343

Math. Proc. Camb. Phil. Soc. (1997), 122, 343

Printed in the United Kingdom

On the limiting absorption and amplitude principles for transmission

problems with unbounded interfaces
By BO ZHANG
Department of Mathematics and Statistics, Brunel University,
Uxbridge UB8 3PH, UK
(Received 4 October 1995)
Abstract
We consider the transmission problem for second-order differential equations with
unbounded interfaces. The spaces B and B are used to prove the limiting absorption
principle for the steady-state problem and to establish the resolvent estimate at low
frequencies for the steady-state operator. These are then applied in showing the
validity of the limiting amplitude principle for such an operator.

1. Introduction
In this paper we mainly deal with the asymptotic behaviour (as t +) of the
equation
(t, x) R+

Wtt C 2 . (1 aW ) = f (x)eit ,

(11)

under the transmission conditions

W1 (t, x) = W2 (t, x),
1
(1) (a(x)W1 (t, x)) . (x)

(t, x) R+

1
(2) (a(x)W2 (t, x)) . (x),

(12)
+

(t, x) R (13)

and the initial conditions

W (0, x) = Wt (0, x) = 0,

x .

(14)

Here , are two non-zero complex number, R+ = (0, ), = 1 x 2 and j (j =

1, 2) are two unbounded domains in Rn (n > 2) separated by an infinite surface
C 2 , (x) = (1 (x), . . . , n (x)) is the unit normal to at x drawn in the
direction from 1 to 2 , W = (W1 , W2 ), Wt = W/t, is a positive constant, a(x) is
a real-valued symmetric and positively definite matrix which tends to the identity
matrix I as r = |x| , C and are two positive functions tending to Cj > 0 and
j > 0, respectively, as r in j (j = 1, 2), f is an L2 function and (j) (x) is the
value of (x) when x approaches the interface from j with j = 1, 2.
Throughout this paper we assume that the following assumptions hold.
> 0.
(A1)
(A2) There is a d > 0 such that d < C(x) < 1/d, d < (x) < 1/d, and
|(x) j | + || + |C(x) Cj | = O(|x|1 )
as |x| in j (j = 1, 2) for some 0 < < 1.

344

Bo Zhang

(A3) a(x) = [akl ]nn is a real-valued symmetric matrix with akl C 2 (Rn ), and
there is a constant C0 > 0 such that
. (a(x)) =

n
X

k, l=1

akl (x)k l > C0 ||2 ,

x, Rn .

1
) as |x| in Rn .
(A4)
i, j, k |k aij | +
i,j,k,l |k l aij | = O(|x|
(A5) There is a constant R0 > 0 such that for all x with |x| > R0

(x) . (a(x)x) = 0.
Under the above assumptions we prove in this paper the validity of the limiting
absorption principle (LABP) for the operator A = C 2 . (1 a) associated with
the problem (11)(14), and the validity of the limiting amplitude principle (LAMP)
for the problem (11)(14). The last statement means that the solution W = (W1 , W2 )
of the problem (11)(14) has the property that if f L2 , then
W = u+ exp (it) + o(1)
in the norm of L
2 () with >
the problem

3
2

(15)

as t +, where u+ = (u1 , u2 ) is the solution of

. (1 au+ ) 2 u+ = f,
u1
1
(1) (au1 ) .

= u2 ,
=

in ,

(16)

on

(17)

1
(2) (au2 ) . ,

on ,

(18)

which is given by LABP:

u+ = R( 2 + i0)f = lim R( 2 + i )f
0+

(19)

with > 21 . Here (x) = C 2 (x)1 (x), R(z), with Iz  0, is

in the norm of L
2
the resolvent of the self-adjoint realization of the operator A under the transmission
conditions (17) and (18) in a weighted L2 space, and L2 = L2 () is the weighted
L2 space defined in the next section. Define q = q(x) = qj = Cj2 1
j (> 0) for x in
j (j = 1, 2). Then by (A2)
|(x) q| = O(|x|1 )

(110)

as |x| in .
The limiting absorption and amplitude principles for A have been proved in [3,
10, 12] for a paraboloid-type interface in the case when a = I, = 1, = = 1 and
C(x) = Cj in j , and in [9, 11, 14] in the case when a = I, = 1 and is a local
perturbation of a hyperplane by introducing a radiation condition to the problems.
There is a large literature about these two principles (see, for example, the above
work and [1, 2, 4, 5, 6, 7, 13] and the references there).
The limiting amplitude principle for the problem (11)(14) is proved based on
a general theory developed by Eidus [2, chapter 1] for studying the asymptotic
behavior, for large time, of solutions to the Cauchy problem for the abstract wave
equation. By this theory, we first have to prove the limiting absorption principle for
the associated steady-state problem (16)(18) with any > 0 instead of 2 . Then
we show that the solution u of (16)(18) with any > 0 instead of 2 is Hoelder

