Arch. Rational Mech. Anal.

154 (2000) 135156

Digital Object Identier (DOI) 10.1007/s002050000102
Existence of Solutions
for the Ericksen-Leslie System
Fang-Hua Lin & Chun Liu
Communicated by D. Kinderlehrer
Abstract
We study the general Ericksen-Leslie system, which describes the owof liquid
crystal materials. The dissipation property of the system is established and is used
to prove the global existence of weak solutions. We also study the existence of
classical solutions and the asymptotic stability of the solutions.
1. Introduction
In many experiments and earlier theoretical works on nematic liquid crystals,
the samples are treated as consisting of slowmoving particles. Hence one approach
is to study the behavior of the director eld n in the absence of the velocity elds.
Unfortunately, the ow velocity does disturb the alignment of the molecules. More
importantly, the converse is also true, that is, a change in the alignment will induce
velocity. This velocity will in turn affect the time evolution of the director eld. In
this process, we cannot assume that the velocity eld will remain small even when
we start with zero velocity eld.
The hydrodynamic theory of liquid crystals was established by Ericksen [E1,
E2] and Leslie [Le] in the 1960s. The Ericksen-Leslie system consists of the
following equations, see [Le][dG]:

t
+(v ) = 0, (1.1)

dv
dt
= F
i
+
ji,j
, (1.2)

1
d
i
dt
=
1
G
i
+g
i
+
ji,j
, (1.3)
which include the uid density , the velocity vector v = (v
1
, v
2
, v
3
) and direction
vector d = (d
1
, d
2
, d
3
). The equations represent the conservation of mass, linear
momentum and angular momentum respectively. We can see that the anisotropic
136 Fang-Hua Lin & Chun Liu
feature of liquid crystal materials is exhibited in (1.3) and its nonlinear coupling in
(1.2). The velocity v also satises the incompressible condition:
v
i,i
= 0. (1.4)
In the above system, we have the following constitutive relations:

ji
= P
ij

F
d
k,j
d
k,i
+

ji
, (1.5)

ji
=
j
d
i
+
F
di, j
, (1.6)
g
i
= d
i

j
d
i,j

F
d
i
+ g

i
, (1.7)
where F
i
is the external body force, G
i
is the external director body force, and ,
come from the restriction of the direction vector |d| = 1. Also
2F = k
22
d
i,j
d
i,j
+(k
11
k
22
k
24
)d
i,i
d
j,j
+(k
33
k
22
)d
i
d
j
d
k,i
d
k,j
+k
24
d
i,j
d
j,i
, (1.8)

ji
=
1
d
k
d
p
A
kp
d
i
d
j
+
2
d
j
N
i
+
3
d
i
N
j
+
4
A
ij
+
5
d
j
d
k
A
ki
+
6
d
i
d
k
A
kj
, (1.9)
g

i
=
1
N
i
+
2
d
j
A
ji
, (1.10)
and

i
=

d
i
=
d
i
t
+(v )d
i
, (1.11)
N
i
=
i
+
ki
d
k
, (1.12)
N
ij
=
i,j
+
ki
d
k,j
, (1.13)
where
2A
ij
= v
i,j
+v
j,i
, 2
ij
= v
i,j
v
j,i
. (1.14)
In the above system, the constants have to satisfy the following (see Leslie
[Le]):

1
=
2

3
,
2
=
5

6
= (
2
+
3
). (1.15)
The above relation was achieved from the hydrodynamic point of view in order
to guarantee the entropy condition, that is, the second law of thermodynamics. The
last part of it,
2
= (
2
+
3
), is called Parodis condition. It is derived from the
Onsager reciprocal relation.
One of the methods used to relax the nonlinear constraint |d| = 1 is to consider
a form of penalization, that is, instead of the restriction |d| = 1, we add the term
1

2
(|d|
2
1)
2
in F. After we discuss the cases for each , we then take the limit of
0. This technique has been successfully used in a lot of other places (see [CS,
LL1, LL2]). It also ts well with the general theory of Landaus order parameter
(cf. [E3, Le]).
Existence of Solutions for the Ericksen-Leslie System 137
To further simplify the calculation, we take
1
= 0,
j
= 0, = 0, F
i
= 0,
= constant and F = |d|
2
+
1

2
(|d|
2
1)
2
. The assumption that F
i
= 0 is
equivalent to the assumption that the exterior forces are conservative. The choice of
F here corresponds to the equal constants case for the static Oseen-Frank energy.
Of course we can have more general Oseen-Frank energy and the treatment will be
exactly the same.
Taking into consideration the above assumptions, the system (1.1)(1.4) be-
comes:

dv
i
dt
=
ji,j
,
g
i
+
ji,j
= 0,
v
i,i
= 0,
(1.16)
and we will still have the following:

ji
= P
ij

F
d
k,j
d
k,i
+

ji
,

ji
=
F
d
i,j
,
g
i
=
F
d
i
+ g

i
,
2F = |d|
2
+
1

2
(|d|
2
1)
2
,

ji
=
1
d
k
d
p
A
kp
d
i
d
j
+
2
d
j
N
i
+
3
d
i
N
j
,
+
4
A
ij
+
5
d
j
d
k
A
ki
+
6
d
i
d
k
A
kj
,
g

i
=
1
N
i
+
2
d
j
A
ji
(1.17)
together with the condition (1.11)(1.15).
Remark. The constant in (1.7) is the Lagrange multiplier used in the original
system to satisfy the constraint |d| = 1. So it is natural to set it to 0 since we are
using the penalization argument. The advantage of this argument is the reduction of
the degree of nonlinearity and simplication of the proof of the existence theorems.
Moreover, here we use the condition k
ii
= 1 for convenience in notation.
The above system was derived from the macroscopic point of view and was
very successful in understanding the coupling between the director eld and the
velocity eld, especially in liquid crystals of nematic type. It also provides a tool to
describe the motion of the defects. However, the model is so complicated that only
some special cases have been treated theoretically or numerically. see [Le, dG].
In [LL1], we studied the following simplied system with suitable boundary and
initial condition: Suppose that is a domain in R
3
or R
2
with smooth boundary
and
v
t
+(v )v v +P = (d d),
d
t
+(v )d = (d f (d)),
v =0,
(1.18)
138 Fang-Hua Lin & Chun Liu
with the boundary condition
v(x, t ) = 0, d(x, t ) = d
0
(x), for (x, t) (0, ), (1.19)
and with initial data
v(x, 0) = v
0
(x) with v
0
= 0, d(x, 0) = d
0
(x), for x . (1.20)
The rst equation is in fact the equation for the conservation of linear momen-
tum. Most of it is exactlythe same as the Navier-Stokes equationwhichdescribes the
ow of a normal isotropic uid. Note that, for the general Ericksen-Leslie system,
this part will also be anisotropic. The last term((dd))
i
=
j
(
i
d
k

