Documente Academic
Documente Profesional
Documente Cultură
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)
(t, x) R+
1
(2) (a(x)W2 (t, x)) . (x),
(12)
+
(t, x) R (13)
x .
(14)
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
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)
. (1 au+ ) 2 u+ = f,
u1
1
(1) (au1 ) .
= u2 ,
=
in ,
(16)
on
(17)
1
(2) (au2 ) . ,
on ,
(18)
(19)
(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
E(R) = \(R)
(R) = {x ||x| < R}
|| 6 k}
X Z
||6k
! 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
< + ,
(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
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
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
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)
(32)
S(R)
If
X
i, j
|aij ij | = O(|x| )
(33)
au)} h1 u (au)(ax)]
= [2R{rh[x (au)](
where a =
(34)
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)
(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)
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
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
349
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
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
(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
{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)
(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)
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)
(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)
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 )
(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)
(431)
S(R)
(432)
S(R)
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)
(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)
(54)
353
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)]
(56)
i, j, k, l
E(R)
(1)
(2)
(1)
(2)
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
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
(511)
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,
(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
355
kR( i ) R( i0)k11 6 C 10
(63)
any 0 < 6 1
(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)
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).