345

Transmission problems with unbounded interfaces

continuous in . Finally we study the asymptotic behaviour of u as 0. These

properties will be examined precisely in what follows.
2. Preliminaries
We use the following abbreviations
(R) = {x ||x| < R},
S(R) = {x ||x| = R},
e (R) = \(R),

E(R) = \(R)
(R) = {x ||x| < R}

(R1 , R2 ) = {x |R1 < |x| < R2 }

(R1 , R2 ) = {x |R1 < |x| < R2 },

B(R) = {x Rn ||x| < R}.

Let be a real number. For an open set V Rn and a non-negative integer k We

define the weighted Sobolev space Hk (V ) by
Hk (V ) {u H k (V w B(R))R > 0| < x > u L2 (V ),

|| 6 k}

with the weighted norm

kuk2k, , V

X Z

||6k

< x >2 | u|2 dx,

< x >= (1 + |x|2 ) 2 ,

where = (1 , . . . , n ), j > 0 is an integer, = 11 nn , j = /xj , and

H k (V ) is the standard Sobolev space with the usual inner product (, )k,V . We write
H0 (V ) = L2 (V ). We also define the spaces B(V ) and B (V ) by

! 12
Z

X
<
|v(x)|2 dx
Rj
B(V ) = v L2 (VR )R > 0|kvkB(V ) =

Vj
j=1

and


Z
2
1
B (V ) = v L2 (VR )R > 0|kvkB (V ) = sup R

R>1

VR

|v(x)| dx

< + ,

where R0 = 0, Rj = 2j1 , j > 1, Vj = {x V |Rj1 < |x| < Rj } and VR = {x

V ||x| < R}. When = 0 or k = 0, the subscript or k will be omitted from the
norm and inner product. When V = , V will be also omitted from the norm,
inner product, space and integral. We remark that for every > 12 the continuous
inclusions
L2 (V ) B(V ) L2 (V ) B (V ) L
2 (V )

(21)

hold [14].
2j for x in j (j = 1, 2) and to
It is convenient to define = (x) = j = ()
introduce the weighted Hilbert space H0 = L2 (; dx) with the inner product
Z
(u, v)0 =
uv dx.

R
Evidently, the inner product (, )0 is equivalent to the usual one (, ) = uv dx by
(A2). In order to obtain a Hilbert space formulation of the problem (11)(14) with
a self-adjoint operator, we introduce the following Hilbert space
H = {u = (u1 , u2 ) H 1 () w H0 |u1 = u2

on }

346

Bo Zhang

with the norm kukH

by

P2
1
= ( j=1 kuj k2j, j ) 2 , and the bilinear functional on H H defined

a(u, v) =

v 1 (au1 )dx +

1 v 2 (au2 )dx

for u, v H. Define the operator A by

D(A) = {u H|h H0 , v H, a(u, v) = (h, v)0 } H0
and Au = h. Thus
1
D(A) = {u = (u1 , u2 ) H|1
(1) (au1 ) . = (2) (au2 ) .

on , . (1 au) L2 }

and
Au = 1 . (1 au)

for u D(A).
P2
Remark 21. By (A1) and (A3), a(u, u) > C1 j=1 kuj k20, j > 0 for u H and
u  0, A is symmetric and positive with C\[0, ) (A), the resolvent set of A, and
D(A) = H0 . Furthermore, we claim that A is closed and self-adjoint. Indeed, it easily
follows that A is closed. Since A is symmetric, A A . Hence, from the fact that
D(A ) = D(A) R(A + i) R(A i) [8, p. 17], we obtain that D(A) = D(A ) and
therefore, A = A , that is, A is self-adjoint. The operator A is the unique self-adjoint
realization of the operator 1 . (1 a) with the transmission conditions (17)
(18).
Remark 22. By elliptic regularity estimates together with (A2) and (A3) it follows
that D(A) H 2 ().
Define the function W (t) H0 on R+ by W (t)(x) = W (t, x) for all x . Then
Wt (t, x) can be thought of as dW (t)/dt, the derivative of the H0 -valued function W (t),
and the problem (11)(14) can be written in the following Hilbert space formulation:
find W (t) H0 such that
d2 W
+ AW = eit f,
dt2
dW
(0) = 0,
W (0) =
dt

t>0

(22)
(23)

where f H0 .
Remark 23. Since A is self-adjoint and positive, the spectral theorem holds
for A. For > 0 and f H0 , the problem (22)(23) has a unique solution
1
W C 2 (R+ , H0 ) w C 1 (R+ , D(A 2 )) w C(R+ , D(A)) which is given by