j
d
k
)
indicates also the anisotropic property of the system. The third equation represents
the property of incompressibility of the liquid, while the second equation is associ-
ated with conservation of the angular momentum. The term f (d) =
1

2
(|d|
2
1)d
is simply the Ginzburg-Landau approximation of the constraint |d| = 1 for small .
There are many interesting features of this system. For instance, the director
eld satises the maximum principle, that is, it will not achieve maximal vector
length in any interior point of the space-time domain. If we plug a semi-stationary
solution (d, 0) into the system, we nd that d not only needs to satisfy the second
equation
d
t
= (d f (d)) (1.21)
but also an extra constraint dd = 0. Not all weak solutions of the second
equation, particularly in the 0
+
limit, will satisfy this constraint. This in
fact requires more regularity of d, that is, d has to satisfy the energy-monotonicity
inequality. In the static limiting case (harmonic maps), this is called a stationary
harmonic map by Schoen & Uhlenbeck [SU]. Although the above system is
highly simplied compared to the system (1.16), (1.17), it captured some of the
basic properties of the original model. Among them, the most important one is the
following so called Basic Energy Inequality:
1
2
d
dt
_

(|v|
2
+|d|
2
+2F(d)) dx =
_

(|v|
2
+ |d f (d)|
2
) dx
(1.22)
for all t (0, T ].
We [LL1] were able to prove local existence of the classical solutions and global
existence of the weak solutions in spaces L
2
(0, T ; H
1
()) L

(0, T ; L
2
()) for
v and L
2
(0, T ; H
2
()) L

(0, T ; H
1
()) for d. For any xed , it was also
shown that the global classical solution exists in the case where the uid viscosity
is large. In the general case, we [LL2] proved that the one dimensional space-
time Hausdorff measure of the singular set of the suitable weak solutions is zero.
Moreover, Liu & Walkington established a numerical code using nite element
methods to study the simplied system [LW]. Many new features, especially the
interaction of the defects and the owelds, were demonstrated in their simulations.
In this paper we establish similar results to the Ericksen-Leslie system (1.16),
(1.17) with the same boundary and initial condition (1.19), (1.20). First, we establish
Existence of Solutions for the Ericksen-Leslie System 139
the energy law for the system. Then we study the existence of the weak and strong
solutions under some conditions.
The paper is written in the following way: in Section 2, we establish the basic
energy inequality. Using this energy law, in Section 3, we establish the existence
of the global weak solution in the case
2
= 0. This condition can be relaxed
somewhat, see the discussion in Section 6. In Section 4, we prove the local existence
of the classical solutions and the global existence under some conditions on the
viscosity constants. The method involves using a higher order energy estimate. In
Section 5, we study the asymptotic behavior of the weak solutions as the time t goes
to innity. We also establish the eventual regularity of the solutions. We discuss
the material constants in Section 6. Finally we end the paper with the result on the
convergence of the penalized problem in Section 7.
2. Basic Energy Inequality
In this section, we formally derive the energy inequality. Throughout this sec-
tion, we will assume the solution (v, d) is smooth. This will allow us to take all the
derivatives and we will adjust the assumption in the later sections.
Theorem 2.1. Suppose that (v, d) is a smooth solution of the system(1.16), (1.17)
with the initial and boundary conditions (1.19) and (1.20). If at the same time

1
=
2

3
,
2
=
5

6
= (
2
+
3
), (2.1)
and also
1
0, then the ow satises the following dissipative relation:
d
dt
_

2
|v|
2
+
1
2
|d|
2
+F(d)
_
dx
_

1
(A
kp
d
k
d
p
)
2
+
4
(A
ij
)
2
_
dx

(
5
+
6
)|Ad|
2
dx
+
_

1
|N|
2
dx (2.2)
+
_

2
2
N
j
A
i,j
d
i
dx.
Remark. The second part of the condition
5

6
= (
2
+
3
) was rst found
by Parodi in the 1970s. The assumption
1
0 has also been used by Leslie &
Ericksen, see [Le, E3]. It provides the necessary condition for the dissipation in
the direction eld.
Proof. We multiply the rst equation of (1.16) by v and integrate over the domain
. We have:

2
d
dt
_

|v|
2
dx =
_

ji,j
v
i
dx =
_

d
k

i
d
k
v
i

ji
v
j,i
dx.
Here we use integration by parts and the fact that v is divergence free. Notice also
that v vanishes on the boundary of .
140 Fang-Hua Lin & Chun Liu
Then we substitute the expression of

ji
from (1.7), and use the fact that A
ij
=
A
ji
, to give

2
d
dt
_

|v|
2
dx =
_

d
k

i
d
k
v
i
dx
_

(
1
(A
kp
d
k
d
p
)
2
+
4
(A
ij
)
2
) dx

2
d
j
N
i
v
i,j
+
3
d
i
N
j
v
i,j
_
dx

5
d
j
d
k
A
ki
v
i,j
+
6
d
i
d
k
A
kj
v
i,j
_
dx
(2.3)
We rewrite the second equation of (1.16) as
(
1
d
i,t
+
1
v
k
d
i,k
+
1

ki
d
k
+
2
A
ji
d
j
) = d
i
f
i
(d), (2.4)
where f
i
(d) =
i
(F(d)) =
1
4
2
(1 +|d|
2
)d
i
.
Next, we multiply (2.4) by
1

