Vous êtes sur la page 1sur 6

Journal of Non-Newtonian Fluid Mechanics 200 (2013) 3–8

Contents lists available at SciVerse ScienceDirect

Journal of Non-Newtonian Fluid Mechanics


journal homepage: http://www.elsevier.com/locate/jnnfm

The analytic solution of Stokes for time-dependent creeping flow around a sphere:
Application to linear viscoelasticity as an ingredient for the generalized
Stokes–Einstein relation and microrheology analysis
Jay D. Schieber a,b,⇑, Andrés Córdoba a, Tsutomu Indei a
a
Department of Chemical and Biological Engineering, Center for Molecular Study of Condensed Soft Matter, Illinois Institute of Technology, 3440 S. Dearborn St., Chicago, IL, USA
b
Department of Physics, Center for Molecular Study of Condensed Soft Matter, Illinois Institute of Technology, 3440 S. Dearborn St., Chicago, IL, USA

a r t i c l e i n f o a b s t r a c t

Article history: Analytic expressions for the transient stream function, transient flow field, and transient pressure field
Available online 20 September 2012 for creeping flow around a sphere are derived. An analytic expression for the total force on the sphere
is also found. The approach is essentially that of Stokes from 1856. Aside from the (essentially trivial)
Keywords: generalization to linear viscoelastic fluids, there is nothing novel in the derivation. Our purpose is to
Creeping flow (1) point out that Stokes, not Basset or Boussinesq derived it first, (2) show how simple the derivation
Linear viscoelasticity is, which may be compared to the more famous solution of Landau and Lifshitz, (3) show an application
Sphere
of the correspondence between creeping flow and linear viscoelastic flow solutions, and (4) provide suf-
Correspondence principle
Generalized Stokes-Einstein relation
ficiently detailed notes so that the derivation might be given in a graduate fluid dynamics or transport
phenomena lecture.
Ó 2012 Elsevier B.V. All rights reserved.

1. Introduction vectors, and appears to rely on the reader’s ability to visualize


things like the curl of the curl of a vector. Perhaps not surprisingly,
Most fluid mechanics or transport phenomena textbooks dis- folklore says that Landau was unable to teach this ability to his stu-
cuss the analytic solution for creeping (or Stokes) flow around a dents. Perhaps such a derivation is straightforward to the student
sphere at steady state in an incompressible fluid. Many of these who is skilled in classical electrodynamics. Berker [4] gives a solu-
books also discuss the stream function in several orthogonal coor- tion in the Encyclopedia of Physics, but does not reveal all of the
dinates, including spherical. However, we are aware of only one logic to obtain it. It appears similar to Basset’s solution, but with
textbook that discusses the analytic solution for Stokes flow significant differences.
around a sphere that is moving in a time-dependent fashion—that Credit is often given to Basset for his solution to the force on the
of Leal [19]. That solution is rather complete. However, a rough sphere in 1888 [2]. However, the force on a sphere of arbitrary dis-
form of the mathematical solution is given as a starting point, placement was found by Boussinesq three years earlier [6,7]. In
the stream function is not used, and displacements of the sphere fact, Stokes solved the problem of a sphere moving in an oscillatory
from the center of the coordinate system are neglected. Also, rele- motion along a straight line at low Reynolds number. Therefore, he
vant to our point here, no connection to Stokes’ solution is made. really just found the solution of the general motion problem in the
Even the comprehensive Stokes-flow text by Happel and Brenner frequency domain. To do so, Stokes used the stream function,
[13], which otherwise contains a wealth of information about which he had invented in 1842 [25]. Before attacking the problem
stream functions, does not cover the problem. (See a summary of of arbitrary bead motion, Landau and Lifshitz also first solved the
a small sample of such books in Table 1.) problem for oscillatory motion, but, apparently unlike Stokes, they
In some ways this situation might seem surprising, since the recognized that this is the solution in Fourier space. Since Stokes
solution was first found by Stokes in 1856 [26] using the stream went on to find the force on the oscillating sphere, it seems unfair
function. Also, the famously creative (and famously complex) book to call the steady-state force on the sphere the ‘‘Stokes force’’ and
by Landau and Lifshitz (LL) on fluid mechanics gives an approach the transient part the ‘‘Basset’’ (or ‘‘Boussinesq’’) force. We resur-
for the analytic solution, which is elegant in many ways. However, rect here the simple solution of Stokes by putting it into slightly
the approach of LL relies heavily on the properties of the curl of more contemporary language.
Like Stokes, we use the stream function, but also take advantage
⇑ Corresponding author. of Fourier transforms (of which Stokes might not have been
E-mail address: schieber@iit.edu (J.D. Schieber). aware). Hence, we can find the flow field around a transiently

