenvalue λmax of the rate of strain

tensor Γ = 12

(

∇u+ (∇u)T)

. In simple shear, λmax = 0.5; during the pair collisions

λmax exceeded 1 only for some of the constitutive models and then only during the

period when 2 . ∆x . 4, the final stage of the collision process. In all cases, the

spatial position where λmax reaches its largest value is the origin. Fig. 5.11 shows

λmax at the origin during pair collisions with ∆y0 = 0.5 under various conditions.

For the C = 0 SK model, there is only a modest increase in λmax in the interval

2 . ∆x . 4 and almost no change when polymers are present. For the NH model,

there is a distinct peak in this interval, but again only a small change in the presence

Page 85: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

61

of polymers. For the C = 10 SK case, however, there is a very large increase in λmax

in this interval, and the increase is dramatically reduced (to the level found in the NH

case) when polymers are present. From these results we can see that (1) the largest

deformation rates arise as fluid is drawn into the region between capsules as they

move apart from one another after colliding, (2) the effect is largest for the nearly

area-incompressible case, and (3) this phase of the capsule interaction process is the

one that most strongly stretches the polymers and is accordingly the one that is most

affected by their presence. As the capsules leave the collision, they draw fluid and

chains into the growing gap between them. The uniaxial extension generated in this

region stretches the polymer chains, and the stretching works against the separation

of the capsules as they move around one another. This resistance allows the capsules

to “roll” around one another more than they would in the absence of the polymer

and thus leads to a smaller net relative displacement in y.

5.5 Discussion

At sufficient concentration, the interactions of capsules play become important. In

suspensions, capsules collide with each other leading to substantial velocity fluctu-

ations that drive the diffusive motions of cells and solutes. The dynamics of pair

collision of cells are strongly affected by the presence of drag reducing polymer ad-

ditives, for reasons that are not understood. Results for pair collisions show that

at high Ca, for neo-Hookean capsules or for Skalak capsules with a small penalty for

area change, there is almost no effect of polymer additives on the collision trajectories,

while for Skalak capsules with a substantial energy penalty for area change, there is a

Page 86: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

62

substantial effect – in this case the net displacement of the capsules in the wall-normal

(gradient) direction is substantially decreased from the Newtonian case. This effect

strongly depends on the proximity of the capsules at the time of collision – the closer

the capsules are at the time of collision, the effect of polymer gets more substantial.

This effect of polymer is found to depend on the area incompressibility of SK cap-

sules. SK capsules that have strong resistance to area change show substantial effect

of polymers. This effect originates in the increased rate of deformation in the gap

between departing capsules that is observed with membranes that strongly resist area

change. The uniaxial extension generated in this region stretches the polymer chains,

and the stretching works against the separation of the capsules as they move around

one another. At low Ca, where the deformations are small, the effect of polymer on

the collision dynamics of NH and SK capsules are substantial. These observations

are concrete indications that the presence of small amounts of polymer additives can

substantially change the flow of suspensions of model cells, and that the membrane

properties of the cells play an important role in the observed changes.

Page 87: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

63

Chapter 6

Suspensions of capsules in Couette

flow

6.1 Theory for suspension of capsules in shear flow

In this section, we present a simple theory for the suspensions of capsules subjected

to a simple shear flow. We consider capsules of radius a confined between two parallel

walls separated by a distance By and subjected to a shear flow with shear rate γ as

shown in Fig. 6.1. The direction of flow is taken to be along x with y and z as

the wall-normal and the vorticity direction respectively. The walls are at y = 0 and

y = By. The volume fraction of the capsules is denoted by φ.

6.1.1 Wall-induced migration of a single capsule

In a bounded flow, a deformable particle (capsule, droplet, vesicle) exhibits cross-

stream migration away from the nearest wall. Fig. 6.2(a) shows the schematic of a

Page 88: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

64

γ⋅

y = By

y

xz y = 0

2a

wall

wall

Figure 6.1: Schematic of suspensions of fluid-filled elastic capsules in Couette flow.

deformable capsule near a wall. To the leading order, the particle can be described

by a point dipole44. The migration of a deformable particle in shear flow away from

a single wall can be described as

umig = Kγa(N1 −N2)

ηγa2

(

1

y2

)

, (6.1)

where umig is the migration velocity, y is the distance from the wall, N1 and N2 are

the first and second normal stress differences of an individual capsule and K is a

coefficient given as

K =3

64πa3n, (6.2)

Page 89: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

65

where n is the number density of the capsule. The above expression includes the

effect of only the nearest wall. The effects of both walls on the migration of a capsule

can be approximated as

umig = Kγa(N1 −N2)

ηγa2

[

1

y2− 1

(By − y)2

]

. (6.3)

Note that eq. 6.3 is a superposition of the individual effect of both walls. It should

be noted that the expressions for the migration velocity (eqs. 6.1 and 6.3) assumes

the particle to be a point dipole and does not consider the finite size of the particle.

Hence the assumption is strictly valid for y/a � 1.

6.1.2 Shear-induced diffusion

For suspensions of capsules, the interactions of capsules become important. Colli-

sion of capsules is asymmetric and the capsules depart on streamlines that are more

widely separated than the ones before the collision74. Fig. 6.2(b) shows a schematic

of a trajectory of a capsule undergoing a collision process in the shear (x− y) plane.

As observed form the figure, the capsule gets displaced from its original streamline

after the collision. Random collisions lead to random motions perpendicular to the

streamlines. From dimensional and physical point of view, self-diffusivity of a di-

lute suspension of capsules is expected to be proportional to the product of collision

frequency (proportional to γφ, where φ is the local volume fraction) and the mean

squared displacement per collision (proportional to a2). The self-diffusivity Dself can

be written as

Dself = γφa2fself, (6.4)

Page 90: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

66

where fself is a coefficient that depends on the deformability, size and shape of capsules.

6.1.3 Model for steady-state distribution

In suspensions, shear-induced diffusion and wall migration work against one another,

and a nonuniform distribution of capsules is expected at steady state. Consider an

initially homogeneous distribution of capsules subjected to shear flow. The local flux

of capsules in the wall-normal direction (y-direction) is

jy = umigφ− ∂ (Dsφ)

∂y, (6.5)

where Ds is the short time diffusivity of the particles and takes the same functional

form as Dself in eq. 6.4 i.e. Ds = γφa2fs. Note that this equation does not take the

form of Fick’s law. This is because in general the diffusivity is position dependent.

The evolution equation of the volume fraction of capsules can be stated as

∂φ

∂t= −∂jy

∂y= − ∂

∂y

(

umigφ− ∂ (Dsφ)

∂y

)

. (6.6)

This equation is a Fokker-Planck equation and is equivalent to writing the evolu-

tion equation for the probability of finding a capsule at a given y−position in the

dilute limit; with the assumption that the motion of capsules is a continuous Markov

process85,86 and is a good approximation for time scales longer than about γ−1.

At steady state, the net flux is zero and the diffusive and convective (due to

migration) fluxes balance each other. The volume fraction balance becomes

∂ (Dsφ)

∂y= umigφ. (6.7)

Page 91: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

67

⋅γ By

y

wall

wall

(a)

net finaldisplacementafter thecollision

γ⋅ y x

(b)

Figure 6.2: (a) Schematic of migration of an isolated capsule away from the nearestwall in a Couette flow. (b) Schematic showing pair collision of the capsules. The linesshows the trajectory of a capsule undergoing a collision process in the shear (x− y)plane.

Page 92: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

68

6.1.4 Steady-state capsule-depleted layer near a single wall

For simplicity we consider the case of semi-infinite domain where the effect of one

wall is only important. Starting from eq. 6.7 and substituting eq. 6.1 gives

φ∂φ

∂y=

(

K(N1 −N2)

ηγ

a

2fs

)

φ

(

1

y2

)

. (6.8)

There exists two solutions of eq. 6.8. One solution is φ = 0, describing a capsule

free layer near the walls, denoted by y = δ ; δ being the thickness of the capsule free

layer. The other solution can be found by the integration of eq. 6.8 as

φ = φb

(

1− δ

y

)

, (6.9)

where

δ =K

2fsφb

(N1 −N2)

ηγa, (6.10)

and φb is the bulk volume fraction. The equation provides a closed-form analytical

expression for the thickness of the capsule-depleted layer. An important observation

from the above equation is that the thickness of the capsule-depleted layer scales as

δ ∼ φ−1. The above equation also shows a direct relationship of δ with the parameters

of wall-induced migration (δ ∼ K (N1−N2)ηγ

) and inverse relationship with the parameter

of shear-induced diffusion (δ ∼ f−1s ) – indicating that wall-induced migration favors

the formation capsule-depleted layer where as shear-induced diffusion opposes it. In

terms of Ca dependance, both (N1−N2)ηγ

and fs are functions of Ca and so δ will depend

on Ca – the nature of this dependence will be determined by which of these two

competing mechanisms has a stronger relationship with Ca. We will show, through

Page 93: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

69

our numerical simulations in sections 6.2 and 6.3, an explicit dependance of (N1−N2)ηγ

,

fs and δ on Ca and then compare our results with the above prediction in section 6.4.

6.2 Migration of a single capsule in a Couette Flow

6.2.1 Newtonian fluid

In this section we report the results of wall-induced migration of an isolated NH cap-

sule in a Couette flow in a Newtonian fluid as shown in Fig. 6.2(a). The capsule is

initially placed at a distance of 2a from the bottom wall (y = 0) and is then subjected

to the shear flow. The effect of box length Bx in the flow direction on single capsule

migration was studied and the results are found to converge for Bx ≥ 10a. For all

the simulations of single capsule migration we use the simulation domain to be a box

of size 16a, 10a and 10a in x, y and z directions respectively. The dimensionless

time step used in all our simulations is 5 × 10−3. The capsule surface is discretized

into 1280 triangular elements corresponding to 642 nodes. Fig. 6.3(a) shows the

trajectory of the center of mass of a capsule in the wall-normal direction as a func-

tion of time at different values of Ca. The figure shows that the capsule migrates

away from the wall towards the center of the channel with the rate of migration

increasing with an increase in Ca. As Ca increases, the capsule becomes more de-

formable – the capsule gets more stretched in the x−direction than in the y−direction.

The shape of the capsule can be characterized by the Taylor deformation parameter,

D = (Lmax − Lmin) / (Lmax + Lmin) where Lmin and Lmax are the smallest and largest

dimensions of the capsule in the shear (x − y) plane. Note that D is zero when the

shape of the capsule is spherical and increases as the shape of the capsule gets more

Page 94: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

70

0 50 100 150 200 250 300

0

1

2

3

4

5

t∗

y/a

Center

Wall

Ca = 0.10Ca = 0.142Ca = 0.30Ca = 0.60

(a)

2 2.5 3 3.5 4 4.5 50

0.1

0.2

0.3

0.4

0.5

0.6

y/a

D

Ca = 0.10Ca = 0.142Ca = 0.30Ca = 0.60

(b)

Figure 6.3: Migration of a capsule in a Newtonian fluid in a Couette flow. (a)Trajectory of the center of mass of a capsule y as a function of time t∗ in the wall-normal direction. The walls are at y = 0 and y = By = 10a. (b) Capsule deformationD as a function of center of mass of a capsule y.

Page 95: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

71

0 50 100 150 200 250 3000

0.005

0.01

0.015

0.02

0.025

t∗

(N1−

N2)/

(ηγ)

Ca = 0.10Ca = 0.142Ca = 0.30Ca = 0.60⋅

(a)

0.1 10.01

0.1

1

Ca

(N1−

N2)/

(ηγ)

∼ Ca0.60

(b)

Figure 6.4: (a) Difference between first (N1) and second (N2) normal stress differencesas a function of y. (b) N1 − N2 evaluated at y = 2.5a (quarter channel height) asa function of Ca. The symbols represent the simulation results and the dashed linerepresents the exponential fit.

Page 96: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

72

deformed. Fig. 6.3(b) shows plots of Taylor deformation parameter as a function of

channel height for a capsule at various values of Ca. As the flow starts, D attains its

steady state value within a short time interval (t∗ ≤ 5) implying that capsules attain

their equilibrium shape quickly before moving significantly in the flow. To study the

effect of capsule mesh resolution on single capsule migration, a simulation with higher

capsule mesh resolution (5120 triangular elements) was performed for a capsule at

Ca = 0.60 and compared with the results of capsule with 1280 triangular elements.

The deviation in the deformation parameter for the latter was within ±1.3% of that

of the former. For all simulations of single capsule migration, we use 1280 triangular

elements to discretize the capsule surface.

Fig. 6.4(b) shows the variation of (N1 − N2) with y for a capsule at various

Ca. Here N1 and N2 are the first and second normal stress differences defined as

N1 = τxx − τyy; N2 = τyy − τzz. Here τ is the extra stress due to the presence of

the capsule. The more deformable a capsule (higher Ca), the larger (N1 − N2) it

can exhibit and higher is its tendency to migrate away from the wall. In the limit

of small deformations (low Ca), one expects the normal stress difference to vary as

(N1 −N2)/(ηγ) ∼ Ca1.0, however, in the intermediate range of Ca (between 0.1 and

1.0), our simulation results suggest (N1−N2)/(ηγ) ∼ Ca0.60 as shown in Fig. 6.4(b).

6.2.2 Validation of dipole approximation

As noted above we presented a simple theoretical expression (section 6.1.1) for the

migration velocity of an isolated capsule assuming the capsule to be a point dipole.

This assumption is only valid when the capsule is far away from the wall i.e. y/a � 1.

Page 97: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

73

0 20 40 60 80 1000

1

2

3

4

5

t*

y/a

y0 = 1.1a

y0 = 1.2a

y0 = 1.5a

y0 = 2.0a

y0 = 3.0a

(a)

0 50 100 150 200 250 3002

2.5

3

3.5

4

4.5

t∗

y/a

Ca = 0.10Ca = 0.142Ca = 0.30Ca = 0.60

(b)

Figure 6.5: Validation of a point dipole approximation. (a) Trajectory of the centerof mass of an isolated capsule y at Ca = 0.30 as a function of time t∗ in the wall-normal direction for different values of initial condition y0. The walls are at y = 0and y = By = 10a. Symbols are simulation results and lines are the fits using eq.6.13.(b) Trajectory of a capsule as a function time for different values of Ca.

Page 98: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

74

2 2.5 3 3.5 4 4.5 50

0.01

0.02

0.03

0.04

0.05u m

ig/(

γa)

y/a

Ca = 0.10Ca = 0.142Ca = 0.30Ca = 0.60

(a)

0 0.2 0.4 0.6 0.80

0.1

0.2

0.3

0.4

0.5

k

Ca

SimulationTheory

(b)

Figure 6.6: (a) Migration velocity umig as a function of center of mass of a capsule y.(d) Comparison of the numerical value of the slope k obtained from simulations (byfitting eq. 6.13) and the theoretical value obtained using eq. 6.12, at different valuesof Ca.

Page 99: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

75

In this section we report a validation of this approximation. Rewriting eq. 6.3 as

umig =dy

dt= k[

1

y2− 1

(By − y)2] , (6.11)

where

k = Kγa(N1 −N2)

ηγa2 =

3(N1 −N2)

64πηn. (6.12)

Eq. 6.11 can be solved analytically to obtain

1

4(y3−y30)−

1

8By(y4−y40)−

By

16(y2−y20)−

B2y

16(y−y0)−

B3y

32ln (

By − 2y

By − 2y0) = kt , (6.13)

Note that eq. 6.13 is implicit in y and explicit in t. Fig. 6.5(a) shows the trajectory

of the center of mass of a capsule at Ca = 0.30 in the wall-normal direction as a

function of time for different values of initial condition y0. Symbols are the results

from simulation and lines are the predictions obtained by fitting eq. 6.13. The figure

shows that the dipole approximation of the capsule holds good for y0/a ≥ 2.0. For

reference we also show the simulation results (symbols) of the trajectories of capsule

at different Ca from Fig. 6.3(a) along with the fits (lines) obtained from eq. 6.13.

Again a good match is observed. Fig. 6.6(a) also shows the migration velocity umig of

a capsule as a function of as a function of center of mass of a capsule y at various Ca.

Symbols are the results from simulation and the lines are the fits obtained from first

fitting eq. 6.13 to obtain k and then using k in eq. 6.11. The results show an excellent

match between our simulation results and the fits obtained from the assumption. It

should be noted that the theoretical predictions do not predict a zero velocity of the

capsule at t = 0 which is actually the case with our simulations and hence does not

Page 100: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

76

match with the simulation results at initial times.

Finally, Fig. 6.6(b) shows a comparison between the numerical value of the slope

k, obtained from fitting eq. 6.13, and theoretical value of k, obtained from using eq.

6.12, at different Ca. The results show that the at low Ca, where the deformations are

small, there is a good match between the numerical and the theoretical values of k but

at high Ca, where the deformations are substantial, the corresponding values deviate

from each other – the theory over predicts the simulation results. This indicates

that the capsule is not a true dipole and higher order singularities are also present.

These singularities become significant at high Ca and must be considered to make

quantitative predictions.

6.2.3 Polymer fluid

We now turn to the effect of polymer on the migration of an isolated capsule. For all

the simulations of suspensions of capsules in polymer solution we use the simulation

domain to be a box of size 16a, 10a and 10a in x, y and z directions respectively. The

results for the polymer solution are found to converge for the box of size Bx ≥ 16a.

In our simulation with polymers, the polymer molecules are first equilibrated for a

period of 50 time units during which the capsules are held fixed and rigid, while the

polymer molecules are stretched to their steady-state values. After the equilibration,

the capsules are allowed to deform and translate according to the flow. The polymer

molecules are initially randomly uniformly distributed in the simulation domain. For

polymers, we use β = 0.9985, 0.997, 0.994 and 0.988, which correspond to number

density n of 0.0235µm−3, 0.047µm−3, 0.094µm−3 and 0.188µm−3, or a mass fraction

of 0.157 ppm, 0.314 ppm, 0.628 ppm and 1.256 ppm of PEO respectively. The numbers

Page 101: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

77

0 50 100 150 2002

2.5

3

3.5

4

4.5

t∗

y/a

NewtonianPolymer (β = 0.997, Wi = 20)

Ca = 0.10Ca = 0.142Ca = 0.30Ca = 0.60

(a)

0 0.2 0.4 0.6 0.80

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Ca

D

NewtonianPolymer (β = 0.994, Wi = 20)

(b)

Figure 6.7: Migration of a capsule in a Newtonian (solid lines) and polymer (dashedlines, β = 0.994,Wi = 20) solutions. (a) Trajectory of the center of mass of a capsuley as a function of time t∗ in the wall-normal direction. (b) Steady state capsuledeformation D as a function of Ca.