1
d and integrate over the domain . Using
integration by part again, we get
1
2
d
dt
_

|d|
2
dx =
_

(v
k
d
i,k
d
i
+
ki
d
k
d
i
+

2

1
A
ji
d
j
d
i
+
1

1
(d
i
f
i
(d))d
i
) dx.
(2.5)
Again we need to point out that both v and d
t
vanish on the boundary.
We also have the following time derivative of the integral of F(d):
d
dt
_

F(d) dx =
_

f
i
(d)d
i,t
dx
=
_

_
f
i
(d)v
k
d
i,k
f
i
(d)
ki
d
k


2

1
f
i
(d)A
ji
d
j

1
(d
i
f
i
(d))f
i
(d)
_
dx
(2.6)
and we nd that the rst term in (2.6) is equal to zero because of the fact that v is
divergence free;
_

f
i
(d)v
k
d
i,k
dx =
_

v
k

k
F(d) dx = 0.
Existence of Solutions for the Ericksen-Leslie System 141
Finally, we add (2.3), (2.5) and (2.6) together:
d
dt
_

2
|v|
2
+
1
2
|d|
2
+F(d)
_
dx
=
_

1
(A
kp
d
k
d
p
)
2
+
4
(A
ij
)
2

1
(d
i
f
i
(d))
2
_
dx
+
_

_
(d
i
f
i
(d))
ki
d
k
+

2

1
(d
i
f
i
(d))A
ji
d
j
_
dx

2
d
j
N
i
v
i,j
+
3
d
i
N
j
v
i,j
_
dx

5
d
j
d
k
A
ki
v
i,j
+
6
d
i
d
k
A
kj
v
i,j
_
dx.
(2.7)
Now we estimate (2.7) term by term:
_

_
(d
i
f
i
(d))
ki
d
k

_

2
d
j
N
i
v
i,j
+
3
d
i
N
j
v
i,j
__
dx
=
_

_
(d
i
f
i
(d))
ki
d
k
+(
2

3
)d
i
N
j
v
i,j
2
2
d
j
N
i
A
i,j
_
dx
=
_

_
(d
i
f
i
(d))
ki
d
k

1
N
j
(A
ij
+
ij
)d
i
2
2
d
j
N
i
A
i,j
_
dx.
Here we have used (1.14) and (1.15).
Now we make use of (2.4) to give
_

_
(d
i
f
i
(d))
ki
d
k
(
2
d
j
N
i
v
i,j
+
3
d
i
N
j
v
i,j
)
_
dx
=
_

_
(d
i
f
i
(d))
ki
d
k
+
_

2
A
ji
d
j
(d
i
f
i
(d))
_

ij
d
j
+(
1
2
2
)d
i
N
j
A
i,j
)
_
dx
=
_

(
2
A
ji
d
j

ij
d
j
(
2
+
3
)d
i
N
j
A
i,j
)dx.
(2.8)
Notice we have used the fact that
ij
=
ji
.
The other terms in (2.7) can be grouped in the following way:

(
5
d
j
d
k
A
ki
v
i,j
+
6
d
i
d
k
A
kj
v
i,j
) dx
=
_

_
(
5

6
)d
j
d
k
A
ki
v
i,j
2
6
d
i
d
k
A
kj
A
ij
_
dx
=
_

2
(A
ij
+
ij
)A
ki
d
j
d
k
2
6
d
i
d
k
A
kj
A
ij
_
dx
=
_

2
|Ad|
2

2
A
ji
d
j

ij
d
j
2
6
|Ad|
2
_
dx.
(2.9)
Here again we have used the relation (1.15).
142 Fang-Hua Lin & Chun Liu
From the above, we see that the terms
2
A
ji
d
j

ij
d
j
in (2.8) and (2.9) will
cancel each other. So what is left in (2.7) is:
d
dt
_

2
|v|
2
+
1
2
|d|
2
+F(d)
_
dx
=
_

1
(A
kp
d
k
d
p
)
2
+
4
(A
ij
)
2

1
(d
i
f
i
(d))
2
_
dx
+
_

1
(d
i
f
i
(d)) A
ji
d
j
_
dx

_
(
2
+
3
)N
j
A
i,j
d
i
+(
5
+
6
)|Ad|
2
_
dx.
(2.10)
Applying equation (2.4) and the Parodis relation, we have
1

1
(d
i
f
i
(d))
2
+

2

1
(d
i
f
i
(d)) A
ji
d
j
(
2
+
3
) N
j
A
i,j
d
i
=
1

1
(d
i
f
i
(d))
_
d
i
f
i
(d) +
2
A
ji
d
j
_
(
2
+
3
)N
j
A
i,j
d
i
= (d
i
f
i
(d)) (N
i
) +
2
N
j
A
i,j
d
i
= N
i
(
1
N
i
) +2
2
N
j
A
i,j
d
i
=
1
|N|
2
+2
2
N
j
A
i,j
d
i
.
Since
1
0, we prove the theorem,
d
dt
_

2
|v|
2
+
1
2
|d|
2
+F(d)
_
dx

_

1
(A
kp
d
k
d
p
)
2
+
4
(A
ij
)
2
_
dx
_

(
5
+
6
)|Ad|
2
dx
+
_

1
|N|
2
dx +
_

2
2
N
j
A
i,j
d
i
dx. (2.11)
Remark. From the above calculation, we can see that the relation (1.15)

1
=
2

3
,
2
=
5

6
has a new meaning, it actually gives the existence of the Lyaponov-type function.
However, the role of the Parodis condition is not necessary. We will see the usage
of Parodis condition again in Section 4 when we prove the existence of the classical
solutions.
Remark. In the case that the density is not a constant, with (1.1), the above
energy law (2.11) still holds. See [Liu].
Existence of Solutions for the Ericksen-Leslie System 143
3. Weak Solutions
One consequence of the penalization termis that we are able to reduce the order
of nonlinearity in (2.4). However, at the same time we lose the control of the norm
of the vector d. In the case of harmonic maps and the simplied system (1.18), we
can regain this by the maximal principle [LL1, CS].
For system(1.16), (1.17), this maximal principle only holds under the condition

