You are on page 1of 24

Journalof Non-Newtonian Fluid Mechanics

ELSEVIER J. Non-Newtonian Fluid Mech., 69 (1997) 47-70

Spherical Couette flow of a viscoelastic fluid Part II: Numerical study for the inner sphere rotation
H. Y a m a g u c h i *, H. Matsui
Department of Mechanical Engineering, Doshisha University, Kyoto 610-03, Japan
Received 20 March 1996; revised 3 September 1996

Abstract
Numerical analysis of the spherical Couette flow of a viscoelastic fluid during rotation of the inner sphere is performed using a constitutive equation from either the Giesekus model or the Oldroyd-B model. The numerical solutions for the flow field are obtained using the finite-difference method with a decoupling technique. The transient torque characteristics associated with the flow field that results from rotation of the inner sphere are also calculated. Two basic flow behaviors, which are obtained from either the Giesekus model or the Oldroyd-B model, are compared with Newtonian flow behavior. The numerical simulation reveals that the shear-thinning effect on the shear viscosity strongly influences the flow characteristics in the equatorial region in the Giesekus model. At a critical Reynolds number, the flow becomes unstable, forming elongated Taylor-G6rtler vortices of different sizes. With the Oldroyd-B model it was shown that at a high Deborah number, elastic instability gives rise to a substantial change in the flow mode in the transient process of the flow behavior. 1997 Elsevier Science B.V.

Keywords: Giesekus model; Numerical study; Spherical Couette flow; Taylor-G6rtler vortex

1. Introduction

The flow characteristic of a fluid between two concentric bodies is of interest in fluid engineering and in basic research in fluid dynamics, in which the effects of rotation play an important role. Cylindrical Couette flow, or Taylor-Couette flow, which deals with the motion of fluid between two concentric rotating cylinders, is a type of flow configuration that has been investigated in both Newtonian and non-Newtonian fluid dynamics. With Newtonian fluids, when the rotational velocity of the inner cylinder increases and at a critical Reynolds number (or Taylor number), hydrodynamic instability due to centrifugal force causes the flow form a
* Corresponding author. 0377-0257/97/$17.00 1997 Elsevier Science B.V. All rights reserved. PII S 0 3 7 7 - 0 2 5 7 ( 9 6 ) 0 1 505-4

48

H. Yamaguchi, H. Matsui / J. Non-Newtonian Fluid Mech. 69 (1997) 47-70

cellular structure [1] of vortices in the gap space and various flow modes appear in super critical flows [2,3]. Other works report on the behaviour of non-Newtonian fluid in cylindrical Couette flow. Lockett et al. [4] showed analytically that the shear thinning characteristic of a non Newtonian fluid has a significant effect on both the critical Taylor number and the critical wave number, reporting that the variation in radial distribution of an effective viscosity is responsible for the occurrence of a minimum value for the critical Taylor number with respect to departure from Newtonian rheology. Several authors report an elastic effect on non-Newtonian fluids in cylindrical Couette flow [5-7]. Shaqfeh et al. [6] conducted a linear stability analysis using the Oldroyd-B model and demonstrated that a vortex generated near the inner cylinder propagates toward the outer cylinder. They also suggested that the negative second normal stress differences may have a stabilizing effect for very small gap ratios, and further showed that increasing the solvent's relative contribution to the viscosity has a stabilizing influence. Phenomena similar to those associated with cylindrical Couette flow are also present in the flow between two concentric rotating spheres or so-called spherical Couette flow. However, with spheres, centrifugal force is a function of latitude, resulting in the existence of different types of flow modes. Numerous works discuss spherical Couette flow both theoretically and experimentally for Newtonian fluids. Wimmer [8] and Nakabayashi [9] give ample details on various flow phenomena relating to the formation of Taylor-G6rtler vortices. Schrauf [10] deals with the instability of Newtonian fluids as it relates to estimating the critical Reynolds number as well as to predicting the flow field. More recently, Mamun and Tuckerman [11] examined asymmetry and Hopf bifurcation in spherical Couette flow of Newtonian fluids, and presented bifurcation diagrams together with torque characteristics. Through our experiments, we have studied unique features of spherical Couette flow of dilute polyacrylamide-water solutions, which we present in Part I [12]. We present flow mode diagrams, discuss the critical conditions necessary for the onset of flow instability, and discuss the effect of varying the gap ratios and the solution properties of the fluids. Based on the results of our flow visualization, we have suggested that the onset of flow instability is strongly dependent on the regions: elastic instability in the polar region and centrifugal instability in the equatorial region. It is thought that the flow instability occurring in spherical Couette flow has simultaneously two distinct characteristics of cylindrical Couette flow [7]; namely, (i) the modifying effect of low levels of elasticity on the centrifugal instability that occurs at high Reynolds numbers; and (ii) the effect of high levels of elasticity on the flow at vanishingly small Reynolds numbers. In the present study, a numerical investigation is carried out for spherical Couette flow of a viscoelastic fluid, using constitutive equations from the Giesekus model [13] and the Oldroyd-B model which can be reduced from the Giesekus model. We perform numerical transient simulations using the finite difference method for low to moderate Reynolds numbers with different elastic numbers where the flow mode changes from weak secondary flow to the appearance of Taylor G6rtler vortices. We pay special attention to flow characteristics, which would be influenced by elastic effects or inertial effects. We provide a qualitative discussion for the results obtained from the experiments [12].

H. Yamaguchi, H. Matsui / J. Non-Newtonian Fluid Mech. 69 (1997) 47 70

49

2. Numerical analysis

2.1. Governing equations


In the present investigations, flow is assumed to be incompressible, isothermal, and without the body force. The equations governing the flow are as follows, V.v = 0, Dr P Dt Vp + V.v, (1) (2)

where Eq. (1) is the continuity equation and Eq. (2) is the m o m e n t u m equation (Cauchy's equation): v the velocity vector, p the density, t the time, p the pressure, and the stress tensor, which is derived from the constitutive equation shown below. In the present study, the Giesekus model [13] is used as the constitutive equation for dealing with a viscoelastic fluid possessing shear-thinning viscous characteristics as well as elastic characteristics. The Giesekus model is well defined for polymer melts or highly concentrated polymer solutions, but it is k n o w n [14] that the model can be applied to dilute polymer solutions as well. The differential form of the Giesekus model is thus written in the following constitutive equation,
V ~1 ~'p -]- Zl'r p -~- ~ - - (~'p" Tp) = qp]~,

qp

(3)

where 9 is the rate-of-strain tensor, qp the polymer viscosity, ~ the mobility factor (which relates to the anisotropic Brownian m o t i o n or hydrodynamic drag factor) and v the upper convective derivative [14]. Taking into account the polymer contribution tp and the solvent contribution v~ for the stress tensor and assuming Newtonian characteristics for the solvent fluid, i.e., zs = r/sy~, Eq. (3) can be reduced to the following formula by setting v = v~ + %.

