J.

Differential Equations 213 (2005) 204 – 233

www.elsevier.com/locate/jde

Existence and global stability of traveling curved

fronts in the Allen–Cahn equations
Hirokazu Ninomiyaa , Masaharu Taniguchib,∗
a Department of Mathematics and Informatics, Ryukoku University, Seta Otsu 520-2194, Japan
b Department of Mathematical and Computing Sciences, Tokyo Institute of Technology, O-okayama
2-12-1-W8-38, Tokyo 152-8552, Japan

Received March 23 2004; revised June 7 2004

Available online 10 August 2004

Abstract
This paper is concerned with existence and stability of traveling curved fronts for the
Allen–Cahn equation in the two-dimensional space. By using the supersolution and the sub-
solution, we construct a traveling curved front, and show that it is the unique traveling wave
solution between them. Our supersolution can be taken arbitrarily large, which implies some
global asymptotic stability for the traveling curved front.
© 2004 Elsevier Inc. All rights reserved.
Keywords: Traveling wave; Allen–Cahn equation; Curved front; Stability

1. Introduction

Traveling wave solutions have been intensively studied in various mathematical mod-
els motivated by many chemical or physical studies on them. See [10] for traveling
wave solutions in the Belousov–Zhabotinsky reactions, and see [5] for the filamen-
tary vortex of the Ginzburg–Landau equation. Traveling wave solutions in many other
systems are treated in [14], for instance.

∗ Corresponding author. Fax: +81-3-5734-3209.

E-mail addresses: ninomiya@math.ryukoku.ac.jp (H. Ninomiya), masaharu.taniguchi@is.titech.ac.jp
(M. Taniguchi).

0022-0396/$ - see front matter © 2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.jde.2004.06.011
 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 205

In this paper, we consider a solution u(·, ·, t) ∈ L∞ (R2 ) of the Allen–Cahn

equation

ut =
u + f (u) (x, y) ∈ R2 , t > 0,

(1.1)
u(x, y, 0) = u0 (x, y) (x, y) ∈ R2 .

Here, a nonlinear smooth function f is of bistable type and a given initial function u0
is bounded. The following is the standing assumptions on f :
(A1) f (1) = 0, f (−1) = 0, f  (1) < 0 and f  (−1) < 0.

1
(A2) −1 f (s) ds > 0.
(A3) f (s) < 0 and f  (s) < 0 for s > 1. f (s) > 0 and f  (s) < 0 for s < −1.
A typical example of f is

f (u) = −(u + 1)(u + a)(u − 1), (1.2)

where a ∈ (0, 1) is a given number.

Now, we state previous works of traveling wave solutions. Bonnet and Hamel [1]
showed the existence of the traveling wave solutions when the reaction term f is of
the "ignition temperature" type. Hamel and Monneau [8] shows the uniqueness of the
traveling front of the corresponding singular limit problem. Hamel and Nadirashvili [9]
studied traveling wave solutions when f is the Fisher-KPP type. Fife [6] studied the
traveling wave solutions as a singular limit problem by putting a small coefficient to
the diffusion term.
Under our assumptions on f , the constant states −1 and 1 are asymptotically stable.
The region of the state 1 is getting larger and larger and finally it covers the whole
space. When the state 1 propagates, we can observe the characteristic profiles. In the
one-dimensional space, one of the typical solutions is a traveling wave solution which
never changes its shape. Setting u(x, t) =
() and  = x − kt, we have

−
 − k
 − f (
) = 0,
 () < 0 (−∞ <  < ∞),

(−∞) = 1,
(+∞) = −1. (1.3)
√ √
For the nonlinearity (1.2), we have k = 2a and
() = − tanh(/ 2).
The Allen–Cahn equation (1.1) is related to the following interface equation
given by

V = H + k, (1.4)

where V is a normal velocity, H is the curvature, and k is a given constant. See [6]
or [3] for example. For this interface model (1.4), traveling fronts have been studied
by [2,5,11,12] and so on.
206 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233

If the interface is represented by the graph y = v(x, t), Eq. (1.4) is reduced to

vxx
vt = + k 1 + vx2 , x ∈ R, t > 0, (1.5)
1 + vx2

If the traveling front is represented by (x) + ct for the suitable coordinate where c
is the speed of the traveling front. Then (x) and c satisfy

xx
c= + k 1 + 2x . (1.6)
1 + 2x

The following is the explicit expression for the traveling curved fronts of (1.6).

Proposition 1.1. (Ninomiya and Taniguchi [11, Propositions 1.1, 2.5]). For c > k >
0, there exists a unique solution (x; c) of (1.5) with asymptotic lines y = m∗ |x|,
where
√
c2 − k 2
def
m∗ = > 0.
k

The graph of y = (x; c) can be parametrized by  = arctan x (x; c) as

 
 2 c+k 
x(; c) = + arctanh tan
c m∗ c c−k 2
 
1 2(c2 − k 2 ) m∗
y(; c) = log + arctan m∗ .
c c(c cos  − k) c

for  ∈ (−arctan m∗ , arctan m∗ ). The graph is strictly convex with

xx (x; c) > 0 for all x ∈ R.

The graph of (x; c) is "V-shaped" and connects two asymptotes.

The asymptotic stability of the traveling curved front in (1.5) is discussed in [5,12].
The traveling curved front is proved to be asymptotically stable, if the initial per-
turbation decays at infinity. See [12, Theorems 1.1 and 4.1]. By the observation for
the interface model, we can expect that a "V-shaped" traveling wave solution of (1.1)
exists.
We study traveling wave solutions. We assume that the solutions travel towards y-
direction without loss of generality. We put

u(x, y, t) = w(x, y − ct, t), z = y − ct.

