Applied Mathematical Sciences, Vol. 2, 2008, no.

43, 2145 - 2156

Controllability of Non-Newtonian Fluids
under Homogeneous Extensional Flow
Lynda Wilson, Hong Zhou
1
, Wei Kang
Department of Applied Mathematics
Naval Postgraduate School, Monterey, CA 93943-5216, USA
Hongyun Wang
Department of Applied Mathematics and Statistics
University of California, Santa Cruz, CA 95060, USA
Abstract
In the discipline of non-Newtonian materials, the ability to control
viscoelastic stresses is very desirable in ascertaining important prop-
erties of the inuenced materials. We apply the nonlinear geometric
control theory to examine the controllability of various popular consti-
tutive models with imposed homogeneous extensional ow. The subse-
quent constitutive laws considered here include the Phan-Thien-Tanner
model, the Johnson-Segalman model, the Giesekus model and the Doi
model. This paper provides the rst analysis on the eect of extensional
ow on these models.
1 Introduction
For the design of certain materials, controllability of viscoelastic uids is
a signicant characteristic. In practice, the shape of an extrudate can be
controlled by varying the size or shape of an orice, whereas the advance of
the free surface can be controlled by varying the inow into a mold.
Previous theoretical studies from Renardy [[Renardy, 2005a], [Renardy, 2005b],
[Renardy, 2007]] investigated the controllability of ows of linear viscoelas-
tic uids for the multi-mode Maxwell models, the controllability of the ho-
mogeneous shear ow of viscoelastic uids with several dierent constitutive
models, and the controllability of nonhomogeneous shear ow of an upper
convected Maxwell uid. Very recent studies have examined the control-
lability of the upper convected Maxwell model under various homogeneous
1
Corresponding author, hzhou@nps.edu
2146 Lynda Wilson, Hong Zhou, Wei Kang and Hongyun Wang
ows [[Zhou et al., 2007]]. Furthermore, the reachable set for the upper con-
vected Maxwell model under imposed extensional ow was precisely specied
in [[Zhou et al., 2007]] even though it is usually very challenging for general
cases. In most of these studies the state of the system is characterized by the
viscoelastic stresses while the control input is in the form of the ow rate or
the body force.
The primary purpose of this paper is to further these earlier studies to
consider the controllability of three dierent model systems under the imposed
homogeneous extensional ow. Like shear ow, the extensional ow is also very
important in physical applications. For example, extensional ow can be used
to locally approximate the ow away from boundaries in extrusion manufactur-
ing or in thin lm and sheet manufacturing; an extension-dominated ow oc-
curs in industrial wire-coating processes along the wire-coating region beyond
the die. Following earlier studies, we restrict our attention to two-dimensional
models. This is mainly due to its mathematical simplicity and is also physi-
cally motivated by monolayer thin ims. Admittedly two-dimensional models
are ideal situations but we cant hope to understand the controllability of more
realistic cases unless we understand the controllability of simpler ones.
For readers convenience, we now present on overall basic denition and
description of weak controllability adopted from [[Isidori, 1995]].
Let
x = f(x) +
m

i=1
g
i
(x)u
i
(1)
be a general nonlinear control system, where x is the state variable, and u
i
R,
i = 1, , m, are the control variables. Let M denote the manifold of state
variables. A point x
1
in M is reachable from a point x
0
in M if there exist
piecewise continuous input functions, u
i
=
i
(t), such that the trajectory, x(t),
of (1) with initial state x
0
reaches x
1
in nite time. The global reachability is
usually hard to prove for nonlinear control systems. A feasible solution is to
seek weak controllability. The system (1) is weakly controllable within some
open subset S M if for each point x
0
S, there is an open neighborhood
U
0
of x
0
so that the set of points reachable from x
0
along trajectories inside
U
0
contains at least an open subset of M.
In (1), f(x) and g
i
(x) (i = 1, , m) are vector elds in M. Under the
Lie bracket operation, [f, g], the space of smooth vector elds on M forms a
Lie algebra. This Lie algebra, which is the smallest subalgebra containing the
vector elds f, g
1
, , g
m
, is called the Control Lie Algebra.
A powerful sucient condition on the weak controllability of a nonlin-
ear control system is the contollability rank condition (CRC) [[Isidori, 1995],
[Hermann and Krener, 1977]]. Namely, a control system of the form (1) is
weakly controllable on an open set S if it satises the controllability rank
Controllability of non-Newtonian uids 2147
condition on S, i.e.,
dim(
C
(x)) n (2)
for all x S, where n is the dimension of the manifold M, and