Page 102: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

78

0 0.2 0.4 0.6 0.80

0.005

0.01

0.015

0.02

Ca

(N1−

N2)/

(ηγ)

NewtonianPolymer (β = 0.994, Wi = 20)

(a)

0 0.005 0.01 0.015 0.020

0.005

0.01

0.015

0.02

0.025

0.03

(N1−N

2)/(ηγ)

u mig

| (y =

2.5

a)/(

γa)

NewtonianPolymer (β = 0.994, Wi = 20)

(b)

Figure 6.8: (a) N1 − N2 evaluated at y = 2.5a as a function of Ca. (b) Migrationvelocity umig evaluated at y = 2.5a (quarter channel height) as a function of N1−N2.The symbols represent the simulation results and the dashed line represents the linearfit.

Page 103: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

79

of polymer chains in the simulation domain for these cases are 1016, 2032 4064 and

8128 respectively. Ex ranges from 4.3 to 34.4. We use Wi = 5, 10 and 20 for the

polymers. Fig. 6.7(a) shows the trajectory of the center of a capsule in wall-normal

direction in Newtonian (circles) and polymer (squares, β = 0.994,Wi = 20) solutions

in a Couette flow. Symbols are simulation results and lines are the fits using eq. 6.13.

Because of the stochastic nature of the simulations with polymers, several realizations

with different initial polymer positions and different random number sequences were

run. The differences between realizations were negligible. In particular, the standard

deviation of the center of mass of a capsule in y−direction evaluated at y = 2.5a for

the realizations was less than 3.0% of capsule radius. The figure shows that polymer

slows down the migration and the effect gets substantial for more deformable capsules

(high Ca). For example, the difference in the capsule’s center between the Newtonian

and the polymer case at t∗ = 300 is 0.5a at Ca 0.10 and gets as big as 1a at Ca

= 0.60. It should be noted that the effect of polymer does not change the final steady

state position of the capsule i.e. the capsule in the polymer case still has a tendency

to migrate towards the center but at a slower rate than in the Newtonian case. The

figure also shows a comparison between the simulation results (symbols) and the fits

(lines) obtained from using the point dipole approximation of capsule (eq. 6.13) for

the polymer case. Although a small discrepancy is observed at initial times (t∗ ≤ 5),

the dipole assumption shows a good match with the numerical simulation results for

later times (t∗ > 5) – indicating that the dipole approximation holds good for the

polymer case as well. Fig. 6.7(b) shows the steady state deformation parameter of the

capsule as a function of Ca in a Newtonian (solid lines) and polymer (dashed lines,

β = 0.994,Wi = 20) solutions. The primary observation is that polymer results in

Page 104: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

80

changing the equilibrium shapes of the capsule – the capsule in the polymer solution

is less deformed than in the Newtonian fluid i.e. the capsule gets less stretched in

the shear plane. This is also reflected in Fig. 6.8(a) which shows that the polymer

case results in lower normal stress differences of the capsules than in the Newtonian

case. As noted above, the effect of polymer on (N1 − N2) and D increases for more

deformable capsules however the difference is small to probably affect the dynamics

of single capsule. Fig. 6.8(b) shows the migration velocity umig evaluated at y = 2.5a

as a function of the normal stress differences of the capsules (at different Ca) in

Newtonian (black) and polymer (blue) solutions. The migration velocity is obtained

using point dipole assumption (eq. 6.11). The figure shows a linear relationship of

umig with (N1 − N2) for both Newtonian and polymer solutions. Interestingly, the

slope of the linear fit between migration velocity and (N1 −N2) for the polymer case

is significantly lower in value than the Newtonian case the origin of which is not

understood at present. The effect of polymer concentration 1 − β at fixed Wi of

20 on the trajectory and migration velocity umig|y=2.5a of a capsule (Ca = 0.30) is

shown in Fig. 6.9(a) and Fig. 6.10(a) respectively. The polymer concentration is

expressed in terms of 1 − β, which is a linear function of concentration in the dilute

regime considered here. The figures show that the effect of polymer concentration

on migration of an isolated capsule is substantial even at such low concentrations.

The effect of WI at fixed polymer concentration β = 0.994 on the trajectory and

migration velocity umig|y=2.5a of a capsule (Ca = 0.30) is shown in Fig. 6.9(b) and

Fig. 6.10(b) respectively. The figure shows that the effect of polymer gets substantial

when the polymer chains stretch significantly in flow (higher Wi). For all simulations

of suspensions of capsules, we use the value of β to be 0.994 and 0.988 and the value

Page 105: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

81

0 50 100 150 2002

2.5

3

3.5

4

4.5

t∗

y/a

NewtonianPolymer (β = 0.997, Wi = 20)

Newtonian β = 0.997 β = 0.994 β = 0.988

(a)

0 50 100 150 2002

2.5

3

3.5

4

4.5

t∗

y/a

NewtonianPolymer (β = 0.997, Wi = 20)

NewtonianWi = 5Wi = 10Wi = 20

(b)

Figure 6.9: Effect of (a) polymer concentration expressed as 1−β at fixed Wi (= 20)and (b) Wi at fixed β (= 0.994) on the trajectory of an isolated capsule (Ca = 0.30)in the wall-normal direction of in a Couette flow. Symbols are simulation results andlines are the fits.

Page 106: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

82

0 0.005 0.01 0.0150

0.005

0.01

0.015

0.02

0.025

0.03

1− β

u mig

| (y =

2.5

a)/(

γa)

(a)

0 5 10 15 200

0.005

0.01

0.015

0.02

0.025

0.03

Wi

u mig

| (y =

2.5

a)/(

γa)

(b)

Figure 6.10: Migration velocity umig evaluated at y = 2.5a as a function of (c) 1−β atfixed Wi (= 20) and (d) Wi at fixed β (= 0.994) for an isolated capsule (Ca = 0.30).

Page 107: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

83

of Wi to be 20. This is done so as to keep the computation cost reasonable while still

allowing the effects of polymer to be substantial.

It should be noted that the phenomenon of wall-induced migration in shear flows

is also observed in dilute polymer solutions and for highly extensible polymers it can

lead to depletion layer at steady state that can be much larger than the size of the

polymer molecule44 (depletion layer thickness scales as ∼ Wi2/3). However, the time

to reach a steady state is much larger than the time period considered here (it is

several orders of magnitude of the polymer relaxation time). We have studied the

temporal variation of the distribution of polymers in our simulations of single capsule

migration and found the variation in the distribution to be negligible – polymers were

uniformly distributed throughout the domain. The observed effects of polymer on

single capsule migration are in contrast to the theoretical predictions38,87 on migration

of deformable drops (analogous to deformable capsules) in viscoelastic fluids – their

predictions suggest that viscoelasticity enhances single particle migration away from

the walls in a Couette flow. It should be noted that the theoretical predictions are

based on the ideal limit where there is separation of length scales between polymer

molecules and drops so that the polymeric fluid can be treated using a continuum

approach ( second-order fluid). Additionally, the predications are derived in the limit

of Wi � 1 i.e. the fluid is weakly non-Newtonian. However the situation under

consideration is qualitatively different – here there is no separation of length scales

between polymer chains and capsules. Furthermore, we are interested in the regime

of high Wi where there is significant stretching of polymer molecules by the flow.

Interestingly, numerical simulation results of a single rigid sphere in viscoelastic fluid

in Couette flow by Avino et al.88 showed that viscoelasticity (at Wi ∼ 1) induce

Page 108: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

84

particle motion towards the wall. They made a simple heuristic argument made that

the viscoelasticity leads to normal stress differences that are asymmetric under the

presence of a wall and this asymmetry leads to the cross-stream migration towards

the wall.

6.3 Suspensions of capsules in a Couette flow

6.3.1 Newtonian solution

We now consider the dynamics of dilute suspensions of capsules in a Couette flow

in a Newtonian fluid in a Couette flow. Simulation box of two different sizes are

considered – a cubic box of size 10a and 16a respectively. The volume fraction of

capsules considered is φ = 0.10 and φ = 0.20 which corresponds to simulations of 25

to 100 capsules. The capsules are initially distributed randomly in the computational

domain and are then subjected to a shear flow for a time period of 400 to 500 time

units. Previous relevant numerical studies have either considered two-dimensional34,89

or have considered three-dimensional simulations of capsules at high mesh resolution

for smaller time period70,69. In our case, we are interested in simulations of capsules

for a time period that is much larger than the previous studies which when coupled

with the simulation of large number polymer molecules becomes computationally too

expensive to obtain results in a reasonable time window. To study the effect of capsule

mesh resolution on suspension dynamics, a simulation of suspensions of capsules at

Ca = 0.142 with a mesh resolution of 1280 triangular elements was performed in a

Newtonian flow in a cubic box of size 10a for a time period of 200 units and results

were compared with that of capsules with 320 triangular elements. The deviation in

Page 109: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

85

(a)

0 100 200 300 400 5000

2

4

6

8

10

t∗

y/a

Wall

Wall

(b)

Figure 6.11: (a) Snapshots of the suspensions of capsules (Ca = 0.60,φ of 0.10) ina Newtonian fluid at t∗ = 1 (left) and t∗ = 300 (right) in a Newtonian fluid in acubic box of size 10a . (b) Trajectories of the center of mass of capsules (Ca = 0.60,φ = 0.10) in the wall-normal direction as a function of time. The walls are at y = 0and y = By = 10a.

Page 110: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

86

the average distance of center of the capsules from the channel center in wall-normal

direction < y−ycenter > for the latter was within±5% of that of the former indicating

reasonable accuracy of the latter. For all simulations of suspensions of capsules, we

use 320 triangular elements to discretize the capsule surface. These parameters of

capsules were chosen to keep the computational cost moderate while still capturing

the suspension dynamics with a reasonable accuracy.

Fig. 6.11(a) shows the snap shot of the suspensions (φ = 0.10) of capsules (Ca

= 0.30) in a cubic box of size 10a at time t∗ = 1 and t∗ = 300. Instantaneous shapes

and distribution of the capsules can be observed from the figure. Since the flow has

a constant shear rate the shapes of the capsules are similar. Fig. 6.11(b), shows the

trajectories of the center of mass of capsules (φ = 0.10, Ca = 0.60) in the wall-normal

direction as a function of time. It should be noted that due to the finite size of

the capsules, the initial distribution of the center of mass of capsules lie in a range

1a < y < 9a. Starting with a uniform distribution, the capsules are found to migrate

towards the center of the channel. This is also observed in the snapshot of capsules

at t∗ = 300. One can also observe fluctuations in the trajectories which are due to

the hydrodynamic interactions due to finite concentration of capsules. Unlike a single

capsule, the capsules in suspensions do not migrate continuously towards the center

but reach a steady state showing accumulation of the capsules in core of the channel

and a capsule-depleted layer near the walls as observed from the figure.

The net motion of the capsules towards the center of the channel can be further

quantified by examining the time evolution of the average distance of the capsule’s

center from the center of the channel < |y − ycenter| >. Fig. 6.12(a) shows < |y −

ycenter| > for the capsules as a function of time in a Newtonian fluid at different

Page 111: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

87

0 100 200 300 4001.6

1.7

1.8

1.9

2

2.1

2.2

2.3

t∗

<|y

−y ce

nter|>

/a

Ca = 0.08Ca = 0.142Ca = 0.30Ca = 0.60

(a)

0 1 2 3 4 50

0.01

0.02

0.03

0.04

y/a

frac

tion

NH, Ca = 0.142NH, Ca = 0.30NH, Ca = 0.60

(b)

Figure 6.12: (a) Average distance from the centerline < |y − ycenter| > of suspensions(φ = 0.10) of capsules in a Newtonian fluid in a a cubic box of size 10a as a functionof time t∗. (b) Steady state distribution of capsules (φ = 0.10) as a function of y.The walls are at y = 0, 10a and y = 5a is the channel centerline. The walls are aty = 0, 10a and y = 5a is the channel centerline.

Page 112: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

88

Ca. The value of < |y − ycenter| > initially decreases and then reach a steady state

indicating initial migration of capsules towards the center of the channel leading

to an accumulation of the capsules in the core of the channel at steady state. We

consider the time at which the steady state is reached as the time after when the

change in average distance from the center is less than 5%. For all simulations of the

suspensions of capsules, we take this time as t∗ ≥ 200 as observed from the figure.

The figure also shows that the extent to which the capsules move towards the center

of the channel increases with an increase in Ca indicating that as the deformability

of the capsules increases, the capsules have a greater tendency to migrate away from

the walls and move closer toward the center of the channel. The collective motion of

capsules towards the center of the channel is also observed in a large channel size as

shown in Fig. 6.13(a).

Fig. 6.12(b) shows the steady state distribution of capsules, represented as the

areal fraction of capsules, at various values of Ca as a function of channel height.

The fraction is calculated by dividing the channel height into bins of equal size and

finding the fraction of triangular elements, that discretize the membrane surface,

in those bins. The bin size was taken to be 0.2a. Since we assume these triangular

elements to be of equal size and mass, the fraction of triangular elements is equivalent

to the areal fraction of the capsules. The distribution was extracted by averaging over

a large time window over which the suspensions remain statistically stationary. We

take this time window to be 200 ≤ t∗ ≤ 400. Furthermore, to avoid the influence

of initial configuration of capsules on the steady state, 5 simulations with different

initial configurations were used and the steady state distribution was averaged over all

those configurations. The standard deviation, due to different initial configurations,

Page 113: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

89

0 100 200 300 400 5003.25

3.3

3.35

3.4

3.45

3.5

t∗

<|y

−y ce

nter|>

/a

Ca = 0.142

(a)

0 2 4 6 80

0.01

0.02

0.03

0.04

y/a

frac

tion

NH, Ca = 0.142

(b)

Figure 6.13: (a) Average distance from the centerline < |y − ycenter| > of suspensionsof capsules (φ = 0.10, Ca = 0.60) in a Newtonian fluid in a cubic box of size 16a asa function of time t∗. (b) Steady state distribution of capsules (φ = 0.10, Ca = 0.60)as a function of y. The walls are at y = 0, 16a and y = 8a is the channel centerline.

Page 114: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

90

0 1 2 3 4 50

0.01

0.02

0.03

0.04

y/a

frac

tion

φ = 0.10φ = 0.20

Figure 6.14: The effect of volume fraction φ on the steady state distribution of cap-sules at Ca = 0.60 as a function of y. The walls are at y = 0, 10a and y = 5a is thechannel centerline.

in the steady-state distribution was observed to be less than 4%. Fig. 6.12(b) shows

that at steady state the distribution of the capsules is inhomogeneous – capsules have

higher concentration near the center of the channel and a capsule-depleted region

near the walls. The thickness of the depleted layer as well as the concentration

near the center of the channel increases with an increase in Ca indicating that more

deformable capsules move farther away from the walls and accumulate closer towards

the channel center. The figure also shows the layer formation of capsules near the

walls, as shown by the off-center peaks in the distribution. The layer formation has

also been observed by Li and Pozrikidis89 in their simulations of suspensions of liquid

drops in wall bounded flows. They observed that the layer formation was a result of

using confined channel size. We have studied the effect of box size on the steady state

Page 115: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

91

distribution. Fig. 6.13(b) shows the steady state distribution of capsules (Ca = 0.60)

in a cubic box of size 16a. The figure shows that formation of layers of capsules near

the walls is diminished indicating that layer formation is a result of using small size

of the channel.

The effect of volume fraction φ of capsules (at Ca = 0.60 ) on the steady state

distribution is shown in Fig. 6.14. The steady state distribution profiles at differ-

ent volume fraction are qualitatively similar to each other – higher concentration of

capsules near the center of the channel and a capsule-depleted region near the walls.

However the figure shows that volume fraction has a strong effect on the capsule

depleted region i.e. increasing the concentration of capsules decreases the thickness

of the capsule-depleted region near the walls.

6.3.2 Polymer solution

We turn our attention towards the effect of polymer on the dynamics of suspensions

(φ = 0.10) of capsules in a Couette flow. All the simulations of suspensions of capsules

in polymer solutions, we use the value of β to be 0.994 and 0.988 and the value of

Wi to be 20. It is based on our results from section ?? which showed substantial

effect of polymer on the dynamics of single capsule migration with these parameters.

As in the case of single capsule migration under shear flow in the polymer fluid, the

polymer molecules are first equilibrated for 50 time units during which the capsules

remain unaltered. After the equilibration, the capsules are allowed to deform and

translate according to the flow. The polymers are randomly uniformly distributed in

the entire simulation domain. Fig. 6.15(a) shows the snap shot of the suspensions

(φ = 0.10) of capsules (Ca = 0.30) in a polymer solution (β = 0.994,Wi = 20) in a

Page 116: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

92

(a)

0 100 200 300 4001.5

2

2.5

t∗

<|y

−y ce

nter|>

/a

NewtonianPolymer (β = 0.994, Wi = 20)

Ca = 0.142Ca = 0.30Ca = 0.60

(b)

Figure 6.15: (a) Snapshots of the suspensions of capsules (Ca = 0.30) at t∗ = 10 in apolymer (β = 0.994,Wi = 20) solutions in a cubic box of size 10a. Polymer moleculesare shown as thin black lines (b) Average distance from the centerline < |y−ycenter| >of suspensions of capsules in Newtonian (solid line) and polymer(dashed lines, β =0.994,Wi = 20) solutions as a function of time t∗.

Page 117: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

93

0 1 2 3 4 50

0.01

0.02

0.03

0.04

y/a

frac

tion

NewtonianPolymer (β = 0.994, Wi = 20)

Ca = 0.142Ca = 0.30Ca = 0.60

(a)

0 0.1 0.2 0.3 0.4 0.5 0.6 0.70.02

0.022

0.024

0.026

0.028

0.03

Ca

frac

tion

NewtonianPolymer (β = 0.994, Wi = 20)

(b)

Figure 6.16: (a) Steady state distribution of capsules as a function of y in Newtonian(solid line) and polymer (dashed line, β = 0.994,Wi = 20) solutions in a cubic boxof size 10a. (b) Steady state distribution of capsules in the “bulk” (2.5a ≤ y ≤ 7.5a)region as a function of Ca in Newtonian (solid line) and polymer(dashed line, β =0.994,Wi = 20) solutions.