 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 207

Then, we obtain

wt − wxx − wzz − cwz − f (w) = 0, (x, z) ∈ R2 , t > 0,

(1.7)
w|t=0 = u0 in R2 .

We denote the solution of (1.7) with w(x, z, 0; u0 ) = u0 (x, z) by w(x, z, t; u0 ).

We seek for v(x, z) with

def
L[v] = −vxx − vzz − cvz − f (v) = 0 in R2 . (1.8)

The traveling wave with speed c in y-direction becomes a stationary solution of (1.7).
Then, two functions
(k(z ± m∗ x)/c) satisfy (1.8) and these are called planar traveling
fronts. Since the maximum of subsolutions is also a subsolution, it turns out that
    
k k
v − (x, z) = max

def
(z − m∗ x) ,
(z + m∗ x)
c c
 
k
=
(z − m∗ |x|)
c

is a subsolution of (1.7). This v − (x, z) is strictly monotone decreasing in z.

Now, we are searching a traveling wave solution towards y-direction. If v − is given
as an initial condition, it seems to converge to a traveling curved front. Actually, we
have the following theorem.

Theorem 1.2 (Existence). There exists a traveling wave solution u(x, y, t) = v∗ (x, y −
ct) of (1.1) with

lim sup |v∗ (x, z) − v − (x, z)| = 0,

R→∞ x 2 +z2 >R 2

v − (x, z) < v∗ (x, z).

The traveling wave solution v∗ is asymptotically stable if the given perturbation

decays at infinity.

Theorem 1.3 (Stability). Let u0 (x, y) satisfy

lim sup |u0 (x, y) − v − (x, y)| = 0, (1.9)

R→∞ x 2 +y 2 >R 2

v − (x, y) ≤ u0 (x, y). (1.10)

208 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233

speed c

Fig. 1. The contour lines of the traveling curved front v∗ .

Then the solution u(x, y, t; u0 ) of (1.1) satisfies

lim u(x, y, t; u0 ) − v∗ (x, y − ct)L∞ (R2 ) = 0.

t→∞

This theorem immediately implies the uniqueness of such traveling wave solutions.

Corollary 1.4 (Uniqueness). Let v(x, z) be a solution of (1.8) with

lim sup |v(x, z) − v − (x, z)| = 0, (1.11)

R→∞ x 2 +z2 >R 2

v − (x, z) ≤ v(x, z) f or (x, z) ∈ R2 . (1.12)

Then v(x, z) ≡ v∗ (x, z) holds true. Here v∗ (x, z) is given by Theorem 1.2.

Hereafter, if v(x, z) is a solution of (1.8) and satisfies (1.11) and (1.12), then v(x, y −
ct) is called a traveling curved front. See Figs. 1 and 2 for the shape of the traveling
curved front.

Remark 1.5. Fife [6] showed the construction of traveling curved fronts by using the
singular perturbation framework, which corresponds to our case when a in (1.2) is
sufficiently close to 0. Theorem 1.2 studies the general case.
 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 209

Fig. 2. The bird’s-eye view of v∗ .

2. Existence of a traveling curved front

There exists a positive constant 1 (0 < 1 < 41 ) with

−f  (s) ≥ 1 for s < −1 + 21 or s > 1 − 21 ,

where

min{−f  (−1), −f  (1)} > 0.

def 1
1 = 2

Since
() is monotone decreasing in , we can define positive constants A and B by

1 1

(−A) = 1 − ,
(B) = −1 + ,
2 2

respectively. Then

1 1
−1 + <
() < 1 −
2 2

is equivalent to −A <  < B.

Lemma 2.1. (Brazhnik [2], Ninomiya and Taniguchi [11,12]) There exist positive con-
stants 1 , 2 , Ki (i = 1, . . . , 4) and ± with


max{|
 ()|, |
()|} ≤ K1 exp(−1 ||), (2.1)


max{| ( )|, | ( )|} ≤ K2 sech(2 ),

c
K3 sech(2 ) ≤
− k ≤ K4 sech(2 ),
1 +  ( )2
210 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233

m∗ | | ≤ ( ),

− ≤ ( ) ≤ +

for any  ∈ R and ∈ R, where

def c(( ) − m∗ | |)
( ) =
.
c − k 1 +  ( )2

We note that
√
def c c2 − k 2
2 = cm∗ = >0
k

and that the curvature of  = ( ) is calculated as


 ( ) c
=
− k.
(1 +  ( )2 )3/2 1 +  ( )2

Theorem 2.2. There exist a positive constant

0 and a positive function 0 (
) so that,
for 0 <
<
0 and 0 <  ≤ 0 (
),
 
+ z − (x)/)
def
v (x, z;
, ) =

+
sech(2 x),
1 +  (x)2

is a supersolution of (1.8) with

lim sup |v + (x, z;

, ) − v − (x, z)| ≤ 2
, (2.2)
R→∞ x 2 +z2 >R 2

v − (x, z) < v + (x, z;

, ) f or (x, z) ∈ R2 , (2.3)

−(v + )z (x, z;
, ) > 0 f or (x, z) ∈ R2 . (2.4)

For simplicity, we often write v + (x, z) instead of v +

(x, z;
, ). We prove this the-
orem in Section 3. For the meaning of (z − (x)/)/ 1 +  (x)2 , see Fig. 3.
Next, we prove the existence of the traveling curved front.
Proof of Theorem 1.2. Theorem 2.2 guarantees the existence of a supersolution
v + (x, z;
, ). By the definition of w(x, z, t; v ± ), we have

v − (x, z) < w(x, z, t; v − ) < w(x, z, t; v + ) < v + (x, z) for x, z ∈ R, t > 0. (2.5)
 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 211

Fig. 3. The construction of the supersolution v + .