0377-0257/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jnnfm.2012.08.002
4 J.D. Schieber et al. / Journal of Non-Newtonian Fluid Mechanics 200 (2013) 3–8

Table 1
Some textbooks in fluid mechanics or transport phenomena, and which subjects relevant to this article are covered. We use the abbreviations: Ex. = Example, § = section,
w = stream function, WWW = Welty, Wicks and Wilson.
Textbook Stream function St.st flow sphere Transient flow
Sphere Notes
Landau and Lifshitz [18] §10, p. 22, Cartesian §20 §24, Problems never use w
a
Berker [4] §4 §66 §69
Gary Leal [19] §7.C §7.F Problem 7-7 See introduction
b
Happel and Brenner [13] Ch. 4 §4-17 Not covered
Bird et al. [5] §4.2 Ex. 4.2-1 Not covered
Deen [10] §6.8 Ex.8.4-2 Not covered w, General coord’s.
Kundu [17] §6.8 §8.12 Not covered
Acheson [1] §7.2 §7.2 Not covered
Denn [11] Ch.14 §12.3 Not covered
Milne-Thompson [22] §§4.30-41, Ch. 6 §22.20 Not covered
Bennet and Myers [3] p. 120 Not covered Not covered w Cartesian only
WWW [27] §10.2 Not covered Not covered w Cartesion only
Whitaker [28] §3.6 Not covered Not covered w Cartesion only
Geankoplis [12] §3.9B Not covered Not covered
Janna [16] Not covered Not covered Not covered
a
An integral solution is given, but the logic behind the derivation is missing.
b
The textbook by Happel and Brenner is especially noteworthy for the general analytic properties given for stream functions.

displaced sphere (at low Reynolds number), the pressure field, and an equivalent single PDE involving only the stream function. By
the total force on the sphere. As a result, we find not only the taking the Fourier transform of the equation, we turn all differen-
Stokes force, but also the ‘‘Basset’’ force, and the added inertia to tial operations in time into purely algebraic expressions. Thus, the
the sphere from the surrounding fluid. The original solution by Bas- single PDE with a single dependent variable now has only two
set also uses the stream function, but since it does not exploit the independent variables (r and h).
Fourier transform, it is more complicated than the solution pro- A symmetry argument of Landau and Lifshitz requires that the
posed here, in our view. In other words, this article is intended solution be separable, with a known angular dependence. There-
to help clarify the historical record,1 resurrect an elegant and useful fore the h dependence of the stream function is known, and we
solution, and aid in pedagogy, but not to provide any fundamentally are left with a single, linear, fourth-order ordinary differential
new results. The solution can be given in a single lecture of a grad- equation in the r-dependent part of the stream function which de-
uate fluid mechanics or transport phenomena course. Table 1 might pends on position r (and the frequency x, whose dependence is
be helpful to see if the textbook in your course is suitable for this strictly algebraic). This equation is solved in two steps, each step
example. Moreover, because of the established correspondence be- a second-order, linear, equidimensional ODE. This last step and
tween creeping Newtonian flow and linear viscoelasticity, the solu- the use of Fourier transforms was not used by Stokes.
tion also applies to the problem of a sphere moving in a
viscoelastic fluid in the limit of small Reynolds number and linear 2. Problem formulation
viscoelasticity [8,15,29,30,20]. Thus the example is also suitable for
a graduate course in rheology. We consider a spherical bead of radius R that moves only in the
The approach relies on knowledge of stream functions, Fourier z direction. It moves with time-dependent velocity u(t), keeping
transforms, and the solution of linear ordinary differential equa-
the Reynolds number, Re :¼ qguR of the flow much less than unity.
tions (ODEs). We give a short summary of the necessary properties 0

