a

r
X
i
v
:
m
a
t
h
/
0
5
0
4
5
3
6
v
1


[
m
a
t
h
.
A
P
]


2
6

A
p
r

2
0
0
5
HIGH FREQUENCY ANALYSIS OF HELMHOLTZ EQUATIONS:
CASE OF TWO POINT SOURCES
ELISE FOUASSIER
Abstract. We derive the high frequency limit of the Helmholtz equation with source
term when the source is the sum of two point sources. We study it in terms of Wigner
measures (quadratic observables). We prove that the Wigner measure associated with the
solution satises a Liouville equation with, as source term, the sum of the source terms
that would be created by each of the two point sources taken separately. The rst step,
and main diculty, in our study is the obtention of uniform estimates on the solution.
Then, from these bounds, we derive the source term in the Liouville equation together
with the radiation condition at innity satised by the Wigner measure.
AMS subject classications. 35Q60, 35J05, 81S30.
1. Introduction
In this article, we are interested in the analysis of the high frequency limit of the
following Helmholtz equation
(1.1) i

+ u

+
n(x)
2

2
u

= S

(x), x R
3
with
S

(x) = S

0
(x) +S

1
(x) =
1

3
S
0
_
x

_
+
1

3
S
1
_
x q
1

_
where q
1
is a point in R
3
dierent from the origin.
In the sequel, we assume that the refraction index n is constant, n(x) 1.
The equation (1.1) modelizes the propagation of a source wave in a medium with
scaled refraction index n(x)
2
/
2
. There, the small positive parameter is related to
the frequency =
1
2
of u

. In this paper, we study the high frequency limit, i.e.

the asymptotics 0. We assume that the regularizing parameter

is positive,
with

0 as 0. The positivity of

ensures the existence and uniqueness

of a solution u

to the Helmholtz equation (1.1) in L

2
(R
3
) for any > 0.
The source term S

models a source signal that is the sum of two source signals

concentrating respectively close to the origin and close to the point q
1
at the scale
. The concentration proles S
0
and S
1
are given functions. Since is also the
scale of the oscillations dictated by the Helmholtz operator +
1

2
, resonant inter-
actions can occur between these oscillations and the oscillations due to the sources
S

0
and S

1
. On the other hand, since the two sources are concentrating close to
two dierent points in R
3
, one can guess that they do not interact when 0.
These are the phenomena that the present paper aims at studying quantitatively.
We refer to Section 3 for the precise assumptions we need on the sources.
In some sense, the sign of the term i

prescribes a radiation condition at

innity for u

. One of the key diculty in our problem is to follow this condition

1
2 ELISE FOUASSIER
in the limiting process 0.
We study the high frequency limit in terms of Wigner measures (or semi-classical
measures). This is a mean to describe the propagation of quadratic quantities, like
the local energy density [u

(x)[
2
, as 0. The Wigner measure (x, ) is the
energy carried by rays at the point x with frequency . These measures were intro-
duced by Wigner [14] and then developed by P. Gerard [6] and P.-L. Lions and T.
Paul [9] (see also the surveys [3] and [7]). They are relevant when a typical length
is prescribed. They have already proven to be an ecient tool in the study of
high frequencies, see for instance [2], [4] for Helmholtz equations, P. Gerard, P.A.
Markowich, N.J. Mauser, F. Poupaud [7] for periodic media, G. Papanicolaou, L.
Ryzhik [11] for a formal analysis of general wave equations, L. Erd os, H.T. Yau [5]
for an approach linked to statistical physics, and L. Miller [10] for a study in the
case with sharp interface.
Such problems of high frequency limit of Helmholtz equations have been studied
in Benamou, Castella, Katsaounis, Perthame [2] and Castella, Perthame, Run-
borg [4]. In [2], the authors considered the case of one point source and a general
index of refraction whereas in [4], they treated the case of a source concentrating
close to a general manifold with a constant refraction index. In the present paper,
we borrow the methods used in both articles.
In the case of one point source, for instance S

0
only, with a constant index of re-
fraction, it is proved in [2] that the corresponding Wigner measure
0
is the solution
to the Liouville equation
0
+

0
(x, ) +
x

0
(x, ) = Q
0
(x, ) =
1
(4)
2
(x)([[
2
1)[

S
0
()[
2
,
the term 0
+
meaning that is the outgoing solution given by

0
(x, ) =
_
0

Q
0
(x +t, )dt.
In particular, the energy source created by S

0
is supported at x = 0. Similarly,
the energy source created by the source S

1
is supported at x = q
1
. Thinking of
the orthogonality property on Wigner measures, one can guess that the energy
source generated by the sum S

0
+S

1
is the sum of the two energy sources created
asymptotically by S

0
and S

1
.
Indeed, we prove in this paper that the Wigner measure associated with the
sequence (u

) satises
(1.2) 0
+
(x, ) +
x
(x, ) = Q
0
+Q
1
,
where Q
0
and Q
1
are the source terms obtained in [2] in the case of one point
source. However, our proof does not rest on the mere orthogonality property.
Let us now give some details about our proof. Our strategy is borrowed from [2].
First, we prove uniform estimates on the sequence of solutions (u