Page 118: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

94

cubic box of size 10a at time t∗ = 10. The distribution of the capsules and polymer

molecules (thin black lines) can be observed from the figure. Since Wi of the polymers

is high, the flow causes significant stretching of the polymer molecules as shown in

the figure. As noted above, the parameters of polymers are so chosen that there

is no separation of length scales – the relative size of stretched polymer chains and

capsules is comparable. As noted in the case of single capsule migration under shear

flow in the polymer fluid, here also we find the distribution of polymer molecules to

be uniform throughout the entire time of simulation.

Fig. 6.15(b) shows the average distance of the capsules from the center of the

channel at different Ca as a function of time in Newtonian (solid lines) and polymer

(dashed lines, β = 0.994,Wi = 20) solutions in a cubic box of size 10a. The primary

observation is that polymer suppresses the net movement of capsules towards the

center of the channel leading to a steady state value of < |y − ycenter| > that is larger

than that in the Newtonian case. This suggests that under the influence of polymer,

capsules accumulate less in the central region of the channel at steady state. This

difference in the steady state values increases with an increase in Ca indicating that

the effect of polymer gets significant as the capsules become more deformable.

The effect of polymer (β = 0.994,Wi = 20) on the steady state distribution of

capsules is shown in Fig. 6.16(a). The figure shows that polymer changes the distri-

bution of capsules. To better understand the effect of polymer, we show the average

distribution (fraction) of capsules in the “bulk” region, denoted by 2.5a ≤ y ≤ 7.5a

in Fig. 6.16(b) in Newtonian and polymer solutions. The average concentration of

the capsules in this region is lower than the corresponding Newtonian case. These

results indicate that polymer results in redistribution of capsules having lower con-

Page 119: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

95

0 1 2 3 4 50

0.01

0.02

0.03

0.04

y/a

frac

tion

NewtonianPolymer (β = 0.994)Polymer (β = 0.988)

Figure 6.17: Steady state distribution of capsules (Ca = 0.60) in Newtonian andpolymer (Wi = 20) solutions with different values of β.

centration of capsules in the central region and correspondingly, by conservation of

capsules, higher concentration of capsules in the wall region. Fig. 6.3.2 shows the

effect of polymer concentration, expressed in terms of 1− β, on the steady state dis-

tribution of NH capsules (Ca = 0.60). As noted in the figure the effect of polymer

on steady state distribution gets enhanced on increasing the concentration i.e. it

results in higher and lower concentration of capsules in the wall and the bulk region

respectively.

6.3.3 Capsule-depleted layer

The existence of the cell-free layer in the flow of blood has been observed both in in

vitro 33,90 and in vivo 91,92,93 experiments. The existence of this layer is the cause of

Page 120: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

96

0 0.2 0.4 0.6 0.80

0.2

0.4

0.6

0.8

1

1.2

Ca

δ/a

δ/a = 0.005+1.364×(Ca)0.397

δ/a = 0.015+0.957×(Ca)0.341

δ/a = 0.010+0.808×(Ca)0.291

NewtonianPolymer (β = 0.994)Polymer (β = 0.988)

(a)

0 0.1 0.2 0.3 0.4 0.50

1

2

3

4

5

flow rate (ml/min)

cell−

free

laye

r (µm

)

δ = 0.50 + 4.55×(flow rate)0.37374

Kameneva et al., 2004

(b)

Figure 6.18: (a) Dependence of capsule-depleted layer thickness on Ca for suspensionsof capsules in Newtonian and polymer (Wi = 20) solutions with different values of βin a cubic box of size 10a. Symbols are the simulation results and lines are the fits.The standard deviation is based on results from different initial configurations. (b)Experimental data (symbols) on the thickness of cell-free layer as a function of flowrate for a suspensions of RBCs from Kameneva et al.7.

Page 121: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

97

lower apparent viscosity of blood in small vessels. The thickness of this layer is of the

same order as the cell dimension. Fig. 6.18(a) shows the dependence of the capsule-

depleted layer thickness δ as a function of Ca for suspensions of capsules Newtonian

and polymer (β = 0.994,Wi = 20) solutions in a cubic box of size 10a. The capsule-

depleted layer thickness is calculated from the steady state distributions of capsules

by finding bins with a non-zero value closest to the top and bottom wall and then

averaging over their corresponding distances from the walls. The standard deviation

in δ is based on averaging over different initial configurations. Symbols are the results

from the simulations and lines are the fits. Consider the Newtonian result first. The

figure shows that the capsule-depleted layer thickness follows a scaling relationship

with Ca as δ ∼ (Ca)n, where n is around 0.40 for the Newtonian case. The power-

law dependence of δ with flow rate (∝ Ca) has been observed experimentally33,90

with n reported to between 0.30 and 0.50. Of particular interest is the work done

by Kameneva et al.7, who studied the dependence of cell-free layer with flow rate

in their microchannel experiment of blood. Fig. 6.18(b) shows their experimental

data (symbols) along with our fit (line) to their data. Interestingly, the fit shows

the scaling of cell-free layer with flow rate as δ ∼ (flow rate)0.37 which is close to our

Newtonian results. Freund34, in his 2-D simulations of RBCs in microvessel showed

a similar scaling behavior as δ ∼ (flow rate)0.43. These observations along with our

numerical results suggest that the relationship of the cell-free layer thickness with flow

rate (∝ Ca) weakly depends of the shape of cell and the geometry of the vessel. The

existence of the capsule-depleted layer can also be observed from the visualizations of

the suspensions of capsules (Ca = 0.60) in Fig. 6.11(a) which shows that a uniformly

distributed capsules (snapshot on the left) migrates towards the channel center leading

Page 122: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

98

to an inhomogeneous distribution (snapshot on the right) with a capsule-depleted

layer near the walls.

The effect of polymers on the capsule-depleted layer is shown in Fig. 6.18(a). The

primary observation is that the polymer reduces the thickness of the capsule-depleted

layer significantly compared to the Newtonian case and the discrepancy increases

with the increase in Ca. In this case, δ also scales with Ca as δ ∼ (Ca)n with a value

of n that is smaller than that of the Newtonian case and decreases on increasing

the polymer concentration (∝ 1 − β). This is in a qualitative agreement to the

experimental observations of Kameneva et al.7 who showed that the addition of small

amount of polymer (PEO) in blood resulted in a significant reduction in the cell-free

layer thickness and the reduction increased with increasing the flow rate (∝ Ca). It

should be noted that the concentration of polymer used in their experiments (10 ppm)

is order of magnitude larger than the one we are simulating (0.63 ppm). Nevertheless,

the deviation of the results of the polymer case from the Newtonian one observed from

our simulations is a reflection of the fact that the effect of polymer is substantial even

at such low concentrations.

6.3.4 Diffusion at steady state

Having studied the migration of a single capsule and behavior of suspensions capsules

in Newtonian and polymer solutions, what remains to complement the theoretical

prediction with simulation results is to study the diffusion of the capsules. Here

we address the effect of Ca and polymers on the diffusive behavior of the capsules.

We calculate the mean squared displacement of the capsules after the steady state

is reached. This is done to eliminate the effect of wall-induced migration. The

Page 123: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

99

0 10 20 30 40 500

0.1

0.2

0.3

0.4

t∗ − t0∗

<(y

− y

0)2 >/a

2

Polymer (β = 0.994, Wi = 20)Newtonian

Ca = 0.08Ca = 0.142Ca = 0.30Ca = 0.60

(a)

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0

0.001

0.002

0.003

0.004

0.005

Ca

Ds/(

γa2 )

NewtonianPolymer (β = 0.994, Wi = 20)

(b)

Figure 6.19: (a) Mean squared displacement of suspensions of capsules in the wall-normal direction at steady state in Newtonian and polymer ( β = 0.994,Wi = 20)solutions as a function as a function of time t∗ and the corresponding short-time dif-fusivities (b) in the wall-normal direction as a function of Ca. The standard deviation(error bar) is based on results from different initial configurations.

Page 124: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

100

0 1 2 3 4 5 0

0.001

0.002

0.003

0.004

y/a

Ds(y

)/(γ

a2 )

Polymer (β = 0.997, Wi = 20)

Newtonian

Ca = 0.08Ca = 0.142Ca = 0.30Ca = 0.60

Figure 6.20: Short-time diffusivities in the wall-normal direction as a function of y inNewtonian (solid line) and polymer(dashed line, β = 0.994,Wi = 20) solutions.

subscript ‘ss’ denotes the value at steady state. Fig. 6.19(a) shows the mean squared

displacement of the suspensions of capsules at different Ca in Newtonian (solid lines)

and polymer (dashed line) solutions. The figure shows that the polymers suppress

the diffusive motions of the capsules. This is also reflected in Fig. 6.19(b) which

shows the short time diffusivity Ds in wall-normal direction of the suspensions of

capsules in Newtonian (solid lines) and polymer (dashed line) solutions as a function

of Ca. The short time diffusivity Ds is calculated by averaging the diffusive motions

of all the capsules after steady state for a time period of t − tss = 50. The effect

of polymers is to suppress the diffusive motions of the capsules resulting in lower

diffusivities than in the Newtonian case. This is in agreement to the simulation

results of pair collisions of elastic capsules in polymer solutions by Pranay et al.74.

They showed that polymers suppress the net final displacements of the capsules after

Page 125: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

101

the collisions. Since in suspensions the random motions due the collisions drive the

diffusive motion of the capsules, the effect of polymer will be to suppress this diffusive

motion reducing the mean squared displacement and the corresponding diffusivities.

The dependance of short time diffusivity on y for capsules at different Ca is shown

in Fig. 6.3.4. As observed from the figure the short time diffusivity is an increasing

function of the local volume fraction of the capsules – showing a maxima near the

center where the capsules are accumulated and having a zero value near the walls

since there are no capsules there. Fig. 6.19(b) also shows that the diffusivity is a

weak function of Ca for the Newtonian case. The weak dependance of diffusivity on

Ca has also been observed for the case of suspensions of deformable drops45 which

is analogous to deformable capsules. Interestingly, the figure shows a non-monotonic

behavior of the diffusivity with Ca around Ca = 0.10. This is in excellent agreement

with the numerical simulation results of pair collision of elastic capsules by Pranay

et al.74. They showed that at low Ca, where the deformations are small, the net

final displacement of the capsules after the collision decreases as Ca decreases. This

is because at low Ca, as Ca decreases, the capsules approach the rigid particle limit

and the net displacement of rigid particles is zero due to the reversibility of Stokes

flow. At high Ca, where the deformations of the capsules are significant, each capsule

becomes more elongated in the flow direction but less so in wall-normal direction as

Ca increases and accordingly the net displacement due to collision decreases. The

net final displacement of the capsules after collision shows a non-monotonic behavior

with Ca around Ca = 0.10.

Page 126: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

102

6.4 Comparison with the theoretical model

In section 6.1 we presented a prediction about the dependance of the capsule-free layer

thickness on two competing mechanisms – wall-induced migration and shear-induced

diffusion (eq. 6.10) as

δ ∼ K (N1 −N2)

fsφbηγ. (6.14)

Eq. 6.14 shows a scaling predcition δ ∼ φ−1. Our numerical simulation results

(Fig. 6.14) on suspensions of elastic capsules at two different volume fraction (φ =

0.10 and φ = 0.20) show that the the thickness of the capsule depleted layer is a

decreasing function of the volume fraction of the capsules. Eq. 6.14 shows δ ∼(

(N1−N2)ηγ

)(

1fs

)

≈ umig

Ds(eqs. 6.3 and 6.4) – indicating that the wall-induced migration

favors formation of thicker capsule free layer where as shear-induced diffusion opposes

migration leading to a thinner capsule-free layer. Numerical simulation results on

suspensions of capsules (Fig. 6.19(b)) show that diffusivity is a weak function of Ca

where as results on single capsule migration show that in the intermediate range of

Ca (between 0.1 and 1.0) the normal stress difference scales (Fig. 6.4(b)) as (N1 −

N2)/(ηγ) ∼ Ca0.60. This leads to a prediction δ ∼ (Ca)0.60. In suspensions, the

effect of wall-induced migration gets weak due to finite concentration and finite size

of capsules, so it would be expected that the exponent of the scaling to be less than

0.60. Our simulation results on suspensions of capsules show δ ∼ (Ca)0.40. This

is in a good agreement to the experimental7 and numerical34 observations. These

observations indicate that the capsule-depleted layer thickness can be determined

primarily by the mechanism of wall-induced migration.

Numerical simulation results on the effect of polymer on the single capsule migra-

Page 127: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

103

0 0.1 0.2 0.3 0.4 0.5 0.6 0.70

0.2

0.4

0.6

0.8

1

Ca

δ p/a

SimulationTheory

Figure 6.21: Comparison of the thickness of the capsule free layer for suspensions(φ = 0.10) of capsules in polymer ( β = 0.994,Wi = 20) solution obtained fromsimulations and predicted from theory (eq. 6.16) as a function of Ca.

tion show that polymer suppresses the migration of capsules. In suspension this would

be expected to lead to lower thickness of the capsule-free layer. Numerical simulation

results on shear-induced diffusion of suspensions of capsules show that polymer sup-

presses the diffusive motions of the capsules which would be expected to lead larger

thickness of the capsule-free layer. But since the influence of migration is dominant

in the dynamics of suspensions, the net effect of the polymer on the suspensions of

capsules would be expected to suppress the collective motion of capsules towards the

center of the channel and to decrease the thickness of the capsule-free layer. In fact

we can use the theoretical model to show that the balance of wall-induced migration

and shear-induced diffusion leads to a smaller capsule-free layer thickness in the poly-

mer case. The thickness of the capsule-free layer for the suspensions of capsules in

Page 128: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

104

polymer solutions from eq. 6.14 can be expressed as

(

δPδN

)

=

(

KP

KN

)(

(N1 −N2)P(N1 −N2)N

)(

fs,Nfs,P

)

, (6.15)

where subscripts “N” and “P” denote results for Newtonian and polymer (β =

0.994,Wi = 20) solutions respectively. Numerical simulation results on single capsule

migration have shown thatKP/KP ∼ 0.42 (Fig. 6.8(b)) and (N1 −N2)P / (N1 −N2)N ∼

0.95 (Fig. 6.8(a)). Numerical simulation results on the diffusivity of capsules in sus-

pensions show that fs,N/fs,P = Ds,N/Ds,P ∼ 1.8 (Fig. 6.19(b)). These values yields a

prediction(

δPδN

)

= 0.72. (6.16)

Indeed our numerical simulation results on the dynamics of suspensions show that

polymer leads to a redistribution of capsules at steady state with a reduction in the

thickness of the capsule-free layer. Fig. 6.21 shows the comparison of δP obtained

from simulations and predicted from theory as a function of Ca. As observed from

the figure the theoretical predictions are reasonably accurate considering the fact the

the prediction was derived from the basic knowledge of the wall-induced migration

and shear-induced diffusion of capsules. Our results show that the thickness of the

capsule-free layer under the influence of polymer also scales with the Ca as δ ∼ (Ca)n.

The exponent of this scaling is less than that of the Newtonian case indicating that

the effect of polymer becomes substantial at high Ca. This is in a good qualitative

agreement to the experimental observations7.

Page 129: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

105

6.5 Discussion

In the present work, we have studied the suspensions of elastic capsules in Couette

flow in Newtonian and polymer solutions. We have presented a theory which suggests

that the competition between the effect of wall-induced migration and shear-induced

diffusion leads to a steady-state distribution of capsules with a capsule-depleted layer

near the walls. The thickness of the capsule-free layer is derived analytically and is

found to be dependent on the capillary number. Numerical simulations is performed

to study the motion of elastic capsules in polymer solutions. Capsule membranes

are modeled using a neo-Hookean (NH) constitutive model and polymer molecules

are modeled as bead-spring chains with finitely extensible non-linearly elastic (FENE)

springs; parameters were chosen to loosely approximate 4000 kD poly(ethylene oxide).

Simulations are performed with a novel Stokes flow formulation of the Immersed

Boundary Method (IBM) for the capsules, combined with Brownian dynamics for the

polymer molecules. The polymer chains interact hydrodynamically with one another

and with the capsules. An approximate treatment of the Brownian motion of the

polymer beads is used, allowing substantial simplification in methodology. Results

for an isolated capsule indicate that the wall-induced migration depends strongly on

the capillary number. Numerical simulation of suspensions of capsules in Newtonian

fluid illustrates the inhomogeneous distribution of capsules at steady-state with the

formation of capsule-depleted layer near the walls. The thickness of this layer is found

to be strongly dependent on the capillary number. The shear-induced diffusivity,

on the other hand, show a weak dependence on capillary number. These results

indicate that the mechanism of wall-induced migration is the primary source for

determining the capsule-depleted layer thickness of capsules in suspensions. These

Page 130: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

106

results are in qualitative agreement to the theoretical model. Numerical simulation

results on the effect of polymer show that both the wall-induced migration and the

shear-induced diffusive motion of the capsules are suppressed under the influence of

polymer. Since the influence of migration is dominant in the behavior of suspensions

of capsules, the net effect of the polymer on the suspensions of capsules is to suppress

the collective motion of capsules towards the center of the channel leading to a lower

thickness of the capsule-free layer and a redistribution of the capsules at steady state.

These observations are representative of the fact that the presence of small amount of

polymers can significantly influence the motion of deformable particles in suspensions.