2
= 0. Although the physical meaning of this relation is unclear, at least it is very
interesting from the analysis point of view.
Throughout this section and next section, we will also assume that
1
,
4
and

5
+
6
are positive numbers,
1
< 0 and
2
= 0. Some discussion of how to
relax these conditions on the constants can be found in Section 6.
Theorem 3.1. In the case of
2
= 0, the systemof (1.16), (1.17) obeys the maximal
principle, that is: if |d
0
| 1 on the boundary of the domain, then |d| 1 in the
whole domain.
The proof of the theorem is obvious from (2.4), by multiplying the equation by
d. An immediately consequence of the theoremis that we can get all the dissipation
we need by letting
4
, which is the viscosity of the uid, be large.
In order to discuss the existence of the weak solutions, we dene the following
spaces:
H
1
0
() = closure of C

0
(, R
3
) in the norm
_
_

|v|
2
dx
_1
2
,
H
1
() = the dual of H
1
0
(),
H
2
() = {v L
2
(, R
3
) : v
x
i
, v
x
i
x
j
L
2
(, R
3
), for 1 i, j 3},
V = C

0
(, R
3
) {v : divv = 0},
H = closure of V in L
2
(, R
3
),
V = closure of V in H
1
0
(),
V

= the dual of V,
where is a bounded smooth domain in R
3
, or R
2
.
We proceed to prove the following existence theorem for the weak solutions.
Theorem 3.2. Under the assumptions that v
0
(x) L
2
, and that d
0
(x) H
1
()
with d
0
|

H
3/2
(), the initial-boundary problem (1.19), (1.20) of system
(1.16), (1.17) has a global weak solution (v, d) such that
v L
2
(0, T ; H
1
()) L

(0, T ; L
2
())
d L
2
(0, T ; H
2
()) L

(0, T ; H
1
())
(3.1)
for all T (0, ).
144 Fang-Hua Lin & Chun Liu
We can use a Galerkin procedure, originally used for the standard Navier-Stokes
equation but modied for the different coupling systems [LL1]. Here we want to
use a slightly different scheme that was used by Lions in [Lions] and Liu [Liu].
Since space V is separable, we can consider V
m
V which are m dimensional
subspace with all components in C
1
. Also we have: for all v V there exists a
sequence v
m
V
m
such that v
m
v.
We will look for the approximate solutions such that for any t (0, T ), v
m

V
m
,
(v
mt
, v) +((v
m
)v
m
, v) = ((d d) v) (

, v) (3.2)
for any v V
m
. Here

is dened as in (1.9) with d and v being replaced by d

m
and v
m
.
Moreover, the equation of d will remain the same as (2.4):

1
d
mt

1
(v
m
)d
m

1

m
d
m
= d
m
F(d
m
). (3.3)
We will try to solve (3.2) and (3.3) with initial conditions:
v
m
(x, 0) = v
m0
(x), d
m
(x, 0) = d
0
(x), (3.4)
where v
m0
is the projection of v
0
in V
m
which guarantees that v
m0
= 0 and the
boundary condition:
v
m
(x.t ) = 0, d
m
(x, t ) = d
0
(x), (3.5)
where x .
System (3.2)(3.5) is a nonlinear differential system. We notice rst that for
given v
m
L
2
(0, T ; H
1
()) L

(0, T ; L
2
()) we can solve d
m
L
2
(0, T ;
H
2
()) L

(0, T ; H
1
()) locally in time because (3.3) is a parabolic equa-
tion. Second, we can plug this d
m
back into (3.2) and get a nonlinear differ-
ential equation for v
m
, and this equation will have local existence in space of
L
2
(0, T ; H
1
()) L

(0, T ; L
2
()). A xed point theorem argument will lead to
the following lemma. For details, we refer to Lemma 2.2 in [LL1].
Lemma 3.1. There exists a T
0
> 0 depending on the initial and boundary condition
(3.4), (3.5) and the domain such that there is a weak solution of (3.2) and (3.3).
Moreover, the solution is smooth, hence the weak solution is unique.
Remark. The smoothness of the approximate solutions comes from the regularity
of the parabolic equations (see [LSU]) and a bootstrapping argument (see [LL1]).
The proof of the uniqueness of the approximate solutions by the energy estimate of
the difference between two different solutions is straightforward (see also [LL1]).
The following standard a priori estimates will be sufcient to prove the global
existence of system (3.2)(3.5).
Existence of Solutions for the Ericksen-Leslie System 145
Lemma 3.2. Let v
m
and d
m
be the solution we get from the previous argument,
then
d
dt
_

2
|v
m
|
2
+
1
2
|d
m
|
2
+F(d
m
)
_
dx

_

1
(A
mkp
d
mk
d
mp
)
2
+
4
(A
mij
)
2
_

(
5
+
6
)|A
m
d
m
|
2
dx +
_

1
|N
m
|
2
dx
for all t (0, T
0
]. In particular, one has
sup
0t T
0
_

2
|v
m
|
2
+
1
2
|d
m
|
2
_
dx
+
_
T
0
0
_

_
(
1
(A
mkp
d
mk
d
mp
)
2
+
4
(A
mij
)
2
_
dx
+
_
T
0
0
_

_
(
5
+
6
)|A
m
d
m
|
2
+
1
|N
m
|
2
_
dx
C
_
v
0
, d
0
, T
0
, , |F|
L

()
, |f |
L

()
_
.
(3.6)
The proof of this lemma follows from replacing v be v
m
in (3.2), then mul-
tiplying (3.3) by
1

1
d
m
, integrating by parts, and nally adding the results and
using the same computation as in Section 2. We want to point out that since the
approximate solution (v
m
, d
m
) is smooth, all the steps, especially those involving
integration by parts, will be rigorous.
Combining Lemma 3.1 and Lemma 3.2 gives us the global existence of a weak
solution of the nite dimensional approximation system (3.2)(3.5).
Lemma 3.3. For any m > 0, T (0, ), and smooth enough initial and boundary
conditions v
0
and d
0
, the problem(3.2)(3.5) possesses a unique classical solution
v
m
and d
m
in Q
T
.
Here Q
T
represents the parabolic cylinder {(x, t ) : x , t (0, T )}.
Next, we can take a subsequence if necessary and get:
(1) v
m
v, weakly in L
2
(0, T ; V), weak-star in L