Since v + is a supersolution and v − is a subsolution, w(x, z, t; v + ) is monotone de-

creasing in t and w(x, z, t; v − ) is monotone increasing in t. The limit functions

def
v ∗ (x, z) = lim w(x, z, t; v + ),
t→∞
(2.6)
def
v ∗ (x, z) = lim w(x, z, t; v − ),
t→∞

exist in L∞ (R2 ) with

L[v∗ ] = 0, L[v ∗ ] = 0,
v − (x, z) ≤ v∗ (x, z) ≤ v ∗ (x, z) ≤ v + (x, z;
, ) for (x, z) ∈ R2 . (2.7)

For the detail, see Sattinger [16].

Theorem 2.2 and (2.5) imply that, for any 0 <
≤
0 and 0 <  ≤ 0 (
), there exist
a positive constant R and a supersolution v + (x, z;
, ) satisfying

v − (x, z) ≤ v∗ (x, z) ≤ v + (x, z;

, ) ≤ v − (x, z) + 2
when x 2 + z2 > R 2 .

Thus, the limit function v∗ is a traveling curved front of (1.1).

Since
∈ (0,
0 ] can be taken arbitrarily small, (2.2) gives

lim sup |v ∗ (x, z) − v∗ (x, z)| = 0. (2.8)

R→∞ x 2 +z2 >R 2
212 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233

By the definition, v ∗ may depend on

and . Corollary 1.4 or Lemma 4.6 says that
v ∗ is independent of
and  and it equals v∗ .

3. Proof of Theorem 2.2

Set

def
= x,
def z − (x)/
 =
,
1 +  (x)2
def
( ) =
sech(2 ).

By the chain rule, we have


  
x = −  −  , (3.1)
1 + 
2
1 + 
2


2  + 2   32   ( − 1)
2 2 2 2

xx = − + + . (3.2)
1 +  (1 +  )2 (1 +  )3/2
2 2 2

Using these equalities and (1.3), we get



() c
 () 
L[v + ] = − 
− (

( )  x ) x −
− f (
() + ( )) − 2  ( )
1 +  ( ) 2
1 +   ( )2
   
1   c
= 1− − x
() − xx
() + k −

2

 ()
1 +   ( )2 
1 +  ( ) 2


+f (
()) − f (
() + ( )) − 2  ( )
= I1 + I2 + I3 + I4 ,

where
 
def 1 
I1 = 1 − 
− x
(),
2
1 +  ( ) 2

I2 = −xx
 (),
def
 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 213
 
 c 
− k 
 (),
def
I3 = −  
1 + 
2

def 
I4 = −f (
+ ) + f (
) − 2  .

By (3.1) and (3.2), we have

  
   2 2 
2   
I1 = − 2 + 
(),
 1 + 2
(1 +  )3/2 
2

 

 +   3  ( − 1)
2 2 2 2

I2 = − −  +  +
 (),
1 +  (1 +  )2 (1 +  )3/2
2 2 2

By Lemma 2.1, we can easily show

|I1 | ≤ K5  sech(2 ),
|I2 | ≤ K6  sech(2 ),
I3 ≥ −K3
 () sech(2 ) > 0,

for 0 <  ≤ 1. Assume

1
0 <
<
0 ≤ . (3.3)
2

Lemma 2.1 implies

I4 ≥ 1  − K7 2
sech(2 ) = (1 − K7 2 )
sech(2 ),

when
() ≤ −1 + 1 /2 or
() ≥ 1 − 1 /2. Then we have L[v + ] ≥ 1 /2 > 0 if

 
1 1
0 <  ≤ min 1, ,
. (3.4)
4K7 4(K5 + K6 )

If −1 + 1 /2 ≤
() ≤ 1 − 1 /2, namely, −A ≤  ≤ B, then we have

I3 ≥ K3 p sech(2 ),
|I4 | ≤ K8  + K7 2
sech(2 ) ≤ (K8 + K7 )
sech(2 ),
214 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233

where

(−
 ()) > 0.

def
p= min (3.5)
−A≤≤B

Take
and  so small that

2 K3 p
1
> (K5 + K6 ) + K8
+ K7 
(3.6)

holds. Then we have

L[v + ] ≥ 1
2 min{1
, K3 p} sech(2 ) > 0 in R2

under (3.3), (3.4) and (3.6). Thus v + is a supersolution.

Next, we will show the inequality (2.3) in x ≥ 0. For simplicity of notation, set

def k def z − m∗ |x|

1 = (z − m∗ |x|), 2 =
.
c 1 +  (x)2

Recall that

z − (x)/
v + − v − =
() −
(1 ) + ( ), =

1 +  (x)2

and that

1
m∗ |x| ≤ (x) for 0 <  < 1, x ∈ R.


If  ≤ 1 , then (2.3) holds true apparently. Assume that  > 1 . We have

 
1 k ( ) − m∗ | |
 − 1 =
− (z − m∗ |x|) −
.

1 +  ( ) 2 c  1 +  ( )2

This equality and Lemma 2.1 imply

( ) 
z − m∗ |x| > ≥ − > 0. (3.7)
 

Thus, we have

0 < 1 <  ≤ 2
 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 215

and

v + − v − =
() −
(2 ) +
(2 ) −
(1 ) + ( )

≥
(2 ) −
(1 ) + ( )
 
1 k
= −
− (z − m∗ |x|)|
 (2 + (1 − )1 )| + ( )
1 +  ( )2 c
 
1 k c1 
= −
− |
(2 + (1 − )1 )| + ( )
1 +  ( )2 c k

for some  ∈ (0, 1). By (2.1), we see

|
 (2 + (1 − )1 )| ≤ K1 exp(−1 |2 + (1 − )1 |) ≤ K1 exp(−1 |1 |).