of the first two subjects in the appendix, but assume some minimal Here the fluid has (constant) density q and (constant) Newtonian
working knowledge in the derivation. Solution to the problem viscosity g0. The bead is suspended in a fluid sufficiently far from
accomplishes two things: it illustrates the power of mathematics any walls that it may be treated as an infinite sea. Once we have
in fluid mechanics, and proves a very useful result used in many found the solution, we can specify how far away the container
fields, such as suspensions, aerosols, microrheology, and diffusion walls need to be. The linearized Navier–Stokes equation (LNSE) is
in biological media. written as [5]
Our approach is the following. The problem begins as a partial @
differential equation with four dependent variables (the three
q v ¼ g0 r2 v  $p þ qg; ð1Þ
@t
components of velocity and pressure), and four independent vari- where v(r, t) is the fluid velocity at spatial coordinate r and time
ables (the three position coordinates and time). However, since @
t; $  @r is the vector differential operator, and g is the gravity
the governing equations are linear, any two solutions can be vector. We wish to keep the origin of our coordinate system in
summed to create a new solution. Therefore we need only solve the center of the bead, which is not an inertial frame of reference
the problem of a sphere moving in one direction to generate a solu- for a bead moving with arbitrary acceleration. So instead we
tion for arbitrary 3D motion. If we pick this direction to be along consider an equivalent problem. We hold the bead fixed in space
the z-coordinate, there are then only three dependent variables and move our container walls with arbitrary velocity uðtÞdz , where
(vr, vh and pressure p), and three independent variables (r, h, and dz is the unit vector in the (Cartesian coordinate) z direction. We
time t). Here we use the notation of Bird, Stewart and Lightfoot’s assume ‘‘sticky’’ or no-slip boundary conditions so that the velocity
Transport Phenomena (BSL) [5]. As with any two-dimensional, of the fluid is that of the bead at its surface
incompressible flow, we can use stream functions to go from two
coupled PDEs involving the velocity components and pressure, to vðr ¼ R; tÞ ¼ 0: ð2Þ
Far from the sphere, the velocity should move with the
container
1
Zwanzig and Bixon were correct in their description of the history, but did not
give the relevant citations.
vðr ! 1; tÞ ! uðtÞdz : ð3Þ
J.D. Schieber et al. / Journal of Non-Newtonian Fluid Mechanics 200 (2013) 3–8 5

2
Since we are moving the container with arbitrary time-depen- wðr; h; tÞ ¼ f ðr; tÞ sin h: ð12Þ
dent displacement, we are introducing a uniform pressure gradient
in the same way that gravity does. Except here the ‘‘gravity force’’ The argument originates with Stokes [26], who noted both that
is time dependent. Therefore it is useful to introduce a perturba- this separation fits the boundary conditions on w, and that the
tion velocity v 0 in our non-inertial reference frame operator E2 preserves the angular dependence. Happel and Brenner
[13] have found a very general solution for the steady-state prob-
v 0 :¼ v þ uðtÞdz ; ð4Þ lem where the velocity far from the sphere is arbitrary. Their solu-
and a modified pressure [5, p. 50] (also sometimes called ‘‘dynamic’’ tion is an infinite sum of terms, each separable in r and h. The radial
or ‘‘equivalent’’ pressure [10, p. 230]) functions are all polynomials in r and the h functions involve
2
Legendre polynomials in cos h [10]. The sin h term here is just
du
P :¼ p þ qgh  qz : ð5Þ the first in the expansion of Happel and Brenner. Stokes made a
dt similar observation in a footnote, and writes ‘‘It was somewhat
The second term involves the height h above the origin, and is a in this way that I first obtained the form of the function w.’’
typical method to eliminate the hydrostatic pressure and gravity Landau and Lifshitz make an even more subtle argument [18
from the incompressible Navier–Stokes equation. The third term p.64], based on two observations. First, the resulting velocity field
is new, and is introduced to eliminate the similar pressure field must depend linearly on the bead velocity u ¼ udz , or v 0h must go as
that arises from our accelerating container. One sees that this trick u sin h. Secondly, they look at a second (primed) coordinate system
works when we put Eqs. (4) and (5) into the linearized Navier– where the z0 -axis points in the direction opposite to z, and ask how
Stokes equation (LNSE, or ‘‘time-dependent Stokes equation’’), Eq. their solution should change. For a point fixed in space one sees
(1), to obtain our governing equation for perturbation velocity that r 0 ¼ r; h0 ¼ p  h; dr0 ¼ dr , and dh0 ¼ dh . Note then that
@ 0 sin h0 ¼ sin h. The stream function must be symmetric so that
q v ¼ g0 r2 v 0  $P: ð6Þ wðr; hÞ ¼ wðr; p  hÞ. Or, this means that w can depend on sin h, or
@t
its square, but not on cos h. Looking now at the h component of
Note that Eq. (6) is exactly of the form of the LNSE, but without
Eq. (10), we see that the only form for the stream function that sat-
gravity, and the names of the velocity and pressure have changed
isfies these requirements is Eq. (12).
slightly. The artifactual accelerations on the left cancel the contri-
We can now exploit Stokes’s observation about the operator E2
bution to the modified pressure. Therefore, any mathematical
in Eq. (9), and how it preserves the angular dependence
tricks that we can use to solve the LNSE could also be used here. !
The boundary conditions (BCs), Eqs. (2) and (3) become 2 2 @2 2 2 2
E f ðr; tÞ sin h ¼ sin h  f ¼ sin hE2r f ; ð13Þ
v0 ðr ¼ R; tÞ ¼ dz uðtÞ ¼ uðtÞ½dr cos h  dh sin h @r 2 r 2
ð7Þ
v0 ðr ! 1; tÞ ! 0: where, for convenience, we introduce the radial operator
 2 
Therefore, we seek a solution to Eqs. (6) and (7) to find the per- E2r :¼ @r@ 2  r22 . Putting the separability form, Eq. (12) into our
turbation velocity and modified pressure as functions of r, h and t. stream function evolution equation, Eq. (8) gives us
Once these are known, the original velocity field can be found from
Eq. (4). @f
qE2r ¼ g0 E4r f : ð14Þ
@t
3. Stream function solution Finally, we can turn this into a linear ordinary differential equa-
tion by taking the Fourier transform of both sides of this equation
Since Eq. (6) has exactly the same form as the LNSE, and because to obtain
our problem is two-dimensional and incompressible, we can use
the stream function formulation qixE2r f ½x ¼ g0 E4r f ½x; ð15Þ
@w where the Fourier transform is defined as
qE2 ¼ g0 E4 w; ð8Þ Z
@t 1
f ½x  F ½f ðtÞ :¼ f ðtÞ expðixtÞdt; ð16Þ
where wðr; h; tÞ is the axisymmetric stream function, and we use the 1
differential operator pffiffiffiffiffiffiffi
2   and i :¼ 1. We indicate the transformed variable by the square
@ sin h @ 1 @ brackets around the argument x. We divide each side of Eq. (15)
E2 :¼ 2 þ 2 ; ð9Þ
@r r @h sin h @h 2
by g0 , define k :¼  qgix, and rearrange slightly to obtain
0