Z

i
1
it
W (t) =
+ sin [ ()t] cos [ ()t] dP ()f,
e
2

where P () is the spectral family of A (see [2, p. 113]).

3. A uniqueness theorem

In this section we establish a uniqueness theorem which plays an important role in

proving LABP and LAMP. As an application, we also prove the absence of eigenvalues
for A in this section.

Transmission problems with unbounded interfaces

347

Theorem 31. Let k > 0 and let u H ((R)) for all R > 0 be a solution of the
homogeneous equation
(1 au) k 2 u = 0,

(31)

satisfying the transmission conditions (17)(18) and the condition

Z
[|u|2 + |u|2 ]ds = 0.
lim inf
R

(32)

S(R)

If
X
i, j

|aij ij | = O(|x| )

(33)

as |x| , then u = 0 a.e. in , where ij is the Kronecker delta.

Proof. The theorem can be proved similarly as in the proof of theorem 3.1 in [14]
by making use of (A2)(A5), (31)(33) and the transmission conditions (17)(18)
together with the basic identity

2R{ . (1 au)rh[x . (au)]}

1

au)} h1 u (au)(ax)]
= [2R{rh[x (au)](

+{h1 (ax) + r h1 [x (ax)] + h1 (ax)}u (au)

X
alk xi l aij k uj u
+h1 u {[a (ax)]u} 2h1 R
i, j, k, l

2h |au| 2rr h |x (au)|2 ,

where a =

[i akl ]ni, k, l=1 .

(34)

(See [15] for details.) The proof is complete.

From Theorem 31 the absence of the eigenvalues for A easily follows and we have
the following result.
Corollary 31. The operator A has no eigenvalues.
4. A priori estimates
Let z = + i with > 0 and  0, f L2 (), and u = R(z)f, where R(z) =
(A z)1 is the resolvent of A in H0 . Then u = (u1 , u2 ) H 2 () and
(1 au) zu = f,
u1
1
(1) (au1 ) .

= u2 ,
=

in

(41)

on

(42)

1
(2) (au2 ) . ,

on .

(43)

In this section we shall establish some a priori estimates for u which is necessary in
the proof of LABP and LAMP. To this end, we make the following assumptions:
X
|aij ij | = O(|x|1 )
(44)
i, j

and
|x|
as |x| .

i, j, k, l

|k l aij | +

i, j, k

|k aij | + || = (|x|2 )

(45)

348

Bo Zhang

We begin with the following result which can be obtained by using LaxMilgram
theorem and elliptic regularity estimates.
Lemma 41. For any f L2 (), there is a constant C independent of z, u and f such
that for any > 0
kuk2, 6 C(kf k0, + kuk0, ).

(46)

integrating over , taking the real and imaginary parts

Multiplying (41) by u,
of both sides of the equation thus obtained and making use of (A3), (42) and (43),
we can easily obtain the following lemma.
Lemma 42. There is a constant C, independent of z, u and f, such that
Z
| | (|u|2 + |u|2 )dx 6 C(|f |, |u|)0 .

(47)

The following results play an important role in establishing the a priori estimate
for u.
Theorem 41. (i) Assume that (44) holds. Let f B(), n > 2, 0 6 6 1 and
0 < | | 6 1 for some positive numbers 0 , 1 and 1 . Then
kukB + kukB 6 C(kf kB + kuk1, (R1 ) )

(48)

where C is a constant independent of z, u and f, and R1 is a fixed number.