). We also study
the limiting behaviour fo the rescaled solutions
d1
2
u

(x) and
d1
2
u

(q
1
+ x).
The obtention of these rst two results is the key diculty in our paper. It relies
HIGH FREQUENCY ANALYSIS OF HELMHOLTZ EQUATIONS: CASE OF TWO POINT SOURCES3
on the study of the sequence (a

) such that
i

+ a

+a

= S
1
_
x
q
1

_
.
Using the explicit formula for the Fourier tranform of a

, we prove that a

is uni-
formly bounded in a suitable space and that a

0 as 0 weakly. We would
like to point out that our analysis, based on a study in Fourier space, strongly rests
on the assumption of a constant index of refraction.
Second, our results on the Wigner measure then follow from the properties proved
in [2]. They are essentially consequences of the uniform bounds on (u

): we write
the equation satised by the Wigner transform associated with (u

), and pass to
the limit 0 in the various terms that appear in this equation. The only dicult
(and new) term to handle is the source term.
Third, we prove an improved version of the radiation condition of [2]. Our argu-
ment relies on the observation that is localized on the energy set [[
2
= 1, a
property that was not exploited in [2].
The paper is organized as follows. In Section 2, we recall some denitions and
state our assumptions. Section 3 is devoted to the proof of uniform bounds on the
sequence of solutions (u

) and of the convergence of the rescaled solutions. Then,

in Section 4, we establish the transport equation satised by the Wigner measure
together with the radiation condition at innity. In the appendix, we recall the
proof of some results established in [2] that we use in our paper.
2. Notations and assumptions
In this section, we recall the denitions of Wigner transforms and of the B,
B

norms introduced by Agmon and H ormander [1] for the study of Helmholtz
equations. Then, we give our assumptions.
2.1. Wigner transform and Wigner measures. We use the following denition
for the Fourier transform:
u() = (T
x
u)() =
1
(2)
3
_
R
d
e
ix
u(x)dx.
For u, v o(R
3
) and > 0, we dene the Wigner transform
W

(u, v)(x, ) = (T
y
)(u
_
x +

2
y
_
v
_
x

2
y
_
),
W

(u) = W

(u, u).
In the sequel, we denote W

= W

(u

).
If (u

) is a bounded sequence in L
2
loc
(R
d
), it turns out that (see [6], [9]), up to
extracting a subsequence, the sequence (W

(u

)) converges weakly to a positive

Radon measure on the phase space T

R
3
= R
3
x
R
3

called Wigner measure (or

semiclassical measure) associated with (u

):
(2.1) (

c
(R
6
), lim
0
W

(u

), ) =
_
(x, )d
We recall that these measures can be obtained using pseudodierential opera-
tors. The Weyl semiclassical operator a
W
(x, D
x
) (or Op
W

(a)) is the continuous

4 ELISE FOUASSIER
operator from o(R
d
) to o

(R
d
) associated with the symbol a o

(T

R
d
) by Weyl
quantization rule
(2.2) (a
W
(x, D
x
)u)(x) =
1
(2)
d
_
R
d

_
R
d
y
a
_
x +y
2
,
_
f(y)e
i(xy)
ddy.
We have the following formula: for u, v o

(R
d
) and a o(R
d
R
d
),
(2.3) W

(u, v), a)
S

,S
= v, a
W
(x, D
x
) u)
S

,S
,
where the duality brackets ., .) are semi-linear with respect to the second argument.
This formula is also valid for u, v lying in other spaces as we will see in Section 3.
2.2. Besov-like norms. In order to get uniform (in ) bounds on the sequence
(u

), we shall use the following Besov-like norms, introduced by Agmon and H orman-
der [1]: for u, f L
2
loc
(R
3
), we denote
|u|
B
= sup
j1
_
2
j
_
C(j)
[u[
2
dx
_
1/2
,
|f|
B
=

j1
_
2
j+1
_
C(j)
[f[
2
dx
_
1/2
,
where C(j) denotes the ring x R
3
/2
j
[x[ < 2
j+1
for j 0 and C(1) is the
unit ball.
These norms are adapted to the study of Helmholtz operators. Indeed, if v is
the solution to
iv + v +v = f
where > 0, then Agmon and H ormander [1] proved that there exists a constant
C independent of such that
|v|
B
C|f|
B
.
Perthame and Vega [12] generalised this result to Helmholtz equations with general
indices of refraction.
We denote for x R
3
, [x[ =
_

3
j=1
x
2
j
and x) = (1 + [x[
2
)
1/2
.
For all >
1
2
, we have
(2.4) |u|
L
2

:= |x)

u|
L
2 C()|u|
B
.
We end this section by stating two properties of these spaces that will be useful
for our purpose (the reader can nd the proofs in [1]). The rst proposition states
that, in some sense, we can dene the trace of a function in B on a linear manifold
of codimension 1.
Proposition 2.1. There exists a constant C such that for all f B, we have
_
R
|f(x
1
, .)|
L
2
(R
2
)
dx
1
C|f|
B
.
The second property gives the stability of the space B by change of variables in
Fourier space.
HIGH FREQUENCY ANALYSIS OF HELMHOLTZ EQUATIONS: CASE OF TWO POINT SOURCES5
Proposition 2.2. Let
1
,
2
be two open sets in R
3
, :
1

2
a (
2
dieo-
morphism, (
1
c
(R
3
). For all u B, we denote
Tu = T
1
_
( u )
_
.
Then
|Tu|
B
C||
C
1
b
||
C
2
b
|u|
B
.
2.3. Assumptions. We are now ready to state our assumptions. Our rst assump-
tion, borrowed from [2], concerns the regularizing parameter

> 0.
(H1)

for some > 0.