(0, T ; H).
(2) d
m
d, weakly in L
2
(0, T ; H
2
), weak-star in L

(0, T ; H
1
).
By Theorem 3.1, we can see that in order to prove v and d is a weak solution
of the original system (1.16), (1.17), we only need to prove:
(a) v
m
v strongly in L
2
(0, T ; H);
(b) d
m
d strongly in L
2
(0, T ; H
1
).
This canbe done inmanyways byusingdifferent types of compactness theorems
(see, for example, [T] Chapter 3, Section 2) by using Fourier transform or [Lions]
by using time-difference quotients. See also [LL1, Liu] for some other applications.
(a) and (b) will provide the convergence of the term vv and dd in (1.16) and
(1.17) respectively.
All these give the proof of Theorem 3.2.
146 Fang-Hua Lin & Chun Liu
Remark. We want to point out that even with the material density not a constant,
with an extra equation (conservation of mass) (1.1), we can still prove the above
results using the method in [Lions, Liu].
4. Strong Solutions
For convenience of notation, we will set = 1 in this section. Remember we
still have
2
= 0.
As we did in [LL1], by looking at the scaling property of the system(1.6), (1.7),
we see that one can assign a scaling dimension to each quantity as follows: each
x
i
has dimension 1, time variable t has dimension 2, v has dimension 1, P has
dimension 2, d has dimension 0, f (d) has dimension 2,

x
i
has dimension 1
and

t
has dimension 2. And the scaling dimension order of the basic energy law
is 2. The scaling property is very important to the different energy inequalities
and the discussion of the partial regularity property of the solutions.
In the previous section, we proved that the weak solutions of the general
Ericksen-Leslie equation exist in time. Due to the fact that (1.16) has Navier-Stokes
equation as its subsystem, we cannot expect to prove the global existence of the
classical solutions at this stage. Instead, we will prove the following theorems.
Theorem 4.1. If dim = 2 and v
0
(x) H
1
(), d
0
(x) H
2
(), the problem
(1.6), (1.7) has a unique global classical solution (v, d).
Remark. In fact we only need to prove that
v L

(0, T ; H
1
()), d L

(0, T ; H
2
()). (4.1)
Then by the Sobolev embedding theorem and the regularity results of Navier-
Stokes equations [Se] and parabolic equations ([LSU] Chapter 3, Section 10), and
a bootstrapping argument, we can proof (v, d) is in fact a classical soliton. For detail,
see [LL1] (See also the remark after the proof of Theorem 3.5 and the remark after
Theorem 4.1). We also want to point out that the solution obtained will be unique.
The method here is to nd a new energy inequality involving the following
quantities:
A
2
(t ) =
_

_
|v|
2
+|d f (d)|
2
_
dx. (4.2)
By the scaling argument used before, we can see this new energy is of higher order
than those in (2.11) and the same as the terms on the right-hand side of (2.11).
Using the same method as in Section 2, together with the relation (1.15), we
calculate the following estimates:
1
2
d
dt
A
2
=
_

_
(v, v
t
) +(d f, d
t
f

(d)d
t
)
_
dx
=
_

1
(
q
A
kp
d
k
d
p
)
2
+
4
(
k
A
ij
)
2

1
((d
i
f
i
(d)))
2
R
_
dx
Existence of Solutions for the Ericksen-Leslie System 147

(
5
+
6
)|Ad|
2
dx +
_

_
(d
i
f
i
(d))(
ij
d
j
)
_
dx
+
_

_
(
2
+
3
)d
i
N
j
A
ij
_
dx +
_

_
(
2

3
)d
i
N
j

ij
_
dx
+
_

(c
1
|Add||Add| +c
2
|Ad||Ad| (4.3)
+c
3
_
|
2
v||d||(d
i
f
i
(d))| +|v||d||(d
i
f
i
(d))|
_
+c
4
|(d
i
f
i
(d))||A|d|) dx
+
_
_
vvv +vd (d f (d))
+(d f (d))(f

(d))(d f (d) vd)

+(d f (d)) (vd v (d) 2v
2
d)
_
dx.
Next, we use the relation (1.15) in Theorem 2.1,
1
=
2

3
to cancel the
terms involving the derivatives of
ij
, . The relation
2
=
2
+
3
= 0 allows
us to cancel the term with the highest order derivatives, that is, the termd
i
N
j
A
ij
.
From here, we can see that Parodis condition is necessary to establish the higher
order energy law.
With the assumption that v = d
t
= 0 on the boundary of , which is the reason
for the choice of the term A, and the fact that v is divergence free, we have that
(4.3) is equal to (see [LL1])
1
2
d
dt
A
2
=
_

_
(v, v
t
) +(d f, d
t
f

(d)d
t
)
_
dx
=
_

1
(
q
A
kp
d
k
d
p
)
2
+
4
(
k
A
ij
)
2

1
((d
i
f
i
(d)))
2
_
dx

(
5
+
6
) |Ad|
2
dx+
_

_
c
1
|Add||Add|+c
2
|Ad||Ad|
+c
3
_
|
2
v||d|| (d
i
f
i
(d)) | +|v||d|| (d
i
f
i
(d)) |
_
+c
4
|(d
i
f
i
(d)) Ad|
_
dx (4.4)
+
_
_
vvv f

(d)(d f (d)) (d f (d))

_
dx
+2
_

(d f (d))
x
j
v
j
x
k
d
x
k
dx.
Now we only need to estimate (4.4) term by term. We have
_

vvv dx v
L
2 v
L
4 v
L
4 Cv
3
2
L
2
v
L
2
v
2
L
2
+
C

3
v
4
L
2
.
(4.5)
148 Fang-Hua Lin & Chun Liu
All the other terms are of the same or lower scaling order than this term, except
the last one, which can be established as (see [LL1]):
_