that arises from eliminating pressure in the axisymmetric LNSE. 2


ðE2r þ k ÞE2r f ¼ 0: ð17Þ
Once the stream function is known, the (perturbation) velocity field
can be found from This particular definition for k might seem strange, but it has
physical meaning that is clear later, and makes the math now a
1 @w 1 @w
v 0 ¼ dr þ dh : ð10Þ bit easier.
r 2 sin h @h r sin h @r Before continuing, it is remarkable to notice how far the
Hence, the BCs for the stream function become problem has been simplified. We began with the general 3D spa-
tial- and time-dependent problem with three partial differential
1 @w 1 @w
r ! 1; ! 0; 2 ! 0; equations (PDEs) involving four dependent and four independent
r @r r @h
variables. Now it is a 1D, ordinary differential equation with one
@w
r ¼ R; ¼ uðtÞR2 sin h cos h; ð11Þ dependent and one independent variable. Once this is solved, all
@h the other objects can be found analytically in the frequency
@w 2 domain. For a Newtonian fluid, all inverses back into the time
¼ uðtÞR sin h:
@r domain can be found analytically.
At this point, for the steady state problem, many texts note the To solve our ODE, we first define the new function
form of the boundary condition at the bead surface to argue for a
g½r; x :¼ E2r f ½r; x; ð18Þ
separation of variables solution with an explicit form
6 J.D. Schieber et al. / Journal of Non-Newtonian Fluid Mechanics 200 (2013) 3–8

whose evolution equation is given by Eq. (17) to be 3R ikR


which yields A ¼  2ik e u½x (or A0 ¼  3ikR eikR u½x) and C 1 ¼ 2kR2
2
! h i
2 2 2
2 d 2 2 3ð1  ikRÞ  k R . Hence the final expression for the stream func-
ðE2r þ k Þg ¼ 2
 þ k g ¼ 0: ð19Þ
dr r2 tion (in the frequency domain) is
" 2  #
The form of the solution for this ODE comes from noticing that R 3ð1  ikRÞ  k R2 3R ikðrRÞ 1
the third term in the parentheses requires just an integrating w½r; h; x ¼ 2
 e 1  u½x
r 2k 2ik ikr
factor, and that the first two terms are just an equidimensional
2
operation. Hence, we insert the general form2  sin h ð29Þ
X
1
At this point, one could take the inverse Fourier transform to
g ¼ eikr An r n ; ð20Þ find the stream function in the time domain. Then it is useful to
n¼1
recognize that the solution is in the form w½r; h; x ¼
into Eq. (19), and find  xu½x sin2 h, where the overbar here means taking the one-
h½r;
X
1 h i sided Fourier transform. Then, one could use the convolution the-
n1 2 Rt
An ðn  2Þðn þ 1Þr n2 þ 2iknr ¼ 0: ð21Þ 0
orem to obtain wðr; h; tÞ ¼ sin h 1 hðr; t  t 0 Þuðt 0 Þdt . However, it is
n¼1 easier to wait until all desired solutions are found (velocity, pres-
Each power of r must be individually zero, so An ¼ 0 for sure, force on the sphere) before returning to the time domain.
n–  1; 0; 2. The remaining terms are then
2ikA1 2A0 4. Velocity, pressure and stress fields, and the hydrodynamic
  2 þ 4ikA2 r ¼ 0: ð22Þ
r2 r force on the sphere
Once again we group like powers of r to find that A2 ¼ 0 and
A1 ¼ A0 =ik. Hence, our solution for g is One can find the perturbation velocity v 0 ½x from the stream
  function in the usual ways by using Eq. (10)
1 
g½r; x ¼ A0 eikr 1  : ð23Þ R 3ð1  ikrÞeikðrRÞ þ ikRð3  ikRÞ  3
ikr v 0r ¼ u½x cos h
r 2
ðkrÞ
Our evolution equation for f is now just the definition for g, Eq. 
(18) and we use our solution above R 3ð1 þ ikrðikr  1ÞÞeikðrRÞ þ ikRð3  ikRÞ  3
v 0h ¼ 2
u½x sin h:
  r 2ðkrÞ
1
E2r f ½r; x ¼ A0 eikr 1  : ð24Þ ð30Þ
ikr
As a linear ODE, we can find the solution as a sum of the homo- The velocity field can be found from the perturbation velocity
geneous fh and particular fp solutions. The homogeneous solution is from Eq. (4). To find the modified pressure, one first takes the Fou-
found by noticing that it is equidimensional rier transform of Eq. (6)
 