This assumption is technical and is used to get a radiation condition at innity in
the limit 0. Next, in order to get uniform bounds on u

, we assume that the

source terms S
0
and S
1
belong to the natural Besov space that is needed to actually
solve the Helmholtz equation (1.1).
(H2) |S
0
|
B
, |S
1
|
B
< .
It turns out that, in order to compute the limit of the energy source, we shall need
the stronger assumption
(H3) x)
N
S
0
L
2
(R
3
) and x)
N
S
1
L
2
(R
3
) for some N >
1
2
+
3
+1
.
3. Bounds on solutions to Helmholtz equations
In this section, we rst establish uniform bounds on the sequence (u

) that will
imply estimates on the sequence of Wigner transforms (W

). It turns out that we

shall also need to compute the limit of the rescaled solutions w

0
and w

1
dened
below in order to obtain the energy source in the equation satised by the Wigner
measure .
Before stating our two results, let us dene these rescaled solutions. Following [2]
and [4], we denote
(3.1)
_
w

0
(x) =
d1
2
u

(x),
w

1
(x) =
d1
2
u

(q
1
+x).
They respectively satisfy
_
i

0
+ w

0
+w

0
= S
0
(x) +S
1
_
x
q1

_
,
i

1
+ w

1
+w

1
= S
0
_
x +
q1

_
+S
1
(x).
We are ready to state our results on u

, w

0
and w

1
.
Proposition 3.1. Assume S
0
, S
1
B. Then, the solution u

to the Helmholtz
equation (1.1) satises the following bound
|u

|
B
C(|S
0
|
B
+ |S
1
|
B
),
where C is a constant independent of .
Proposition 3.2. Let w

0
and w

1
be the rescaled solutions dened by (3.1). Then,
the sequences (w

0
) and (w

1
) are uniformly bounded in B

and they converge weakly-

in B

to the outgoing solutions w

0
and w
1
to the following Helmholtz equations
_
w
0
+w
0
= S
0
w
1
+w
1
= S
1
,
6 ELISE FOUASSIER
i.e. w
0
and w
1
are given in Fourier space by
w
j
() =

S
j
()
[[
2
1 +i0
=
_
p.v.
_
1
[[
2
1
_
+i([[
2
1)
_

S
j
(), j = 0, 1.
Remark. The Helmholtz equation w + w = S does not uniquely specify the
solution w. An extra condition is necessary, for instance the Sommerfeld radiation
condition. When the refraction index is constant equal to 1, this condition writes
(3.2) lim
r
1
r
_
Sr

w
r
+iw

2
d = 0.
Such a solution is called an outgoing solution.
Alternatively, still assuming that the refraction index is constant, the outgoing
solution to the Helmholtz equation may be dened as the weak limit w of the
sequence (w

) such that
iw

+ w

+w

= S(x).
We point out that the two points of views are equivalent in the case of a constant
index of refraction (which is not true for a general index of refraction).
We prove the two propositions in the following two sections. As we will see in the
proofs, our main diculties are linked to the rays that are emitted by the source at
0 towards the point q
1
(and conversely). Hopefully, the interaction between those
rays is destructive and not constructive.
3.1. Proof of Proposition 3.1. In the sequel, C will denote any constant inde-
pendent of .
The scaling invariance
|u

|
B
|w

0
|
B
,
makes it sucient to prove bounds on w

0
. Since w

0
is a solution to
i

0
+ w

0
+w

0
= S
0
(x) +S
1
_
x
q
1

_
we may decompose w

0
=

w

0
+a

, where

w

0
and a

satisfy
_
i

0
+

0
+

w

0
= S
0
(x),
i

+ a

+a

= S
1
_
x
q1

_
.
First, we note that the bound |

0
|
B
C|S
0
|
B
is established in Agmon-H orman-
der [1] (see also Perthame-Vega [12]). Hence, the proof of Proposition 3.1 reduces
to the proof of the following lemma.
Lemma 3.3. If a

is the solution to
i

+ a

+a

= S
1
_
x
q
1

_
then a

is uniformly (in ) bounded in B

:
|a

|
B
C|S
1
|
B
HIGH FREQUENCY ANALYSIS OF HELMHOLTZ EQUATIONS: CASE OF TWO POINT SOURCES7
Proof. We want to prove that
v B, [a