(d f (d))
x
j
v
j
x
k
d
x
k
dx
(|(d f (d))|
2
L
2
+|v|
2
) +
C

3
|v|
2
(|d f (d)|
2
+1).
Here we use the interpolating inequality for L
4
in the 2-dimensional case. All
the other terms can be estimated in the same way. After all, by taking sufciently
small in above calculations, we arrive at
d
dt
A
2
(t ) C
1
A
4
(t ) +C
2
A
2
(t ) (4.6)
for some constants C
1
, C
2
depending only on f, , d
0
, v
0
.
From the basic energy law (2.11) and (3.7), we also have
_
T
0
A
2
(t )dt < M(T ) (4.7)
for some constant M(T ) which depends also on f, , d
0
, v
0
.
Combining (4.6) and (4.7), one deduces that A
2
(t ) is uniformly bounded on
[0, T ], for any T , And this proves (4.1) which in turn yields Theorem 4.1.
In the case that the space dimension is 3, we will have a different interpolating
inequality for the estimate of the L
4
norms by the L
2
and H
1
norms. However, (4.3)
and (4.4) will still hold. Instead of (4.6), we will have
dA
2
dt
C

(A
6
+1). (4.8)
The consequence of this a priori estimate is the following local existence of classical
solutions.
Theorem 4.2. For system (1.16), (1.17), there is a time T

such that A
2
(t ) is uni-
formly bounded for all t [0, T

]. Thus we obtain the unique classical solution in

[0, T

].
In fact, we can get the following result by estimating the terms in (4.4) more
carefully.
Theorem 4.3. If dim = 3, with smooth enough boundary and initial conditions,
there is a unique classical solution of system (1.16), (1.17), if the following in-
equality holds:

4

4
(
i
,
i
, v
0
, d
0
). (4.9)
Existence of Solutions for the Ericksen-Leslie System 149
We refer to [LL1] for details. For instance,
_

vvvdx |v|
L
2 |v|
L
4 |v|
L
4 C|v|
7
4
L
2
|v|
L
2
C|v|
L
2 |v|
2
L
2
+|v|
L
2
and
_

(d f (d))
x
j
v
j
x
k
d
x
k
dx

1
2
4
(A
2
+C)|v|
2
+
C

4
| (d f (d))|
2
A
2
+A
C

4
+C.
If we set

A
2
= A
2
+1, we have
1
2
d
dt

A
2

_

4
C
1

1
2
4

A
2
_
B
2

_
C
2

C
3

A
2

4
_
B
2
+C
4

A
2
, (4.10)
where C
i
, i = 1, 2, 3, 4 are positive constants depending on the initial data and
other
i
s.
Also
B
2
=
_

1
(
q
A
kp
d
k
d
p
)
2
+
4
(
k
A
ij
)
2

1
((d
i
f
i
(d)))
2
_
dx

(
5
+
6
) |Ad|
2
dx.
(4.11)
Also by (2.9),
_
t +1
t

A
2
(t )dt M (4.12)
for all t [0, T 1] and some constant M.
We can employ an argument we used in [LL1]. The inequality (4.11) implies
that, for all t [0, T 1], there is a t
1
[t, t +1], such that

A
2
(t
1
) 2M. (4.13)
Next, we shall assume that
4
is so large that

1
2
4
2C[

A
2
(0) +4M
1
2
+1]. (4.14)
Then, initially there is some T
0
> 0 such that

4
C
1
2
4

A
2
(t ) 0, 1
C

A
2
(t )

4
0 (4.15)
for all t [0, T
0
]. We assume T

is the largest such T

0
, then we claim T

= T .
150 Fang-Hua Lin & Chun Liu
Given (4.10), (4.14) and (4.15), we have for all t with 0 t T

A
2
(t )

A
2
(0) +2C
_
t
0

A
2
(t )dt

A
2
(0) +2CM, (4.16)
which shows that (4.15) is always true. This shows that T

= T if T < 1.
However, in the case T > 1, we can use (4.13) to get a t

[T

1
2
, T

] such
that

A
2
(t

) 4M,
and thus the inequality (4.15) is valid at t

in the strict sense by our choice of

4
in
(4.14). Now we repeat the above reasoning with t replaced by t t

to conclude a
contradiction if T

< T .
And this proves Theorem 4.3.
5. Asymptotic Stability
Before we begin the discussion of the asymptotic behavior of the solutions, we
want to point out the following uniqueness result:
Lemma 5.1. Suppose that (v
i
, d
i
) with i = 1, 2 are two solutions of system(1.16),
(1.17) with the same initial and boundary conditions. If one of the solutions, say
(v
1
, d
1
), is smooth, then the two solutions must be identical.
The proof of the lemma can be carried out by estimating the energy of the
difference of the two solutions (v
1
v
2
, d
1
d
2
). We refer to [LL1] and [Se] for
the detail.
Let us consider a global weak solution (v, d) of (1.16), (1.17). Because of the
energy law (2.9), we have that
__

2
|v|
2
+
1
2
|d|
2
+F(d)
_
dx
_
(t )
+
_
t
0
__

1
(A
kp
d
k
d
p
)
2
+
4
(A
ij
)
2
+(
5
+
6
)|Ad|
2
+
1
|N|
2
_
dx
_
dt

__

2
|v|
2
+
1
2
|d|
2
+F(d)
_
dx
_
(0) = C
0
(5.1)
for all t (0, ).
We can see immediately that for each
1
> 0, there is a T
1
(0, ) such that
_
_

[(
1
(A
kp
d
k
d
p
)
2
+
4
(A
ij
)
2
+(
5
+
6
)|Ad|
2
+
1
|N|
2
]dx
_
(T
1
)
1
(5.2)
Existence of Solutions for the Ericksen-Leslie System 151
and
_

T
1
_
_

[(
1
(A
kp
d
k
d
p
)
2
+
4
(A
ij
)
2
+(
5
+
6
)|Ad|
2
+
1
|N|
2
]dx
_
dt
1
(5.3)
We can state the following results:
Theorem 5.1. There is an
2
> 0 such that if
__