(ii) Assume that (44) and (45) hold. Let f L2 , n > 3, 0 < 6 1 and 0 < | | 6 1 .
Then

kukB + kukB + kuk 6 C(kf k + kuk1, (R1 ) )

(49)
where > 23 .

integrating over E(r1 )

Proof. Multiplying (41) by r1 {2x . (au)+[
. (ax)1]u},
and making use of the basic idendity (34) with h(r) = r1 , we are able to obtain,
by integrating by parts together with (A1)(A5) and (42)(44), that for sufficiently
large r1 (> R0 )
Z
(n 3)(n 1)
2

x)
|x(au)|

2
1 r1 [u(au)
x(a
]dx +
2
E(r1 )
Z
Z
2
2

x)q|u|

1 r3 |u|2 dx +
1 [2|x(au)|
+ x(a

E(r1 )
S(r1 )
Z
Z
n1

x)]ds

1 r2 [|u|2 + rr (|u|2 )]ds 6 C

u(au)
x(a
+
2
S(r1 )
E(r1 )
{r1 [(|z| + P (r))|u|2 + (|z| + Q(r))|u|2 ] + |f |(|u| + |u|/r)}dx

Z
Z
1
2
2

r
(N (r)|u| + |u| )ds + 2 I
u x(au)dx,
+
S(r1 )

(410)

E(r1 )

where P (r) = Q(r) = N (r) = 1 when only (44) holds, and P (r) = r1 , Q(r) = r3
and N (r) = r2 when both (44) and (45) hold. Make use of (21), Lemma 42 and

Transmission problems with unbounded interfaces

349

the fact that |z| < + | | to deduce

Z
(n 3)(n 1)
2

x)
|x(au)|

2
1 r1 [u(au)
x(a
]dx +
2
E(r1 )
Z
Z
2
2

x)q|u|

1 r3 |u|2 dx +
1 [2|x(au)|
+ x(a

E(r1 )
S(r1 )
Z
n1

x)]ds

1 r2 [|u|2 + rr (|u|2 )]ds

u(au)
x(a
+
2
S(r1 )

/2
6 C(1 ) r1 [( + P (r1 ))kuk2B + ( + m)kuk2B + M (r1 )k < x >1 uk2B ]

Z
1
2
2
+
r
(N (r)|u| + |u| )ds + (|f |, |u| + |u|)
(411)
S(r1 )

for r1 > t with sufficiently large t > 1, where m = 1 and M (r1 ) = 0 when only (44)
holds, and m = 0 and M (r1 ) = r11 when both (44) and (45) hold. Integrating (411)
multiplied by R1 with respect to r1 from t to R gives



Z
Z
n1
2t
1
1
2

x)

u(au)
+ q|u| dx +
x(a
1
R (t,R)
r
2R (t,R)
 Z
(n 3)(n 1) 1
1 r2 [|u|2 + rr (|u|2 )]dx +
1 r2 |u|2 dx
2
R (t,R)

Z
Z
t
+
1 r3 |u|2 dx
1 r3 |u|2 dx 6 C(1 ){(|f |, |u| + |u|)
R E(t)
E(R)
+R/2 [( + P (R))kuk2B + ( + m)kuk2B + M (R)k < x >1 uk2B ]}. (412)

By integrating by parts the second integral on the left-hand side of (412) and making
use of (A1)(A4) it is obtained that

 Z
Z
(n 3)(n 1)
1
2
2
2
2
r |u| dx +
[|u| + |u| ]dx +
C
R (R)
2R
S(R)

Z

(R t)1 r3 |u|2 dx 6 C(1 ) R/2 [( + P (R))kuk2B

E(R)

+( + m)kuk2B + M (R)k < x >1 uk2B ]


Z
t
3
2
1
2
r |u| dx + (|f |, |u| + |u|) + R kuk1, (4t) .
+
R (t,R)

(413)

(i) Assume that (44) holds. Let f B, 0 < 0 6 6 1 and n > 2. Then by (21)
and (413) we have for R > t > 1
Z
1
[|u|2 + |u|2 ]dx 6 C(0 , 1 ){t/2 [kuk2B + kuk2B ]
R (R)
+(|f |, |u| + |u|) + kuk21, (4t) },

(414)

which together with Schwarzs inequality implies the estimate (48) for sufficiently
large t with R1 = 4t.
(ii) Assume that (44) and (45) hold. Let f L2 with > 32 , n > 3 and 0 < 6 1 .