~+21~+aZl(~'~)-a22(~'~+P'~)=r/o r/0

P + 22~ + ,~2- a-r- (~ z,

where 21 is the relaxation-time constant, 22 the retardation-time constant, r/0 the zero shear viscosity r/0 = q~ + r/p and a constant, which can be obtained as r/,~+ r/p
a = ~ (5)

qp

In Eq. (4), 2z is obtained from the relationship 22 = 21(r/J(qs + r/p)). For our examination of a dilute viscoelastic fluid, the constants appearing in the above relations are set at ~ = 0.1 and r/~/r/o= 0.1, which are c o m m o n values of fitting rheological data from Fig. 2 of Part I [12]. These constants are used in calculations as reference constants. The Oldroyd-B model constitutive equations can be reduced from Eqs. (4) and (5) by setting ~ = 0. In Fig. 1, the rheological characteristics are depicted for the Giesekus model and the reduced Oldroyd-B model with different relaxation-time constants. Fig. l(a) shows the shear viscosity and Fig. l(b) shows the first normal stress difference against, the simple shear rate i.

50

H. Yamaguchi, H. Matsui / J . Non-Newtonian Fluid Mech. 69 (1997) 47 70

In the present numerical simulation, a spherical coordinate system with an inner sphere radius r~ and an outer sphere radius r2 is used (Fig. 2). The flow is assumed to be axisymmetric (0/~3 -- 0), where the prevailing flow is described in the meridional plane. Thus, the velocity components ( v , V o , r e ) , in the spherical gap can be expressed in terms of the stream function , = ,(r, 0) and the angular velocity function q) = @(r, 0) as follows, in forms which satisfy the continuity Eq. (1):
V~ --

1 0~, r 2 sin 0 c~0'

1
Vo - - r - sin -

~?, 0 &'

v -

- sin r 0"

(6)

Eqs. (2), (4), and (6) are then nondimensionalized as follows,

102

......

........

........

........

........

e3 ca_.

lO 0

10 -2
,,,,I

,,,,,,,]

......

ill

Illlll,I

, ,ll,I

10-2'

10

~' [l/s]

10 2

(a) Shear viscosity


...... I ........ I ........ I ........ I .....

~'10 4

a_
z

10 0

~ ,.,.~-'-~

10 -4
........ l ........ I , l ,,l.d , , 'llllll ........ l

10 -2

10

102

[1/s]
(b) First normal stress difference Oldroyd-B Giesekus . . . . . . . . . . . . . . ), 1=0.01 ), 1=1.0 ' ......... 2 1=0.01 ), 1=1.0 ' ...... .... ;l 1=0.1 ). 1=0.1

Fig. 1. Rheological characteristics of the Oldroyd-B and Giesekus models.

H. Yamaguchi, H. Matsui /J. Non-Newtonian Fluid Mech. 69 (1997) 47 70


Z
/k

51

>Y

Fig. 2. Spherical coordinate system.

r* 00"= O0 UL'

v* ~,.=

Vr

I)0

-- l)0

Ill* =

~1

~@

~ L,

% r/o U'

U t* = t--

L'

R e = - pUL -,
r/o

De=21 U

L'

M=).2U L'

(7)

where the governing dimensionless parameters Re and De are the Reynolds number (based on the zero shear viscosity) and Deborah number respectively. In the present study, the elastic number E = De/Re is defined, and is measure of the relative importance of elastic to inertial effects. The characteristic values of velocity U and length L are taken to be U = rtco and L = r2 - rl respectively, where co is the angular velocity of the inner sphere rotation. In order to obtain the numerical solutions, the angular velocity function ~* is calculate with the ~b component of Eq. (2), while the q~ component ( of the vorticity vector V v is calculated with the q~ component of the vortice transport equation, which can be obtained by operating the rotation to Eq. (2). The complete set of equations to be solved is as follows: q~ component of the momentum, q5 component of the velocity transport equation, the decoupled stress components, and the Poisson equation for the stream function:

0-7 + vr 7 7 +

r 00
r 00

- - Re -

+--&-r + r ~ 0 - +

r 0 r cot 0 )

(8)

& + vr 0-7 =Re

0(

Of + vo 0(

2op(O~op 2( (vr + v0 cot 0) + r3 sin 0 ~ r

l{D2(+rsinO
- cot 0

_1(02 02 ) _1(02 0 IS) --r ~-05+~ 7~+ r a - ; T 0 + c o t 0 7 + 7 +

Too

+ -

r r ~-0

~rr

+ - -

r 2 sin 2 0

-+

Or 2

r 2 002

r 2 ~--0 ~0

'

(9)

52

H. Yamaguchi, H. Matsui / J . Non-Newtonian Fluid Mech. 69 (1997) 4 7 - 7 0

- = r~ -- 2 2v, 2. r,~ Or'


(OVo 10v,. ~ro = ~r0 -- \ - b 7 + r 00

Ovo + r Voo = too -- 2 \ 71 ~

% ~ = t e e -- 2

+ vo

r cot 0 ' (lo)

~j)
r J'

(0 2
(=

1 02

cot0~0 )
r2

~r 2 + r 2 00 2

"

The component form of the constitutive equation with the Giesekus model (Eq. (4)) is shown in the Appendix. It should be noted that in Eqs. (8)-(10) and in the Appendix, the asterisk indicating dimensionless parameters is dropped for the sake of clarity. The stress components are calculated from the constitutive equation by taking the difference of the Newtonian stress components, and the vorticity transport Eq. (9) is solved using the decoupling technique proposed by Cochrane et al. [15]. By adopting the decoupling technique, the elliptic form of the vorticity transport equation for a linear system is recaptured, and thus Eq. (9) can be solved as in the Newtonian case by the successive over relaxation (SOR) iterative method.
2.2. B o u n d a r y conditions

Boundary conditions are such that the inner sphere rotates with an angular velocity co while the outer sphere is kept stationary. No-slip conditions are imposed both at the inner sphere surface and the outer sphere surface, and there are axisymmetric conditions at the axis of rotation. Below, the boundary conditions are shown for (a) the axis of rotation, (b) the inner sphere surface and (c) the outer sphere surface where for each boundary (i) the flow field conditions and (ii) the stress field conditions are listed. The stress components are calculated using the constitutive equation (Appendix A) at both surfaces with boundary conditions of (ii). (a) 1/fl < r* < (1 + f l ) / f l , 0 = O, rc (axis of rotation)

(i) ,=0,
(ii) vo = O,

q~=0,
v~ = 0 ,

(--0,
OU0 Or = O, OUqb Or = O, 02V0 Or 2 - O, 02U~b & 2 - O.