! 1
d
2
2 $P½x ¼ qix 2
r2 þ 1 v0 ½x: ð31Þ
E2r fh ¼ 2
 fh ¼ 0 ð25Þ k
dr r2
After inserting Eq. (30) into Eq. (31), we obtain the two PDEs
P n
so the general solution fh ¼ n Bn r has only two terms that survive
@P R qix½ikRð3  ikRÞ  3
¼ u½x cos h; ð32Þ
B1 @r r ðikr Þ
2
fh ¼ þ B2 r2 : ð26Þ
r
and
The particular solution is found by noticing that, since g was
2
constructed to satisfy Eq. (19), so that E2r g ¼ k g. Hence @P qix½ikRð3  ikRÞ  3
2
¼R 2
u½x sin h: ð33Þ
fp ¼ g=k and our solution is @h 2ðikr Þ
 
B1 1 These may be integrated to find the modified pressure within a
f ½r; x ¼ þ B2 r 2 þ Aeikr 1  ; ð27Þ
r ikr constant to be
2
where A :¼ A0 =k , and is just some new as-yet-unknown constant. qix½ikRð3  ikRÞ  3
P ¼ P0  R 2
u½x cos h: ð34Þ
We find these constants from the boundary conditions. 2ðikr Þ
The distant BC, the first line in Eq. (11), says that both f 0 =r ! 0
and f =r 2 ! 0 as r ! 1. Either way requires that B2 ¼ 0. Taking the Note that this expression has removed both the hydrostatic
Fourier transform of the second and third lines of Eq. (11) give us pressure from gravity, and artificial pressure that arose from accel-
the remaining necessary conditions to find the constants erating our container, while holding the sphere fixed. Therefore, at
 distances far from the sphere, the modified pressure becomes P 0 , a
df  constant.
f ½r ¼ R; x ¼ u½xR2 ; and ¼ u½xR ð28Þ
dr r¼R The stress field is found from the constitutive equation.
qix h i
s½x ¼ 2
$v ½x þ ð$v ½xÞy : ð35Þ
k
However, if we are interested only in the total force on the
2
We note in passing that eikr also works as an integrating factor. However, if the sphere, we can use a general expression derived by Happel and
imaginary part of k is positive, this factor grows to infinity as r ! 1, and cannot be a Brenner [13, Eq. (4-14.16)] for axisymmetric flow around an axi-
solution. Note from our definition of k2 that there are two roots, and that we have symmetric body. Although the stream function is used to find v 0 in-
chosen the root with positive imaginary part. Had we chosen differently, we would
have to use the other integrating factor, and our current integrating factor would
stead of v, gradients of these two quantities are the same, so the
diverge at infinity and therefore be discarded. We would ultimately obtain the same stream function can be used to find stress. As a result, the expres-
results. sion of Happel and Brenner can still be used to find for a sphere
J.D. Schieber et al. / Journal of Non-Newtonian Fluid Mechanics 200 (2013) 3–8 7