, v)[ C|S
1
|
B
|v|
B
.
Using Parsevals equality, we write
(3.3) a

, v) =
_
R
3
e
i
q
1

S
1
()

v()
[[
2
+ 1 i

d.
To estimate this integral, we shall distinguish the values of close to or far from
two critical sets: the sphere [[
2
= 1 (the set where the denominator in (3.3)
vanishes when 0) and the line collinear to x
0
(the set where we cannot
apply directly the stationary phase theorem to (3.3)).
More precisely, we rst take a small parameter ]0, 1[, and we distinguish in
the integral (3.3), the contributions due to the values of such that [
2
1[
or [
2
1[ . Let (

c
(R) be a truncation function such that () = 0 for
[[ 1. We denote

() =
_
||
2
1

_
. We accordingly decompose
a

, v) =
_
R
3
e
i
q
1

S
1
()

v()

()
[[
2
+ 1 i

d +
_
R
3
e
i
q
1

S
1
()

v()(1

())
[[
2
+ 1 i

d
= I

+II

.
First, since the denominator is not singular on the support of

, we easily bound
the rst part with the L
2
norms
[I

[
||
L

S
1
|
L
2| v|
L
2,
and using B L
2
, we obtain the desired bound
(3.4) [I

[ C|S
1
|
B
|v|
B
.
Let us now study the second part II

where the denominator is singular. Up to

a rotation, we may assume q
1
= [q
1
[e
1
, where e
1
is the rst vector of the canonical
base. We make the polar change of variables
=
_
_
_
r sin cos
r sin sin
r cos
.
Remark. In order to make the calculations easier, we write this paper in di-
mension equal to 3, but the proof would be similar in any dimension d 3.
Hence, q
1
= [q
1
[r sin cos , and we get
II

=
_
e
i
|q
1
|

r sin cos
r
2
+ 1 i

S
1

v(1

)
_
((r, , ))r
2
sin drdd
Now, we distinguish the contributions to the integral dd linked to the values close
to, or far from, the critical direction =

2
, = 0 (which corresponds to the case
collinear to q
1
). To that purpose, let > 0 be a small parameter and denote
K =
_
(r, , )

1
_
r
2
1

_
,= 0,
_


2

_
,= 0,
_

_
,= 0
_
.
8 ELISE FOUASSIER
Then K is a compact set. Let k (

c
be such that (1

)k(, ) is a localization
function on K. We write
II

=
_
e
i
|q
1
|

r sin cos
r
2
+ 1 i

S
1

v)((r, , ))(1

(r))k(, )r
2
sin drdd
+
_
e
i
|q
1
|

r sin cos
r
2
+ 1 i

S
1

v)((r, , ))(1

(r))(1 k(, ))r

2
sin drdd
II

= III

+IV

.
To estimate the contribution III

, we apply the stationary phase method. We

denote =

2
. The phase function is g
r
(, ) = r cos cos so
g
r

= r sin cos = 0 at (, ) = (0, 0),

g
r

= r cos sin = 0 at (, ) = (0, 0),

and the Hessian at the point (0, 0) is
D
2
g
r
(0, 0) =
_
r 0
0 r
_
,
which is invertible at any point in K. Since K is a compact set, we can apply the
Morse lemma: there exists a nite covering (
j
)
j=1,n
(n N) of K such that on
each set
j
, there exists a (

change of variables (, ) (
j
,
j
) such that
g
r
(, ) = r r

2
j
2
r

2
j
2
.
Moreover, we can write (1

)k =

n
j=1

j
where
j
(

c
and supp(
j
)
j
.
Then, we make the changes of variables

j
=
_
r
2

j
,

j
=
_
r
2

j
. Finally, we
decompose
j
=
1
j

2
j
. Thus, we obtain, for the contribution III

, the formula
(3.5) III

=
n

j=1
_
e
i
x
1

(r+
2
j
+
2
j
)
r + 1 +i

T
1
j
S
1
(r,

j
,

j
)

T
2
j
v(r,

j
,

j
)drd

j
d

j
,
where
T
1
j
S
1
:= T
_
(
1
j

S
1
) (r, (
j
,
j
), (
j
,
j
))
_
,
T
2
j
v := T
_
r + 1 +i

r
2
+ 1 i

(
2
j
v) (r, (
j
,
j
), (
j
,
j
))

2
r

d
d(r, , )

d(, )
d(
j
,
j
)

_
.
As a rst step, using Proposition 2.2, we directly get T
1
j
S
1
B with
|T
1
j
S
1
|
B
C|S
1
|
B
.
As a second step, we study T
2
j
v. Since for r close to 1,

r + 1 +i

r
2
+ 1 i

1,
we recover, from Proposition 2.2,
T
2
j
v B and |T
2
j
v|
B
C|v|
B
.
HIGH FREQUENCY ANALYSIS OF HELMHOLTZ EQUATIONS: CASE OF TWO POINT SOURCES9
Now, we apply Parsevals equality with respect to the r variable in the formula (3.5)
III

=
n

j=1
_
e
i
|q
1
|

(
2
j
+
2
j
)
r + 1 +i

T
1
j
S
1
(.
[q
1
[

, ., .)