C
(x) = span{X(x)|Xis a vector eld in the control Lie algebra}.
Note that the big advantage of the CRC is that it is purely algebraic and it
does not require integrations of the dierential equations.
2 The Phan-Thien-Tanner Model
The Phan-Thien-Tanner (PTT) model is one of the most applied dieren-
tial type of the nonlinear viscoelastic constitutive equations. It contains two
parameters that control nonlinearity and has the ability to t, to some extent,
the shear and elongational properties of the viscoelastic materials indepen-
dently. Thus, the Phan-Thien-Tanner model is physically more realistic than
the constant shear viscosity Oldroyd-B model or upper-convected Maxwell
model.
The Phan-Thien-Tanner model has the constitutive equation
[[Phan-Thien and Tanner, 1977]]

T(v)TT(v)
T
+T+(trT)T = 2D, (3)
where T is the stress tensor, v is the velocity vector, v is the velocity gradient
tensor, is the relaxation rate, is a constant, the notataion tr stands for the
trace of the tensor, is the elastic modulus and D is the rate-of-deformation
tensor.
We consider 2-D homogeneous viscoelastic uids and denote the stress ten-
sor and the imposed extensional ow with rate (t) by
T =
_
T
11
T
12
T
12
T
22
_
, v = ( (t)
x
2
, (t)
y
2
), (4)
respectively, where T
11
is the rst normal stress dierence, T
22
the second
normal stress dierence and T
12
the shear stress. Suppose the control input is
denoted by which is closely related to the velocity. Then the general dynamic
problem of the PTT model (3) takes the form

T = F( (t), T), T(0) = T

0
, T(t
final
) = T
1
, (5)
where T
0
and T
1
are the given initial and nal states. The state of the system
(5) is characterized by viscoelastic stress T with three components T
11
, T
22
and T
12
.
2148 Lynda Wilson, Hong Zhou, Wei Kang and Hongyun Wang
For the extensional ow in (4), the velocity gradient is
v =

(t)
2
0
0
(t)
2

, (6)
which implies that the rate-of-strain tensor is
D =
1
2
[v + (v)
T
] =
(t)
2
_
1 0
0 1
_
. (7)
Applying the velocity gradient (6) and the rate-of-strain tensor (7) to the PTT
system (3), we have in component form that

T
11
= [ +(T
11
+T
22
)]T
11
+ ( +T
11
) (t)

T
22
= [ +(T
11
+T
22
)]T
22
( +T
22
) (t)

T
12
= [ +(T
11
+T
22
)]T
12
(8)
For mathematical convenience, the following notation is introduced:
x =

x
1
x
2
x
3

T
11
T
22
T
12

. (9)
Then the system (8) can be rewritten as
dx
dt
=

f(x) +g(x)u, (10)
where

f(x) = [ +(x
1
+x
2
)]x, g(x) =

+x
1
( +x
2
)
0

, u = (t). (11)
The rst Lie bracket is dened by
[

f, g] = g

f

f g =

x
1
(x
1
x
2
) +( +x
1
+x
2
)
x
2
(x
1
x
2
) ( +x
1
+x
2
)
(x
1
x
2
)x
3

. (12)
Similarly, the second Lie bracket is given by
[

f, [

f, g]] = [

f, g]