Z p  Z 
2 @P
 2pRqix p  viscoelasticity (LVE) by simply replacing g0 with the complex vis-
F½x ¼ pR2 sin h  dh þ sin hE2 w dh: ð36Þ
0 @h r¼R k
2
0 r¼R cosity g . That is because the LVE constitutive equation for stress
Z t h i
Note that this expression does not require the full solution for y 0
s¼ Gðt  t0 Þ $v ðt0 Þ þ ð$v ðt 0 ÞÞ dt ð40Þ
the pressure field, or stress tensor. Since the h-dependence is rela- 1
2 2 2
tively simple for @P @h
, Eq. (33), and E w ¼ sin ¼ sin hg ¼ hE2r f has the same form as the Newtonian constitutive equation in Fou-
2 3ikR ikðrRÞ
1
Rp 3
 sin h 2 e 1  ikr u½x, we require only 0 sin hdh ¼ 4=3. rier space
Hence, the force on the sphere in the frequency domain is h i
( ) s½x ¼ g ðxÞ $v ½x þ ð$v ½xÞy : ð41Þ
6pRqix 6pR2 qx 2 3
F½x ¼ 2
þ  pR qix u½x: ð37Þ 2
Then k ¼ qix=g ¼ qx2 =G ðxÞ, where G⁄ is the dynamic 
k k 3
modulus of the medium, and all solutions through Eq. (37) also
The first term inside the curly brackets is the steady-state hold for a viscoelastic medium in the linear limit. (Note that this
Stokes drag, and the third term is the fluid inertia added to the is one of the reasons to define k2 the way that we did.) Such a sim-
bead inertia, first found by Poisson (per Stokes). The second term ple correspondence between creeping flow and LVE was first
is often attributed to Basset [2], but was found first by Stokes in exploited by Zwanzig and Bixon in 1970 [30], though Lee had no-
the frequency domain [26], and by Boussinesq in the time domain ticed a similar correspondence in 1955 between LVE and elasticity
[6]. Stokes claims that the third term, which describes the added [20]. We have recently exploited this isomorphism in generalizing
mass to the bead from surrounding fluid, was first found by Simeon the analysis of microrheology experiments to higher frequencies
Denis Poisson in 1831. The drag from the Basset force arises from [8,15]. In general, it is not always possible to pass analytically to
inertial shear waves that radiate away from the bead at high fre- the time domain from the frequency domain for the hydrodynamic
quencies [14]. force on the sphere.
2
For a purely Newtonian fluid recall that, k ¼ qix=g0 , where Note that k has dimensions of inverse length, and has real part
g0 is the viscosity. In the frequency domain, the force can then equal in magnitude to its imaginary part for a purely Newtonian
be written as fluid. Importantly, the real and imaginary parts of k have different,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi frequency dependent, magnitudes for a viscoelastic fluid. From the
2 second term in the stream function solution, Eq. (29), the exponent
F½x ¼  6pRg0 þ 6pR2 qixg0 þ ixpR3 q u½x: ð38Þ
3 ðeikðrRÞ Þ represents shear waves traveling from the bead. The real
Note that we can now write the force and velocity as vectors. part of the argument (the imaginary part of k) gives a decay length
This is because our problem is linear, so we can just superimpose for wave propagation. The imaginary part of the argument (the real
the solutions for each of the components, and these all look exactly part of k) gives the period of the wave. Both the period and the de-
like the z-component solution we just found. Using the convolution cay length are frequency dependent, through the dynamic modu-
theorem, the force on the sphere as a function of time can then be lus G⁄. If the distance between the particle and container walls is
written as comparable to the decay length, the particle will hydrodynamically
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi interact with its wave reflections [9]. If the size of the particle com-
Z t
g0 q duðt0 Þ 0 parable to the wave period, Basset forces will be important [14].
FðtÞ ¼ 6pRg0 uðtÞ  6pR2 dt
1 pðt  t0 Þ dt0
2 3 duðtÞ Acknowledgments
 pR q ; ð39Þ
3 dt
The authors are grateful to the Army Research Office (Grant
for any given history of bead velocity. Note that we could have just Nos. W911NF-09-1-0378, W911NF-09-2-0071 and W911NF-11-
as easily found the bead velocity as a function of a given history of 2-0018) for financial support.
forces from Eq. (38).
Appendix A. Fourier transforms
5. Linear viscoelastic fluids
Depending on the pre-factor used, there are different ways to
All viscoelastic materials obey the theory of linear viscoelastic- define the Fourier transform. Here, for the two-sided Fourier trans-
ity (LVE) in the limit of either small strain or small strain-rate, such form we use
that the material stays very near equilibrium. For synthetic poly- Z 1
mers it is typically safe to apply LVE for strains of approximately f ½x  F ½f ðtÞ :¼ f ðtÞeixt dt; ð42Þ
1
10%, whereas semi-flexible biopolymers such as actin filaments
have shown nonlinearity for strains as small as two percent. The whereas other conventions use a factor of 21p or p1ffiffiffiffi
2p
in front of the
valid range of strain rates is determined by the dimensionless integral. In our case, the inverse transform is defined as
Weissenberg number We, defined as the product of strain rate Z 1
1
times the longest relaxation time of your material. For We  1, f ðtÞ ¼ f ½xeixt dx: ð43Þ
2p 1
one can safely assume LVE. If one is performing passive microrhe-
ology [21], where the motion of the sphere is driven completely by In other words, having complete information about the Fourier
the equilibrium Brownian forces, LVE is appropriate by definition. transformed function f[x] allows one to completely recover the
Experiments also use ‘‘active microrheology’’ where the bead is original function in the time domain, f(t). We also use the one-
moved by an external, usually oscillatory, force. If the maximum sided Fourier (or Fourier–Laplace) transform
displacement of the bead in such an experiment is much less than Z 1
the bead radius, one can also assume linear viscoelasticity. A re- f ½x  F
 ½f ðtÞ :¼ f ðtÞeixt dt: ð44Þ
0
view of the technique, including the issue of nonlinear viscoelastic-
ity is given by Squires and Mason [24]. Note the change in notation and the lower limit on the integral.
The approach of staying in the frequency domain for our solu- We can use integration by parts to obtain a simple expression
tion here has the added advantage of giving the solution for linear for the Fourier transform of the derivative of a function
8 J.D. Schieber et al. / Journal of Non-Newtonian Fluid Mechanics 200 (2013) 3–8