(b) r* = 1/fl, (i) , = O, (ii)


Ur ~ O,

0 __<0 < rc (inner sphere surface)


02/

-- r 2 sin 2 O,
U0 = O, V~ = re)

( = 0r2, sin 0, OVr - ~ = 0,


Ou 0 ~0 = O, 021.)r 002 ~- 0 02!.) 0 ~02 -- O.

(c) r* = (1 + f l ) / f l , (i) ~, = 0, (ii) Vr = O,


=

0 < 0 < Tc (outer sphere surface) U, 0, ~"- - 0r2,


V~ = O, Ov~ 00 -- O, O Vo 0 0 ~-- O, 02U r 0 0 2 ~- O, 02U 0 0 0 2 -- O.

VO = O,

(11)

H. Yamaguchi, H. Matsui / J . Non-Newtonian Fluid Mech. 69 (1997) 47 70

53

2.3. Numerical method and procedure


Eqs. (8)-(10) and the constitutive equation (Appendix A) are discretized in order that they can be solved by the finite difference method. In Fig. 3, the mesh configuration used in the finite difference calculations is depicted. It is noted here that Fig. 2 is appropriately scaled for the r direction to provide a clear view of the gap space and that hereafter all results obtained from the calculations are presented for this view of spherical gap. The size of the mesh used in the numerical simulation is 31 x 101 for i (r directional position index) and j (0 direction position index) respectively, and the mesh is equally spaced in each direction. In order to ensure resolution for the numerical solutions, some trial runs were performed using different sizes of meshes. Representative calculations for a Newtonian fluid showed that the results obtained with a 31 x 101 mesh system made an approximately 0.09% difference to the torque coefficient Cm compared with a 51 x 181 mesh system. Since the difference due to mesh size was minimal, a 31 x 101 mesh system was used in all calculations. The finite difference equations for Eqs. (8)-(10) are formulated with a finite difference of second-order accuracy in space, and the time marching procedure (explicit Euler method) is used to solve the transient equations. In each time step, the Poisson equation (10) is solved using the (SOR) technique with a relaxation factor of 1.60. The convergence of the Poisson equation is determined with a maximum relative error of less than 1.0 x 10 - 4 . It is known [8-11] that in Newtonian fluids, the transition from stable simple Couette flow with a weak secondary flow in the meridional plane to the appearance of TG vortices (Taylor-G6rtler vortices) occurs at a critical Reynolds number due to centrifugal instability in the vicinity of the equatorial plane. At the same time, for a given critical Reynolds number,

Fig. 3. Mesh system.

54

H. Yamaguchi, H. Matsui /J. Non-Newtonian Fluid Mech. 69 (1997) 47-70

there exists the possibility that some different flow modes, such as a single pair of TG vortices or two pairs of TG vortices will appear, depending upon the gap ratio It = (r2 - rl)/r~ and the acceleration of the inner sphere rotation. Also as the Reynolds number increases, the prevailing flow states for postcritical Reynolds numbers are different [8,9] depending on the flow mode at the critical Reynolds number. In regard to these points, M a m u n and Tuckerman [11] reported from their instability analysis that when the inner sphere is started up from rest, then for a gap ratio/~ = 0.1538 and a Reynolds number range 183 < Re < 191, the only possible flow model is two pairs to TG vortices. (There are four different flow modes that could appear as a result of solution bifurcation when the Reynolds number gradually changes within that range, but for a given Reynolds number within the range, the only flow mode possible when started up from a state of rest is two pairs of TG vortices.) Thus in the present study, in accordance with the analytical results obtained by M a n u m and Tuckerman [11], typical numerical simulations were performed for Reynolds numbers Re = 1, 30, 100 and 185, with the gap ratio/q = 0.1538. The critical Reynolds number was Rec = 125, and below Rec there was no flow transition possible, whereas for Re = 185, the only possible flow mode was two pairs of TG vortices when the inner sphere was started up from rest and the fluids were Newtonian. Direct simulations for non-Newtonian fluids such as P A A - w a t e r solutions used in Part I [12] cannot be carried out for high Reynolds numbers since De was very high for the solution used in the experiment. For the P A A - w a t e r solutions, only cases with low Reynolds numbers are simulated and presented. In order to make a qualitative examination of flow phenomena and the characteristics of viscoelastic fluids, the rheological parameters for the Giesekus model are set at ~ = 0.1, V/s= 1 x 10- 3 (Pa s) and V/p= 9 x 10-3 (Pa s). The viscoelastic characteristics of dilute polymer solutions can typically be expressed as shown in Fig. 1. In the present calculations, when the Reynolds number is Re = 1, the Stokes flow approximation is made so that the convective terms in the governing equations are totally ignored. In order to obtain flow characteristics associated with the prevailing flows, the torque coefficient Cm, which is defined as follows, is calculated by integrating the shear stress vr+ on the inner sphere surface: = 27r f)~ pr2~o2 Zr~ sin 2 0 dO.

Cm

(12)

2.4. Results and discussion

In Fig. 4, calculation results with E = 0 (Newtonian case) are depicted for Reynolds numbers
Re = 1, 30, 100 and 185. Fig. 4(a) shows the steady state stream function; Fig. 4(b) shows the

time variation for the torque coefficient, and Fig. 4(c) shows the representative velocity profiles Vr, VO and v~ corresponding to the flow states (Fig. 4(a)) at the equator 0 = 90 and in the northern hemisphere 0 = 45 . In Fig. 4(a) (1), no secondary flow is evident in the gap space, because for Re = 1, the Stokes approximation (no inertia effect) is assumed. The resultant calculation shows that the circumferential velocity V+ distribution is almost linear (see Fig. 4 (c) (1) and (2)), indicating that the prevailing flow is simple spherical Couette flow. The torque coefficient for time variation reaches asymptotic values (see Fig. 4(b)) which can be calculated from

H. Yamaguchi, H. Matsui /J. Non-Newtonian Fluid Mech. 69 (1997) 47-70

55

( 1 ) R e = l (Time=0.50sec)
0.0000 e+00,
.