It follows from (3.7) that

k k ( ) k
1 = (z − m∗ |x|) ≥ ≥ −.
c c c

Using (2.2) and the above inequalities, we have

 
1 k c 1 K 1
v+ − v− ≥ −
− exp(−1 |1 |) +
sech(2 )
1 +  ( )2 c k
 
K1 K4
≥ − 2 sup 0 exp(−0 ) sech(2 ) +
sech(2 )
2
k 1 1 0 >0
4K1 K4 c
≥ −  sech(2 ) +
sech(2 ).
e2 k 2 21 −

Define
0 and 0 (
) by

def 1 K3 p

0 = min 1, , > 0,
2 4K8
  
def 1 K3 p 1 e2 k 2 21 −
0 (
) = min 1, , ,
,
> 0.
4K7 4(K5 + K6 + K7 ) 4(K5 + K6 ) 4K1 K4 c

Then, we see that v + ≥ v − for x ≥ 0 if 0 <

≤
0 and 0 <  ≤ 0 (
).
In the case where x < 0, (2.3) can be also proved similarly.
216 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233

Next we consider (2.2). The mean-value theorem implies

v + − v − =
() −
(1 ) + ( ),
=
 (3 )( − 1 ) + ( ),

where

def z − ( )/ k
3 =  + (1 − )1 = 
+ (1 − ) (z − m∗ |x|) (3.8)

1 +  ( ) 2 c

with some  ∈ (0, 1). Since 0 < ( ) ≤

, we will show that, for some large R1 > 0
and R2 > 0,

|
 (3 )( − 1 )| ≤

if | | ≥ R1 or |z| ≥ R2 . By the definition of 3 , we have

    
( ) (1 − )km∗ | | k 1 k
z = 3 +
+ +
− . (3.9)

 1 +  ( ) 2 c c 1 +  ( )2 c

This implies

 (3 )( − 1 )
 
1 k
=
−
 (3 )z
1 +  ( )2 c
 
1 ( ) km∗
−
− | |
 (3 ). (3.10)
 1 +  ( )2 c

By Lemma 2.1, we have

sup |
 (3 )| ≤ K1 , sup |
 (3 )3 | ≤ K9 (3.11)

3 ∈R 3 ∈R

with some positive constant K9 . It follows from (3.9)–(3.11) that if R1 is so large, then

sup |
 (3 )( − 1 )| ≤

| |≥R1
 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 217

holds true. Next, we consider the case where | | ≤ R1 and |z| ≥ R2 . By (3.8), we have

lim inf sup |3 | = ∞. (3.12)

z→∞ | |≤R1

Taking R2 large enough and using (3.9)–(3.12), we have

sup sup |
 (3 )( − 1 )| ≤

.
|z|≥R2 | |≤R1


If R = R12 /2 + R22 , we see (2.2). Finally, (2.4) follows immediately from (1.3) and
the definition of v + (x, z;
, ). This completes the proof.

4. Uniqueness of the traveling curved fronts

Let w (j ) (x, z, t) be the solution of

(j )
wt + L[w (j ) ] = 0 in R2 , t > 0,

(j )
w(j ) (x, z, 0) = w0 (x, z) in R2 ,

for j = 1, 2. We have the following fact.

Lemma 4.1 (Continuity in initial data). If

(j )
−2 ≤ w0 (x, z) ≤ 2 for (x, z) ∈ R2 , j = 1, 2,

then
(2) (1)
sup |w (2) (x, z, t) − w (1) (x, z, t)| ≤ eMt sup |w0 (x, z) − w0 (x, z)|
(x,z)∈R 2
(x,z)∈R 2

where
M = sup |f  (s)|.
def
(4.1)
|s|≤2

Proof. Let ŵ be a solution of

w
t − w
xx − w
zz − c
wz − M w
=0 in R2 , t > 0,

(2) (1)
w
(x, z, 0) = sup |w0 (x, z) − w0 (x, z)| in R2 .
(x,z)∈R 2
218 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233

def
 ≥ 0 and w̃ = ŵ − (w (2) − w (1) ) satisfies
Then we have w

w̃t − w̃xx − w̃zz − cw̃z − f  (w (2) + (1 − )w (1) )w̃

= (M − f  (w (2) + (1 − )w (1) ))ŵ ≥ 0 in R2 , t > 0,
w̃(x, z, 0) ≥ 0 in R2

with some  ∈ (0, 1). The maximum principle, see [15] for example, yields

(2) (1)
w(2) (x, z, t) − w (1) (x, z, t) ≤ ŵ(x, z, t) = eMt sup |w0 (x, z) − w0 (x, z)|
(x,z)∈R2

for (x, z) ∈ R2 , t > 0. Similarly, we have

(2) (1)
w (1) (x, z, t) − w (2) (x, z, t) ≤ ŵ(x, z, t) = eMt sup |w0 (x, z) − w0 (x, z)|
(x,z)∈R 2

for (x, z) ∈ R2 , t > 0. This completes the proof.

Let w1 (t) be defined by

w1 (t) = f (w1 (t)) for t > 0,

 
w1 (0) = min −1, inf u0 (x, z) ≤ −1.
(x,z)∈R2

Similarly, we define w2 (t) by

w2 (t) = f (w2 (t)) for t > 0

 
w2 (0) = max 1, sup u0 (x, z) ≥ 1.
(x,z)∈R2

Then w1 (t), w2 (t) are solutions to (1.7) with w1 (0) ≤ u0 (x, z) ≤ w2 (0). The compari-
son principle yields

w1 (t) ≤ w(x, z, t) ≤ w2 (t) for (x, z) ∈ R2 , t > 0. (4.2)

Sending t → ∞, we have

−1 ≤ lim inf w(x, z, t) ≤ lim sup w(x, z, t) ≤ 1 for (x, z) ∈ R2 .

t→∞ t→∞
 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 219

The strong maximum principle, see [7] for instance, implies the following strict in-
equalities.