Page 131: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

107

Chapter 7

Conclusion

Blood flow in the microcirculation is strongly affected by the presence of drag reducing

polymer additives, for reasons that are not understood. The present work represents

an initial step toward understanding these observations, and more broadly, shedding

light on the dynamics of complex multiphase fluids. Simulations of simple particle

motions and pair collisions of nominally spherical elastic capsules in shear flow are

performed using a novel immersed boundary method for Stokes flow; the suspended

polymer molecules are simulated as bead-spring trimers with parameters chosen to

model poly(ethylene oxide). The polymer chains interact hydrodynamically with one

another and with the capsules. An approximate treatment of the Brownian motion

of the polymer beads is used, allowing substantial simplification in methodology and

savings in computation time. Because the simulations are performed at high Peclet

number, this approximation does not introduce substantial error into the results. The

dynamics of isolated capsules in shear are found to be almost completely unaffected

by the polymer additives at the very dilute concentrations considered here. This

Page 132: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

108

result is a reflection that the flow field in this case is shear-dominated, and thus

does not lead to substantial polymer stretching. Results for pair collisions are more

complex. At high Ca, for neo-Hookean capsules or for Skalak capsules with a small

penalty for area change, there is almost no effect of polymer additives on the collision

trajectories, while for Skalak capsules with a substantial energy penalty for area

change, there is a substantial effect – in this case the net displacement of the capsules

in the wall-normal (gradient) direction is substantially decreased from the Newtonian

case. This effect strongly depends on the proximity of the capsules at the time of

collision, concentration of polymer and the area incompressibility of SK capsules and

originates in the increased rate of deformation in the gap between departing capsules

that is observed with membranes that strongly resist area change. At low Ca, where

the deformations are small, the effect of polymer on the collision dynamics of NH and

SK capsules are substantial.

A simple theoretical model is presented to describe the cross-stream migration

of deformable capsules in suspensions which comes from a balance of two competing

mechanisms – wall-induced migration and shear-induced diffusion. The wall-induced

migration favors continuous migration of capsules away from the vessel wall where

as shear-induced diffusion opposes migration. The balance between these two effects

leads to an inhomogeneous distribution at steady state with higher concentration of

capsules near the center and a capsule-depletion layer near the walls. The thick-

ness of this layer depends on both mechanisms. Results for an isolated capsule near

a wall indicate that the wall-induced migration depends strongly on the capillary

number. Numerical simulation of suspensions of capsules in Newtonian fluid illus-

trates the inhomogeneous distribution of capsules at steady-state with the formation

Page 133: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

109

of capsule-depleted layer near the walls. The thickness of this layer is found to be

strongly dependent on the capillary number. The shear-induced diffusivity, on the

other hand, show a weak dependence on capillary number. These results indicate

that the mechanism of wall-induced migration is the primary source for determining

the capsule-depleted layer thickness of capsules in suspensions. Numerical simulation

results show that both the wall-induced migration and the shear-induced diffusive

motion of the capsules are suppressed under the influence of polymer. Results on sus-

pensions of capsules illustrate that the net effect of polymers is to reduce the thickness

of the capsule-depleted layer resulting in a redistribution of capsules at steady state.

The results are in qualitative agreement to the experimental observations. These

observations are concrete indications that the presence of small amounts of polymer

additives can substantially change the flow of suspensions of model cells, and that the

membrane properties of the cells play an important role in the observed changes.

Page 134: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

110

Chapter 8

Current work: Suspensions of

capsules in a pressure-driven flow

8.1 Overview

We consider the motion of spherical NH capsules of nominal radius a, suspended in

in Newtonian fluid, in a pressure-driven flow as shown in Fig. 8.1. The simulation

box is periodic in the flow (x) and vorticity (z) directions. The wall-normal direction

is y. The walls are at y = 0 and y = By. The undisturbed flow field in x -direction

is a fully developed Poiseuille (parabolic) flow given by

u(y) = 4U

[

(

y

By

)

−(

y

By

)2]

, (8.1)

where U is the channel centerline velocity. The capillary number of the capsule is

defined as Ca = ηsγwa/G, where γw is the shear rate at the wall given as: γw = 4U/By.

Time is non-dimensionalized with By/U as t∗ = tU/By. The dimensionless time step

Page 135: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

111

y

z

2a

y = By

xy = 0

Figure 8.1: Schematic of suspensions of fluid-filled elastic capsules in a pressure-drivenflow.

used in our simulations is 5 × 10−3. In all our simulations we use the size of the

simulation domain as a slit of height 10a, 16a and 40a. Considering the nominal

radius of the capsule to be 4µm, the size of the slit corresponds to lengths 40µm,

64µm and 160µm respectively. To avoid wrinkling instabilities56,6, the capsules are

initially preinflated such that the radius of the prestressed capsule a is related to the

radius of the unstressed capsule a0 as a = 1.05a0. We will consider two cases: the

dynamics of an isolated capsule place near a wall and suspensions of capsules in a

pressure-driven flow in a slit geometry as shown in Fig. 8.1. The numerical details of

these two situations are outlined below.

Page 136: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

112

8.2 Migration of an isolated capsule in a pressure-

driven Flow

In this section we report the results of wall-induced migration of an isolated NH

capsule in a pressure-driven flow in a Newtonian fluid. For all the simulations of single

capsule migration, unless otherwise specified, we use the simulation domain to be a

cubic box of size 10a. The capsule is initially placed at a distance of 2a (y/By = 0.20)

from the bottom wall (y = 0) and is then subjected to the shear flow. For all the

simulations of single capsule migration in pressure-driven flow the capsule surface

is discretized into 1280 triangular elements corresponding to 642 beads. Fig. 8.2(a)

shows the trajectory of the center of mass of the capsule in the wall-normal direction as

a function of time at different values of Ca. The figure shows that the capsule migrates

continuously away from the wall towards the center of the channel with the rate of

migration increasing with an increase in Ca. As shown in the figure for a time period

of 300 time units, the capsule with Ca = 0.60 almost reaches the center of the channel

y = 5a. As Ca increases, the capsule becomes more deformable – the capsule gets more

stretched in the x−direction than in the y−direction. The shape of the capsule can be

characterized by the Taylor deformation parameter, D = (Lmax − Lmin) / (Lmax + Lmin)

where Lmin and Lmax are the smallest and largest dimensions of the capsule in the shear

(x−y) plane. Fig. 8.2(b) shows plots of Taylor deformation parameter as a function of

channel height for NH capsule at various values of Ca. As the flow starts, D attains

its steady state value within a short time interval (t∗ ≤ 5) implying that capsules

attain their equilibrium shape quickly before moving significantly in the flow. It

should be noted that since the flow is pressure-driven, the shear rate is not constant

Page 137: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

113

0 50 100 150 200 250 3000

1

2

3

4

5

t∗

y/a

Ca = 0.10Ca = 0.142Ca = 0.30Ca = 0.60

(a)

2 2.5 3 3.5 4 4.5 50

0.1

0.2

0.3

0.4

0.5

y/a

D

Ca = 0.10Ca = 0.142Ca = 0.30Ca = 0.60

(b)

Figure 8.2: Migration of NH capsules in a Newtonian fluid in a pressure-driven flow.(a) Trajectory of the center of mass of a capsule y as a function of time t∗ in the wall-normal direction. The walls are at y = 0 and y = By = 10a. (b) Capsule deformationD as a function of center of mass of a capsule y.

Page 138: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

114

2 2.5 3 3.5 4 4.5 50

0.01

0.02

0.03

0.04

0.05

0.06

y/a

(N1−

N2)/

(ηγ w

)

Ca = 0.10Ca = 0.142Ca = 0.30Ca = 0.60

(a)

0 50 100 150 200 250 300

0.2

0.25

0.3

0.35

0.4

0.45

0.5

t∗

y/B

y

2a/By = 0.20

2a/By = 0.05

NH, Ca = 0.142SK, Ca = 0.60

(b)

Figure 8.3: (a) Difference between first (N1) and second (N2) normal stress differencesas a function of y. (b) The trajectory of an isolated capsule in a Newtonian fluid fordifferent channel heights. Solid lines are simulation results for channel of heightBy = 10a and dashed are for By = 40a.

Page 139: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

115

along the wall-normal direction. The shear is zero at the center and maximum (in

magnitude) at the walls. The figure shows that as the capsule move towards the

center in the low shear region, the deformation parameter decreases and approaches

to zero (as seen for Ca = 0.60) at the center. Fig. 8.3(a) shows the variation of

the normal stress difference difference N1 −N2 of an individual capsule as a function

of channel height. The the first and second normal stress differences are defined as

N1 = τxx − τyy; N2 = τyy − τzz. Here τ is the extra stress due to the presence of

the capsule. As observed from the figure the more deformable a capsule (higher Ca),

the larger (N1 − N2) it can exhibit and higher is its tendency to migrate away from

the wall. The figure also shows that normal stress difference decreases as the capsule

move towards the lower shear rate region.

It should be noted that migration of an isolated capsule near a single wall is

qualitatively different for pressure-driven flows than for Couette flow. The shear rate

is constant in the case of Couette flow where as the shear rate varies across the channel

height for the case of pressure-driven flows. Theoretical predictions38,87 of migration

of a deformable drop (analogous to a deformable capsule) suggest that in the case

of linear unidirectional shear flow (Couette flow) the wall-induced migration has the

dominant effect (the effect is O(ε4) where ε = 2a/By) . However, in the case of

quadratic unidirectional shear flow (Poiseuille flow) the dominant contribution comes

from the curvature (the effect is O(ε3)). In fact the migration of a deformable particle

is faster in the pressure-driven flow than in the Couette flow having same shear rates

at the walls. The effect of confinement (ε = 2a/By) on migration is shown in Fig.

8.3(b). The figure shows that the effect of channel height on the migration capsule

is significant. Increasing the channel size (decreasing ε ) decreases the the migration

Page 140: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

116

rate.

8.3 Suspensions of capsules in a pressure-driven

Flow

We now consider the dynamics of dilute suspensions of NH capsules in a pressure-

driven flow in a Newtonian fluid as shown in Fig. 8.1. Simulation box of three

different sizes are considered – a cubic box of size 10a, a cubic box of size 16a and a

slit of size 15a,40a and 6.5 in x, y and z directions respectively. The volume fraction

of capsules considered is φ = 0.10 which corresponds to simulations of 25, 100 and 100

capsules in the corresponding simulation domains. For all simulations of suspensions

of capsules, we use 320 triangular elements to discretize the capsule surface which

corresponds to 162 nodes per capsule. These parameters of capsules were chosen to

keep the computational cost moderate while still capturing the suspension dynamics

with a reasonable accuracy (see chapter 6). The capsules are initially distributed

randomly in the computational domain and are then subjected to a the Poiseuille

flow for a time period of 300 time units. The net motion of the capsules towards the

center of the channel is quantified by examining the time evolution of the average

distance of the capsule’s center from the center of the channel < |y − ycenter| >. Fig.

8.4(a) shows < |y − ycenter| > for the capsules as a function of time in a Newtonian

fluid at different Ca. The value of < |y−ycenter| > initially decreases and then reach a

steady state indicating initial migration of capsules towards the center of the channel

leading to an accumulation of the capsules in the core of the channel at steady state.

We consider the time at which the steady state is reached as the time after when the

Page 141: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

117

0 50 100 150 200 250 3001

1.5

2

2.5

t∗

<|y

−y ce

nter|>

/a

NH, Ca = 0.08NH, Ca = 0.142NH, Ca = 0.30NH, Ca = 0.60

(a)

0 50 100 150 200 250 3000.5

0.6

0.7

0.8

0.9

1

t∗

<u>

/U

−−− <U>/U

NH, Ca = 0.08NH, Ca = 0.142NH, Ca = 0.30NH, Ca = 0.60

(b)

Figure 8.4: (a) Average distance from the centerline < |y − ycenter| > of suspensions(φ = 0.10) of NH capsules in a Newtonian fluid in a a cubic box of size 10a as afunction of time t∗. (b) Average velocity of the center of mass of capsules in flow(x)direction as a function of time t∗. The dashed line represents the average velocityof the undisturbed flow (〈U〉/U = 2/3).

Page 142: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

118

change in average distance from the center or average center of mass of the capsules,

is less than 5%. For all simulations of the suspensions of capsules, we take this time

as t∗ ≥ 200 as observed from the figure. The figure also shows that the extent to

which the capsules move towards the center of the channel increases with an increase

in Ca indicating that as the deformability of the capsules increases, the capsules

have a greater tendency to migrate away from the walls and move closer toward the

center of the channel. Comparison of the present results to the results of suspensions

in Couette flow (chapter 6, Fig. 6.12(a)) indicate that the migration of capsules

towards the center of the channel in pressure-driven flow is faster than that of the

Couette flow (both cases have same shear rate at the walls) indicating that curvature

enhances the net migration of capsules.

Migration of capsules towards the center and accumulation at the core of the

channel can also be quantified by examining the average velocity of the capsule’s

center. Fig. 8.4(b) shows the average velocity of the center of mass of capsules in

flow (x)direction as a function of time. The dashed line represents the average velocity

of the undisturbed flow given as 〈U〉 = 2/3U . It should be noted that the average

velocity of the capsules in the flow direction is proportional to the flow rate of the

capsules. The figure shows that as the flow starts, the capsules migrates towards the

center and hence their average velocity increases (in pressure-drive flow the velocity of

the fluid is maximum at the center). Interestingly, the average velocity of the capsules

is more than the average velocity of the undisturbed flow indicating that capsules get

accumulated away from the walls of the channel and start flowing with the faster

moving fluid near the center. The figure also shows the effect of deformability on the

average velocity of the capsules. More deformable capsules (high Ca) migrate more

Page 143: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

119

0 2 4 6 8 100

0.2

0.4

0.6

0.8

1

y/a

u/U

NH, Ca = 0.08NH, Ca = 0.142NH, Ca = 0.30NH, Ca = 0.60

(a)

0 2 4 6 8 100

0.01

0.02

0.03

0.04

0.05

0.06

y/a

frac

tion

NH, Ca = 0.08NH, Ca = 0.142NH, Ca = 0.30NH, Ca = 0.60

(b)

Figure 8.5: (a) Steady-state distribution of the velocity of capsule’s center (symbols)in flow direction (x) as a function of y in a Newtonian fluid in a cubic box of size10a. The dashed lines are the quadratic fits to the symbols. The solid line (black)represents parabolic profile of the undisturbed velocity. (b) Steady state distributionof capsules as a function of y. The walls are at y = 0 and y = 10a.

Page 144: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

120

towards the center and hence have higher average velocity in the flow direction (∝

flow rate).

The steady-state distribution of the velocity of capsule’s center (symbols) in flow

direction (x) as a function of y is shown in Fig. 8.5(a). Symbols are the simulation

data for the velocity of capsule’s center at various Ca and the dashed lines are the

corresponding quadratic fits. For reference the parabolic profile of the undisturbed

flow is also shown (solid black line). The primary observation is that suspensions of

capsules leads to the blunting of the velocity profiles. As noted in chapter 3.2, the

capsule membranes are required to move with the fluid velocity in order to satisfy

no-slip boundary condition. Hence the velocity of the center of capsules is directly

related to the velocity of the fluid elements. The figure shows that at steady-state the

capsule accumulate near the center of channel and move slowly than the corresponding

fluid element in the undisturbed flow. This leads to the blunting of the velocity

profiles. The blunting of the velocity profiles has also been observed in suspensions of

rigid particles43, emulsions of deformable liquid droplets89 and suspensions of elastic

capsules94,70. The shear-induced particle drift in this leads to the blunting of the

velocity. The deformability of the capsule paly an important role in the blunting of

the velocity profiles. More deformable capsules cause less blunting of the velocity and

vice versa.

Fig. 8.5(b) shows the steady state distribution of capsules, represented as the

fraction of capsules, at various values of Ca as a function of channel height. The

fraction is calculated as an areal fraction of membrane surface as described in chap-

ter ??. Fig. 8.5(b) shows that at steady state the distribution of the capsules is

inhomogeneous – capsules have higher concentration near the center of the channel

Page 145: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

121

0 50 100 150 200 250 3000.12

0.14

0.16

0.18

0.2

0.22

0.24

t∗

<|y

−y ce

nter|>

/By

2a/By = 0.20

2a/By = 0.125

2a/By = 0.05

Figure 8.6: Average distance from the centerline < |y − ycenter| > of suspensions(φ = 0.10) of NH (Ca = 0.142) capsules in a Newtonian fluid for different channelheights.

and a capsule-depleted region near the walls. The thickness of the depleted layer as

well as the concentration near the center of the channel increases with an increase

in Ca indicating that more deformable capsules move farther away from the walls

and accumulate closer towards the channel center. The figure also shows the layer

formation of capsules near the walls, as shown by the off-center peaks in the distri-

bution. The layer formation has also been observed by Li and Pozrikidis89 in their

simulations of suspensions of liquid drops in wall bounded flows. They observed that

the layer formation was a result of using confined channel size. We have studied the

effect of box size on the steady state distribution of capsules in Couette flow (chapter

??). Results show formation of layers diminishes in large box indicating that layer

Page 146: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

122

formation is an artifact of using small size of the channel.

The effect of confinement (2a/By) on suspensions of capsules is shown in Fig. 8.6.

As observed from the figure, increasing the channel height decreases the migration

rate of the capsules towards the center i.e. the capsules in a bigger channel take longer

time to reach a steady state than in the case of smaller one. The capsules in a channel

of height 40a is far from reaching the steady state even at a time period of 300 time

units where as the capsules in a channel of height 10a reach a steady state around

150 time units. If we assume migration to be a dominant phenomenon in suspension

dynamics then the time it takes to reach a steady state should scale as the time it

takes a capsule to migrate to the capsule-free layer. Theoretical predictions38,87 of

migration of a deformable drop in a pressure-driven flow show that migration velocity

scales as O(ε3) where ε = 2a/By. Hence, the time to reach a steady state tss should

scale as tss ∼ (By/2a)3. For example, in the figure, the time it takes to reach a

steady state for capsules in a slit geometry of height 40a should be 64 times the time

the capsules would take to reach a steady state in a slit of height 10a (for the same

volume fraction). Similar scaling was obtained for the suspensions of rigid particles

in pressure-driven flow43.

8.4 Discussion and future work

Numerical simulation results show that migration of an isolated capsule in a pressure-

driven flow in a Newtonian fluid depends strongly on the Ca. The effect of curvature

enhances the migration rate of the capsules. The results of suspensions of elastic

capsules in pressure-driven flow show that capsules collectively migrates towards the

Page 147: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

123

center of the channel and reach a steady state showing an inhomogeneous distribution

– having higher concentration at the core of the channel and a depletion layer near

the walls. The migration of capsules leads to the blunting of the velocity profiles. The

Ca has a strong influence on the suspension dynamics – more deformable capsules

migrates more towards the center leading to a thicker depletion layer near the walls

and more concentration of capsules near the center.