350

Bo Zhang

Then it follows from (413) that

Z
Z
1
r2 |u|2 ds 6 C(1 )
[|u|2 + |u|2 ]dx +
R (R)
S(R)

{t/2 [kuk2B + kuk2B + k < x >1 uk2B ] + (|f |, |u| + |u|) + kuk21, (t1) } (415)

for R > t2 , where t1 = max (4t, t2 ). Multiply (415) by T 1 and integrate the inequality thus obtained with respect to R from t1 to T to yield
Z
1
(1 + r2 )1 |u|2 dx 6 C(1 ){t/2 [kuk2B + kuk2B + k < x >1 uk2B ]
T (T )
+(|f |, |u| + |u|) + kuk21, (t1) }

(416)

for T > t1. From (415) and (416) it follows on using Schwarzs inequality that
kuk2B + kuk2B + k < x >1 uk2B 6 C{k < x > f k2B + kuk21, (t1) }

(417)

for sufficiently large t, which combined with (21) implies the required estimate (49)
with R1 = t1. The theorem is thus proved.
From the inequalities (413) and (417) and Theorem 41 the following result easily
follows.
Lemma 43. There exists a constant r1 > 0 such that for all R > r1
Z
1
[|u|2 + |u|2 ]dx 6 C{R/2 kuk21, (R1 ) + R/2 kf k2B },
R (R)

(418)

if (44) holds, f B, 0 < 0 6 6 1 and n > 2; and

Z
Z
1
r2 |u|2 ds
[|u|2 + |u|2 ]dx +
R (R)
S(R)


Z
1
3
2
/2
/21
2
/2
2
6C R
r |u| dx (419)
( + R
)kuk1, (R1 ) + R kf k +
R (1, R)
if (44) and (45) hold, f L2 , 0 < 6 1 and n > 3; where and R1 are the same as
in Theorem 41.
By using a compact argument combined with Lemma 43 and Theorem 31 the
term kuk1, (R1 ) in Theorem 41 can be removed from the right-hand side of (48) and
(49).
Theorem 42. (i) Assume that (44) and (45) hold. Let f L2 with > 23 , n > 3,
0 < 6 1 and 0 < | | 6 1 . Then there is a constant C > 0, independent of z, u and
f, such that

(420)
kukB + kukB + kuk 6 Ckf k .
(ii) Assume that (44) holds. Let f B(), n > 2, 0 < 0 6 6 1 and 0 < | | 6 1 .
Then
kukB + kukB 6 Ckf kB .

(421)

Proof. (i) To prove (420) it is enough to show that there is a constant C > 0 such
that
kuk1, (R1 ) 6 Ckf k

(422)

Transmission problems with unbounded interfaces

351

for all (0, 1 ] and | | (0, 1 ]. If (422) were false, then there would be sequences
{zm } with zm = m + im , {vm } and {fm } such that
kvm k1, (R1 ) = 1,

kfm k 0

(423)

and zm z = + i with 0 6 6 1 and 0 6 | | 6 1 , as m , where

vm = R(zm )fm . By Theorem 41,

(424)
kvm kB + m kvm kB + kvm k 6 C(kfm k + 1) < ,
which combined with Rellichs selection theorem and standard elliptic estimates
asserts that there is a subsequence of {vm }, denoted by {vm } again, such that vm
2
strongly converges to some function v in H
() and |vm | strongly converges to

1
|v| in L2 () with > 2 as m . Clearly, v satisfies (41)(43) with f = 0, and
from (424)

kvkB + kvkB + kvk 6 C.

(425)
Next, we prove that v = 0 a.e. This contradicts (423), and the required estimate
(420) then follows. There are three cases to be distinguished.
Case 1. > 0 and = 0.
By Lemma 43, vm satisfies (419) with replaced by m and f by fm , and on
passing the limit as m in the inequality thus obtained it follows that v satisfies
the condition (32) of Theorem 31. Thus v = 0 a.e. by Theorem 31.
Case 2. > 0 and | | > 0.
In this case (47) can be used to yield that v = 0 a.e.
Case 3. z = 0, that is, = = 0.
In this case,
. (1 av) = 0,

v1
1
(1) (av1 ) .

in

= v2 ,
=

(426)

on

(427)

1
(2) (av2 ) . ,

on ,

(428)

2
() with > 23 , and by Lemma 43 we have
where v = (v1 , v2 ) H


Z
Z
Z
1
|v|2 dx +
r2 |v|2 ds 6 CR1 1 +
r3 |v|2 dx
R (R)
S(R)
(1, R)

(429)

for R > r1. Multiply (429) by R1 and integrate the inequality thus obtained with
respect to R from t to T with t > r1 to get


Z
Z
3
2
1
3
2
1+
r |u| dx 6 Ct
r |v| dx .
(1, T )

(t,T )

For sufficiently large t we have


Z
Z
3
2
1
1+
r |u| dx 6 Ct
(t, T )

(1, t)

|v| dx ,
1

which, on letting T , implies < x > 2 u L2 (), where < x >= (1 + |x|2 ) 2 .
This together with (429) asserts that |v| L2 ().