__

1
(A
kp
d
k
d
p
)
2
+
4
(A
ij
)
2
+(
5
+
6
)|Ad|
2
+
1
|N|
2
_
dx
_
_
(0)
2
(5.4)
then one of the following two cases has to be true:
(1) (1.16), (1.17) has a unique global classical solution in time with the property
||v(t )||
H
1
()
0 and ||d f(d)||
L
2 (t) 0 (5.5)
as t ;
(2) there is a T

(0, ), such that

E(T

) < E(0)
2
, (5.6)
where
E(t ) =
__

2
|v|
2
+
1
2
|d|
2
+F(d).
_
dx
_
(t ) (5.7)
Proof. First we recall the estimate (4.8):
dA
2
dt
C

(A
6
+1),
where, (4.2),
A
2
(t ) =
_

_
|v|
2
+|d f (d)|
2
_
dx
and C

is a constant depending on the domain and the material constants

i
s.
If for some
3
> 0,
__

__

1
(A
kp
d
k
d
p
)
2
+
4
(A
ij
)
2
+(
5
+
6
)|Ad|
2
+
1
|N|
2
_
dx
_
_
(0)
3
,
(5.8)
then we can see that A
2
(t ) will be uniformly bounded up to some T
0
= T
0
(
3
, C

).
Note that T
0
(, C

) T
0
> 0 if
1
.
If the case (2) is not true, then by (5.1)
_

0
__

1
(A
kp
d
k
d
p
)
2
+
4
(A
ij
)
2
+(
5
+
6
)|Ad|
2
+
1
|N|
2
_
dx
_
dt
2
(5.9)
152 Fang-Hua Lin & Chun Liu
Now for any T (0, ), there is a T
1
[T, T +T
0
] such that
_

__

1
(A
kp
d
k
d
p
)
2
+
4
(A
ij
)
2
+(
5
+
6
)|Ad|
2
+
1
|N|
2
_
dx
_
(T
1
)

2
T
0
. (5.10)
We can choose
2
so small that

2
T
0
<
3
. Then we see the conclusion of (1)
comes from the same argument as the proof of Theorem 4.3. The nal statement of
(1) follows from the energy law (2.11).
We can state the following results as the consequence of this theorem:
Theorem 5.2. Let (v, d) be a global weak solution of (1.16), (1.17). Then there is
a T
0
(0, ) such that (v, d) is smooth in [T
0
, ). Moreover, ||v(t )||
H
1
()
+
||d f (d)||
L
2 (t ) 0 as t goes to innity.
This theoremshows that the systemhas the property of eventual regularity. This
is the rst step in the study of the absorbing set of the weak solutions.
We point out that, in the 0
+
limit, the liquid crystals often possess defects
for all time even for energy minimizing static congurations. Here the regularity
results are for the positive case.
Theorem 5.3. Let d

H
2
() be an absolute minimizer of the functional
E(d)
1
2
_

_
|d|
2
+F(d)
_
dx. (5.11)
Then there is an > 0, which may depend on the system (1.16), (1.17), i.e., the
parameters
i
,
i
s and the data v
0
, d
0
, such that if ||d
0
d

||
H
2
()
+||v
0
||
H
1
()
<
, then the original system (1.16), (1.17) has a unique global solution (v, d) with
v(x, t ) 0 in H
1
(), as t +. Moreover, for any sequence t
i
+,
d(x, t

i
)

d in H
2
() for a subsequence {t

i
}, where

d is a critical point of (5.11).
6. Further Discussions of the Constants
In the above sections, we mainly discussed the case where the constants on the
right-hand side of (2.11) are positive, that is

1
< 0 (6.1)
and

4
> 0,
1
> 0 and
5
+
6
> 0. (6.2)
The condition (6.1) is always assumed. However, in the case
2
= 0, where the
maximal principle holds, we can see the dissipation will be valid if
4
is positive
and big enough related to
1
and
5
+
6
.
Existence of Solutions for the Ericksen-Leslie System 153
In the case where
4
and
1
are both positive,

1
(A
kp
d
k
d
p
)
2
+
4
(A
ij
)
2
+(
5
+
6
)|Ad|
2
=
1
(
i
f
2
i
)
2
+
4
(
i

i
) +(
5
+
6
)(
2
i
f
2
i
), (6.3)
where
i
are the eigenvalues of A and f = (f
i
) is the corresponding components
of d after the transform.
When we look at the energy identity (2.11), we see that the only term on the
right-hand side that might be nonnegative, under the assumptions (6.1), (6.2), is the
integral of 2
2
N
j
A
i,j
d
i
. Of course, if we have

1
0,
especially
(
1
)
1
2
(
5
+
6
)
1
2

2
,
we will still get the relation of dissipation. (At this time, the maximal principle will
not be satised, so the theorems of the existence will be somewhat different.)
Unfortunately most materials do not have these properties (cf. [BC]).
In the case of |d| = 1 (which we do not have in (1.6), (1.7)), we can do the
following decomposition and rearrange the terms:
Dene the matrix B in terms of A as
B = A +
1
2
(dAd)I d Ad Ad d +
1
2
(dAd)d d. (6.4)
Remark. This procedure was used by Ericksen in [E3] and Landau in [LaL].
It is easy to check that B is symmetric and has zero trace. The most important
point is that
trA
2
= trB
2
+2|Ad|
2

1
2
(dAd)
2
. (6.5)
So we see that we only need
(
1
)
1
2
(
5
+
6
+2
4
)
1
2

2
,

1

1
2

4
0
(6.6)
to make the system dissipative.
In fact, we can do even better. Since |d| = 1, we can dene the projection
P = I d d,
since
|Ad|
2
= |PAd|
2
+(dAd)
2
.
Then we can decompose trA
2
more like the following
trA
2
= trB
2
+2|PAd|
2
+
3
2
(dAd)
2
. (6.7)
154 Fang-Hua Lin & Chun Liu
The advantage here is that
Nd = 0
which in turn gives
_

2
2
N
j
A
i,j
d
i
dx =
_

2
2
N
j
PA
i,j
d
i
dx.
The condition becomes
(
1
)
1
2
(
5
+
6
+2
4
)
1
2

2
(6.8)
and

1
+
3
2

4
0 (6.9)
and samples of some of the most commonly used nematic liquid crystals, PAA,
5CB and MBBA, all satisfy this relation (cf. [BC]).
7. The Limiting Case
We see that in both [LL2] and Section 1 of this paper a penalized term
1