  Z 1
@f ðtÞ @f ðtÞ ixt and
F ¼ e dt;  
@t 1 @t wðr; h; tÞ 1 @w 1 @w
Z 1 Z 1 v ¼ $  d/ ¼ dr 2 þ dh ; ð51Þ
@  @ r sin h r sin h @h r sin h @r
¼ f ðtÞeixt dt  f ðtÞ eixt dt
1 @t 1 @t ð45Þ
Z 1 Using these expressions in the linearized Navier–Stokes equa-
1
¼ f ðtÞeixt 1 þ ix f ðtÞeixt dt tion, neglecting
1
@ the terms
@
nonlinear in velocity and performing
the operation @h dr  @r rdh  to both sides of the results gives us
¼ ixf ½x;
Eq. (8).
where we used integration by parts to get the second line, the fun-
damental theorem of calculus to obtain the third, and the definition References
of the Fourier transform to obtain the fourth line.
[1] D.J. Acheson. Elementary Fluid Mechanics, Oxford, 1990.
[2] A.B Basset, A Treatise of Hydrodynamics with Numerous Examples, vol. II,
Appendix B. Stream functions Deighton, Bell and Co.,, 1888.
[3] C.O. Bennett, J.E. Myers, Momentum, Heat and Mass Transfer, McGraw-Hill,
1974.
We use a general curvilinear coordinate system with coordi- [4] Ratip Berker, Intégration des équations du mouvement d’un fluide visqueux
nates (a, b, c). These have unit coordinate vectors defined by incompressible, Strömungsmechanik II of Handbuch der Physik, vol. VIII,
  Springer-Verlag, Berlin, 1963, pp. 1–384 (Chapter 1).
1 @r [5] R. Byron Bird, Warren E. Stewart, Edwin N. Lightfoot, Transport Phenomena,
da :¼ ; a; b ¼ a; b; c; ð46Þ second ed., John Wiley and Sons, New York, 2002.
ha @a b–a
[6] Joseph Boussinesq, Sur la résistance qu’oppose un liquide indéfini en repos,
where r is the position in space, as in the main body of the manu- sans pesanteur, au mouvement varié d’une sphère solide qu’il mouille sur
toute sa surface quand les vitesses restent bien continues et assez faibles pour
script, and ha is called the ‘‘scale factor’’ defined by que leurs carrés et produits soient négligeables, Comptes rendu de l’Académie
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi des Sciences Paris 100 (1885) 935–937.
   u 3  
 @r  u X @xi [7] Joseph Boussinesq, Théorie Analytique de la Chaleur, vol. II, Gauthier-Villars,
ha :¼  ¼t ; ð47Þ 1903.
@a b–a  i¼1
@a b–a [8] Andrés Córdoba, Tsutomu Indei, Jay D. Schieber, Elimination of inertia from a
generalized Langevin equation: applications to microbead rheology modeling
and makes the vector on the left side of Eq. (46) dimensionless. and data analysis, J. Rheol. 56 (2012) 185.
[9] Andrés Córdoba, Jay D. Schieber, Tsutomu Indei, The effects of hydrodynamic
Then, the vector differential operator $ can be written in our interaction and inertia in determining the high-frequency dynamic modulus of
orthogonal coordinate system as a viscoelastic fluid with two-point passive microrheology, Phys. Fluids 24
(2012) 073103.
X 1 @ [10] William M. Deen, Analysis of Transport Phenomena, second ed., Oxford
$¼ da : ð48Þ University Press, 1998.
a¼a;b;c
ha @a
[11] Morton M. Denn, Process Fluid Mechanics, Prentice Hall, 1980.
[12] Christie John Geankoplis, Transport Processes and Separation Process
This is the general way that one obtains the Navier–Stokes equa- Principles, Prentice Hall, 2003.
tion in curvilinear coordinates, such as cylindrical, spherical or oth- [13] John Happel, Howard Brenner, Low Reynolds Number Hydrodynamics, second
ers [23]. If a flow is two-dimensional, then the velocity field has ed., Martinus Nijhoff, 1973.