352

Bo Zhang

On the other hand, by (411), (417), (423), (21), (A1)(A3) and Schwarzs inequality it can be seen that vm satisfies


Z
Z
2
2
/2
1
2
/2
2
r |vm | ds 6 C R
(m + R ) +
(430)
|vm | ds + R kfm k
S(R)

S(R)

for R > t with sufficiently large t > 1. Letting m in (430) leads to



Z
Z
2
2
1/2
2
r |v| ds 6 C R
+
|v| ds ,
S(R)

(431)

S(R)

which in conjunction with the fact that |v| L2 () guarantees

Z
[r1 |v|2 + r|v|2 ]ds = 0.
lim inf
R

(432)

S(R)

integrate over and make use of (427)(428), (432) and

Multiply (426) by v,
partial integration to obtain
Z
1 v . (av)dx = 0.

This combined with (A1)(A3) implies vj = 0 a.e. in j , that is, vj = Cj a.e. in j ,

where Cj is a constant (j = 1, 2). However (432) asserts that Cj = 0 (j = 1, 2). Hence
v = 0 a.e. in , and the estimate (420) is thus proved.
(ii) The estimate (421) can be proved similarly. In fact, in this case only Cases 1
and 2 above should be distinguished since z  0. The proof is complete.
5. More a priori estimates

In order to prove the limiting absorption and amplitude principles we need more a
priori estimates for u = R(z)f with Iz  0, which will be established in this section.
To this end, we introduce the following interpolation inequality for which a proof
can be found in [3].
Lemma 51. Let 1 < 2 < 3 and let v Hm3 () with m > 0 an integer. Then
1/q
kvkm, 2 6 kvk1/p
m, 1 kvkm, 3

(51)

where p = (3 1 )/(3 2 ) and q = (3 1 )/(2 1 ).

By combining Lemma 51 with Lemma 42 and Theorem 42 (ii) the following
result can be obtained.
Lemma 52. Assume that (44) holds. Let u = R(z)f with f B(), z = + i,
0 < 0 6 6 1 , and 0 < | | 6 1 , and let n > 2. Then
kuk1, 6 C| |(1)/2 kf kB ,

(52)

where 0 6 6 21 .
Proof. By Lemma 42, Theorem 42 (ii) and Schwarzs inequality we have
kuk21, 0 6 | |1 (|f |, |u|)0 6 C| |1 kf k2B .

(53)

On the other hand, Theorem 42 (ii) together with (21) gives

kuk1,1 6 Ckf kB .

(54)

353

Transmission problems with unbounded interfaces

By using (53) and (54) in conjunction with Lemma 51 the required estimate (52)
follows on taking 1 = 1, 2 = and 3 = 0. The proof is complete.
Lemma 53. Assume that (44) holds. Let u(j) = R(z)f (j) with f (j) L12 (), z = + i,
0 < 0 6 6 1 , and 0 < | | 6 1 , and let n > 2. Then
|(u(1) , u (2) )| 6 C| |0 kf (1) k1 kf (2) k1 ,

(55)

where C, 0 are positive constants independent of z, u(j) and f (j) , = ()2j in j with
j = 1, 2 and 21 < 0 < 1.
Proof. It follows from the definition of R(z) that u(j) H 2 () satisfies (41)(43)
with u, f replaced by u(j) , f (j) respectively, where j = 1, 2. Multiply (41) with u(j)
and f (j) replacing u and f, respectively, by [x . (au(3j) )], integrate it over E(R)
with R > R0 and make use of (A5), the transmission conditions (42)(43) and the
basic identity
. (1 au)[x . (av)] + . (1 av)[x . (au)]

= . {[x . (av)](1 au) + [x . (au)](1 av) 1 u . (av)(ax)}

+{1 . (ax) + 1 . (ax)}u . (av) + 1 u . {[a (ax)]v}

X
alk xi l aij [k vj u + k uj v] 21 (au)(av).
1

(56)

i, j, k, l

By using partial integration and CauchySchwarz inequality together with (A2),

(A4), (110) and (44) it can be obtained that
Z

Z





1
(1)
(2)
(1) (2)
(1) (2)



2|z| 
u u dx 6 (n 2) 
[ u (au ) zu u ]dx + C
E(R)

E(R)

(1)

(2)

(1)