2
(|d|
2

1)
2
was added to F to relax the restriction |d| = 1. After all we have done, it is
natural to take the limit of 0. Of course, this will be where the singularity
appears most of the time.
We will end the paper by treating the simplied system (1.18), (1.19) in the
following theorem.
Theorem 7.1. Suppose (v
i
, d
i
) are all weak solutionof (1.18), (1.19) anduniformly
bounded in spaces:
v
i
L
2
(0, T ; H
1
()) L

(0, T ; L
2
()),
d
i
L

(0, T ; H
1
())
for all T (0, ). Then there exists a subsequence (v
i
, d
i
) such that
v
i
v strongly in L
2
(0, T ; L
2
()),
v
i
v weakly-* in L

(0, T ; L
2
()),

t
v
i

t
v weakly in L
2
(0, T ; H
1
()),
d
i
d strongly in L
2
(0, T ; H
1
()),
d
i
d weakly-* in L

(0, T ; H
1
()),

t
d
i

t
d weakly in L
2
(0, T ; L
2
()),
and (v, d) is a solution of
v
t
+(v )v v +P = M,
d
t
+(v )d = (d |d|
2
d),
|d| = 1,
and M is a measure-valued tensor.
Existence of Solutions for the Ericksen-Leslie System 155
We point out that all the results will be true in the case of
2
= 0 for the
generalized Ericksen-Leslie system.
The proof comes from following two lemmas. The rst one is developed in the
literature of harmonic mapping when the target space in a sphere. It has been used
before by Chen [CS], Shatah [Sh], Struwe [St] and many others. The second one is
a compactness theorem due to Lions and Aubin. It is a Arzela-Ascoli theorem type
of result which was also used in the theory of Navier-Stokes equations [T].
Lemma 7.1. The following two systems are equivalent:
u
t
= u |u|
2
u,
and
|u| = 1,
u
t
u (u u) = 0.
Lemma 7.2. Let Y = {v L

0
(0, T ; X
0
), v
t
L

1
(0, T ; X
1
)} with norm
v
Y
= v
L

0
(0,T ;X
0
)
+v
L

1
(0,T ;X
1
)
,
where X
0
X X
1
are Hilbert spaces. Suppose that the injections X
0
X
X
1
are continuous, and the injection from X
0
into X is compact. Here
0
,
1
> 1.
Then the injection from Y into L

0
(0, T ; X) is compact.
With these two lemmas, the proof of the theorem is somewhat standard. The
convergence of the second equation comes after the exact same argument as that
for the heat ow of harmonic maps and Lemma 7.1 (see [CS]). The convergence
of the rst equation follows from Lemma 7.2 and the similar argument in [LL1]
and [T].
References
[BC] L. M. Blinov &C. G. Chigrinov, Electrooptic Effects in Liquid Crystal Materials,
Springer-Verlag (1994).
[CS] Y. M. Chen & M. Struwe, Existence And Partial Regularity For Heat Flow For
Harmonic Maps, Math. Z. 201, (1989) 83103.
[dG] P. G. de Gennes, The Physics Of Liquid Crystals, Oxford (1974).
[E1] J. Ericksen, Conservation Laws For Liquid Crystals, Trans. Soc. Rheol. 5 (1961)
2234.
[E2] J. Ericksen, Continuum Theory Of Nematic Liquid Crystals, Res. Mechanica 21
(1987) 381392.
[E3] J. Ericksen, Liquid Crystals with Variable Degree of Orientation, Arch Rational
Mech. Anal. 113 (1991) 97120.
[LSU] O. A. Ladyzenskaya, N. A. Solonnikov &N. N. Uraltseva, Linear And Quasi-
linear Equations Of Parabolic Type, Trans. Of Math. Mon. Vol 23, AMS (1968).
[LaL] L. D. Landau & E. M. Lifshitz, Theory of Elasticity, Pergamon Press (1986).
[Le] F. Leslie, Theory Of Flow Phenomenum In Liquid Crystals, (Vol 4) Brown (Ed.)
A.P. NewYork (1979) 181.
[LL1] F. H. Lin & C. Liu, Nonparabolic Dissipative Systems Modeling The Flow Of
Liquid Crystals, Comm. Pure Appl. Math. 48 (1995) 501537.
156 Fang-Hua Lin & Chun Liu
[LL2] F. H. Lin & C. Liu, Partial Regularities Of The Nonlinear Dissipative Systems
Modeling The FlowOf Liquid Crystals, Discrete and Continuous Dynamic Systems
2 (1996) 123.
[Lions] J. L. Lions, On Some Questions in Boundary Value Problems of Mathematical
Physics. Contemporary Development in Continuum Mechanics (1978) 285346.
[Liu] C. Liu, Dynamic Theory for Incompressible Smectic-A Liquid Crystals. Discrete
and Continuous Dynamic Systems 6 (2000) 591608.
[LW] C. Liu & N. Walkington, Approximation of liquid crystal ows. SIAM J. Numer.
Anal. 37 (2000) 725-738.
[SU] J. R. Schoen & K. Uhlenbeck, A Regularity Theory for Harmonic Maps, J. Diff.
Geom. 17 (1982) 307333.
[Se] J. Serrin, On The Interior Regularity Of Weak Solutions Of Navier-Stokes Equa-
tions, Arch. Rational Mech. Anal. 9 (1962) 187195.
[St] M. Struwe, Geometric Evolution Problems, IAS/Park City Math. Series Vol 2
(1996) 257333.
[Sh] J. Shatah, Weak Solutions and Development of Singularities in SU(2) -model.
CPAM 41 (1988) 459469.
[T] R. Temam, Navier-Stokes Equations, North Holland (1977).
Courant Institute of Mathematical Sciences
NewYork University
NewYork, NY 10012
and
Department of Mathematics and Center for Materials Physics
Penn State University
University Park, PA 16802
(Accepted January 15, 2000)
Published online September 12, 2000 c Springer-Verlag (2000)