f[

f, g] =

x
1
(x
1
x
2
) +[ +(x
1
+x
2
)]
x
2
(x
1
x
2
) [ +(x
1
+x
2
)]
(x
1
x
2
)x
3

.
(13)
Controllability of non-Newtonian uids 2149
Utilizing the vectors from g(x), the two Lie brackets (12) and (13), one con-
tructs a matrix. By investigating the rank of this matrix, the weak control-
lability can be characterized. More specically, if the matrix is full rank or
nonsingular, then the system is weakly controllable; if the matrix is not full
rank or singular, then this approach oers no insight to the weak controllability
of the system and one has to seek more advanced tools.
We now compute the determinant and obtain
det
_
g, [

f, g], [

f, [

f, g]]
_
= 2x
3
[ +(x
1
+x
2
)](x
1
x
2
)
2
. (14)
The determinant does not equal zero when x
3
= 0, x
1
+x
2
= / and x
1
= x
2
.
It can be concluded that the system of the PTT model under extensional ow
satises the CRC and thus is weakly controllable when T
12
= 0, T
11
+ T
22
=
/ and T
11
= T
22
. This result can be summarized geometrically as follows.
Given R
3
= {(x
1
, x
2
, x
3
)|x
1
, x
2
, x
3
R}, there exist three surfaces:
S
1
= {(x
1
, x
2
, x
3
)|x
3
= 0}
S
2
= {(x
1
, x
2
, x
3
)| +(x
1
+x
2
) = 0}
S
3
= {(x
1
, x
2
, x
3
)|x
1
= x
2
}
such that the PTT model (3) under extensional ow is weakly controllable at
all points in R
3
\ {S
1
S
2
S
3
}.
3 The Johnson-Segalman model
The Johnson-Segalman model characterizes the behavior of non-Newtonian
uids, including special cases of Newtonian and Maxwell uids. Additionally,
it is a viscoelastic uid model which was developed to allow non-ane defor-
mations [[Johnson and Segalman, 1977]].
The Johnson-Segalman (JS) model gives a viscoelastic constitutive equa-
tion in the form

T
a + 1
2
_
(v)T+T(v)
T

a 1
2
_
(v)
T
T+T(v)

+T = 2D. (15)
Here a is a parameter describing polymer slip where 1 < a < 1; When a = 1,
the model (15) reduces to the Oldrody-B model.
Analogous to the previous model, we now apply the gradient of the homo-
geneous extensional ow (6) to the JS system (15) which leads to

T
11
= T
11
+ ( +aT
11
) (t)

T
22
= T
22
( +aT
22
) (t)

T
12
= T
12
(16)
2150 Lynda Wilson, Hong Zhou, Wei Kang and Hongyun Wang
Note that the third equation in (16) can be solved exactly such that T
12
=
T
12
(0)e
t
. Consequently, the Johnson-Segalman model (16) is not weakly
controllable under extensional ow. Nonetheless, the state space has a stable
invariant subspace T
12
= 0. It is in this subspace that all trajectories of the
system, under any control input, asymptotically move toward the subspace
T
12
= 0. As a result, the decisive behavior of the control system (16) can be
characterized by a reduced subsystem on this stable subspace:

T
11
= T
11
+ ( +aT
11
) (t),

T
22
= T
22
( +aT
22
) (t),
(17)
which can be put in a compact form
dx
dt
=

f(x) +g(x)u,
where
x =
_
x
1
x
2
_
=
_
T
11
T
22
_
,

f(x) = x, g(x) =
_
+ax
1
( +ax
2
)
_
, u = (t).
(18)
We compute the Lie bracket and nd
[

f, g] =
_

_
.
It follows immediately that
det
_
g, [

f, g]
_
= a(x
2
x
1
). (19)
The determinant does not vanish when x
1
= x
2
. Therefore, the subsystem
(17) of the Johnson-Segalman model under extensional ow satises CRC and
is weakly controllable when T
11
= T
22
. In geometrically terms, the system (17)
is weakly controllable at all points in R
2
\ S
1
where
R
2
= {(x
1
, x
2
)|x
1
, x
2
R}, S
1
= {(x
1
, x
2
)|x
1
= x
2
}.
In [[Zhou et al., 2007]] we have characterized the set of reachable states for
a subsystem of the upper-convected Maxwell model under extensional ow.
We have found that for the system

T
11
= ( (t))T
11
+ (t)