(2)

ku k1,/2 ku k1,/2 + Rku k1,S(R) ku k1,S(R) +

2
X
j=1

(j)

kf k1 ku

(3j)

k . (57)

On the other hand, integrating by parts combined with (41)(43) with u, f replaced
with u(j) , f (j) respectively, where j = 1, 2, implies
Z
Z
Z
1
(1)
(2)
(1) (2)
(58)
u (au )dx = z u u dx + f (2) u(1) dx

for sufficiently large R > R0 . From (57) and (58) we obtain, on using Cauchy
Schwarz inequality and applying standard elliptic estimates to the integrals over
S(R) together with (110), Theorem 42 (ii) and (21), that
Z



|z|  u(1) u(2) dx 6 CR

)
(
2
X
ku(1) k1, /2 ku(2) k1, /2 + kf (1) k1 kf (2) k1 +
kf (j) k1 ku(3j) k , (59)
j=1

where CR is a positive constant depending on R. Now combining (59) with Lemma

52 and the fact that 0 < < 1 gives the required estimate (55) with 0 = 1 /2.
This completes the proof.

354

Bo Zhang

Theorem 51. Assume that (44) holds. Let u = R(z)f with f L12 (), z = + i,
0 < 0 6 6 1 , and 0 < | | 6 1 , and let n > 2. Then we have

du

(510)
6 C| |0 kf k1 ,
dz
2, 1
where

1
2

< 0 < 1 is the same as in Lemma 53.

Proof. By the first resolvent identity we have

dR(z)
= R2 (z).
dz

(511)

Hence for any g L12 () H0 ,



dR(z)

f, g = (R(z)f, R(z)g)
dz
since A is self-adjoint in H0 , where is as defined in Lemma 53. By Lemma 53




 dR(z) f, g  6 C| |0 kf k1 kgk1 .


dz

Taking g =< x >2 (dR(z)/dz)f in the above inequality and using (A2) we obtain
that


dR(z)
0


(512)
dz f 6 C| | kf k1 .
1

Next, let v = du/dz. Then by (511) v = R(z)u. Therefore (510) follows from
Lemma 41, Theorem 42 (ii) and (512), and the theorem is thus proved.
6. The limiting absorption and amplitude principles

Theorems 42 and 51 will be used in this section to justify the validity of the
limiting absorption and amplitude principles for the operator A.
Theorem 61. Assume that (44) holds. Let n > 2. Then for each > 0 there exist
the operators R( i0) in L(L12 (), L1
2 ()) such that
lim R( i )f = R( i0)f

0+

(61)

2
() for all f L12 (). Furthermore, the limit R( i0) are local
in the norm of H1
Hoelder continuous in in the strong topology of L(L12 (), L1
2 ()), that is,

kR( i0) R( i0)k11 6 C| |

(62)

for any and in an arbitrarily fixed compact interval of (0, ), where L(X, Y ) is the
Banach space of all bounded linear operators from X into Y, k k denotes the operator
norm when considered as an operator from L2 () into L2 (), 0 < < 1 and C, do not
depend on and .
Proof. (i) The existence of the operators R(i0) and the limits (61) easily follows
from Theorem 51.
(ii) It is sufficient to prove (62) for | | small. Theorem 51 implies that for

Transmission problems with unbounded interfaces

355

kR( i ) R( i0)k11 6 C 10

(63)

any 0 < 6 1

uniformly in in the required region.

On the other hand, from Theorem 51 we have for any 0 < 6 1
kR( i ) R( i )k11 6 C 0 | |.

(64)

On taking = | |, (63) and (64) yield (62) with = 10 . The proof is complete.
Making use of Theorems 61 and 42 together with the same argument as used by
Saito in proving Theorem 43 in [12], we can prove that the limits (61) also exist in
2
the norm of H
() with > 12 for all f B().
Theorem 62 (Limiting absorption principle). Assume that (44) holds. Let n > 2.
Then for each > 0 there exist the operators R( i0) in L(B(), B ()) such that
lim R( i )f = R( i0)f

0+
2
() with >
in the norm of H

1
2

(65)

for all f B(). Moreover, we have

ku kB + ku kB + ku k2, 6 Ckf kB ,

(66)

where u = R( i0)f.
Proof. It is enough to assume that 21 < 6 1. We prove the theorem only for the
+ case. The case can be treated similarly.
Let R( + i0) be the operator in L(L12 , L1
2 ) obtained in Theorem 61, and we first
1
prove (65) for f L2 (). Let {m } be an arbitrary positive sequence such that
m 0 as m , and let um = R( + im )f. Then by Theorem 42 (ii) and Lemma
41
kum kB + kum kB + kum k2, 6 Ckf kB ,

(67)

which combined with Rellichs selection theorem and standard elliptic estimates
implies that there is a subsequence of {um }, denoted by {um } again, such that um
2
strongly converges to a function u in H
(). On the other hand we find from (61)
that the sequence {um } itself converges to R( + i0)f in L1
2 (). As a result we have
u = R( + i0)f and by (67)
kukB + kukB + kR( + i0)f k2, 6 Ckf kB .