As a part of the future work, we would like to investigate the effect of long chain

drag reducing additives on the dynamics of suspensions of capsules in pressure-driven

flows. Since most of the experiments of DRAs on blood flow pertains to the case of

pressure-driven flow, it would be of high interest to perform simulations to observe

the effects, at least qualitatively, on suspensions of capsules in pressure-driven flow.

The main focus of the study would be to gain a mechanistic understanding on the

effect of polymers on concentration distribution, flow resistance, capsule-depletion

layer and velocity profiles of the capsules at steady state.

Page 148: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

124

Chapter 9

Future work

9.1 Suspensions of Red Blood Cells

With the current numerical method and model, a variety of interesting issues can

be addressed. Until now the focus of the study have been on the dynamics of elastic

capsules which were nominally spherical in shape. The spherical capsules were chosen

for the study because of the simplicity of shape. The short term goal now is to

incorporate the model to study the dynamics of suspensions of fluid-filled elastic

capsules which have the same nominal shape as that of the red blood cell i.e. capsules

having biconcave discoid shapes. As a starting point, we will use a discoid shape as

the equilibrium shape of the capsules. The analytical form of the discoid shape is

given as95,96,70

x = R0x′, y =

1

2R0

(1− r2)(

C0 + C2r2 + C4r

4)

, z = R0z′, (9.1)

Page 149: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

125

Figure 9.1: A schematic showing suspensions of RBCs in the microcirculation.

where x, y, z are the coordinates of the cell surface, x′, y′, z′ are the coordinates of

a unit sphere, R0 is a constant to preserve the volume, r2 = x′2 + z′2, C0 = 0.32,

C2 = 2.003 and C4 = −1.123. Fig. 9.1 illustrates the basic situation of interest. As

noted in our methodology (chapter 3.2), we use two different models for the capsule

membrane – neo-Hookean (NH) model and Skalak (SK) model. The SK model is more

accurate representation of RBCs than NH model as it can be parametrized to yield

a strong resistance to area change relative to its resistance for shear deformation

as is the case for RBCs60. The Skalak model contains a shear modulus G and an

additional parameter C associated with the energy penalty for area change. Typically,

C � 1 indicating approximate area-incompressibility. For modelling the suspensions

of RBCs, SK model will be used.

Page 150: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

126

In our model capsule membranes are assumed to have no bending. It is because the

membrane is assumed to have zero thickness that displays no resistance to bending.

In reality, the RBCs have membranes that have a finite thickness (∼ 4nm) and hence

have small resistance to bending. For example, for an RBC, R ∼ 1µm and G ∼

10−6J/m32. The bending modulus kb of the membrane is kb ∼ 10−20J97. The relative

contribution of shear deformation over bending can be estimated as R2G/kb ≈ 10−2−

10−3. Thus for modelliing spherical capsules, where there are no sharp curvatures,

the contribution of bending to the total membrane force can neglected. However, an

RBC can display sharp curvature depending on the flow geometry. For example, for

bending energy to have the same order as the elastic energy, κ/κ0 ∼ 103, where κ is

the local mean curvature and κ0 is the curvature in the undeformed state. In most

physical conditions, for example in suspensions of RBCs in confined geometry, the

value of κ/κ0 can be of that order indicating the need to incorporate bending in the

existing model. Here we outline the implementation of bending in our model.

For a closed membrane Γ, bending energy called Canham-Helfrich energy98,99 is

given by:

E =KB

2

Γ

(2κH)2dS +KB

Γ

κGdS, (9.2)

where κH , κG are the mean and gaussian curvature of the surface, and EB and EB are

bending moduli. Gauss-Bonnet theorem ensures that the second term is a constant

when no topological changes are involved. Force density due to bending is given by

first variation of the free energy Eq. 9.2:

fbi = KB

(

2∆sκH + 2κH(κ2H − κG)

)

n, (9.3)

Page 151: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

127

where ∆sH is the Laplace-Beltrami operator over the surface. The resultant mem-

brane force density is the combination of bending and mechanical tension, fci = fei +fbi ,

where fei is the force density due to elasticity of membrane as discussed in chapter

3.2. The details of the methodology is provided by Boedec et al.100 in their work on

vesicle dynamics simulation.

9.2 Drug delivery

In recent years, colloid, polymer and lipid chemistry have been harnessed to create

a great diversity of particles based on silica, polymer gels, polymeric micelles or

liposomes101,102,103,104,105,106,107. Furthermore, a great variety of targeting moities (

antibodies, aptamers, vitamins, targeting peptides) have been conjugated to these

particles. Some interesting drug delivery particles are shown in Figure 9.2.

Recent years have seen increase in shape of these particles other than spherical.

Evidence shows that wormlike “filomicelles” developed by Discher and coworker106

stay longer in blood than spherical counterpart. Mitragotri and coworkers104 synthe-

sized nonspherical particles and observed similar results. DeSimone and coworkers108

developed a variant of imprint lithography to generate particles of wide variety of

shapes. With advances in chemistry and nanotechnology for particle synthesis, there

is a growing realization that shape of particle is a important determinant of their life

in blood stream and their specificity. Numerical complexity will augment in studying

behavior of these complex shape particles in an already complex blood circulation.

Principles which govern the distribution of such particles in the circulation are not

yet known and will be investigated as a part of the future work.

Page 152: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

128

Figure 9.2: (a) Spheres. (b) Rectangular disks. (c) Rods. (d) Worms. (e) Oblateellipses. (f) Elliptical disks. (g) UFOs. (h) Circular disks. (Scale bars: 2 µm.)104

We can use the same theoretical approach developed in chapter ?? to make a

prediction of the evolution of a drug delivery particle in suspensions of deformable

particles (capsules/RBCs). If we simply take the evolution of the concentration c of

the drug delivery particle to obey the same Fokker-Planck equation (eq. 6.6) given

above for a dilute suspensions of capsules, then the balance of shear-induced diffusion

and hydrodynamic migration of the particle can be expressed as

∂c

∂t= −∂jy

∂y= − ∂

∂y

(

vmig,pc−∂ (Ds,p(y)c)

∂y

)

. (9.4)

Here vmig,p is cross-stream drift and Ds,p(y) is the short time diffusivity of the drug

delivery particle. The above equation is equivalent to writing the evolution equation

for the probability of finding a particle at a given y−position in the dilute limit; with

the assumption that the motion of particle is a continuous Markov process85,86 and is a

good approximation for time scales longer than about γ−1. In suspensions of capsules

with drug delivery particles, the short time diffusivity of the particle will be dominated

by the shear-induced diffusion because of the presence of capsules. Here Ds,p takes

Page 153: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

129

rigid particles −marginated

critically marginatedparticles

highly deformable particles−not marginated

Figure 9.3: A schematic showing hypothesized distributions of drug delivery particlesin the microcirculation.

the same functional form as Dself for capsules in eq. 6.4. Consider the limiting case of

a rigid particle, which is a good approximation for a leukocyte, platelet or for many

proposed drug delivery particles. For rigid particles, the wall-induced hydrodynamic

migration effect vanishes and there are no other obvious mechanisms for drift (at low

Reynolds number), so at steady state, the flux balance simply reduces to

0 = − ∂

∂y

(

∂ (Ds,p(y)c)

∂y

)

, (9.5)

which gives the solution as

c = Kp/Ds,p(y). (9.6)

Here Kp is a constant, and Ds,p(y) is an increasing function of capsule concentra-

Page 154: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

130

tion, so near the wall, where the capsule concentration is lowest, this simple equation

predicts that the particle concentration is highest, which is indeed the case. This

idea can be generalized to make predictions for finding the ideal location of the drug

delivery particles in suspensions of capsules/RBCs based on the physical properties

of the particle such as Ca, size, aspect ratio. For example, a relatively rigid parti-

cle in a flowing suspension of capsules/RBCs, as determined by its size, shape and

mechanical properties through capillary number, will most likely be found in the cap-

sule depletion layer. In contrast, a highly flexible particle will have larger migration

velocity than an capsule and so it will probably not be found in the depletion layer

but rather migrate into the concentrated layer of capsules where it will then diffuse

by shear-induced diffusion. A schematic of the hypothesized prediction is shown in

Fig. 9.3. Therefore as a part of future work, we want to extend the current model

to incorporate flow of mixtures of particles having different physical properties and

then to characterize the mechanical properties of drug delivery particles in flowing

suspensions of capsules/RBCs.

9.3 Leucocyte margination under the influence of

polymers

Finally, we would want to extend the our current model and methodology to facilitate

the study of mixture of particles (capsules, RBCs) with different physical properties

along with the effect of polymers. As noted above (chapter 2.1) the migration of RBCs

towards the center of a vessel is assumed to increase the probability of less deformable

leucocytes, such as white blood cells and platelets, to be found near the blood vessel

Page 155: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

131

walls, a phenomenon known as margination. Margination is found to play a key role

in the process of inflammation35,36. On the other hand, microchannel experiments7

have shown that the addition of small amount of these polymer additives to the

suspensions of RBCs resulted in a redistribution of RBCs with a significant reduction

in the thickness of the cell-free layer. A recent experiment by Zhao et al.47 showed

that the addition of small amount of drag reducing polymers (10ppm of 4×106D PEO)

to blood can not only decrease the thickness of the cell-free layer layer but also reduce

the concentrations of leucocytes in these layers. Therefore as a part of future work

we would want to use our methodology and computational approach to examine the

dynamics of suspensions of elastic capsules (having discoid shapes representing RBCs)

with small concentration of leucocytes, modelled as capsules with low Ca, under the

presence of small amount of DRAs. Our simulation results have shown that DRAs

can influence the behavior of suspensions of capsules leading to the decrease in the

thickness of capsule-depleted layer. We now want to extend our study to incorporate

the presence of small concentrations of rigid particles (low Ca capsules) to find out if

we can reproduce the results of Zhao et al.47 qualitatively.

Page 156: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

132

Appendix A

Single Polymer Dynamics

The fluctuation-dissipation theorem implies a nontrivial coupling between the Brow-

nian motion and the configurations of the polymer molecules and capsules. This

coupling is expensive to compute78. In dilute solutions and at high Wi, the effect of

Brownian motion on polymer dynamics gets relatively weak (Weissenberg number is

proportional to particle Peclet number). The Brownian motion of polymer under such

conditions can be approximated by using a free draining approach in the Brownian

term in an unbounded domain – where the contribution of hydrodynamic interaction

to the Brownian motion of polymer is neglected. In this section, we validate our

approximation of neglecting the hydrodynamic interaction in the Brownian term for

the polymers in the limit of high Wi and infinite dilution.

As a test case, we place a single polymer molecule (3-bead chain) in an unbounded

domain under shear flow with shear rate γ. We then study its evolution under the

influence of flow with and without the effect of hydrodynamic interactions in the

Brownian term respectively. The equation of motion of each bead of the chain is

Page 157: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

133

governed by the balance of forces:

0 = fhi + f si + f ei + f bi . (A.1)

Here the superscripts h, s, e and b denote the hydrodynamic drag, spring, external

and Brownian forces respectively. This force balance leads to the following evolution

equation:

dxp = [u∞ +M · fp] dt+B · dW, (A.2)

where fp = f e + f s, M is the mobility tensor expressed as

Mi,j =1

ζδδij + (1− δij)G

R,ξ∞ (xi,xj) , (A.3)

B ·BT = (2kBT )M, (A.4)

u∞ (= γyex) is the externally imposed fluid velocity field, dW is a vector of indepen-

dent random variables, each chosen from a Gaussian distribution with zero mean and

variance dt and GR,ξ∞ (xi,xj) is the free space regularized Green’s function78 given by

GR,ξ∞ (xi,xj) = 1

8πηs

[

δ + xxr2

]

(

erf(ξr)r

)

+ 18πηs

[

δ − xxr2

]

(

2ξπ1/2 e

−ξ2r2)

, (A.5)

x = xi − xj, r = |x|.

Here ξ is the regularization parameter; its reciprocal represents the length scale over

which the force is spread. ξ scales as the size of the bead radius i.e. ξ ∼ a−1p . Eq. A.5

reduces to the Oseen-Burgers tensor as ξ → ∞. Further, the mobility tensor will be

Page 158: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

134

positive definite for ξ−1 ≥ 3ap/√π. We take this value here. The term B · dW in the

evolution equation (eq. A.2) represents the Brownian contribution to the motions of

polymer molecules. The coupling between the Browinian motion and configuration of

polymer molecules is represented in eq. A.4. Note that computation of B is expensive

(scales as O (N3) for N particle system). Neglecting the hydrodynamic interaction in

the Brownian term simplifies eq. A.4 as

B =

2kBT

ζδ, (A.6)

and reduces the evolution equation (eq. A.2) to

dxp = [u∞ +M · fp] dt+√

2kBT

ζδ · dW. (A.7)

The above equation is identical to the evolution equation (eq. 3.25) used for polymer

molecules in bounded domain where the hydrodynamic interaction in the Brownian

term is neglected. Note that computation of B in eq. A.6 now scales as O (N).

The results on the dynamics of single polymer molecule in an unbounded shear

flow as a function of Wi are shown in Fig. A.1. We denote “HI” (solid lines) as the

simulation results including the hydrodynamic interactions in the Brownian term i.e.

using eq. A.2 and eq. A.4; and denote “FD” (dashed lines) as a free-draining model

representing simulations neglecting the hydrodynamic interactions in the Brownian

term i.e. using eq. A.7. Each data point the figures corresponds to averaging over 25

different initial configurations of polymer molecule. The mean properties of the poly-

mer dynamics in the figure are calculated after the polymer molecules are equilibrated

for a time period of 50 units. Fig .A.1(a) shows the steady state RMS end-to-end

Page 159: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

135

0 5 10 15 20

0.01

0.1

1

Wi

<R

0>/L

c

HIFD

(a)

0 5 10 15 200.001

0.01

0.1

1

10

Wi

< τ

p>/(

η γ

n)

HIFD

τxx

τyy

τzz⋅

(b)

Figure A.1: (a) Average steady state RMS end-to-end distance < R0 > and (b) av-erage steady state polymer stress < τ p > as a function of Wi for a single polymermolecule in an unbounded shear flow. x, y and z represents flow, gradient and neutraldirections respectively. “HI” denote simulations including hydrodynamic interactionsin the Brownian term. “FD” represents simulations neglecting hydrodynamic inter-actions in the Browninan term.

distance < R0 > as a function of Wi. The figure shows that the difference in < R0 >

between HI and FD case decreases as Wi increases. Fig. A.1(b) shows the normal

components of steady state polymer stress < τ p > as a function of Wi for HI and FD

cases respectively. Here the expressions for the stress tensor for a bead spring chain,

as given by Kramers-Kirkwood method77, is

τ p = n

beads∑

i=1

(xi − xc) fpi , (A.8)

where xc is the center of mass of a polymer molecule and n is the number density. Here

also the figure shows that the results for HI and FD case converge at high Wi (≥ 5)

indicating that the effect of hydrodynamic interaction in the Brownian motion gets

weak as Wi increases. This shows that the approximation of neglecting hydrodynamic

interaction in the Brownian term holds good at high Wi and low concentration of

Page 160: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

136

polymer.

Page 161: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

137

Appendix B

Comparison of GGEM/IBM with

other methods

B.1 Overview of Conventional IBM Method

The immersed boundary method (IBM) has been used to study fluid-structure inter-

actions, especially in the field of biological fluid dynamics. This section deals with a

brief review on the conventional IBM described by Peskin75. The IBM , which is both

a mathematical formulation and a numerical scheme, involves a mixture of Eulerian

and Lagrangian variables. The coupling between these two types of variables is done

by the Dirac delta function. The Eulerian variables are defined on a fixed Cartesian

mesh, and the Lagrangian variables are defined on a curvilinear mesh that moves

freely through the fixed Cartesian mesh without being constrained to adapt to it in

any way at all. The interaction equations of the numerical scheme involve a smoothed

approximation to the Dirac delta function75,80.

Page 162: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

138

The problem under consideration is motion of liquid capsules and polymer molecules,

immersed in a surrounding fluid. The Eulerian description applies to the fluid and

the Lagrangian description applies to the body (capsules and polymers in our case).

At zero Reynolds number, the fluid obeys Stokes equations. Let x be the Eulerian

coordinate, u be the Eulerian velocity and ρ be the Eulerian force density that the

immersed body applies to the fluid. Let η be the fluid viscosity and P be the pressure.

In the IBM, the Stokes equations hold in all of space, including that occupied by the

body. These equations are :

−∇P (x) + η∇2u(x) = −ρ(x),

∇ · u(x) = 0,(B.1)

with the boundary condition as

u(xW ) = u|W , (B.2)

where xW represents the solid boundary and u|W is velocity at the boundary. All

quantities in eqs. B.1 and B.2 are function of Eulerian variable x.

The Lagrangian variables define the configuration of body immersed in the fluid.

Let xc denote the coordinate of the surface of body and uc denote its velocity. Let

f c be the Lagrangian force applied by the immersed body. A capsule, for example,

comprises a nominally spherical elastic membrane of radius, enclosing a Newtonian

liquid with density equal to that of the outside fluid. The viscosities of the liquids

inside and outside the drop are considered to be identical and equal to ηs. The

membrane is represented as a collection of points (or elements). The forces resisting

deformation are calculated by computing displacements of the vertices of the deformed

Page 163: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

139

element with respect to the un-deformed element. The equation of motion is :

dxci

dt= uc(xc

i), (B.3)

where subscript i denote a point on the capsule membrane.

It is now important to understand how the Eulerian variables interact with the

Lagrangian variables. The interactions can be expressed by using the property of

Dirac delta function as

ρ(x) =∑

δξIBM(|x− xc

i |) f ci , (B.4)

dxci

dt= uc =

δξIBM(|xc

i − x|)u (x) , (B.5)

where δξIBMis a smoothed approximation of delta function and ξ−1

IBMdenotes its

length scale for smoothing. Eq. B.4 shows the conversion of Lagrangian to Eulerian

variables. It states that the Eulerian force density and the Lagrangian force density

are, at corresponding points, equal as densities if integrated over a volume. Eq. B.5

shows the conversion of Eulerian to Lagrangian variables. It states that the Eulerian

velocity, u, and the Lagrangian velocity are equal at corresponding points. This also

comes from the fact that the fluid obeys no-slip condition on the surface of the body.