Lemma 4.2 (A priori estimate). Let v∗ (x, y), v ∗ (x, y) be as in (2.6). Then

−1 < v∗ (x, z) ≤ v∗ (x, z) < 1 for (x, z) ∈ R2 .

holds true. Moreover, one has

−(v∗ )z (x, z) > 0, −(v ∗ )z (x, z) > 0 for (x, z) ∈ R2 .

Proof. We show the latter statement. Recall that w(x, z, t; v ± ) is the solution of (1.7)
with w(x, z, 0; v ± ) = v ± (x, z). Then the derivative wz (x, z, t; v + ) satisfies
 
* *
− − c + f  (w(x, z, t; v + )) wz (x, z, t; v + ) = 0 in R2 , t > 0
*t *z
wz (x, z, 0; v + ) ≤ 0 in R2 .

Then the maximum principle yields

wz (x, z, t; v + ) < 0 in R2 , t > 0.

Sending t → ∞, we have (v ∗ )z (x, z) ≤ 0. The strong maximum principle says

−(v ∗ )z (x, z) > 0 in R2 .

Since v − (x, z) is also strictly monotone decreasing in z, the same argument holds for
v∗ (x, z).

To show the uniqueness we prepare the following lemma.

Lemma 4.3. There exists a positive constant 3 with

−(v∗ )z (x, z) ≥ 3 when − 1 + 1 ≤ v∗ (x, z) ≤ 1 − 1 ,

−(v + )z (x, z) ≥ 3 when − 1 + 1 ≤ v + (x, z) ≤ 1 − 1 . (4.3)

Proof. By (2.2) and (2.7), there exists a constant R with

1
|v∗ (x, z) − v − (x, z)| ≤ for x 2 + z2 ≥ R 2 .
2
220 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233

Then, we have

{(x, z) ∈ R2 | x 2 + z2 > R 2 , −1 + 1 ≤ v∗ (x, z) ≤ 1 − 1 }

 

2  2 1 − 1
⊂ (x, z) ∈ R  x + z > R , −1 +
2 2
≤ v (x, z) ≤ 1 +
2 2
 
 k
= (x, z) ∈ R2  x 2 + z2 > R 2 , −A ≤ (z − m∗ |x|) ≤ B .
c

For C > 0 we set

def
n = {(x, z)|(x − n)2 + (z − m∗ |n|)2 < C 2 },
def
vn (x, z) = v∗ (x − n, z − m∗ |n|),
def
B0 = {(x, z) ∈ R2 |x 2 + z2 < 4C 2 } (4.4)

for n = 0, ±1, ±2 . . .. Choose a positive constant C so large that

  ∞
 k 
(x, z) x 2 + z2 > R 2 , −A ≤ (z − m∗ |x|) ≤ B ⊂ n
c n=−∞

is valid. The function vn is a solution of

−
vn − c(vn )z − f (vn ) = 0 in B0 .

Theorem 2.2 also says that

  
 
lim sup  vn (x, z) −
k (z − m∗ (sgn n)x)  = 0.
n→±∞  c 
(x,z)∈B0

Using the interior Lq estimate for the second derivative of elliptic equations, see [7]
for instance, we have

 
k
vn (x, z) →
(z − m∗ x) in W 2,q (0 ) as n → ∞,
c

 
k
vn (x, z) →
(z + m∗ x) in W 2,q (0 ) as n → −∞,
c
 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 221

for any q > 1. If we take q > 2 and choose a positive constant ∗ with

2
0 < ∗ < 1 − , (4.5)
q

this immediately implies

 
k k
−(vn )z → −
 (z − m∗ x sgn n) in C ∗ (0 ) as n → ±∞.
c c

Thus there exists a large positive integer n0 with

  
1 k k
−(vn )z ≥ inf −
 (z − m∗ |x|)
2 (x,z)∈0 c c

for |n| ≥ n0 . Defining



1   
k k pk
−

def
3 = min inf (z − m∗ |x|) , ,

 2 (x,z)∈0 c c c




inf
 
(−(v∗ )z (x, z)), >0,
cn 2 

x 2 +z2 ≤ R+ k 0

we obtain (4.3) for v∗ .

Next, we will show the second inequality of (4.3). By the definition of v + (x, z;
, )
and
0 , we have 0 <
< 1 /2 and

{(x, z) ∈ R2 | x 2 + z2 > R 2 , −1 + 1 ≤ v + (x, z;

, ) ≤ 1 − 1 }
 
2 2 1 1
⊂ (x, z) ∈ R  x + z > R , −1 +
2 2
≤
() ≤ 1 +
2 2
 !

= (x, z) ∈ R2  x 2 + z2 > R 2 , −A ≤  ≤ B .

Then, we get

−
 ()
−(v + )z = −
 ()z =
.
1 +  ( )2
222 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233

By (3.5) and Lemma 2.1, we have

pk
−(v + )z ≥ .
c

This completes the proof.

Lemma 4.4. Let v̄ be a supersolution to (1.8) with

−v̄z (x, z) > 0, −1 − 1 < v̄(x, z) < 1 + 1 for all (x, z) ∈ R2 ,

−v̄z (x, z) ≥ 3 if − 1 + 1 ≤ v̄(x, z) ≤ 1 − 1 .

Let v be a subsolution to (1.8) with

−v z (x, z) > 0, −1 − 1 < v(x, z) < 1 + 1 f or all(x, z) ∈ R2 ,

−v z (x, z) ≥ 3 if − 1 + 1 ≤ v(x, z) ≤ 1 − 1 .

Then there exists a large positive constant  such that, for any  ∈ (0, 1 /2], w + and
w− defined by

w+ (x, z, t; v̄) = v̄(x, z − (1 − e−t )) + e−t ,

def

w− (x, z, t; v) = v(x, z + (1 − e−t )) − e−t

def

are a supersolution and a subsolution of (1.7), respectively.