[14] Tsutomu Indei, Jay D. Schieber, Andrés Córdoba, Competing effects of particle
only two components, say a and b, and depends on only two spatial and medium inertia on particle diffusion in viscoelastic materials, and their
coordinates (a and b). One can generate any such flow and guaran- ramifications for passive microrheology, Phys. Rev. E 85 (2012) 041504.
tee that the velocity field has zero divergence by writing [15] Tsutomu Indei, Jay D. Schieber, Andrés Córdoba, Ekaterina Pilyugina, Treating
  inertia in passive microbead rheology, Phys. Rev. E 85 (2012) 021504.
1 [16] William S. Janna, Introduction to Fluid Mechanics, PWS-Kent, 1993.
v ¼ $  wða; b; tÞ dc : ð49Þ [17] Pijush K. Kundu, Fluid Mechanics, Academic Press, 1990.
hc [18] L.D. Landau, E.M. Lifshitz, Fluid Mechanics, volume 6 of Course of Theoretical
Physics, Pergamon Press, Oxford, 1959.
Here we use the symbol  to indicate cross product, so that $is [19] L. Gary Leal, Advanced Transport Phenomena, Cambridge University Press,
the curl of the vector field. Note that there are alternative defini- 2007.
tions for w that also work, but lead to different intermediate expres- [20] E.H. Lee, Quart. Appl. Math. 13 (1955) 183.
[21] T.G. Mason, D.A. Weitz, Optical measurements of frequency-dependent linear
sions. One can rewrite the Navier–Stokes equation using Eqs. (48) viscoelastic moduli of complex fluids, Phys. Rev. Lett. 74 (7) (1995) 1250.
and (49). To eliminate pressure, one takes a scalar product operator [22] L.M. Milne-Thompson, Theoretical Hydrodynamics, Dover, 1996.
that eliminates the divergence in this coordinate system. In other [23] P. Moon, D.E. Spencer, Field Theory Handbook, Springer-Verlag, 1971.
  [24] Todd M. Squires, Thomas G. Mason, Fluid mechanics of microrheology, Ann.
@
words, one takes the operation @b ha da  @@a hb db  to both sides of Rev. Fluid Mech. 42 (1) (2010) 413–438.
[25] George Gabriel Stokes, On the steady motion of incompressible fluids, Trans.
the Navier–Stokes equation, which eliminates pressure. This gives Cambridge Philos. Soc. 7 (1842) 439–453.
us the general stream function equation for any 2D incompressible [26] George Gabriel Stokes, On the effect of the internal friction of fluids on the
flow in a general curvilinear, orthogonal coordinate system, for a motion of pendulums, Trans. Cambridge Philos. Soc. 9 (part II) (1856) 8–106.
[27] James R. Welty, Charles E. Wicks, Robert E. Wilson, Gregory L. Rorrer,
fluid either Newtonian or linear viscoelastic. It should be pointed
Fundamentals of Momentum Heat and Mass Transfer, Wiley, 2008.
out that this is not the way that Stokes developed stream functions. [28] Stephen Whitaker, Introduction to Fluid Mechanics, Krieger, 1968.
For spherical coordinates, hr ¼ 1; hh ¼ r, and h/ ¼ r sin h. Note [29] Ke Xu, M. Gregory Forest, Isaac Klapper, On the correspondence between
creeping flows of viscous and viscoelastic fluids, Journal of Non-Newtonian
that your favorite text may reverse the definitions for h and /.
Fluid Mechanics 145 (2–3) (2007) 150–172.
For our definitions, we have an axisymmetric flow (c/), so [30] Robert Zwanzig, Mordechai Bixon, Hydrodynamic theory of the velocity
correlation function, Phys. Rev. A 2 (5) (1970) 2005–2012.
@ 1 @
$ ¼ dr þ dh ; ð50Þ
@r r @h

Vous aimerez peut-être aussi