0.0000

e+00

(2)Re=30 ( T i m e = l 0.0sec) 0.3008 e-01, . -0.3008 e-01

),

( 3 ) R e - 1 0 0 (Time=50.0sec) 0.9953 e-01, . -0.9953 e-01 (a)


Fig. 4.

(4)Re=185 ( T i m e = 2 5 0 . 0 s e c ) 0.2165 e+00, . -0.2165 e+00 , -0.1129e+00,. 0.1129e+00 0.1762 e+00, ,, -0.1762 e+00

56

H. Yamaguchi, H. Matsui / J . Non-Newtonian Fluid Mech. 69 (1997) 47-70


.' '""1 ........ i ........ i ........ i ........ I ........ ] ........
i

102
E O

101

10 e
\N

"\~..\
....... i

~'- . . . . . . . .

03916 0.1185 0.0936

1 0 -1
.... i ........ i ........ i ........ i ........ ] i iiMr.l

10

-4

10

-2

10 ~ - .... Re=30 --- Re=185

10

Time[s]

(B)

----

Re=l , Re=100,

(1)

0
--

=45
'~t .............
'~\

[xl 0 "3] 0

;>r
/,i
1t /

o.1
//~ X

\ \\

/ i ~b

O,E

t\ ~\
\'t '
, '

ti

y ......... y
o

i;

X,,

>
x _j
/ i /

~ > O.Z
o.~

X
. . . . . . . . i

/.

,\

/'

6 ......... r

-0.1

6 ....

'

4-

(2)

=90
]/ ~'\,\

0.06r , "

' 0.05
J / ' y

o.04 t
0.021
0

/'

',

> "~tt

~\~._.

0
, i

>

i//
i . . . . . . . . .

/,

-0.05 o 1
i

(c)

Re=l

....... .....

Re=30 Re= 185

Re= 100,

Fig. 4. (a) Calculation results for stream function (E = 0, Newtonian). (b) Torque coefficient for time variation (E = 0, Newtonian). (c) Velocity profiles (E = 0, Newtonian).

H. Yamaguchi, H. Matsui /J. Non-Newtonian Fluid Mech. 69 (1997) 47 70 Cm = 2re

57 (13)

3fl - 1 (1 +/7) 3 (1 +fl--)g

fo

sin g 0 dO,

where Eq. (13) is obtained analytically by solving the equation of motion for/q = 0.1538 [12]. However, when Re is increased to 30 or 100, the secondary flow associated with circulation motion appears (see Fig. 4(a) (2) and (3)). The circulation motion becomes stronger when Re is increased as the magnitude of the velocity components Vr and Vo (Fig. 4(c)) increases, while the shape of the secondary flow is uncharged. The velocity profiles reveal that the circulation of the secondary flow in the northern hemisphere is in an anti-clockwise direction (see Fig. 4(c)) (1) and Fig. 4(a) (2) and (3)). With increased Re, the non-linearity of the velocity profile v becomes strong (Fig. 4(c)), followed by the appearance of a strong secondary flow. The resultant steady torque coefficient, which agrees well with the experimental value in Part I [12], decreases, as seen in Fig. 4(b). After the onset of hydrodynamic instability, two pairs of T G vortices are generated in the equatorial region (see Fig. 4(a) (4)) with Re = 185. The transient torque curve increases (when the time is approximately 100 in Fig. 4(b)) and reached the asymptotic value of Cm = 0.0936 after time elapses. The steady state torque coefficient is close to the Cm = 0.095 obtained in the experiment in Part I. In the same fashion, the results of the direct simulation for PAA500 with Re* = 1.17 and Re* = 30 [12], which imply Re = 1 and 10.48 (definition on the basis of the zero shear viscosity in Part II), are presented. Fig. 5 shows the results for E = 0.565. As seen in Fig. 5(a) (1) for Re = 1, no vortices exist and symmetric secondary flow appears in each hemisphere. The secondary flow (Fig. 5(a) (1)) circulates in a clockwise direction in the northern hemisphere, which is completely the reverse of the Newtonian case (Fig. 4(a) (1)). The reversal of the circulation of the secondary flow is also shown by the direction of the flow velocity ve, in Fig. 5(c). However, when the Reynolds number is increases (Fig. 5(a) (2)), the secondary flow circulates in the same direction as in the Newtonian case, and this is also indicated by the velocity profile of vo in Fig. 5(c). In Fig. 5(b) for the transient torque characteristic, the torque coefficients reach the steady state values of Cm = 10.336 and 0.3902 with Re = 1 and Re = 10.48 respectively. By comparison with the measured torque in Part I, the calculated torque coefficients are lower than the measured values of Cm = 14.5 and 0.68 with Re* = 1.17 and Re* = 30 respectively. The discrepancy between the calculated value and the measured value is due to the weak non-linearity of the v+ velocity profile as displayed in Fig. 5(c), whereas the measured velocity profile (Fig. 7, Part I [12]) shows stronger non-linearity, as discussed in Part I. It should be remembered that direct simulation for higher Reynolds numbers ( R e = 100 ~ 185), with which the generation of TG-Type vortices is observed in the experiments (Part I), could not be achieved. This is due to the high elastic number (E = 0.565) in the whole region (Re = 1 --~ 85) for the P A A - w a t e r solutions used in the experiments. In the present numerical analysis, convergent solutions are obtained only for E = 0.02 -~ 0.00556 in the range Re = 100 ~ 185. However, the present numerical study is continued within the calculation limit of the elastic number (E = 0.02-0.00556) to gain more qualitative results for viscoelastic fluids. Fig. 6 presents results for non-Newtonian fluids at different Reynolds numbers: Re = 1, 30, 100 and 185, with the elastic number being kept constant at E = 0.0056. As seen in Fig. 6(a) (1), a secondary flow appeared associated with the circulation motion; it is identified from the sign of the stream function (a negative value in the northern hemisphere) and also from the velocity profiles of vo and Vr (note in Fig. 6(c) (1) that the magnitude of vo and Vr is very small). The

58

H. Yamaguchi, H. Matsui / J . Non-Newtonian Fluid Mech. 69 (1997) 47 70

(1) Re:1[Re*=1.17], E=0.565[*=0.482] (Time=5.0sec) -0.1262 e-02, . 0.1262 e-o2 (a)


.' '""! ........ I ........ I .... ""1 ........