T
2
j
vdrd

j
d

j
=
_
e
i
|q
1
|

(
2
j
+
2
j
)
1
{>0}
e
(i)t
T
r
(

T
1
j
S
1
(.
[q
1
[

, ., .))( t,

j
,

j
)
T
r
(

T
2
j
v)(,

j
,

j
)dtdd

j
d

j
.
where 1
{>0}
denotes the characteristic function of the set > 0.
Hence, we obtain
[III

[
n

j=1
_
_
|T
r
(

T
1
j
S
1
(.
[q
1
[

, ., .))())|
L
2 d
_

__
|T
r
(

T
2
j
v)()|
L
2 d
_
[III

[
n

j=1
_
_
|T
1
j
S
1
(
[q
1
[

)|
L
2d
__
_
|T
2
j
v()|
L
2d
_
[III

[
n

j=1
_
_
|T
1
j
S
1
()|
L
2d
__
_
|T
2
j
v()|
L
2d
_
[III

[ C
n

j=1
|T
1
j
S
1
|
B
|T
2
j
v|
B
[III

[ C|S
1
|
B
|v|
B
,
which is the desired estimate.
We are left with the part IV

, which corresponds to the directions that are

not collinear to q
1
. We denote K

the support of (1

)(1k) which is a compact

set. In K

, we can choose as new independent variables

1
= q
1
,
2
= [[
2
1.
More precisely, since
d(
1
,
2
)
d
=
_
q
1
2
_
is of maximal rank 2, there exists a nite covering (

j
)
j=1,m
(m N) of K

such
that in

j
, we can make the change of variables . As before, we denote

j
=
3
j

4
j
some localization functions on

j
such that (1

)(1 k) =

m
j=1

j
.
Thus, for j = 1, . . . , m,
_
e
i
q
1

[[
2
+ 1 +i

S
1
v

j
d =
_
e
i

1
+i

S
1
v

j
)(())

d
d

d.
If we denote
T
3
j
S
1
:= T
1
((
3
j

S
1
) ),
T
4
j
v := T
1
_
(
4
j
v) )

d
d

_
,
10 ELISE FOUASSIER
and if T
1
denotes the Fourier transform with respect to the
1
variable, Parsevals
equality with respect to
1
gives

_
e
i
q
1

[[
2
+ 1 +i

S
1
v

j
d

= (2)
d

_

{t>0}
e
t
(T
1
1
(

T
3
j
S
1
))(x
1
t)
(T
1
1
(

T
4
j
v))(x
1
)e
i2/
dtdx
1
d
2
d
3

C|S
1
|
B
|v|
B
.
Summing over j, we obtain
[IV

[ C|S
1
|
B
|v|
B
,
which ends the proof of the bound
[a

, v)[ C|S
1
|
B
|v|
B
.

3.2. Proof of Proposition 3.2. We prove the result for the sequence (w

0
), the
convergence of the sequence (w

1
) can be obtained similarly. As we did in the proof
of Proposition 3.1, we write w

0
=

w

0
+a

. Since

w

0
is the solution to a Helmholtz
equation with constant index of refraction and xed source, it converges weakly-
to the outgoing solution w
0
to w+w = S
0
. Hence, it suces to show the following
result.
Lemma 3.4. If a

is the solution to
i

+ a

+a

= S
1
_
x
q
1

_
then a

0 in B

.
Proof. The proof of this result requires two steps (using a density argument):
(1) for v B, we have the bound

, v)

C|S
1
|
B
|v|
B
(2) if S
1
and v are smooth, then a

, v) 0.
The rst point is exactly the result in Lemma 3.3. It remains to prove the conver-
gence in the smooth case (the second point above).
We write
a

, v) =
_
R
3
e
i
q
1

S
1
()

v()
[[
2
+ 1 i

d.
We are thus left with the study of
(3.6) R

() =
_
R
3
e
i
q
1

()
[[
2
+ 1 i

d
where =

S
1
v belongs to o(R
3
).
As in the proof of Lemma 3.3, we distinguish the contributions of various values
of . We shall use exactly the same partition, according to the values of close to,
or far from, the sphere [[ = 1 and collinear or not to q
1
. We shall use the same
HIGH FREQUENCY ANALYSIS OF HELMHOLTZ EQUATIONS: CASE OF TWO POINT SOURCES 11
notations for the various truncation functions.
We rst separate the contributions of such that [
2
1[ and [
2
1[
R

() =
_
R
3
e
i
q
1

()

()
[[
2
+ 1 i

d +
_
R
3
e
i
q
1

()(1

())
[[
2
+ 1 i

d
= I

+II

.
In the support of

, since the denominator is not singular, we can apply the

non stationary phase method.
Since q
1
,= 0, we may assume q
1
1
,= 0 and we have
I

=

iq
1
1
_
R
3
e
i
q
1

1
_
()

()
[[
2
+ 1 i

_
d
=

iq
1
1
_
R
3
e
i
q
1

_

1
(()

())
[[
2
+ 1 i

2()