T
22
= ( + (t))T
22
(t)
(20)
Controllability of non-Newtonian uids 2151
the reachable set is R
1
+R
0
where R
1
and R
0
are given by
R
1
=

(T
11
(t
f
), T
22
(t
f
))

_
(
T
11
(t
f
)

+ 1)(
T
22
(t
f
)

+ 1) 1
> e
t
f
[
_
(
T
11
(0)

+ 1)(
T
22
(0)

+ 1) 1]

and
R
0
=

(T
11
(t
f
), T
22
(t
f
))

_
(
T
11
(t
f
)

+ 1)(
T
22
(t
f
)

+ 1) 1
= e
t
f
[
_
(
T
11
(0)

+ 1)(
T
22
(0)

+ 1) 1]
and T
11
(0) = T
22
(0), T
11
(t
f
) = T
22
(t
f
)

Note that if we redene the two parameters and (t) as

new
=

old
a
,
new
(t) =
a
old
(t), then the system (17) has the same form as (20). Therefore, the
reachable set for the subsystem of the Johnson-Segalman model (17) under
extensional ow is R
1
+R
0
where is replaced by /a.
4 The Giesekus Model
Another typical nonlinear viscoelastic uid model is the Giesekus model
[[Giesekus, 1982]]. The constitutive relation for the Giesekus model is given
by

T(v)TT(v)
T
+T+T
2
= 2D. (21)
Applying the gradient of the homogeneous extensional ow (6) to (21), we
obtain

T
11
= T
11
(T
2
11
+T
2
12
) + ( +T
11
) (t)

T
22
= T
22
(T
2
12
+T
2
22
) ( +T
22
) (t)

T
12
= [ +(T
11
+ T
22
)]T
12
(22)
which in turn can be put in the compact form
dx
dt
=

f(x) +g(x)u. (23)
Here
x =

x
1
x
2
x
3

T
11
T
22
T
12

,

f(x) =

x
1
(x
2
1
+x
2
3
)
x
2
(x
2
2
+x
2
3
)
[ +(x
1
+x
2
)]x
3

,
g(x) =

+x
1
( + x
2
)
0

, u = (t).
(24)
2152 Lynda Wilson, Hong Zhou, Wei Kang and Hongyun Wang
The two Lie brackets are computed as follows:
[

f, g] = g

f

f g =

(x
2
1
x
2
3
) + + 2x
1
x
2
(x
2
2) +x
2
3
(x
1
x
2
)x
3

, (25)
and
[

f, [

f, g]] = [

f, g]

f[

f, g] =

(x
2
1
x
2
3
)( + 2) +( + 2x
1
)
(x
2
2
x
2
3
)( 2) ( + 2x
2
)
x
3
(x
1
x
2
)(2 )