(2) Re-10.48[Re*=30], E=0.565[*=0.197] (mime=lO.Osec)

0.1947e-01,
........ I ........ I

-0.1947e-01

102 E O

101

10336

10

"'"
~ . . . . . . . . . . 0.3902

10-1
,.,,I ...... i[l ........ I i i ...... I , ,ii..J ..... lid ........ I

1 0 -4 __ (b)

1 0 -2 10 1 02 Time[s] Re=l[_Re*=l 17] . . . . . . Re=10 48[Re*=30] E=0.565[E"=0.482] , E=0.565[E'~--0.197]

Fig. 5.

H. Yamaguchi, H. Matsui / J. Non-Newtonian Fluid Mech. 69 (1997) 47-70

59

(1) 0 = 4 5
[x10-4] 0 [xl 0 - 2 ]
/

",,

0.6

\
..\

\
'~ ". / "

,,/

>
0.2

<../

6 . . . . . . . . .r

11

........

(2)

0 =90
[x 10 - 3 ]

:./,--,.

[xl 0 - 3 ]
> t :

/-'L

i
k '. k '-. ,, .." ,,'

'>\,,

,"

.........

.........

t . . . . . . . . 0

i 1

Re=1[_Re*=1.17]
(c)

..... ,

Re=10.48[Re*=30] E=O.565[E*=0.197]

E=0.565[E*=0.482]

Fig. 5. (a) Calculaton results for stream function using experimental parameters (PAAS00). (b) Torque coefficient for time variation using experimental parameters (PAA500). (c) Velocity profiles for experimental parameters (PAA500). secondary flow is very weak, but the velocity profiles of vo and vr are similar to those obtained in Fig. 5(c). As in the case, the secondary flow in the northern hemisphere circulated in a clockwise direction, which is completely the reverse of the case with the Newtonian flow (for example, in Fig. 4 for Re -- 1, the circulation is anti-clockwise in the northern hemisphere). This reversal of the secondary flow is also obtained in the previous direct numerical simulation (see Fig. 5(a) (1)). The appearance of the secondary flow for a viscoelastic fluid without the inertia effect is due to the existence of the first normal stress, which appears strongly in the equatorial region because of higher shear rates on the inner sphere surface. The appearance of "Err , which will act on the fluid element in the opposite direction to the centrifugal force is noteworthy. A separate calculation using the Oldroyd-B model for Re = 1 showed a secondary flow similar to that obtained in Fig. 6(a) (1) and 5(a) (1). However, when the Reynolds number increases to R e - - 3 0 and Re = 100, as shown in Fig. 6(a) (2) and (3), the secondary flow in the northern hemisphere becomes anti-clockwise, as is also seen in the velocity profiles of vr and vo (Fig. 6(c)). This is due to the introduction of the inertia term in the governing equations. The inertial effect

60

H. Yamaguchi, H. Matsui / J . Non-Newtonian Fluid Mech. 69 (1997) 47-70

(1)Re=l (Time=5.00sec) -0.1191 e-04, 0.1191 e-04

( 2 ) R e = 3 0 (Time=10.0sec) 0.2982 e-01, -0.2982 e-01

( 3 ) R e = 1 0 0 (Time=80.0sec) 0.1038e+00, -0.1038e+00

(a)
Fig. 6.

(4)Re=185 (Time=250.0sec) 0.2337 e+00, -0.2337 e+00 , -0.1693 e+00,. 0.1693 e+o0 0.1858 e+O0, A -0.1858 e+O

H. Yamaguchi, H. Matsui /J. Non-Newtonian Fluid Mech. 69 (1997) 47-70

61

~b thin line and $eft value : E=0,00278 ~--"


old line and right value : E=0.00556

10 2 E O 10 ~

. ;;~-"",~,
10 o
0,3908,0.3094

10-~
..... I ........ I ........ I ........ I ........ ] ........ I ........ I

01153,0~1C~ 0,0919, 00831

10 -4

10 -2 Time[s] Re=l ,

10 - .... Re=30 --Re=185

102

(b)
(1)

--

---Re=100,

0 =45
'~\ ......... \\~ ... ....

[x104] O "i"
-2

....... ~ f
//'

0.1 0.6

\ '~

/ /

/,'

\
~0.4 0.2

-4
--~

\ \'\ .J'

-0.1
r

(2)

0 =90
0.06 ~0.04
!

'
/ \

0.05

," /,

,~
'\

>
\ / ~, ,J

0.02
/

\
-0.05
" ~ - *.

Z-<-". . . . . . . . . . . . . . . . . . . . . . . .

6 ......... r

6 ........

.i

0
r

(c)

Re=l, Re=100,

.......

Re=30 Re=185

Fig. 6. (a) Calculation results for stream function (E = 0.00556). (b) Torque coefficient for time variation (E = 0.00278 and 0.00556). (c) Velocity profiles (E = 0.00556).