()
1
([[
2
+ 1 i

)
2
_
.
Hence, we obtain the bound
[I

[

[q
1
1
[
_
R
3
_
1

[
1
()[ +
2

2
[
1
[
_
d.
Since
1
() and
1
belongs to o, we have, as 0,
I

0.
Let us now study the second term II

. We use the same changes of variables as

in Section 3.1. It leads to the following formula
II

=
n

j=1
_
e
i
q
1

(r
2
j

2
j
)
r + 1 +i

j
(r,

j
,

j
)

(r,

j
,

j
)drd

j
d

j
,
where

j
(r,

j
,

j
) =
j
(r, (

j
,

j
), (

j
,

j
))
2(r + 1 +i

)
r(r
2
+ 1 i

d(, )
d(
j
,
j
)

(r,

j
,

j
) = (r, (

j
,

j
), (

j
,

j
)),
are still smooth functions that are bounded independently from .
Using Parsevals inequality with respect to the variables (

j
,

j
) for each integral,
we obtain the bound
[II

[ C
n

j=1

_
e
i
|q
1
|

r
e
i(
2
j
+
2
j
)
r + 1 +i

T
j,j
(
j

)drd
j
d
j

.
To obtain the convergence of II

, it remains to study an integral of the following

type
_
|r1|
e
i
|q
1
|

r
w(r)
r + 1 +i

dr, where w o.
This is done in the following lemma.
Lemma 3.5. w o, (0, 1), we have
_
|r|
e
i
|q
1
|

r
w(r)
r +i

dr = iw(0) +O
0
(

).
12 ELISE FOUASSIER
Using this lemma, we readily get the estimate
(3.7) [III

[ C
1
(0, 1),
which proves that III

0 as 0.
There remains to give the
Proof of Lemma 3.5. We write
_

e
i
|q
1
|

r
w(r)
r +i

dr =
_

e
i
|q
1
|

r
w(r)
r
2
+ (

)
2
(r i

)dr
= i

e
i
|q
1
|

r
w(r)
r
2
+ (

)
2
dr +
_

e
i
|q
1
|

r
rw(r)
r
2
+ (

)
2
dr
= I +II.
We have
I = i
_

e
i|q1|y
w(

y)
y
2
+ 1
dy iw(0),
and
II =
_

_
e
i
|q
1
|

r
w(r) w(0)
_
r
r
2
+ (

)
2
dr +
_

w(0)
r
r
2
+ (

)
2
dr .
The last term vanishes because the integrand is odd. Moreover, using the smooth-
ness of w, we easily obtain that for all (0, 1),

e
i
|q
1
|

r
w(r) w(0)

_
r

Thus,

_
e
i
|q
1
|

r
w(r) w(0)
_
r
r
2
+ (

)
2
dr

[r[
1
dr
and the result is proved.
We are left with the study of IV

. We use the same change of variables as in

Section 3.1.
IV

=
m

j=1
_
e
i
q
1

[[
2
+ 1 +i

()
j
()d
=
m

j=1
_
e
i

1
+i

(
j
)(())

d
d

d
= i
m

j=1
_
e
i

1
+i

1
_
(
j
)(())

d
d

_
d.
The integral obviously converges with respect to all the variables except
1
. It re-
mains to prove the convergence with respect to the
1
variable, i.e. the convergence
of
_
()

1
+i

d
1
,
where
=
1
_
(
j
)(())

d
d

_
HIGH FREQUENCY ANALYSIS OF HELMHOLTZ EQUATIONS: CASE OF TWO POINT SOURCES 13
is smooth and compactly supported with respect to . It is a consequence of the
fact that the distribution (x +i0)
1
is well-dened on R by
1
x +i0
= v.p.(
1
x
) i(x).
We conclude that IV

0 and a

, v) 0 as 0.
4. Transport equation and radiation condition on
In this section, we state and prove our results on the Wigner measure associated
with (u

). Since we established the uniform bounds on (u

) and the convergence

of (w

0
), (w

1
), these results now essentially follows from the results proved in [2].
We rst prove bounds on the sequence of Wigner transforms (W

) that allow us to
dene a Wigner measure associated to (u

). Then, we get the transport equation

satised by together with the radiation condition at innity, which uniquely
determines .
4.1. Results.
Theorem 4.1. Let S
0
, S
1
B and > 0. The sequence (W

) is bounded in
the Banach space X

and up to extracting a subsequence, it converges weak- to a

positive and locally bounded measure such that
(4.1) sup
R>0
1
R
_
|x|<R
_
R
3
(x, ) dxd C(|S
0
|
B
+ |S
1
|
B
)
2
.
The Banach space X

is dened as the dual space of the set X

of functions (x, )
such that (x, y) := T
y
( (x, )) satises
(4.2)
_
R
d
sup
xR
d
(1 + [x[ + [y[)
1+
[(x, y)[dy < .
Theorem 4.2. Assume (H1), (H2), (H3). Then the Wigner measure asssociated
with (u