.
(26)
As a result, the determinant is found to be
det
_
g, [

f, g], [

f, [

f, g]]
_
= 2x
3
(x
1
x
2
)( )
2
( 2). (27)
Clearly, the determinant does not vanish when x
3
= 0, = , = 2 and
x
1
= x
2
. We conclude that the system of the Giesekus model under extensional
ow satises the CRC and thus is weakly controllable when T
12
= 0, T
11
= T
22
and the parameters satisfy the constraints = or 2. This result can be
summarized geometrically as follows.
Dene two surfaces
S
1
= {(x
1
, x
2
, x
3
)|x
3
= 0},
S
2
= {(x
1
, x
2
, x
3
)|x
1
= x
2
}.
Then the Giesekus model (21) under extensional ow becomes weakly control-
lable at all points in R
3
\ {S
1
S
2
} provided that the parameters satisfy the
constraints = or 2.
5 The Doi Model
The Doi model for rodlike liquid crystal polymers in a solvent is well-
known for its capability to describe both the isotropic and nematic phases
and phase transition between them. A fundamental element of the model
is the single molecule orientation distribution function. Interactions between
molecules are represented by a mean-eld potential. The rodlike molecules are
also subject to Brownian force due to the fact that they interact with other
rodlike molecules and with the ow. Generally, the model is a microscopic
Smoluchowski equation or Fokker-Planck type equation for the dynamics of the
orientational distribution function coupled with a macroscopic hydrodynamic
equation [[Doi and Edwards, 1986]]. The Smoluchowski equation depicts the
convection, rotation and diusion of the rodlike molecules.
Controllability of non-Newtonian uids 2153
The full Doi orientation tensor theory is developed after the kinetic Smolu-
chowski equation is projected onto a second-moment description using various
closure rules. The major ingredient in this tensor theory is the second-moment
tensor which describes the orientational distribution of the ensemble of rodlike
macromolecules. The orientation tensor is traceless and symmetric. The phys-
ical and practical signicance of the orientation tensor is that it is the basis
for micro-scale light scattering measurements of primary axes (directors),
degrees of molecular alignment (birefringence), and normal and shear stress
measurements. The study of two-dimensional liquid crystal polymers has been
physically inspired by monolayer lms. Thin lms of liquid crystal polymers
are used as alignment layers for liquid crystal displays because of their sta-
bility and nonlinear optical properties. A lot of theoretical and experimental
studies have been devoted to the two-dimensional Doi model (for example, see
[[Lee et al., 2006]] and references therein).
The two-dimensional Doi model is given by [[Lee et al., 2006]]

Q = QQ+a[DQ+QD] +aD2aD : Q(Q+

I
2
) 6D
r
F(Q), (28)
where Q is the orientation tensor such that
Q =
_
Q
11
Q
12
Q
21
Q
22
_
, Q
11
= Q
22
, Q
12
= Q
21
;
is the vorticity tensor such that =
1
2
[v v
T
]; a is a dimensionless
parameter which depends on the molecular aspect ratio; D is the rate-of-strain
tensor; D
r
is the rotary diusivity; F(Q) is dened by
F(Q) = (1
N
2
)QNQ
2
+NQ : Q(Q+
I
2
)
and N is a dimensionless concentration of nematic polymers.
As with the previous models, the gradient of the homogeneous extensional
ow (6) is applied to the Doi system (28), yielding

Q
11
= 6D
r
_
(1
N
2
)Q
11
+ 2NQ
11
(Q
2
11
+Q
2
12
)

+a(
1
2
2Q
2
11
) (t),

Q
12
= 6D
r
Q
12
_
(1
N
2
) + 2N(Q
2
11
+Q
2
12
)

2aQ
11
Q
12
(t).
(29)
Using the nematic relaxation time scale
1
Dr
, the ow eld and orientation dy-
namics of (28) can be non-dimensionalized. The key dimensionless parameters
are then the Peclet number Pe(t) = (t)/D
r
(the shear rate normalized with
respect to nematic relaxation rate) and the dimensionless concentration pa-
rameter N. Rescaling time as

t = tD
r
, the nematodynamic model (28) in the
dimensionless form becomes

Q
11
= 6Q
11
_
1
N
2
+ 2N(Q
2
11
+Q
2
12
)

+a(
1
2
2Q
2
11
)Pe(t),

Q
12
= 6Q
12
_
1
N
2
+ 2N(Q
2
11
+Q
2
12
)

2aQ
11
Q
12
Pe(t).
(30)
2154 Lynda Wilson, Hong Zhou, Wei Kang and Hongyun Wang
The system (30) can be written as
dx
dt
=

f(x) +g(x)u,
where
x =
_
x
1
x
2
_
=
_
Q
11
Q
12
_
,

f(x) = 6
_
1
N
2
+ 2N(x
2
1
+x
2
2
)

x,
g(x) =

a(
1
2
2x
2
1
)
2ax
1
x
2

, u = Pe(t).
(31)
The Lie bracket is calculated to be
[

f, g] =

6a
_
2x
2
1
[1
N
2
2N(x
2
1
+x
2
2
)] +
1
2

N
4
+N(3x
2
1
+x
2
2
)
_
12ax
1
x
2
[1 +
N
2
2N(x
2
1
+x
2
2
)]