62

H. Yamaguchi, H. Matsui /J. Non-Newtonian Fluid Mech. 69 (1997) 47-70

becomes compatible against the elastic effect, and the centrifugal force appears to drive the secondary flow towards the r-direction in the equatorial plane when the rotating speed of the inner sphere is a maximum. With Reynolds numbers R e = 30 and 100, no flow instability appears to change the flow mode of the secondary flow, but in the higher latitude 0 = 45 , the non-linearity of the v+ profiles (Fig. 6(c) (1)) tends to become noticeable. (This non-linear velocity profile of v~ for viscoelastic fluids was also observed in Part I [12] and Ref. [16].) With R e = 100, the secondary flow velocity Vo tends to become larger in the vicinity of the inner sphere, shifting the profile of the secondary flow towards the inner sphere wall (Fig. 6(c) (1)). A large overshoot of the torque coefficient for time variation is obtained as shown in Fig. 6(b), where the overshoot is lost as time elapses and reaches a plateau. The magnitude of the two overshoot decreases when the Reynolds number increases, and the transient torque characteristics for higher R e show slower response together with a smaller overshoot since the hydrodynamic effect due to inertia dominates the flow transition. In Fig. 6(b), the history of the torque coefficients when E - - 0.00278 is also displayed for the sake of comparison. As seen in Fig. 6(b), for E--0.00278, similar transient torque behavior is obtained in which the steady state torque coefficient becomes smaller for a given Reynolds number as the elastic number is reduced from E = 0.00556 to E = 0.00278. This tendency is also supported by the results of the experiments in Part I. It is noted that the equatorial symmetry of the flow breaks down slightly, although the magnitude of the deviation of the velocity, as shown in Fig. 4(c) (2) at 0 = 90 for vr and vo, is extremely small compared to re, and in the present calculation with the Giesekus model, no change of flow mode occurs due to symmetry breaking for this range of elastic numbers (as shown in Fig. 6(c) (2), which is not much different from Fig. 4(c) (2)) with Reynolds numbers below 185. For R e = 185, as shown in Fig. 6(a) (4), the flow transition associated with the appearance of two pairs of TG vortices occurs. This transition is caused by centrifugal instability in the vicinity of the equatorial plane since the decrease in the shear viscosity due to the shear-thinning effect becomes significant for the higher shear rate region. Thus the apparent local Reynolds number in the region of the equatorial plane increases, causing flow instability. With the onset of the critical flow mode (TG vortices), the velocity profile of v+ becomes highly non-linear, as shown in Fig. 6(c) (1) and (2), and equatorial symmetry is largely broken at 0 -- 90 . Fig. 6(b) also reveals that for R e = 185 (with E--0.00556 and E--0.00278), the torque decreases after having overshot, but then tends to increase in the transient process due to the appearance of TG vortices. In supercritical flows as observed in Fig. 6(a) (4), in comparison with the Newtonian case of Fig. 4(a) (4), the neighboring pair of TG vortices at the equatorial plane tend to expand towards the polar region, while the second outer pair is pushed further away, maintaining a shape almost exactly equal to the gap space. The phenomena are due to the effects of shear-thinning viscosity in the vicinity of the equatorial plane, where the region of low viscosity (together with higher first normal stress differences) tends to expand toward the polar region as E is increased. With non-Newtonian fluids, the higher torque at the generation of TG vortices (as compared with the Newtonian case), as shown in Fig. 6(b) for R e = 185 is probably due to the expansion of the region of TG vortices causing the velocity profile of ve to be highly non-linear towards the higher altitude at the inner wall, as indicated in Fig. 6(c). It is speculated that at the polar region the flow state will still be governed by the effect of elasticity, owing to the slow rotational speed near the axis of rotation. In order to examine the interaction of the effect of elasticity in the

H. Yamaguchi, H. Matsui / J . Non-Newtonian Fluid Mech. 69 (1997) 47 70

63

polar region and the effect of shear-thinning viscosity in the equatorial region simultaneously, the elastic number must be increased further. However, in the present investigation with the Giesekus and Oldroyd-B models, we could not get numerical solutions for elastic numbers higher than E = 0 . 0 0 5 5 6 for R e = 185. The reason for this numerical breakdown [17] is an extremely highly concentrated distribution of stresses appearing for higher elastic numbers, resulting in numerical solutions diverging by whatever small increment in the time step or the small mesh spacing are used. In Fig. 7, as a representative case, flow characteristics (steady state stream function) when the Reynolds number is taken as the infinitesimal value of R e - - 1 x 10 4 while the Deborah numbers are taken as the large values of De = 1 x 108 and De = 3 x 10 ~, are depicted. It is seen that when E is very large, the transient torque becomes infinitely large and the prevailing velocity profile becomes infinitesimal since the Reynolds number is infinitesimal (although the trend for the transient torque characteristics and velocity profiles are similar to those obtained in Fig. 5(b) and (c) with Re = 1). As seen in Fig. 70), the secondary flow appears as it did in the case of Re = 1 in Fig. 5(a) (1) and 6(A) (1). The induced flow obtained in Fig. 7(I) is caused solely by the elastic effects; the effect of inertia is minimal. However, when the Deborah number

(1)

Re = 1 10 -4

(11)

Re = 1 10 - 4

De=

i x l0 s
0.4875 e-14

De=

101
0.4286 e-16 -0.1178 e-16 0.1113 e-16

-0.4875 e-14,

-0.4286 e-16, . 0.1178 e-16,

-0.1113e-16,

Fig. 7. Calculation results for stream function (E ~ oc).

64

H. Yamaguchi, H. Matsui / J . Non-Newtonian Fluid Mech. 69 (1997) 47 70

is further increases to D e = 3 x 10 l, leading to E = 3 x 1014(E~ oo), TG-type vortices are generated, as seen in Fig. 7(II). A pair of largely elongated vortices appears in the equatorial plane and the outer pair of vortices is pushed away towards the polar region, causing hydrodynamic instability even in the case of an infinitesimal Reynolds number where the elastic effect dominates over the inertial effect. It is thought that the flow instability is caused by elasticity [5], while the enlarged pair of vortices is probably affected by the strong shear-thinning characteristic of the Giesekus model after the onset of hydrodynamic instability (due to the infinite viscous effect with very strong shear-thinning character once the meridional velocity component is induced). In order to examine the pure elastic effect, a numerical simulation with the Oldroyd-B model was conducted for Re = 100 in which the flow mode is subcritical (Fig. 4(a) (3)) and there would be no centrifugal instability in the case of Newtonian fluids. It is thought that flow behaviour for a relatively high elastic number with a strong elastic effect (followed by no shear-thinning effect) can be investigated using the Oldroyd-B model. In Fig. 8(a), some representative results of the transient calculations using the Giesekus model (I) and the Oldroyd-B model (II) are compared for Re = 100 with E = 0 . 0 2 . As can be seen in Fig. 8(a) (I), TG vortices were generated with the Giesekus model, whereas with the Oldroyd-B model, there were none (see Fig. 8(a) (II)). With Newtonian fluids at Re = 100, flow instability does not occur (Fig. 4(a) (3)). The flow is stable, with the appearance only of a secondary flow. Thus it is thought that the flow instability in the Giesekus model is caused by the strong shear-thinning effect in the equatorial region where the shear rate is maximum. The destabilizing of the flow of the P A A - w a t e r solutions was also observed in our experiments (Part I), where TG vortices were observed in the equatorial region for low Reynolds numbers. On the other hand, the Oldroyd-B model, which lacks shear-thinning character, does not destabilize the flow. Thus for higher Reynolds numbers with relatively small elastic numbers such as are considered in Fig. 8(a), shear-thinning plays an important role in flow instability. With separate calculations using the Oldroyd-B model, including the inertia term for Re = 185, there is no flow transition and all flows are similar to those in Fig. 8(a) (II), even for E = 0.00556. Thus, the major cause of flow transition in Fig. 8(a) (1) is the centrifugal instability caused by the shear-thinning effect of the fluid, because the shear-thinning effect becomes significant in the equatorial region where the flow is destabilized. In Fig. 8(b), the result of calculations using the Oldroyd-B model with a higher elastic number (E = 0.4) is depicted for Re = 100 in order to show the strong elastic effect against the inertial effect. It is noted that no comparison is possible with the Giesekus model with E = 0.4 since no convergence of the numerical solution can be obtained, as described earlier. As seen in Fig. 8(b), at the initial stage after the start-up of the inner sphere rotation (Fig. 8(b) (1)), two circulation zones appear in each hemisphere. As time elapses, the circulation zone on the inner sphere tends to expand towards the radial direction with its center of circulation point (the position of the highest magnitude of stream function distribution; the position -0.2471 x 10-~ as shown in Fig. 8(b) (2)) shifting toward the equatorial plane, while the circulation zone on the outer sphere is squeezed, its center of circulation zone being shifted toward the polar region. After time elapses, these two circulation zones fuse together in each hemisphere, forming one circulation zone for the secondary flow, with circulation being anti-clockwise, as shown in Fig. 8(b) (3), which is in fact an inertial flow modified by a small amount of elasticity, while the secondary flow with the clockwise circulation (Fig. 8(a) (II)) is derived from the elastic effect. Then a single