) satises the following transport equation

(4.3)
x
=
1
(4)
2
_
(x)[

S
0
()[
2
+(x q
1
)[

S
1
()[
2
_
([[
2
1) := Q(x).
Moreover, is the outgoing solution to the equation (4.3) in the following sense:
for all test function R (

c
(R
6
), if we denote g(x, ) =
_

0
R(x t, )dt, then
(4.4)
_
R
6
R(x, )d(x, ) =
_
R
6
Q(x, )g(x, )dxd.
Remark. Here the support of the test function R contains 0, contrary to [2].
4.2. Proof of Theorem 4.1. This theorem, that is proved in [2], is a consequence
of the uniform estimate on the sequence (u

) in the space B

obtained in Proposition
3.1. We observe that for any > 0,
(4.5) |x)

1
2

(x)|
L
2 C|u

|
B
C|f|
B
,
14 ELISE FOUASSIER
hence, for any function satisfying (4.2), we have
[W

(u

), )[

_
R
6
[u

[(x +

2
y)[u

[(x

2
y)
x +

2
y)
1
2
+0
x

2
y)
1
2
+0
x +

2
y)
1
2
+0
x

2
y)
1
2
+0
[[(x, y)dxdy
C|f|
2
B
_
R
3
sup
xR
3
[x[ + [y[)
1+0
[(x, y)[dy.
So (W

(u

)) is bounded in X

, > 0. We deduce that, up to extracting a subse-

quence, (W

(u

)) converges weak- to a nonnegative measure satisfying

(4.6) [, )[ C|f|
2
B
_
R
3
sup
xR
3
[x[ + [y[)
1+0
[(x, y)[dy.
We refer for instance to Lions, Paul [9] for the proof of the nonnegativity of .
The bound (4.1) is obtained using the following family of functions

(x, y) =
1

3/2
e
|y|
2
/
1
R
(x) R)
and letting 0, R .
4.3. Proof of the transport equation 4.3. This section is devoted to the proof of
the transport equation satised by . We rst write the transport equation satised
by W

in a dual form. Then, we study the convergence of the source term (the
convergence of the other terms is obvious). Finally, choosing an appropriate test
function in the limiting process, we get the radiation condition at innity satised
by . Proving rst a localization property, we improve the radiation condition
proved in [2].
4.3.1. Transport equation satised by W

. W

satises the following equation

(4.7)

+
x
W

=
i
2
1m W

(f

, u

) := Q

.
This equation can be obtained writing rst the equation satised by
v

(x, y) = u

_
x +

2
y
_
u

_
x

2
y
_
.
From the equality

y

x
v

=

2
_
u

(x +

2
y)u

(x

2
y) u

(x

2
y)u

(x +

2
y)

,
we deduce

+i
y

x
v

+
i
2
_
n
2
(x +

2
y) n
2
(x

2
y)

(x, y),
where

(x, y) :=
i
2
_
S

(x +

2
y)u

(x

2
y) S

(x

2
y)u

(x +

2
y)

.
After a Fourier transform, we obtain the equation (4.7).
Then we write the dual form of this equation. Let o(R
6
), we have
(4.8)

, ) W

,
x
) = Q

, ).
By the denition of the Wigner measure , we get

, ) 0 and W

,
x
) ,
x
).
HIGH FREQUENCY ANALYSIS OF HELMHOLTZ EQUATIONS: CASE OF TWO POINT SOURCES 15
Hence we are left with the study of the source term Q

, ).
4.3.2. Convergence of the source term. In order to compute the limit of the source
term in (4.7), we develop
Q

, ) =
i
2
1m
_
W

(S

0
, u

), ) + W

(S

1
, u

), )
_
.
Thus, the result is contained in the following proposition.
Proposition 4.3. The sequences
_
W

(S

0
, u

)
_
and
_
W

(S

1
, u

)
_
are bounded in
o

(R
6
) and for all o(R
6
), we have
lim
0
W

(S

0
, u

), )
S

,S
=
1
(2)
3
_
R
3
w
0
()

S
0
()(0, )d, (4.9)
lim
0
W

(S

1
, u

), )
S

,S
=
1
(2)
3
_
R
3
w
1
()

S
1
()(q
1
, )d, (4.10)
where w
0
and w
1
are dened in Proposition 3.2.
Using Proposition 4.3, we readily get
lim
0
Q

, ) =
i
2(2)
3
1m
__
R
3
w
0
()

S
0
()(0, )d +
_
R
3
w
1
()

S
1
()(q
1
, )d
_
=
1
(4)
2
__
R
3
[

S
0
()[
2
(
2
1)(0, )d
+
_
R
3
[

S
1
()[
2
(
2
1)(q
1
, )d
_
,
which is the result in Theorem 4.2.
Let us now prove Proposition 4.3.
Proof of Proposition 4.3. The two terms to study being of the same type, we only
consider the rst one in our proof. Let o(T