We remark that both v∗ and v − satisfy the assumption on v.

Proof. We have

wt+ + L[w + ] = e−t (−v̄z − 1) + L[v̄ + e−t ]

≥ e−t (−v̄z − 1) − f (v̄ + e−t ) + f (v̄)

 " 1 
−t  −t
= e −v̄z −  − f (v̄ + s e ) ds ,
0
 " 
1
−t −t
wt− −
+ L[w ] = −e −v z −  − 
f (v − s e ) ds ,
0
 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 223

where v̄ = v̄(x, z − (1 − e−t )) and v = v(x, z + (1 − e−t )). For simplicity, let
v be either v̄ or v. By the assumption, we have
"  
1
 −t M
−vz −  − f (v ± s e ) ds ≥  3  − 1 −
0 

when −1 + 1 ≤ v ≤ 1 − 1 . Here M is as in (4.1). For v < −1 + 1 or v > 1 − 1 ,

we have an estimate
" 1
−vz −  − f  (v ± s e−t ) ds ≥ 1 − .
0

We choose  so small and  so large to get

+M
0 <  < 1 , > . (4.6)
3

This immediately implies

wt+ + L[w + ] ≥ 0, wt− + L[w + ] ≤ 0.

Thus, w+ and w − are a supersolution and a subsolution, respectively.

To show Theorem 1.3, we prepare the following two lemmas.

Lemma 4.5. Let w(x, z, t) be the solution of (1.7) with (1.9). Then

lim sup |w(x, z, T ) − v − (x, z)| = 0

R→∞ x 2 +z2 >R 2

holds true for any fixed T > 0.

Lemma 4.6. Let v∗ and v ∗ be as in (2.6). Then

v∗ (x, z) ≡ v ∗ (x, z) f or all(x, z) ∈ R2 .

holds true.
We state the proof of these lemmas later.
Proof of Theorem 1.3. The solution w(x, z, t) of (1.7) satisfies (4.2). We will show
that, for any
∗ > 0, there exists a positive constant T∗ with

sup |w(x, z, t) − v∗ (x, z)| ≤

∗
(x,z)∈R2

for t > T∗ .
224 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233

First, we choose  small enough to satisfy

∗
v ∗ (x, z − ) ≤ v ∗ (x, z) + , 0 <  <
0 (4.7)
3

and fix it where
0 and  are given in Theorem 2.2 and Lemma 4.4, respectively.
Next, we will choose a suitable supersolution. By the argument of (4.2), there exists
T > 0 with


w(x, z, t; v − ) ≤ w(x, z, t) < 1 + for (x, z) ∈ R2 , t ≥ T .
2

Lemma 4.5 shows that


w(x, z, T ) ≤ v − (x, z) + for x 2 + z2 ≥ R 2
2

for some R > 0. If  is small enough,

   
z − ( )/ k (0) 
=
≤ R− ≤
−1 1 −
1 +  ( )2 c 2 2

for x 2 + z2 ≤ R 2 . Namely, choosing  so small to satisfy


(0)
0 <  < min , 0 () ,
2(R − min{
−1 (1 − /2)c/k, 0})

we get


v + (x, z) ≥ 1 − for x 2 + z2 ≤ R 2 ,
2

where v + (x, z) = v + (x, z; , ). Combining these inequalities and (2.5), we obtain

w(x, z, T ) < v + (x, z) +  for (x, z) ∈ R2 .

Then

w(x, z, t + T ; v − ) ≤ w(x, z, t + T ) ≤ w+ (x, z, t; v + )

holds true for t ≥ 0. Using the maximum principle again, we have

w(x, z, t + s + T ; v − ) ≤ w(x, z, t + s + T ) ≤ w(x, z, s; ut ) (4.8)

 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 225

holds true for t ≥ 0 and s ≥ 0 where

ut (x, z) = w + (x, z, t; v + ).
def

Since w(x, z, t; v + ) monotonically converges to v ∗ (x, z) as t → ∞, there exists a

positive constant s1 with

∗
sup |w(x, z, s1 ; v +, ) − v ∗ (x, z − )| ≤ ,
(x,z)∈R 2 3

where

v +, (x, z) = v + (x, z − ).

def

Lemma 4.1 implies

sup |w(x, z, s1 ; u) − w(x, z, s1 ; v +, )| ≤ ems1 sup |u(x, z) − v +, (x, z)| (4.9)
(x,z)∈R 2
(x,z)∈R2

for any function u(x, z). We can take T1 large enough to satisfy

∗
ems1 sup |w + (x, z, t; v + ) − v + (x, z − )| ≤ (4.10)
(x,z)∈R2
3

for t ≥ T1 by the definition of w + . Combining (4.9) and (4.10) with u = ut , we have

|w(x, z, s1 ; ut ) − v ∗ (x, z − )|

≤ |w(x, z, s1 ; ut ) − w(x, z, s1 ; v +, )| + |w(x, z, s1 ; v +, ) − v ∗ (x, z − )|

≤ 23
∗ .

for any t ≥ T1 . Thus,

w(x, z, t + s1 + T ) ≤ w(x, z, s1 ; ut ) ≤ v ∗ (x, z − ) + 23

∗ (4.11)

holds true for t ≥ T1 . By (4.7), (4.8), (4.11) and Lemma 4.6, we have

w(x, z, t; v − ) ≤ w(x, z, t) ≤ v ∗ (x, z) +

∗ = v∗ (x, z) +
∗
226 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233

for (x, z) ∈ R2 and t ≥ s1 + T1 + T . Since v∗ (x, z) = limt→∞ w(x, z, t; v − ), we have

completed the proof.

5. Proof of key lemmas

Now we prove Lemma 4.5.