Let hc be the characteristic element size of the membrane and let hg be the size

of the Eulerian mesh. To avoid leaks, the restriction needed is

hc < hg/2, or hg ∼ hc. (B.6)

Page 164: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

140

In the IBM method by Peskin75, force distributions at moving interfaces or mem-

branes are discretized as distributions of regularized point forces, where the length

scale for smoothing the delta function scales as the Langrangian grid spacing used in

the simulation, i.e.

ξIBM

−1 ∼ hc. (B.7)

This ensures that the force density associated with each node is uniformly spread over

the length scale of the associated elements, thereby preventing fluid from penetrating

the membrane surface, and also preventing the un-physically large fluid (and thus

node) velocity that would be present at the node if ξIBM is made too large (in which

case δ approaches a true delta function). Moreover, this scaling is a natural one so

that δξIBM→ δ as hc → 0.

Eqs. B.6 and B.7 suggest that for the accuracy of the method all the length scales

should be equal i.e. all the length scales should resolve the membrane mesh size

(Lagrangian mesh), hc. A schematic of the length scales are shown in Fig. B.1(a).

Therefore, a conventional IBM requires h3c mesh points (if a uniform mesh is used).

An important remark is that the equations of motion that are derived for a viscous

incompressible fluid containing an immersed elastic boundary are mathematically

equivalent to the conventional equations that one would write down involving the

jump in the fluid stress across that boundary.

B.2 IBM and GGEM

With the same definition of Eulerian and Lagrangian variables defined in previous

section, the approach taken in our present work is to use an O (N) (where N is

Page 165: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

141

hg ∼ ξ−1

IBM

hc ∼ ξ−1

IBM

(a) length scales associated with conventional IBM method

hc

hg ∼ α−1

∼ ξ−1GGEM

(b) length scales associated with IBM/GGEM

Figure B.1: (a) Schematic of different length scales associated with the conventionalimmersed boundary method. (b) Schematic of different length scales associated withIBM/GGEM.

Page 166: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

142

the number of nodes to represent the immersed body) particle-particle/particle-mesh

method that we have recently developed for Stokes flow problems in nonperiodic

geometries, called GGEM (General geometry Ewald-Like Method)78,81. GGEM starts

with splitting the interaction equations for force density into two parts,

ρ(x) = ρl(x) + ρg(x),

ρl(x) =∑

i

[δξGGEM(|x− xc

i |)− gα (|x− xci |)] f ci ,

ρg(x) =∑

i

gα (|x− xci |) f ci ,

(B.8)

where δξGGEM(r) is a regularized delta function and the screening function, gα (r)

is used to split the force density into a “local” and “global” part denoted by the

subscripts l and g. The regularized delta function and the screening functions are

given by

δξGGEM(r) =

ξGGEM

3

π3/2e(−ξGGEM

2r2)[

5

2− ξGGEM

2r2]

(B.9)

gα(r) =α3

π3/2e(−α2r2)

[

5

2− α2r2

]

, (B.10)

where α is the screening parameter. The screening function takes the same mathe-

matical form as the regularizing delta function. The flow field is a superposition of

flow driven by ρl(x) and ρg(x):

u(x) = ul(x) + ug(x), (B.11)

Page 167: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

143

or equivalently in terms of Lagrangian velocities:

uc(xci) = uc

l (xci) + uc

g(xci). (B.12)

The equation of motion is :

dxci

dt= uc(xc

i) = ucl (x

ci) + uc

g(xci). (B.13)

The general idea in GGEM is to separately compute the velocity fields due to

the two force densities. The total velocity is then simply the sum of the individual

contributions.

The local velocity ul(x), which results from the force density ρl(x). This is com-

puted analytically using solutions to Stokes’ equations in an unbounded domain - the

error incurred by assuming an unbounded domain will be canceled out in the global

solution described below. One can show that the modified force distribution, eq. B.8

yields a velocity field given by

ul(x) =∑

i

GRl (x− xc

i) · f ci , (B.14)

where the subscript l indicates the “local” contribution to the velocity. The modified

Green’s function, GRl is given by,

GRl (x) =

1

8πηs

[

δ +xx

r2

]

[

erf (ξr)

r− erf (αr)

r

]

− 1

8πηs

[

δ − xx

r2

]

[

π1/2e−α2r2 − 2ξ

π1/2e−ξ2r2

]

. (B.15)

Page 168: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

144

Because of the presence of the screening function in the local force density, this

function decays rapidly to zero over a distance proportional to α−1. The calculation of

the local velocity field begins by first identifying point forces located within a sphere

of fixed radius 4/α. Any force outside this cut-off radius is ignored. The value of α is

empirically determined. For a Lagrangian point xcj on the immersed body, the local

Lagrangian velocity can be written as

ucl

(

xcj

)

=

n∑

i=1,i 6=j

GR,cl

(

xf − xci

)

· f ci + limx→0

GR,cl (x) · f cj . (B.16)

Here n is the number of Lagrangian points which lie within the cut-off sphere with

center at(

xcj

)

. In eq. B.16, the velocity at Lagrangian point includes a contribution,

a “self-term”, due to the force exerted by the same point on the fluid. For a point

force, this velocity is infinite. However, the Green’s function is regularized, and hence

has a finite limit as the separation vector vanishes.

This gives the local velocity at any point (both Eulerian and Langrangian) in the

system.

The global contribution to the velocity, ug (x), is due to the global part of the force

distribution, ρg (x) in eq. B.8. The solution to the Stokes equations with a forcing

function ρg (x) is calculated numerically, requiring that the total velocity u (x) satisfy

appropriate boundary conditions. The Stokes equation for the global field are:

−∇P (x) + η∇2ug(x) = −ρg(x),

∇ · ug(x) = 0.(B.17)

All quantities in eq. B.17 are function of Eulerian variable x. At points xW on the

Page 169: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

145

bounding walls, boundary conditions has to be satisfied. Therefore, the corresponding

boundary condition for the global velocity field is given by

ug|W = −ul|W + u|W , (B.18)

where u|W is the actual boundary condition and ul|W is the known local component

of the velocity field. Note that by imposition of this boundary condition for ug (x),

we cancel out the error introduced by using unbounded-geometry solutions for the

local problem.

The interaction equation for global force density from Lagrangian to Eulerian

quantities is shown in eq. C.10. The global calculations are performed on Eulerian

mesh having characteristic length as hg. With the same argument as in the previous

section, the constraint on the mesh size is :

hg ∼ α−1. (B.19)

This is the key difference between conventional IBM and GGEM. It is due to the

fact that the interaction equations of global force density, ρg(x), between Lagrangian

and Eulerian quantities are linked by the screening function, gα(x), having a char-

acteristic length scale as α−1. Hence the global calculations can be performed on a

coarse grid depending on α. However, the the interaction equations of local force

density, ρl(x), between Lagrangian and Eulerian quantities are linked by the delta

function, δξ(x), having a characteristic length scale as ξ−1. So to maintain numerical

accuracy in the local calculations, ξ−1 scales as the Lagrangian grid spacing used in

Page 170: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

146

the simulation, i.e.

ξ−1 ∼ hc. (B.20)

Fig. B.1(b) shows the schematic of the length scales associated with GGEM. Fig.

B.1 shows these differences between the two methods.

Now what remains is the interactions from Eulerian variables to Lagrangian vari-

ables. The global velocity in Lagrangian coordinate, ucg(x

ci), can be obtained from

the interpolation of the global velocity in Eulerian coordinate, ug(x), as :

ucg(x

ci) = L(xc

i )ug(x), (B.21)

where L(x) is a interpolation function in three dimensions. In our work we use second

order Lagrange interpolation.

B.3 BIM and GGEM

We turn now to the relationship between IBM/GGEM and boundary integral equation

methods. A Stokes flow u(x) driven by a force density ρ(x) is given by:

u(x)− u∞(x) =

V

G (x,x′) · ρ(x′) dx′, (B.22)

where G (x,x′) is the Green’s function for the geometry and boundary conditions of

interest and u∞(x) is the velocity in the absence of the forcing. If the force density

is localized on an interface S within the fluid (e.g. a capsule membrane), the force

density can be written as:

Page 171: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

147

ρ(x′) = −∫

S

∆t(x′′)δ(x′ − x′′) dS(x′′), (B.23)

where −∆t(x′′) dS(x′′) is the total force exerted by the surface on the fluid at point

x′′. Inserting eq. B.23 into eq. B.22 yields:

u(x)− u∞(x) = −∫

V

S

G (x,x′) ·∆t(x′′)δ(x′ − x′′) dS(x′′) dx′, (B.24)

which simplifies to:

u(x)− u∞(x) = −∫

S

G (x,x′′) ·∆t(x′′) dS(x′′). (B.25)

This is the standard boundary integral equation for an interface immersed in a single

fluid.

In IBM/GGEM, eq. B.23 is discretized by (1) approximating δ(x′−x′′) by δc(x′−

x′′); (2) approximating −∆t(x′′) dS(x′′) by f c(x′′); and (3) approximating the integral

as a sum, where the x′′ are nodal positions xi. This yields the IBM/GGEM force

density

ρ(x′) =

nodes∑

i

δc(x′ − xi)fci . (B.26)

This is identical to eq. 3.26 in the absence of polymer molecules. Inserting this into

eq. B.22 leads to a discretization of the boundary integral equation eq. B.25:

u(x)− u∞(x) =

nodes∑

i=1

GR,c (x,xci) f

ci , (B.27)

Page 172: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

148

where GR,c (x,xci) is the regularized Green’s function that corresponds to δc(r). This

velocity field is the solution to eq. C.1 with the relevant boundary conditions and

force density eq. 3.26 (again with no polymers), and is precisely what is computed

by GGEM. These results establish the relationship between IBM/GGEM and the

boundary integral method. An identical analysis, replacing δc(r) by δcIBM

shows the

same result for IBM in the Stokes flow case.

Page 173: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

149

Appendix C

Implementation Issues-Slit

The problem under consideration is a Stokes flow in a slit geometry of fluid with

velocity u, pressure P , viscosity η. To treat the bounded domains, the present method

starts with restatement of the force-density expression: ρ(x) = ρl(x) + ρg(x). Here

ρl(x) is the “local” density which drives the local contribution of the velocity field,

ul(x), and ρg(x) is the “global” density which drives the global contribution of the

velocity field, ug(x). By linearity u(x) = ul(x) + ug(x). The local velocity ul(x) is

calculated analytically assuming an unbounded domain. The global velocity ug(x) is

calculated numerically on a coarse grid with appropriate boundary conditions. The

Stokes equation for the global field are:

−∇P (x) + η∇2ug(x) = −ρg(x),

∇ · ug(x) = 0,(C.1)

or

− ∂P∂xi

+ η∂2ug,i

∂x2j

= −ρg,i,

∂ug,j

∂xj= 0,

(C.2)

Page 174: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

150

where x1, x2 and x3 are the flow, wall-normal and vorticity directions respectively

and ug,i, are the corresponding velocity components for the global field. We also use

interchangeably x,y,z for xi and ug, vg, wg for ug,i. The box is periodic in x− and z−

directions and bounded by walls in y− direction. At points xwall on the bounding

walls, boundary conditions has to be satisfied. Therefore, the corresponding boundary

condition for the global velocity field is given by

ug|wall = −ul|wall + u|wall, (C.3)

where u|wall is the actual boundary condition and ul|wall is the local component of

the velocity field.

C.1 Different flow profiles:

Different flow behavior can be solved by using different expressions for pressure and

boundary conditions. For example the pressure P can be restated as

P = −Gx+ p, (C.4)

where G is the mean pressure gradient in the flow direction and p is the pressure driven

by the forces that particles exert on the fluid. The following cases show different

expressions for G and boundary conditions:

For Couette flow in a slit geometry between two parallel plates the unperturbed

Page 175: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

151

velocity profile is linear and is given by

u(y) = γy, v = 0, w = 0, (C.5)

where y = ±By/2 represents wall, By is the height of the channel and γ is strain rate.

The expression for G and boundary conditions are given as

G = 0,

ug|wall = −ul|wall ± γBy/2,

vg|wall = −vl|wall,

wg|wall = −wl|wall.

(C.6)

Here, G is zero and the boundary conditions are no-slip at the walls.

For pressure-driven channel flow the unperturbed velocity profile is parabolic and

is given by

u(y) = Ucl

[

1−(

y

By/2

)2]

, v = 0, w = 0, (C.7)

where y = ±By/2 represents wall, By is the height of the channel and Ucl is the

channel centerline velocity. The expression for G and boundary conditions are given

as

G = 8ηUcl

B2y,

ug|wall = −ul|wall,

vg|wall = −vl|wall,

wg|wall = −wl|wall.

(C.8)

Here G is constant and the boundary conditions are no-slip at the walls.

Page 176: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

152

C.2 Assignment: putting ρg on the mesh - collo-

cation

The global calculations are performed on mesh consisting of Mx ×My ×Mz discrete

points. The global density, ρg(x), which drives the global contribution of velocity

field ug(x), is calculated as

ρg(x) =∑

i

g(|x− xi|) fi(xi), (C.9)

where g(r) is the screening function and fi(xi) is the force exerted by the particle i

at position xi. The mathematical form of the screening function is given by

g(r) =α3

π32

e(−α2r2)[

5

2− α2r2

]

, (C.10)

where α is the screening parameter. The screening function, g(r), satisfies

V

g(r) dV =

∫ ∞

0

4πr2g(r) dr = 1, (C.11)

where V is the infinite space. One of the properties of the screening function is that

it decays rapidly to zero over a distance proportional to α−1. We take this domain

to be a sphere of radius 4/α, eq. C.11 then becomes

∫ 4/α

0

4πr2g(r) dr ≈ 1. (C.12)

Page 177: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

153

The calculation of global density, ρg(x), can then be simplified by identifying all the

particle positions within this domain. Contribution of any particle outside this cut-off

radius is ignored. For any point xM on the mesh, the calculation of the global density

is given by

ρg(xM) =

n∑

i

g(|xM − xi|) fi(xi). (C.13)

Here n is the number of particles which lie within the cut-off sphere with center at

xM .

C.3 Solution approach - FFT/finite differences

For problems involving periodic boundary conditions, Fourier techniques is adopted

to guarantee the periodicity of global velocity ug(x). Using Fast Fourier Transforms

(FFT) eqs. C.2 and C.3 can be written as

ikxηp− G

ηδ(k=0) +

(

k2 − ∂2

∂y2

)

ug =1ηρg,x,

1η∂p∂y

+(

k2 − ∂2

∂y2

)

vg =1ηρg,y,

ikzηp+

(

k2 − ∂2

∂y2

)

wg =1ηρg,z,

ikxu+ ∂v∂y

+ ikzw = 0,

(C.14)

with the boundary conditions as:

ug|y=wall = −ul|y=wall + u|y=wall,

vg|y=wall = −vl|y=wall + v|y=wall,

wg|y=wall = −wl|y=wall + w|y=wall,

(C.15)

where k =√

k2x + k2

z , and kx and kz denote the wave numbers in the x- and z-

Page 178: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

154

directions respectively and superscript ∧ refers to the Fourier transforms as:

ug,i(y) =∑

x

z

ug,ie−ikxxe−ikzz, (C.16)

p(y) =∑

x

z

pe−ikxxe−ikzz. (C.17)

C.3.1 Treatment of periodic boundary condition:

GGEM was originally conceived as an Ewald-like approach to solving Stokes flow

problems in non-periodic domains. In this situation, we imagine that the total velocity

u = ul + ug satisfies appropriate boundary conditions, such as no-slip. In this case

the solution to the global problem satisfies the boundary condition

ug|wall = −ul|wall + u|wall. (C.18)

In the case where one or more of the directions for the flow is periodic, there is an

important subtlety to the implementation of GGEM that needs to be understood. For

simplicity let us just consider the case where there is one periodic direction, x (e.g.

the geometry is a tube). In this case the boundary condition on the total velocity is

simply

u(0, y, z) = u(Bx, y, z), (C.19)

where Bx is the periodic box length and thus that

ul(0, y, z) + ug(0, y, z) = ul(Bx, y, z) + ug(Bx, y, z). (C.20)

Page 179: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

155

Rearranging we have then that

ug(0, y, z)− ug(Bx, y, z) = ul(Bx, y, z)− ul(0, y, z). (C.21)

This result is exact.

Now let’s turn to the exact form of the local solution in the case of the periodic

domain. The exact local velocity field (again treating only the x-direction as periodic)

is

ul(x) =

∞∑

j=−∞

N∑

i=1

Gl(x− (xi + jBxex)) · fi. (C.22)

This velocity field satisfies the condition

ul(Bx, y, z)− ul(0, y, z) = 0, (C.23)

allowing the use of the periodic boundary condition for the global velocity (which we

use):

ug(Bx, y, z)− ug(0, y, z) = 0, (C.24)

In practice, however, the local velocity is treated as:

ul(x) =N∑

i=1

Gl(x− x∗i ) · fi, (C.25)

where x∗i is the image of particle i which is closest to x (minimum image convention).

Eq. C.25 does not exactly satisfy eq. C.23 which in turn leads to a small error in

the global solution if we impose eq. C.24 as the boundary condition for the global

problem.

Page 180: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

156

C.3.2 Proper treatment of k = 0:

For k = 0 (kx = 0 and kz = 0) eq. C.14 becomes:

−Gη+(

− ∂2

∂y2

)

ug(k = 0) = 1ηρx(k = 0),

1η∂p(k=0)

∂y+(

− ∂2

∂y2

)

vg(k = 0) = 1ηρy(k = 0),

(

− ∂2

∂y2

)

wg(k = 0) = 1ηρz(k = 0),

∂v(k=0)∂y

= 0,

(C.26)

with

ug(k = 0)|y=wall = −ul(k = 0)|y=wall + u(k = 0)|y=wall,

vg(k = 0)|y=wall = −vl(k = 0)|y=wall + v(k = 0)|y=wall,

wg(k = 0)|y=wall = −wl(k = 0)|y=wall + w(k = 0)|y=wall,

(C.27)

as the boundary condition.