.
Finally, we have
det
_
g, [

f, g]
_
= 12a
2
x
1
x
2
. (32)
The determinant does not equal zero when x
1
= 0 and x
2
= 0. So the Doi
model (30) under extensional ow satises CRC and is weakly controllable
when Q
11
= 0 and Q
12
= 0. In terms of geometric words, the system (30) is
weakly controllable at all points in R
2
\ {S
1
S
2
} where
R
2
= {(x
1
, x
2
)|x
1
, x
2
R}, S
1
= {(x
1
, x
2
)|x
1
= 0}, S
2
= {(x
1
, x
2
)|x
2
= 0}.
6 Concluding Remarks
We have applied the controllability rank condition to the vector elds in the
Phan-Thien-Tanner model, the Johnson-Segalman model, the Giesekus model
and the Doi model to study the controllability of non-Newtonian uids driven
by imposed homogeneous extensional ow. In our control system, the state
variable is the stress for the Phan-Thien-Tanner model, the Johnson-Segalman
model and the Giesekus model and the orientation tensor for the Doi model.
The extensional ow rate corresponds to the available control. We have derived
sucient conditions for the weak controllability of each model.
Acknowledgment
This research was supported in part by the Air Force Oce of Scientic
Research and the National Science Foundation.
Controllability of non-Newtonian uids 2155
References
[Bird et al., 1987] [1] R. B. Bird, R. C. Armstrong and O. Hassager, Dynamics
of Polymeric Liquids, 2nd edition, Wiley (1987).
[Doi and Edwards, 1986] [2] M. Doi and S. F. Edwards, The Theory of Poly-
mer Dynamics, Oxford University Press, New York, 1986.
[Giesekus, 1982] [3] H. Giesekus, A unied approach to a variety of consti-
tutive models for polymer uids based on the concept of conguration
dependent molecular mobility, Rhelo. Acta 21 (1982), 366-375.
[Hermann and Krener, 1977] [4] R. Hermann and A. Krener, Nonlinear con-
trollability and observability, IEEE Trans. Automat Contr. AC-22
(1977), 728-740.
[Isidori, 1995] [5] A. Isidori, Nonlinear Control Systems, 3rd edition, Springer
(1995).
[Johnson and Segalman, 1977] [6] M. W. Johnson and D. Segalman, A model
for viscoelastic uid behavior which allows non-ane deformation, J.
Non-Newtonian Fluid Mech. 2 (1977), 255-270.
[Larson, 1998] [7] R. G. Larson, The Structure and Rheology of Complex Flu-
ids, Oxford (1998).
[Lee et al., 2006] [8] J. H. Lee, M. G. Forest and R. Zhou, Alignment and
rheo-oscillator criteria for sheared nematic polymer lms in the mono-
layer limit, Discrete and continuous dynamical systems-series B 6
(2006), 339-356.
[Nijmeijer and van der Shaft, 1990] [9] H. Nijmeijer and A. J. van der Schaft,
Nonlinear Dynamical Control Systems, Springer (1990).
[Phan-Thien and Tanner, 1977] [10] N. Phan-Thien and R. I. Tanner, A new
constitutive equation derived from network theory, J. Non-Newtonian
Fluid Mech. 2 (1977), 353-365.
[Renardy, 2005a] [11] M. Renardy, Are viscoelastic ows under control or out
of control? Systems & Control Letters 54 (2005a) 1183-1193.
[Renardy, 2005b] [12] M. Renardy, Shear ow of viscoelastic uids as a con-
trol problem, J. Non-Newtonian Fluid Mech. 131 (2005b) 59-63.
[Renardy, 2007] [13] M. Renardy, On control of shear ow of an upper con-
vected Maxwell uid, Z. Angew. Math. Mech. 87 (2007) 213-218.
2156 Lynda Wilson, Hong Zhou, Wei Kang and Hongyun Wang
[Zhou et al., 2007] [14] H. Zhou, W. Kang, A. Krener and H. Wang, Homo-
geneous ow eld eect on the control of Maxwell materials, J. Non-
Newtonian Fluid Mech. 150 (2008) 104-115.
Received: January 26, 2008