(68)

Since {m } is an arbitrary sequence, we conclude that (65) is true for f L12 (). From
(68) and the denseness of L12 () in B() that R( + i0) can be uniquely extended
2
to a bounded operator from B() into H
(). The extension will be denoted by
R( + i0) again. Then, by the use of the denseness of L12 () in B(), (68), Theorem
42 (ii) and Lemma 41, it is easy to see that (65) is true for f B(). The estimate
(66) follows from (68). The theorem is proved.
From Theorems 62 and 42 (i) the following asymptotic behavior easily follows
for the resolvent R( i0) as 0+, which plays an important role in proving the
limiting amplitude principle.

356

Bo Zhang

Theorem 63. Assume that (44) and (45) hold. Let n > 3 and let > 32 . Then
kR( i0)k = O(1)

(69)

as 0 + .
By means of the local Hoelder continuity (62) in (> 0) and the low frequency
behaviour (69) for the resolvent R( i0) the limiting amplitude principle can be
easily shown on following exactly the same argument as in the proof of theorem 11
in [7] (see [7] for details; see also [2, 11, 13]).
Theorem 64. Assume that (44) and (45) hold. Let n > 3 and let W (t) = W (t, x) be
the solution to (11)(14) with f L2 (), > 32 . Then W (t) behaves like
W = exp (it)R( 2 + i0)f + o(1)
strongly in

L
2 ()

(610)

as t +.

Acknowledgements. This work was partly done when the author was in the Department of Mathematics at Keele University, UK.

[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]

REFERENCES
C. O. Bloom. A rate of approach to the steady state of solutions of second-order hyperbolic
equations. J. Differential Equations 19 (1975), 296329.
D. Eidus. The principle of limiting amplitude. Russian Math. Surveys 24 (1969), 97167.
D. Eidus. The limiting absorption and amplitude principles for the diffraction problem with
two unbounded media. Commun. Math. Phys. 107 (1986), 2938.
D. Eidus. Asymptotically periodic solutions of the wave equation in the Hilbert space. Asymptotic Anal. 8 (1994), 217236.
M. Kadowaki. The limiting absorption principle for the acoustic wave operators in two unbounded media. Tsukuba J. Math. 17 (1993), 345362.
M. Kadowaki. The limiting amplitude principle for the acoustic wave operators in two unbounded media. Tsukuba J. Math. 18 (1994), 175191.
K. Kikuchi and H. Tamura. Limiting amplitude principle for the acoustic propagator in
perturbed stratified fluids. J. Differential Equations 93 (1991), 260282.
R. Leis. Initial boundary value problems in mathematical physics (John Wiley, 1986).
G. F. Roach and B. Zhang. A transmission problem for the reduced wave equation in inhomogeneous media with an infinite interface. Proc. Roy. Soc. Lond. A436 (1992), 121140.
G. F. Roach and B. Zhang. On Sommerfeld radiation conditions for the diffraction problem
with two unbounded media. Proc. Roy. Soc. Edinburgh 121A (1992), 149161.
G. F. Roach and B. Zhang. The limiting amplitude principle for the wave propagation problem
with two unbounded media. Math. Proc. Camb. Phil. Soc. 112 (1992), 207223.
Y. Saito. A remark on the limiting absorption principle for the reduced wave equation with
two unbounded media. Pacific J. Math. 136 (1989), 183208.
H. Tamura. Resolvent estimates at low frequencies and limiting amplitude principle for acoustic propagators. J. Math. Soc. Japan 41 (1989), 549575.
B. Zhang. On transmission problems for wave propagation in two locally perturbed halfspaces. Math. Proc. Camb. Phil. Soc. 115 (1994), 545558.
B. Zhang. On the limiting absorption and amplitude principles for transmission problems
with unbounded interfaces. Brunel University, Department of Mathematics and Statistics
Technical Report (1995).