Divergence-free condition of total velocity gives:

V

∇ · u dv = 0, (C.28)

or∫

V

∇ · ul dv +

V

∇ · ug dv = 0, (C.29)

where V is the total volume of the simulation domain. The local velocity, by definition,

is also divergence free:∫

V

∇ · ul dv = 0. (C.30)

Page 181: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

157

Eqs. C.29 and C.30 result in making the global velocity divergence free. Therefore,

V

∇ · ug dv = 0. (C.31)

Using Guass-Divergence theorem on equations C.29, C.30 and C.31 gives:

S

n · u dS = 0, (C.32)∫

S

n · ul dS = 0, (C.33)∫

S

n · ug dS = 0, (C.34)

where n is a unit vector normal to the surface, S. Since the domain under considera-

tion is periodic in x− and z− directions, the surface integral of u and w components

of velocity is zero. Eqs. C.32, C.33 and C.34 reduces to

top

v dS −∫

bottom

v dS = 0, (C.35)

top

vl dS −∫

bottom

vl dS = 0, (C.36)

top

vg dS −∫

bottom

vg dS = 0, (C.37)

(C.38)

where top refers to top wall (y = +By/2) and bottom refers to bottom wall (y =

−By/2). Since the walls are assumed to be impenetrable, there is no net flux across

the walls. This condition give rise to

top

v dS =

bottom

v dS = 0, (C.39)

Page 182: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

158

or,∫

top

vl dS +

top

vg dS =

bottom

vl dS +

bottom

vg dS = 0. (C.40)

The local force density, ρl (x), decays rapidly to zero over a distance proportional

to α−1, where α is the screening parameter. The calculation of local velocity field

begins by identifying interactions located within a sphere of fixed radius 4/α. Any

interactions outside this cut-off distance is ignored. The cut off distance for the local

calculation is always set such that no point has interactions that intercepts both walls

i.e.

4/α < By/2. (C.41)

Therefore, it can be assumed that the local velocity, vl, due to a point force is negligible

at at least on of the walls. For example consider a point near the top wall. Its

contribution to the bottom wall will be negligible. Therefore

bottom

vl dS = 0. (C.42)

Comparing eqs. C.36 and C.42:

top

vl dS = 0. (C.43)

Substituting eqs. C.42 and C.43 in eq. C.40 gives

top

vg dS =

bottom

vg dS = 0. (C.44)

The same analogy can be applied for a point close to the bottom wall to obtain eq.

Page 183: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

159

C.44 and as such eq. C.44 can be obtained for any point inside the domain. Eq. C.44

can be represented in Fourier space as

vg(k = 0)|y=wall = 0. (C.45)

Taking a closer look at equation(C.26)gives

∂v(k = 0)

∂y= 0. (C.46)

This equation, when coupled with the boundary conditions for vg obtained at the

walls (eq. C.45) give rise to

vg(k = 0) = 0. (C.47)

Using the above result, eq. C.26 is then reduced to a system of decoupled equations:

−Gη+(

− ∂2

∂y2

)

ug(k = 0) = 1ηρx(k = 0),

1η∂p(k=0)

∂y= 1

ηρy(k = 0),

(

− ∂2

∂y2

)

wg(k = 0) = 1ηρz(k = 0),

vg = 0,

(C.48)

with

ug(k = 0)|y=wall = −ul(k = 0)|y=wall + u(k = 0)|y=wall,

wg(k = 0)|y=wall = −wl(k = 0)|y=wall + w(k = 0)|y=wall,(C.49)

The above system of decoupled equations can be solved individually to get the cor-

responding values at kx = 0 and kz = 0 mode.

The velocities and pressure in physical space can be evaluated by taking the inverse

Page 184: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

160

Fourier transforms as:

ug,i(x) =∑

kx

kz

ug ,i(y)eikx xe ikz z , (C.50)

p(x) =∑

kx

kz

p(y)e ikx xe ikz z . (C.51)

C.3.3 Finite Difference:

Finite difference scheme can be used to solve for the wall-normal, y−, direction.

Staggered grid can be used to solve for the above problem. On a staggered grid

the scalar variables (pressure, force density) are stored in the cell centers of the

control volumes, whereas the velocity variables are located at the cell faces. This is

different from a collocated grid arrangement, where all variables are stored in the same

positions. For every kx and kz mode, there are My velocity variables and (My − 1)

pressure variables. HereMy refers to the mesh in y− direction. Using finite difference,

eqs. C.14 and C.15 are reduced to

(

−1∆2

y

)

ug(yi−1) +(

k2 + 2∆2

y

)

ug(yi) +(

−1∆2

y

)

ug(yi+1)

+(

ikx2η

)

p(yi−1) +(

ikx2η

)

p(yi) =1ηρg,x(yi) · · · ∀ i = 2 . . .My − 1,

(

−1∆2

y

)

vg(yi−1) +(

k2 + 2∆2

y

)

vg(yi) +(

−1η∆2

y

)

vg(yi+1)

+(

−1η∆y

)

p(yi−1) +(

1∆y

)

p(yi) =1ηρg,y(yi) · · · ∀ i = 2 . . .My − 1,

(

−1∆2

y

)

wg(yi−1) +(

k2 + 2∆2

y

)

wg(yi) +(

−1∆2

y

)

wg(yi+1)

+(

ikz2η

)

p(yi−1) +(

ikz2η

)

p(yi) =1ηρg,z(yi) · · · ∀ i = 2 . . .My − 1,

(

ikx2

)

ug(yi) +(

ikx2

)

ug(yi+1) +(

−1∆y

)

vg(yi) +(

1∆y

)

vg(yi+1)

+(

ikz2

)

wg(yi) +(

ikz2

)

wg(yi+1) = 0 · · · ∀ i = 1 . . .My − 1,

(C.52)

Page 185: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

161

with

ug(y1) = −ul|y=bottomwall + u|y=bottomwall, ug(yMy) = −ul|y=top wall + u|y=top wall,

vg(y1) = −vl|y=bottomwall + v|y=bottomwall, vg(yMy) = −vl|y=top wall + v|y=top wall,

wg(y1) = −wl|y=bottomwall + w|y=bottomwall, wg(yMy) = −wl|y=top wall + w|y=top wall,

(C.53)

as the boundary conditions. Here ∆y is the mesh resolution in y−direction. The

above equations can then be solved as a system of linear equations:

A ·U = Q, (C.54)

Page 186: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

162

where A is a coefficient matrix of size (3My − 1)× (3My − 1), U,and Q are vectors

of size (3My − 1). A, U and Q are defined as

A(i, i− 1) = A(i, i+ 1) = −1∆2

y

A(i, i) = k2 + 2∆2

y

A(i, 3My + i− 1) = A(i, 3My + i) = ikx2η

Q(i) = ρxη

A(My + i,My + i− 1) = A(My + i,My + i+ 1) = −1∆2

y

A(My + i,My + i) = k2 + 2∆2

y

−A(My + i, 3My + i− 1) = A(2My + i, 3My + i) = 1∆yη

Q(My + i) = ρyη

A(2My + i, 2My + i− 1) = A(2My + i, 2My + i+ 1) = −1∆2

y

A(2My + i, 2My + i) = k2 + 2∆2

y

A(2My + i, 3My + i− 1) = A(2My + i, 3My + i) = ikz2η

Q(2My + i) = ρzη

∀ i = 2 . . .My − 1

A(3My + i, i) = A(3My + i, i+ 1) = ikx2

−A(3My + i,My + i) = A(3My + i,My + i+ 1) = 1∆y

A(3My + i, 2My + i) = A(3My + i, 2My + i+ 1) = ikz2

∀ i = 1 . . .My − 1

Page 187: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

163

A(1, 1) = A(My,My) = 1

Q(1) = ug|y=bottomwall, Q(My) = ug|y=top wall

A(My + 1,My + 1) = A(2My, 2My) = 1

Q(My + 1) = vg|y=bottomwall, Q(2My) = vg|y=top wall

A(2My + i, 2My + 1) = A(3My, 3My) = 1

Q(2My + 1) = wg|y=bottomwall, Q(3My) = wg|y=top wall

Boundary Conditions

U =

ug(y1)

...

ug(yMy)

vg(y1)

...

vg(yMy)

wg(y1)

...

wg(yMy)

p(y1)

...

p(yMy−1)

where yi are the mesh points in y− direction. The above system of linear eqs. C.54

can be solved using any linear algebra package for all kx and kz modes except for

k = 0 (kx = 0 and kz = 0) mode.

Page 188: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

164

Using finite difference for k = 0 modes eqs. C.48 and C.49 can be represented as

(

−1∆2

y

)

ug(yi−1, k = 0) +(

2∆2

y

)

ug(yi, k = 0) +(

−1∆2

y

)

ug(yi+1, k = 0)

= 1ηρg,x(yi, k = 0) + G

η· · · ∀ i = 2 . . .My − 1,

(

−1η∆y

)

p(yi−1, k = 0) +(

1∆y

)

p(yi, k = 0)

= 1ηρg,y(yi, k = 0) · · · ∀ i = 2 . . .My − 1,

(

−1∆2

y

)

wg(yi−1, k = 0) +(

2∆2

y

)

wg(yi, k = 0) +(

−1∆2

y

)

wg(yi+1, k = 0)

= 1ηρg,z(yi, k = 0) · · · ∀ i = 2 . . .My − 1,

(C.55)

with

ug(y1, k = 0) = −ul(k = 0)|y=bottomwall + u|y=bottomwall,

ug(yMy,k=0) = −ul(k = 0)|y=top wall + u(k = 0)|y=top wall,

wg(y1, k = 0) = −wl(k = 0)|y=bottomwall + w(k = 0)|y=bottomwall,

wg(yMy,k=0) = −wl(k = 0)|y=topwall + w(k = 0)|y=topwall,

(C.56)

as the boundary conditions. The above system of equations can then be solved for

the velocity fields individually as systems of linear equations:

A0 ·Ux = Qx,

A0 ·Uz = Qz,(C.57)

where A0 is a coefficient matrix of size (My)× (My), Ux, Uz,Qx and Qz are vectors

Page 189: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

165

of size (My). They are defined as follows:

A0(i, i− 1) = A0(i, i+ 1) = −1∆2

y

A0(i, i) =2∆2

y

Qx(i) =ρxη+ G

η

Qz(i) =ρzη

∀ i = 2 . . .My − 1

A0(1, 1) = A0(My,My) = 1

Qx(1) = ug(k = 0)|y=bottomwall, Qx(My) = ug(k = 0)|y=top wall

Qz(1) = wg(k = 0)|y=bottomwall, Qz(My) = wg(k = 0)|y=topwall

Boundary Conditions

Ux =

ug(y1, k = 0)

...

ug(yMy , k = 0)

, Uz =

wg(y1, k = 0)

...

wg(yMy , k = 0)

.

C.4 Interpolation: getting ug at the particle posi-

tions

Many interpolation techniques can be used to get the value of global velocity at any

desired lactation. The interpolation used in this method is quadratic interpolation

using 3 points.

Page 190: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

166

C.4.1 Lagrange interpolation of degree 2 :

Given a set of 3 data points

(x1, f(x1)), (x2, f(x2)), (x3, f(x3)) (C.58)

where no two xi are the same, the Lagrange form of interpolating polynomial of degree

2 is a linear combination

L2(x) = l1(x)f(x1) + l2(x)f(x2) + l3(x)f(x3) (C.59)

of Lagrange polynomial, li, of degree 2

li(x) =

3∏

j=1,j 6=i

(x− xj)

(xi − xj). (C.60)

li is associated with the interpolating point xi in the sense:

li(xi) = 1 ∀ i 6= j,

= 0 ∀ i = j. (C.61)

The Lagrange form of the interpolation polynomial shows the linear character of

polynomial interpolation and the uniqueness of the interpolation polynomial.

C.4.2 Interpolation of global velocity:

Let x = {x, y, z} be the point where the value of global velocity, ug(x, y, z) =

{ug(x, y, z), vg(x, y, z), wg(x, y, z)}, needs to be interpolated. Let xi,j,k = {xi, yj, zk}

Page 191: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

167

be the nearest mesh point and ug(xi, yj, zk) = {ug(xi, yj, zk), vg(xi, yj, zk), wg(xi, yj, zk)}

be the corresponding global velocity at that mesh point. The value of ug(x, y, z) are

obtained by doing one-dimensional interpolation sequentially in all the three direc-

tions using eq. C.59. As the interpolation in each direction requires 3 points, a total

of 3× 3× 3 = 27 mesh points is required to interpolate at one point in three dimen-

sion. The order of interpolation in any direction does not change the result. The

interpolation of ug(xi, yj, zk) in y− direction gives intermediate velocities as :

ug(xi, y, zk) = ly,1(y)ug(xi, yj−1, zk) + ly,2(y)ug(xi, yj, zk) + ly,3(y)ug(xi, yj+1, zk),

(C.62)

where ly,i is a lagrange polynomial for y− direction interpolation

ly,1 =y−yj

yj−1−yj

y−yj+1

yj−1−yj+1,

ly,2 =y−yj−1

yj−yj−1

y−yj+1

yj−yj+1,

ly,3 =y−yj−1

yj+1−yj−1

y−yjyj+1−yj

,

(C.63)

and ug(xi, y, zk) is the interpolation of velocity in y−direction. The velocity, ug(xi, y, zk),

is then interpolated in z−direction to obtain

ug(xi, y, z) = lz,1(z)ug(xi, y, zk−1) + lz,2(z)ug(xi, y, zk) + lz,3(z)ug(xi, y, zk+1), (C.64)

where lz,i is a lagrange polynomial for z− direction interpolation

lz,1 =z−zk

zk−1−zk

z−zk+1

zk−1−zk+1,

lz,2 =z−zk−1

zk−zk−1

z−zk+1

zk−zk+1,

lz,3 =z−zk−1

zk+1−zk−1

z−zkzk+1−zk

,

(C.65)

Page 192: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

168

and ug(xi, y, z) is the interpolation of velocity in z−direction. The velocity, ug(xi, y, z),

is then interpolated in x−direction to obtain velocity at particle position, x, as

ug(x, y, z) = lx,1(x)ug(xi−1, y, z) + lx,2(x)ug(xi, y, z) + lx,3(x)ug(xi+1, y, z), (C.66)

where lx,i is a lagrange polynomial for x− direction interpolation

lx,1 =x−xi

xi−1−xi

x−xi+1

xi−1−xi+1,

lx,2 =x−xi−1

xi−xi−1

x−xi+1

xi−xi+1,

lx,3 =x−xi−1

xi+1−xi−1

x−xi

xi+1−xi.

(C.67)

Eqs. C.62, C.64 and C.66 can be written in a compact form as

ug(x, y, z) =3

p=1

3∑

q=1

3∑

r=1

lx,p ly,q lz,r ug(xi+p−2, yj+q−2, zk+r−2). (C.68)

Page 193: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

169

Bibliography

[1] C. A. Macias, M. V. Kameneva, J. J. Tenhunen, J. C. Puyana, and M. P.

Fink, “Survival in a rat model of lethal hemorrhagic shock is prolonged follow-

ing resuscitation with a small volume of a solution containing a drag-reducing

polymer derived from aloe vera,” Shock 22, 151–156 (2004).

[2] M. V. Kameneva, B. M. Repko, E. F. Krasik, B. C. Perricelli, and H. S.

Borovetz, “Polyethylene glycol additives reduce hemolysis in red blood cell sus-

pensions exposed to mechanical stress,” ASAIO Journal 49, 537–542 (2003).

[3] E. Lac, D. Barthes-Biesel, N. A. Pelekasis, and J. Tsamopoulos, “Spherical

capsules in three-dimensional unbounded Stokes flows: effect of the membrane

constitutive law and onset of buckling,” Journal Of Fluid Mechanics 516, 303–

334 (2004).

[4] S. K. Doddi and P. Bagchi, “Effect of inertia on the hydrodynamic interaction

between two liquid capsules in simple shear flow,” International Journal Of

Multiphase Flow 34, 375–392 (2008).

[5] S. Ramanujan and C. Pozrikidis, “Deformation of liquid capsules enclosed by

Page 194: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

170

elastic membranes in simple shear flow: large deformations and the effect of

fluid viscosities,” Journal Of Fluid Mechanics 361, 117–143 (1998).

[6] E. Lac, A. Morel, and D. Barthes-Biesel, “Hydrodynamic interaction between

two identical capsules in simple shear flow,” Journal Of Fluid Mechanics 573,

149–169 (2007).

[7] M. V. Kameneva, Z. J. J. Wu, A. Uraysh, B. Repko, K. N. Litwak, T. R.

Billiar, M. P. Fink, R. L. Simmons, B. P. Griffith, and H. S. Borovetz, “Blood

soluble drag-reducing polymers prevent lethality from hemorrhagic shock in

acute animal experiments,” Biorheology 41, 53–64 (2004).

[8] H. Alam and P. Rhee, “New developments in fluid resuscitation,” Surgical Clin-

ics of North America 87, 55–72 (2007).

[9] M. Piagnerelli, K. Boudjeltia, M. Vanhaeverbeek, and J. Vincent, “Red blood

cell rheology in sepsis,” Intensive Care Medicine 29, 1052–1061 (2003).

[10] A. Ehrly, Therapeutic Hemorheology (Springer Verlag, Berlin, 1991).

[11] H. Greene, R. Nokes, and L. Thomas, “Drag reduction in pulsed flow of blood,”

Medical Research Engineering 9, 19 (1970).

[12] R. Mostardi, H. Greene, R. Nokes, L. Thomas, and T. Lue, “The effect of drag

reducing agents on stenotic flow disturbances in dogs.” Biorheology 13, 137–141

(1976).

[13] R. Mostardi, L. Thomas, H. Greene, F. VanEssen, and R. Nokes, “Suppression

Page 195: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

171

of atherosclerosis in rabbits using drag reducing polymers.” Biorheology 15,

1–14 (1978).

[14] H. Greene, R. Mostardi, and R. Nokes, “Effects of drag reducing polymers on

initiation of atherosclerosis,” Polym. Eng. Sci 20, 499–504 (1980).

[15] G. Driessen, H. Scheidt-Bleichert, A. Sobota, W. Inhoffen, H. Heidtmann,

C. Haest, D. Kamp, and H. Schmid-Schonbein, “Capillary resistance to flow

of hardened (diamide treated) red blood cells (RBC),” Pflugers Archiv Euro-

pean Journal of Physiology 392, 261–267 (1982).

[16] P. Coleman, B. Ottenbreit, and P. Polimeni, “Effects of a drag-reducing poly-

electrolyte of microscopic linear dimension (Separan AP-273) on rat hemody-

namics,” Circulation research 61, 787–796 (1987).