H. Yamaguchi, H. Matsui / J. Non-Newtonian Fluid Mech. 69 (1997) 47-70 (I)Giesekus model: Re=100

65

j _ / / (1)Time=50.0sec
o12ooe+OO,
.

(2)Time=90.0sec -o1200e+00 01305e+00. 01309 e+01 , . -o.1305e+00 -o 13o9 e + o l

(3)Time=130.0sec 0 1876e+00. o.1781 e + o l .


*

(4)Time=250.Osec -01576e+00 -o 1781 e+Ol * 01925e+09. - 0 1 0 2 9 e+Ol. 9.1524e+Ol.


,

~.1925e+oo

o 1029 e+Ol . -o1524e+01.

(ll)Oldroyd-B model : Re=100

(1)Time=5O.0sec

(2)Tirne=90.Osec

(3)Time=130.0sec

(4)Time=250.0$ec
*

(a)

0.9656 e-01.

-O9656 e-01

0.9675 e-01.

- 0 9 6 7 5 e-Ol

09675 e-ol.

-0.9875 e 01

0 9 6 7 6 e-Ol.

-O.9676 e-01

9
(1)Time-~0.0see 0 2 4 7 1 e-01. -0.9515e-o2. -0.2471 e-01 . 0.9515e-o2 (2)Tirne=50.0sec * 09582 e-02 -01775e-01. * - o 9 5 8 2 e-O2 (3)Tirne=60 0.8640 e - 0 1 0see * 0 8 6 4 0 e-01 (4)Time=120.Osec O 1589 e+oo. -07559e~1. . -0.1589 e+oo 07559(~-Ol 0.1775e-ol

(b)
Fig. 8.

pair of vortices appears in the equatorial plane, as shown in Fig. 8(b) (4). The generation of a single pair of vortices cannot be predicted by the stability analysis [11] of Newtonian fluid flow. It is of interest to note that a similar flow configuration was obtained by Olagunju [18] in his numerical study of viscoelastic fluids between a cone and a plate using the Oldroyd-B model. As has been mentioned in Part I [12], he has shown two distinct regions of secondary flow appearing in the gap space, depending on the Deborah number, the Reynolds number, and the gap space (De = 1.0, Re = 1.5,/~ = 0.2). He thus concluded that for a small enough De or/? an inner vortex is induced by viscoelastic effects and an outer vortex is induced by inertial effects, where in the outer vortex flow is inward near the stationary plate and outward near the rotating

66

H. Yamaguchi, H. Matsui /J. Non-Newtonian Fluid Mech. 69 (1997) 47-70

cone. The numerical results obtained in the present study, particularly in Fig. 8(b) (1) with the flow configuration after a sudden start-up, have m a n y features in c o m m o n with those reported by Olagunju [18], although the flow geometry is different. The results shown in Fig. 8(b) can be further compared phenomenologically with results obtained by Shaqfeh et al. [6], who showed the elastic instability in cylindrical Couette flow. They reported that a vortex which has a standing wave structure is generated on the inner cylinder and propagates toward the radial direction while a new vortex appears on the inner cylinder. These propagation phenomena are relevant to the secondary flow in the transition obtained in the present study of spherical Couette flow. It is thus thought that in spherical Couette flow, the major cause of these p h e n o m e n a at the initial stage after start-up is fluid elasticity. The difference in the propagation direction is due to the geometry of the flow field since the shear rate is dependent upon the latitudinal angle 0 in the spherical geometry. It is speculated further that at later stages of flow-mode transitions such as are shown in Fig. 8(b), the centrifugal effect tends to overcome the elastic effect, leading to TG-type vortices appearing on the equatorial plane. The appearance of a single pair of vortices may be due to the residual effect of elasticity on centrifugal instability. The transient torque coefficient follows a quite different path, as shown if Fig. 8(c), where the transient values of Cm obtained from the resultant flow fields in Fig. 8(a) and (b) are compared. It is seen in Fig. 8(c) that the transient torque curve for the Oldroyd-B model with E = 0.4 shows no overshoot character in contrast with the other two cases. This m a y be because in the early stages after a sudden start-up, there exist two pairs of secondary flow (as described below), which yields a stronger hydrodynamic effect over the elastic effect on the transient torque process, creating a smooth torque change with time. On the other hand, when the Giesekus model is compared with the Oldroyd-B model
...... I ........ I ...... I ........ I ....... i

i0 c

E 0

10-1

01174 01143 0.0962

i lllll

........

........

........

........

10 -2
- Giesekus

10 Time[s]
model(E=O.02), -------

10 2
Oldroyd-B
Oldroyd-B

model(E=O.02)
model(E=0.4)

(c)

Fig. 8. (a) Calculation results for transient stream fuction (E = 0.02); (I) Giesekus model; (II) Oldroyd-B model. (b) Calculation results for transient stream function (E = 0.4); Oldroyd-B model, Re = 100. (c) Torque coefficientfor time variation (Giesekus and Oldroyd-B models, Re = 100).