Proof of Lemma 4.5. For any fixed T > 0, we prove that w(x, z, T ) satisfies (1.11).
We define
 
z − (x)
V (x, z) =

.
1 +  (x)2

Lemma 2.1 implies

lim sup |v − (x, z) − V (x, z)| = 0.

R→∞ x 2 +z2 >R 2

This relation and the assumption on the initial condition give

lim sup |u0 (x, z) − V (x, z)| = 0.

R→∞ x 2 +z2 >R 2

We define

def
W (x, z, t) = w(x, z, t) − V (x, z).

Then, we have

Wt − Wxx − Wzz − cWz − f (W + V ) + f (V ) = h(x, z).

Here

def
h(x, z) = −L[V ]

satisfies

lim sup |h(x, z)| = 0.

R→∞ x 2 +z2 >R 2

Using

−f (W + V ) + f (V ) = −f  (V + W )W
 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 227

for some 0 < (x, z, t) < 1, we obtain

Wt − Wxx − Wzz − cWz − f  (V + W )W = h(x, z) for (x, z) ∈ R2 , t > 0,

W (x, z, 0) = W0 (x, z) for (x, z) ∈ R2 .

Here

def
W0 (x, z) = u0 (x, z) − V (x, z)

satisfies

lim sup |W0 (x, z)| = 0.

R→∞ x 2 +z2 >R 2

# (x, z, t) by
We define W

# ] def
L̃[W = W #t − W
#xx − W
#zz − cW # = |h(x, z)|
#z − M∗ W for (x, z) ∈ R2 , t > 0,
# (x, z, 0) = |W0 (x, z)|
W for (x, z) ∈ R2 .

Here, we define

|f  (s)|.
def
M∗ = sup
|s|≤2+u0 L∞ (R2 )

The maximum principle yields

# (x, z, t) ≥ 0
W for (x, z) ∈ R2 , t ≥ 0.

Then, we have
 2 2

* * * * # − W ) = (M∗ − f  (V + W ))W
#≥0
− 2 − 2 − c − f  (V + W ) (W
*t *x *z *z

for (x, z) ∈ R2 , t > 0 with the initial condition

# − W )|t=0 ≥ 0
(W for (x, z) ∈ R2 .

The maximum principle yields

# (x, z, t) for (x, z) ∈ R2 , t > 0.

W (x, z, t) < W
228 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233

Applying the argument also to (−W ), we obtain

# (x, z, t) for (x, z) ∈ R2 , t > 0.

|W (x, z, t)| < W

Let Z(x1 , x2 , t) be the kernel of L̃, that is,

 
1 x12 + (x2 + ct)2
def
Z(x1 , x2 , t) = exp − + M∗ t .
4t 4t

Then one has

# (x1 , x2 , t) = I (x1 , x2 , t) + J (x1 , x2 , t),

W

where
" t "
def
I = d Z(x1 − 1 , x2 − 2 , t − )h( 1 , 2 ) d 1 d 2 ,
0 R2
"
def
J = Z(x1 − 1 , x2 − 2 , t)W0 ( 1 , 2 ) d 1 d 2 .
R2

For an estimate for the kernel function, one has

 
c1 x12 + x22
|Z(x1 , x2 , t)| ≤ exp −c2 for 0 < t ≤ T ,
t t

where c1 , c2 depend only on T and M∗ . For example, we can take

 
1 c2 T 1
c1 = exp M∗ T + , c2 = .
4 4 8

By direct calculations we have

"
√ √
e−c2 (1 +2 ) |W0 (x1 +
2 2
|J | ≤ c1 t 1 , x2 + t 2 )| d 1 d 2 . (5.1)
R2

There exists 0 < t1 < t < T with

"
I =t Z(x1 − x2 , 1 − 2 , t − t1 )h( 1 , 2 ) d 1 d 2 .
R2
 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 229

Thus we get
"
√ √
e−c2 (1 +2 ) |h(x1 +
2 2
|I | ≤ c1 T t − t1 1 , x2 + t − t1 2 )| d 1 d 2 . (5.2)
R2

Gathering (5.1) and (5.2), we obtain

lim sup # (x1 , x2 , T ) = 0

W
R→∞
x12 +x22 >R 2

and thus

lim sup |W (x, z, T )| = 0.

R→∞ x 2 +z2 >R 2

This complete the proof of Lemma 4.5.

Proof of Lemma 4.6. We prove this lemma by contradiction using a similar argument
as in [4,13]. Assume v∗  ≡ v ∗ . Then the strong maximum principle and

v∗ (x, z) ≤ v ∗ (x, z)

yield

v∗ (x, z) < v ∗ (x, z) for all (x, z) ∈ R2 .

Take  and  as in (4.6). For any 0 <  < 1 /2, we take  large enough so as to get

v ∗ (x, z) ≤ v∗ (x, z − ) + 

using (2.8). Lemma 4.4 says that

w + (x, z − , t; v∗ ) = v∗ (x, z −  − (1 − e−t )) + e−t

def

is a supersolution. Letting t → ∞ in

v ∗ (x, z) ≤ v∗ (x, z −  − (1 − e−t )) + e−t ,

we obtain

v ∗ (x, z) ≤ v∗ (x, z −  − ).

230 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233

Define

 = inf{ | v ∗ (x, z) ≤ v∗ (x, z − )}.

def

Then, we have  ≥ 0 and

v ∗ (x, z) ≤ v∗ (x, z − ). (5.3)

Now, we show

lim sup max{|(v∗ )z (x, z − )|, |(v ∗ )z (x, z − )|} = 0. (5.4)

R∗ →∞ |z−m∗ |x||≥R∗

We show this equality only for v∗ , since the same argument is valid for v ∗ . Assume
the contrary, then there exist
1 > 0 and {(xn , zn )}∞
n=1 with

lim |zn − m∗ |xn || = ∞, −(v∗ )z (xn , zn ) ≥

1 .
n→∞

For simplicity we assume limn→∞ (zn − m∗ |xn |) = ∞. Define

def
vn (x, z) = v∗ (x + xn , z + zn ) in B0 . (5.5)

Because of limR→∞ supx 2 +z2 >R 2 |v∗ (x, z) − v − (x, z)| = 0, we have

lim sup |vn (x, z) + 1| = 0,

n→∞
B0

lim sup |f (vn (x, z))| = 0.

n→∞
B0

Applying the Schauder interior estimate to (5.5), we get

vn → −1 in C ∗ (0 ).