[17] F. Faruqui, M. Otten, and P. Polimeni, “Protection against atherogenesis with

the polymer drag-reducing agent Separan AP-30,” Circulation 75, 627–635

(1987).

[18] D. Smyth, D. Blandford, D. Jones, and P. Polimeni, “Separan AP-273 and renal

function: a novel natriuretic substance,” Canadian Journal of Physiology and

Pharmacology 65, 2001–2003 (1987).

[19] P. Polimeni, D. Bose, R. Bose, M. Otten, and B. Ottenbreit, “Drag-reducing

polymers: A novel category of drugs potentially useful in cardiovascular dis-

eases,” Journal of Applied Cardiology 3, 57–66 (1988).

[20] K. J. Hutchison, J. D. Campbell, and E. Karpinski, “Decreased poststenotic flow

Page 196: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

172

disturbance during drag reduction by polyacrylamide infusion without increased

aortic blood-flow,” Microvascular Research 38, 102–109 (1989).

[21] P. I. Polimeni and B. T. Ottenbreit, “Hemodynamic-effects of a poly(ethylene

oxide) drag-reducing polymer, polyox wsr n-60k, in the open-chest rat,” Journal

Of Cardiovascular Pharmacology 14, 374–380 (1989).

[22] D. D. Smyth and P. I. Polimeni, “Drag-reducing polymers - a novel class of

diuretic and natriuretic compounds,” Cardiovascular Drugs And Therapy 4,

297–300 (1990).

[23] J. L. Unthank, S. G. Lalka, J. C. Nixon, and A. P. Sawchuk, “Improvement

of flow through arterial stenoses by drag reducing agents,” Journal Of Surgical

Research 53, 625–630 (1992).

[24] A. P. Sawchuk, J. L. Unthank, and M. C. Dalsing, “Drag reducing polymers

may decrease atherosclerosis by increasing shear in areas normally exposed to

low shear stress,” Journal Of Vascular Surgery 30, 761–764 (1999).

[25] S. Kaul and A. R. Jayaweera, “Determinants of microvascular flow,” European

Heart Journal 27, 2272–2274 (2006).

[26] N. Antonova and Z. Lazarov, “Hemorheological and hemodynamic effects of

high molecular weight polyethylene oxide solutions,” Clinical Hemorheology

And Microcirculation 30, 381–390 (2004).

[27] J. J. Pacella, M. M. Csikari, M. V. Kameneva, D. Fischer, B. S. Harbison,

and F. S. Villanueva, “Drag-reducing polymers enhance myocardial perfusion

during coronary stenosis,” Circulation 106, 1887 (2002).

Page 197: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

173

[28] J. J. Pacella, M. V. Kameneva, M. Csikari, E. X. Lu, and F. S. Villanueva,

“A novel hydrodynamic approach to the treatment of coronary artery disease,”

European Heart Journal 27, 2362–2369 (2006).

[29] B. Toms, “Proceedings of the International Congress on Rheology,” Vol. II,

North Holland Publishing Company, Amsterdam (1949).

[30] P. Virk, “Drag reduction fundamentals,” AIChE Journal 21, 625–656 (1975).

[31] Y. Fung and B. Zweifach, “Microcirculation: mechanics of blood flow in capil-

laries,” Annual Review of Fluid Mechanics 3, 189–210 (1971).

[32] Y. Fung, Biodynamics: Circulation (Springer-Verlag New York, 1983).

[33] W. Reinke, P. Gaehtgens, and P. Johnson, “Blood viscosity in small tubes: effect

of shear rate, aggregation, and sedimentation,” American Journal of Physiology-

Heart and Circulatory Physiology 253, H540 (1987).

[34] J. B. Freund, “Leukocyte margination in a model microvessel,” Physics Of

Fluids 19, 023 301 (2007).

[35] F. Allison Jr, M. Smith, and W. Wood Jr, “Studies on the pathogenesis of

acute inflammation: II. The action of cortisone on the inflammatory response

to thermal injury,” The Journal of Experimental Medicine 102, 669 (1955).

[36] J. Williamson and J. Grisham, “Electron microscopy of leukocytic margination

and emigration in acute inflammation in dog pancreas,” The American Journal

of Pathology 39, 239 (1961).

Page 198: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

174

[37] S. Doddi and P. Bagchi, “Lateral migration of a capsule in a plane Poiseuille flow

in a channel,” International Journal of Multiphase Flow 34, 966–986 (2008).

[38] L. Leal, “Particle motions in a viscous fluid,” Annual Review of Fluid Mechanics

12, 435–476 (1980).

[39] G. Danker, P. Vlahovska, and C. Misbah, “Vesicles in Poiseuille flow,” Physical

review letters 102, 148 102 (2009).

[40] J. R. Smart and D. T. Leighton, “Measurement of the drift of a droplet due to

the presence of a plane,” Physics Of Fluids A-Fluid Dynamics 3, 21–28 (1991).

[41] D. Leighton and A. Acrivos, “The shear-induced migration of particles in con-

centrated suspensions,” Journal of Fluid Mechanics 181, 415–439 (1987).

[42] R. Phillips, R. Armstrong, R. Brown, A. Graham, and J. Abbott, “A constitu-

tive model for concentrated suspensions that accounts for shear-induced particle

migration,” Physics of Fluids A: Fluid Dynamics 4, 30 (1992).

[43] P. Nott and J. Brady, “Pressure-driven flow of suspensions: simulation and

theory,” Journal of Fluid Mechanics 275, 157–199 (1994).

[44] H. Ma and M. Graham, “Theory of shear-induced migration in dilute polymer

solutions near solid boundaries,” Physics of Fluids 17, 083 103 (2005).

[45] M. King and D. Leighton Jr, “Measurement of shear-induced dispersion in a

dilute emulsion,” Physics of Fluids 13, 397 (2001).

[46] S. Hudson, “Wall migration and shear-induced diffusion of fluid droplets in

emulsions,” Physics of fluids 15, 1106 (2003).

Page 199: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

175

[47] R. Zhao, J. Marhefka, J. Antaki, and M. Kameneva, “Drag-reducing poly-

mers diminish near-wall concentration of platelets in microchannel blood flow,”

Biorheology 47, 193–203 (2010).

[48] R. Larson, The Structure and Rheology of Complex Fluids (Oxford University

Press, New York, 1999).

[49] C. Pozrikidis, “Finite deformation of liquid capsules enclosed by elastic mem-

branes in simple shear-flow,” Journal Of Fluid Mechanics 297, 123–152 (1995).

[50] C. Pozrikidis, “Effect of membrane bending stiffness on the deformation of

capsules in simple shear flow,” Journal of Fluid Mechanics 440, 269–291 (2001).

[51] C. Pozrikidis, “Numerical simulation of the flow-induced deformation of red

blood cells,” Annals Of Biomedical Engineering 31, 1194–1205 (2003).

[52] C. Pozrikidis, “Numerical Simulation of Blood Flow Through Microvascular

Capillary Networks,” Bulletin Of Mathematical Biology 71, 1520–1541 (2009).

[53] D. Barthes-Biesel and J. Rallison, “Time-dependent deformation of a capsule

freely suspended in a linear shear flow,” Journal of Fluid Mechanics 113, 251–

267 (1981).

[54] D. Barthes-Biesel, “Theoretical modeling of the motion and deformation of cap-

sules in shear flows,” Biomaterials Artificial Cells And Immobilization Biotech-

nology 21, 359–373 (1993).

[55] D. Barthes-Biesel, A. Diaz, and E. Dhenin, “Effect of constitutive laws for

Page 200: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

176

two-dimensional membranes on flow-induced capsule deformation,” Journal Of

Fluid Mechanics 460, 211–222 (2002).

[56] E. Lac and D. Barthes-Biesel, “Deformation of a capsule in simple shear flow:

Effect of membrane prestress,” Physics Of Fluids 17, 072 105 (2005).

[57] E. Lac and D. Barthes-Biesel, “Pairwise interaction of capsules in simple shear

flow: Three-dimensional effects,” Physics Of Fluids 20, 040 801 (2008).

[58] W. R. Dodson and P. Dimitrakopoulos, “Spindles, cusps, and bifurcation for

capsules in Stokes flow.” Phys Rev Lett 101, 208 102 (2008).

[59] J. Clausen and C. Aidun, “Capsule dynamics and rheology in shear flow: Par-

ticle pressure and normal stress,” Physics of Fluids 22, 123 302 (2010).

[60] R. Skalak, A. Tozeren, R. P. Zarda, and S. Chien, “Strain energy function of

red blood-cell membranes,” Biophysical Journal 13, 245–280 (1973).

[61] X. Li and K. Sarkar, “Front tracking simulation of deformation and buckling

instability of a liquid capsule enclosed by an elastic membrane,” Journal of

Computational Physics 227, 4998–5018 (2008).

[62] J. Walter, A. Salsac, D. Barthes-Biesel, and P. Le Tallec, “Coupling of finite

element and boundary integral methods for a capsule in a Stokes flow,” Inter-

national Journal for Numerical Methods in Engineering 83, 829–850 (2010).

[63] P. Bagchi, P. C. Johnson, and A. S. Popel, “Computational fluid dynamic sim-

ulation of aggregation of deformable cells in a shear flow,” Journal Of Biome-

chanical Engineering-Transactions Of The Asme 127, 1070–1080 (2005).

Page 201: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

177

[64] J. M. Charrier, S. Shrivastava, and R. Wu, “Free and constrained inflation of

elastic membranes in relation to thermoforming - non-axisymmetric problems,”

Journal Of Strain Analysis For Engineering Design 24, 55–74 (1989).

[65] C. S. Peskin, “Numerical-analysis of blood-flow in heart,” Journal Of Compu-

tational Physics 25, 220–252 (1977).

[66] S. O. Unverdi and G. Tryggvason, “A front-tracking method for viscous, in-

compressible, multi-fluid flows,” Journal Of Computational Physics 100, 25–37

(1992).

[67] S. Jadhav, Y. C. Kit, K. Konstantopoulos, and C. D. Eggleton, “Shear modu-

lation of intercellular contact area between two deformable cells colliding under

flow,” Journal Of Biomechanics 40, 2891–2897 (2007).

[68] A. Kumar and M. Graham, “Accelerated boundary integral method for multi-

phase flow in non-periodic geometries,” submitted (2011).

[69] H. Zhao, A. Isfahani, L. Olson, and J. Freund, “A spectral boundary integral

method for flowing blood cells,” Journal of Computational Physics 229, 3726–

3744 (2010).

[70] S. K. Doddi and P. Bagchi, “Three-dimensional computational modeling of

multiple deformable cells flowing in microvessels,” Phys Rev E 79, 046 318

(2009).

[71] M. Dupin, I. Halliday, C. Care, L. Alboul, and L. Munn, “Modeling the flow of

dense suspensions of deformable particles in three dimensions,” Physical Review

E 75, 66 707 (2007).

Page 202: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

178

[72] E. Ding and C. Aidun, “Cluster size distribution and scaling for spherical par-

ticles and red blood cells in pressure-driven flows at small Reynolds number,”

Physical review letters 96, 204 502 (2006).

[73] R. Macmeccan, J. Clausen, G. Neitzel, and C. Aidun, “Simulating deformable

particle suspensions using a coupled lattice-Boltzmann and finite-element

method,” Journal of Fluid Mechanics 618, 13–39 (2009).

[74] P. Pranay, S. Anekal, J. Hernandez-Ortiz, and M. Graham, “Pair collisions of

fluid-filled elastic capsules in shear flow: Effects of membrane properties and

polymer additives,” Physics of Fluids 22, 123 103 (2010).

[75] C. Peskin, “The immersed boundary method,” Acta Numerica 11, 479–517

(2002).

[76] M. Doi and S. Edwards, The theory of polymer dynamics (Oxford University

Press, New York, 1988).

[77] R. Bird, C. Curtiss, R. Armstrong, and O. Hassager, Dynamics of Polymeric

Liquids, Vol. 2 (Wiley-Interscience, New York, 1987).

[78] J. P. Hernandez-Ortiz, J. J. de Pablo, and M. D. Graham, “Fast computa-

tion of many-particle hydrodynamic and electrostatic interactions in a confined

geometry,” Physical Review Letters 98 (2007).

[79] H. Ottinger, Stochastic Processes in Polymeric Fluids (Springer Verlag, Berlin,

1996).

Page 203: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

179

[80] T. Bringley and C. Peskin, “Validation of a simple method for representing

spheres and slender bodies in an immersed boundary method for Stokes flow

on an unbounded domain,” Journal of Computational Physics 227, 5397–5425

(2008).

[81] J. P. Hernandez-Ortiz, H. B. Ma, J. J. de Pablo, and M. D. Graham, “Concen-

tration distributions during flow of confined flowing polymer solutions at finite

concentration: slit and grooved channel,” Korea-Australia Rheology Journal

20, 143–152 (2008).

[82] A. Sierou and J. Brady, “Accelerated Stokesian dynamics simulations,” Journal

of Fluid Mechanics 448, 115–146 (2001).

[83] H. B. Ma and M. D. Graham, “Theory of shear-induced migration in dilute

polymer solutions near solid boundaries,” Physics Of Fluids 17, 083 103 (2005).

[84] J. P. Hernandez-Ortiz, H. B. Ma, J. J. Pablo, and M. D. Graham, “Cross-

stream-line migration in confined flowing polymer solutions: Theory and simu-

lation,” Physics Of Fluids 18, 123 101 (2006).

[85] C. Gardiner, Handbook of stochastic methods (Springer Berlin, 1985).

[86] D. Koch, “On hydrodynamic diffusion and drift in sheared suspensions,” Physics

of Fluids A: Fluid Dynamics 1, 1742 (1989).

[87] P. Chan and L. Leal, “The motion of a deformable drop in a second-order fluid,”

Journal of Fluid Mechanics 92, 131–170 (1979).

Page 204: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

180

[88] G. D’Avino, P. Maffettone, F. Greco, and M. Hulsen, “Viscoelasticity-induced

migration of a rigid sphere in confined shear flow,” Journal of Non-Newtonian

Fluid Mechanics 165, 466–474 (2010).

[89] X. Li and C. Pozrikidis, “Wall-bounded shear flow and channel flow of suspen-

sions of liquid drops,” International Journal of Multiphase Flow 26, 1247–1279

(2000).

[90] C. Alonso, A. Pries, O. Kiesslich, D. Lerche, and P. Gaehtgens, “Transient rhe-

ological behavior of blood in low-shear tube flow: velocity profiles and effective

viscosity,” American Journal of Physiology- Heart and Circulatory Physiology

268, H25 (1995).

[91] D. Long, Microviscometric analysis of microvascular hemodynamics in vivo

(University of Illinois at Urbana-Champaign, 2004).

[92] D. Long, M. Smith, A. Pries, K. Ley, and E. Damiano, “Microviscometry re-

veals reduced blood viscosity and altered shear rate and shear stress profiles

in microvessels after hemodilution,” Proceedings of the National Academy of

Sciences of the United States of America 101, 10 060 (2004).

[93] H. Lipowsky, S. Usami, and S. Chien, “In vivo measurements of apparant vis-

cosity and microvessel hematocrit in the mesentry of the cat,” Microvascular

Research 19, 297–319 (1980).

[94] G. Breyiannis and C. Pozrikidis, “Simple shear flow of suspensions of elastic

capsules,” Theoretical and Computational Fluid Dynamics 13, 327–347 (2000).

[95] Y. Fung, Biomechanics: mechanical properties of living tissues (Springer, 1993).

Page 205: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

181

[96] E. Evans and Y. Fung, “Improved measurements of the erythrocyte geometry*

1,” Microvascular research 4, 335–347 (1972).

[97] D. Boal, Mechanics of the Cell (Cambridge Univ Pr, 2002).

[98] P. Canham, “The minimum energy of bending as a possible explanation of the

biconcave shape of the human red blood cell*,” Journal of Theoretical Biology

26, 61–81 (1970).

[99] W. Helfrich, “Elastic properties of lipid bilayers: theory and possible experi-

ments,” Z. Naturforsch 28, 693–703 (1973).

[100] G. Boedec, M. Leonetti, and M. Jaeger, “3D vesicle dynamics simulations with

a linearly triangulated surface,” Journal of Computational Physics (2010).

[101] S. Mitragotri and J. Lahann, “Physical approaches to biomaterial design,” Na-

ture materials 8, 15–23 (2009).

[102] P. Decuzzi, S. Lee, M. Decuzzi, and M. Ferrari, “Adhesion of microfabricated

particles on vascular endothelium: a parametric analysis,” Annals of biomedical

engineering 32, 793–802 (2004).

[103] D. Discher and F. Ahmed, “Polymersomes,” Annu. Rev. Biomed. Eng. 8, 323

(2006).

[104] J. Champion, Y. Katare, and S. Mitragotri, “Making polymeric micro-and

nanoparticles of complex shapes,” Proceedings of the National Academy of

Sciences 104, 11 901 (2007).

Page 206: DYNAMICS OF SUSPENSIONS OF ELASTIC CAPSULES IN POLYMER SOLUTION · 2020-02-06 · tions of high molecular weight long-chain polymer molecules known as drag-reducing additives (DRAs)

182

[105] J. Champion, Y. Katare, and S. Mitragotri, “Particle shape: a new design

parameter for micro-and nanoscale drug delivery carriers,” Journal of Controlled

Release 121, 3–9 (2007).

[106] S. Cai, K. Vijayan, D. Cheng, E. Lima, and D. Discher, “Micelles of differ-

ent morphologiesadvantages of worm-like filomicelles of PEO-PCL in paclitaxel

delivery,” Pharmaceutical Research 24, 2099–2109 (2007).

[107] J. Park, G. von Maltzahn, L. Zhang, M. Schwartz, E. Ruoslahti, S. Bhatia, and

M. Sailor, “Magnetic iron oxide nanoworms for tumor targeting and imaging,”

Advanced Materials 20, 1630–1635 (2008).

[108] J. Rolland, B. Maynor, L. Euliss, A. Exner, G. Denison, and J. DeSimone,

“Direct fabrication and harvesting of monodisperse, shape-specific nanobioma-

terials,” J. Am. Chem. Soc 127, 10 096–10 100 (2005).