H. Yamaguchi, H. Matsui / J. Non-Newtonian Fluid Mech. 69 (1997) 47-70

67

for E = 0.02, the purely elastic effect of the Oldroyd-B model shows higher overshoot character than the Giesekus model, indicating that the shear-thinning character of viscosity eases the overshoot of torque by suppressing the elastic effect. The result of the calculations shown in Fig. 8 are verified by reducing the time step and using finer meshes, but the results are the same. Therefore it is certain that the phenomena obtained in the calculation are not derived from numerical instability but from the purely rheological effects of the Oldroyd-B model. Further studies are required to understand the flow transitions of viscoelastic fluids. An instability analysis has to be achieved that includes the inertial effect as well as the elastic effect in the spherical Couette flow configuration. However, that is not within our scope in the present study, and we shall be reporting on detailed flow-transition modes with a bifurcation analysis, including predictions of critical parameters, in future publications.

3. C o n c l u s i o n

A numerical study is conducted to show the flow behavior of a viscoelastic fluid in spherical Couette flow by using the Giesekus model and the Oldroyd-B model. From the results of the calculation, the following conclusions are drawn. (1) At very low Reynolds numbers such as Re = 1 or below ( E ~ re), ignoring the inertia effect, the secondary flow for viscoelastic fluids circulated in the opposite direction to that of Newtonian fluids. This phenomenon is commonly observed in both models. (2) For higher Reynolds numbers, by using the Giesekus model, the effect of shear-thinning viscosity becomes significant in the equatorial region, where the flow becomes unstable, generating two pairs of TG vortices of different sizes. (3) With a relatively high elastic number, the results of calculations using the Oldroyd-B model revealed that in the initial stages after start-up of rotation of the inner sphere, two circulations zones are generated. The inner circulation zone propagates towards the radial direction, lowering its altitude, while the outer circulation zone is squeezed toward the polar region.

Appendix A

The stress components calculated by the Giesekus model in the spherical coordinate system.

l + De ~t+ v~-~r+ r OOj - 2-~-O ( 2 a M + D e ) z ~ - r


= --2

(aM +

--~ VrO

rro+\er
M

r~+ +aDe(z2~+zro+z,~)

t~
I J J

JI

-F-

r ~ l

+ ~, + ~ + ~ +

b~

~
Jr
I

~1~ ~ 1 ~

~ ~
I

+ I~ + a
~

~
~ ~
@

~1~
J--I

~1~

1 r

+ ~l ~
c)

+
+ + ~ + ~ J

~ L
a
+
-~

c~

~ ~. +
+ +
e-t

+
~

--k

~ ---k

b~

e.-/

I
N.

L ~

-k

g~
e-.i

I
II

r~l

II

r ~

~
+
I

+
~ + ~"

~l~
~
+ +
",I ~ ~1 ~''

~1~
~
I

~1~

~l ~
+
+ "~1~ ~ ~ ~-~

+
i

"~1~

"~1+ ~
~ +

~1 ~
~ ~
~ "~1~

~l~
+

~1 ~ "~

+
b~

~
+
~ +

+
+

"~1~

I
+

+
I
0

I~
I
[

I
+
I I

~
I

70

H. Yamaguchi, H. Matsui / J . Non-Newtonian Fluid Mech. 69 (1997) 47-70

References
[1] G.J. Taylor, Stability of a viscous liquid contained between two rotating cylinders, Philos. Trans., 223 (1923) 289. [2] T. Mullin, Chaos in fluid dynamics, in T. Mullin (Ed.), The Nature of Chaos, Oxford University Press, 1993. [3] J.J. Kokrine and T. Mullin, Low-dimensional bifurcation phenomena in Taylor-Couette flow with discrete azimuthal symmetry, J. Fluid Mech., 275 (1994) 375. [4] T.J. Lockett, S.M. Richardson and W.J. Worraker, The stability of inelastic non-Newtonian fluids in Couette flow between concentric cylinders: a finite-element study, J. Non-Newtonian Fluid Mech., 43 (1992) 165. [5] R.G. Larson, E.S.G. Shaqfeh and S.J. Muller, A purely elastic instability in Taylor-Couette flow, J. Fluid Mech., 218 (1990) 573. [6] E.S.G. Shaqfeh, S.J. Muller and R.G. Larson, The effects of gap width and dilute solution properties on viscoelastic Taylor-Couette instability, J. Fluid Mech., 235 (1992) 285. [7] S.J. Muller, E.S.G. Shaqfeh and R.G. Larson, Experimental studies of the onset of oscillatory instability in viscoelastic Taylor-Couette flow, J. Non-Newtonian Fluid Mech., 46 (1993) 315. [8] M. Wimmer, Experiments on a viscous fluid flow between concentric rotating spheres, J. Fluid Mech., 78 (1976) 317. [9] K. Nakabayashi, Transition of Taylor-G6rtler vortex flow in spherical Couette flow, J. Fluid Mech., 42 (1983) 209. [10] G. Schrauf, The first instability in spherical Taylor-Couette flow, J. Fluid Mech., 166 (1986) 287. [11] C.K. Mamun and L.S. Tuckerman, Asymmetry and Hopf bifurcation in spherical Couette flow, Phys. Fluids, 7 (1995) 80. [12] H. Yamaguchi, J. Fujiyoshi and H. Matsui, Spherical Couette flow of a viscoelastic fluid. Part I: Experimental study of the inner sphere rotation, J. Non-Newtonian Fluid Mech., 69 (1997) 29-46. [13] H. Giesekus, A simple constitutive equation for polymer fluids based on the concept of deformation-dependent tensorial mobility, J. Non-Newtonian Fluid Mech., 11 (1982) 69. [14] R.B. Bird, R.C. Armstrong and O. Hassager, Dynamics of Polymeric Liquids, 2nd. edn, Vol. 1, Wiley, New York, 1987. [15] T. Cochrane, K. Walters and M.F. Webster, Newtonian and Non-newtonian flow near a re-entrant corner, J. Non-Newtonian Fluid Mech., 10 (1982) 95. [16] C.J. Rofe, R.K. Lambert and P.T. Callaghan, Nuclear magnetic resonance imaging of flow for a shear-thinning polymer in cylindrical Couette geometry, J. Rheol., 38(4) (1994) 875. [17] M.J. Crochet, A.R. Davies and K. Walters, Numerical simulation of Non-Newtonian Flow, Elsevier, 1984. [18] D.O. Olagunju, Asymptotic analysis of the finite cone-and-plate flow of a non-Newtonian fluid, J. Non-Newtonian Fluid Mech., 50 (1993) 289.

You might also like