Here, ∗ is as in (4.5) and B0 is as in (4.4). Thus we get

lim (vn )z (0, 0) = lim (v∗ )z (xn , zn ) = 0,

n→∞ n→∞

which contradicts the assumption. If we assume limn→∞ (zn − m∗ |xn |) = −∞, we get
the same contradiction. Thus we proved (5.4).
 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 231

We prove  = 0 by contradiction. Assume the contrary. Take R∗ so large that

2 sup max{|(v∗ )z (x, z − )|, |(v ∗ )z (x, z − )|} < 1 (5.6)

|z−m∗ |x||≥R∗ −1

holds true. By (2.8) and the strong maximum principle, we have

v ∗ (x, z) < v∗ (x, z − ) in D,

where

def
D = {(x, z)| |z − m∗ |x| | ≤ R∗ }.

Again by (2.8) we can choose a small positive constant h with


1 
0 < h < min , ,
2 2

v ∗ (x, z) < v∗ (x, z −  + 2h) in D.

In R2 \D we have

" 1
v∗ (x, z −  + 2h) − v∗ (x, z − ) = 2h (v∗ )z (x, z −  + 2s h) ds ≥ −h,
0

using (5.6). This implies

v∗ (x, z − ) ≤ v∗ (x, z −  + 2h) + h, in R2 \D.

Combining the inequalities in D and R2 \D, we get

v ∗ (x, z) ≤ v∗ (x, z −  + 2h) + h in R2 .

Now

w++ (x, z, t) = v∗ (x, z −  + 2h − h(1 − e−t )) + he−t

def
232 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233

is a supersolution by Lemma 4.4. Since

v ∗ (x, z) ≤ w++ (x, z, 0),

holds, the comparison principle yields

v ∗ (x, z) ≤ w++ (x, z, t) for t ≥ 0.

By sending t → ∞, this inequality implies

v ∗ (x, z) ≤ v∗ (x, z −  + h),

which contradicts the definition of . Thus,  = 0 follows and v∗ ≡ v ∗ holds true.

This completes the proof.

Acknowledgments

The authors thank Professor Eiji Yanagida of Tohoku University for stimulating
discussions.

References

[1] A. Bonnet, F. Hamel, Existence of nonplanar solutions of a simple model of premixed bunsen flames,
SIAM J. Math. Anal. 31 (1999) 80–118.
[2] P.K. Brazhnik, Exact solutions for the kinematic model of autowaves in two-dimensional excitable
media, Physica D 94 (1996) 205–220.
[3] X. Chen, Generation and propagation of interfaces for reaction–diffusion equations, J. Differential
Equations 96 (1) (1992) 116–141.
[4] X. Chen, Existence, uniqueness and asymptotic stability of traveling waves in nonlocal evolution
equations, Adv. Differential Equations 2 (1) (1997) 125–160.
[5] K. Deckelnick, C.M. Elliott, G. Richardson, Long time asymptotics for forced curvature flow with
applications to the motion of a superconducting vortex, Nonlinearity 10 (1997) 655–678.
[6] P.C. Fife, Dynamics of Internal Layers and Diffusive Interfaces (CBMS-NSF Reg. Conf. Ser. Appl.
Math. 53) SIAM, Philadelphia, PA, 1988.
[7] D. Gilbarg, N.S. Trudinger, Elliptic Partial Differential Equations of Second Order, Springer, Berlin,
1983.
[8] F. Hamel, R. Monneau, Existence and uniqueness for a free boundary problem arising in combustion
theory, Interfaces Free Boundaries 4 (2002) 167–210.
[9] F. Hamel, N. Nadirashvili, Travelling fronts and entire solutions of the Fisher-KPP equation in RN ,
Arch. Rational Mech. Anal. 157 (2001) 91–163.
[10] V. Pérez-Muñuzuri, M. Gómez-Gesteira, A.P. Muñuzuri, V.A. Davydov, V. Pérez-Villar, V-shaped
stable nonspiral patterns, Phys. Rev. E 51 (2) (1995) 845–847.
[11] H. Ninomiya, M. Taniguchi, Traveling curved fronts of a mean curvature flow with constant driving
force, in: Free Boundary Problems: Theory and Applications I, GAKUTO Internat. Ser. Math. Sci.
Appl. 13 (2000) 206–221
 H. Ninomiya, M. Taniguchi / J. Differential Equations 213 (2005) 204 – 233 233

[12] H. Ninomiya, M. Taniguchi, Stability of traveling curved fronts in a curvature flow with driving
force, Methods Appl. Anal. 8 (2) (2001) 429–450.
[13] T. Ogiwara, H. Matano, Monotonicity and convergence results in order-preserving systems in the
presence of symmetry, Discrete Continuous Dynamical Systems 5 (1) (1999) 1–34.
[14] P. Pelcé, Dynamics of Curved Fronts, Academic Press, New York, 1988.
[15] M.H. Protter, H.F. Weinberger, Maximum Principles in Differential Equations, Springer, Berlin, 1984.
[16] D.H. Sattinger, Monotone methods in nonlinear elliptic and parabolic boundary value problems,
Indiana Univ. Math. J. 21 (11) (1972) 979–1000.