R
d
) and (x, y) = T
1
y
((x, )),
then we have
W

(S

0
, u

), )
S

,S
=
_
S

0
_
x +

2
y
_
u

_
x

2
y
_
(x, y)dxdy
=
_
S
0
(x)w

0
(x +y)((x +
y
2
), y)dxdy.
Hence, using that o(R
2d
), we get

(S

0
, u

), )
S

,S

C
_
x)
N
[S
0
(x)[
[w

0
(x +y)[
x + y)

x +y)

x)
N
y)
k
dxdy
C|x)
N
S
0
|
L
2|w

0
|
B

_
R
3
y
sup
xR
3
x +y)

x)
N
y)
k
dy
for any k 0 and > 1/2, upon using the Cauchy-Schwarz inequality in x.
Then, we distinguish the cases [x[ [y[ and [x[ [y[ : the term stemming from
the rst case gives a contribution which is bounded by C
_
dy
y
k
and the second
contribution is bounded by C
_
dy
y
k
. So, upon choosing k large enough, we obtain
that
[W

(S

0
, u

), )
S

,S
[ C|x)
N
S
0
|
L
2|w

|
B
.
16 ELISE FOUASSIER
Now, in order to compute the limit (4.9), we write
W

(S

0
, u

), ) =
_
S
0
(x)w

0
(x +y)
_

_
x +
y
2
_
, y
_
(0, y)
_
dxdy
+
_
w

0
(x)S
0
(x y)(0, y)dxdy
= I

+II

.
Reasonning as above, we readily get that lim
0
I

= 0. For the second term, we

have
II

=
_
w

0
(x)
_
S
0
(0, .)
_
(x)dx,
hence, since w

0
converges weakly- in B

, it suces to prove that S

0
(0, .) belongs
to B. We denote = (0, .). We have, for > 1/2,
|S
0
|
B
C|S
0
|
L
2

= C
_
x)

[S
0
(x)[
2
dx
C||
L
1
_
x)

[S
0
[
2
[[(x)dx
where we used the Cauchy-Schwarz inequality. Hence, we get
|S
0
|
B
C|x)
N
S
0
|
L
2
_
R
3
y
sup
xR
3
x +y)

x)
N
y)
k
dy,
for any k. As before, this integral converges. Thus, we have established that
S
0


(0, .) belongs to B, which implies that
II

_
S
0
(x)w
0
(x +y)

(0, y)dxdy.

4.4. Proof of the radiation condition (4.4). It remains to prove that satises
the weak radiation condition (4.4).
4.4.1. Support of . In order to prove the radiation condition without restriction
on the test function R (as assumed in [2]), we rst prove a localization property on
the Wigner measure . This property is well-known when u

satises a Helmholtz
equation without source term. It is still valid here thanks to the scaling of S

.
Proposition 4.4. Under the hypotheses (H1), (H2), (H3), the Wigner measure
satises
supp() (x, ) R
6
/ [[
2
= 1.
Proof. Let (

c
(R
6
) and

=
W
(x, D
x
). Let us denote H

=
2
1.
Since u

satises the Helmholtz equation (1.1), we have

(4.11) i

+H

=
2
S

.
Moreover, H

is a pseudodierential operator with symbol [[

2
1. By pseudodif-
ferential calculus,

= Op
W

((x, )([[
2
1)) +O() so, using the denition of
the measure , we get that
lim
0
(

, u

) = lim
0
(Op
W

((x, )([[
2
1))u

, u

)
=
_
(x, )([[
2
1))d.
HIGH FREQUENCY ANALYSIS OF HELMHOLTZ EQUATIONS: CASE OF TWO POINT SOURCES 17
Using the equation (4.11), we write
(

, u

) =
2
(

, u

) i

, u

) =
2
(W

(S

, u

), ) i

, u

).
On the rst hand, Proposition 4.3 gives that lim
0

2
(W

(S

, u

), ) = 0. On the
other hand, (

, u

) is bounded so lim
0

, u

) = 0. Therefore, for any

(

c
(R
6
), we have
_
([[
2
n
2
(x))d = 0, so supp() [[
2
= 1.
4.4.2. Proof of the condition (4.4). Using the previous localization property, in
order to prove the radiation condition (4.4), one may only use test functions R
(

c
(R
6
) such that supp(R) R
6
= 0.
Let R be such a test function. We associate with R the solution g

to

+
x
g

= R(x, ).
By duality, we have
Q

, g

) = W

, R),
so that it suces to establish the following two convergences:
(4.12) lim
0
Q

, g

) = Q, g),
(4.13) lim
0
W

, R) = f, R),
where Q et g are dened in Theorem 4.2.
As before, since R X

for any > 0, the limit (4.13) follows from the weak-
convergence of W

in X

.
On the other hand,
Q

, g

) = 1m
_
R
6
S
0
(x)w

0
(x +y)

([x +
y
2
], y)dxdy (4.14)
+1m
_
R
6
S
1
(x)w

1
(x +y)

(q
1
+[x +
y
2
], y)dxdy,
so Q

, g

) is the sum of two terms of the same type. Such a term has been studied
in [2], where the following result is proved.
Proposition 4.5. Assume (w

) is bounded in B

and that (w

) converges weakly-
in B

to w
0
. Assume S
0
satisfy (H3). Let R (

c
(R
6
) be such that supp(R)
R
6
= 0. Let g

be the solution to

+
x
g

= R(x, )
and g(x, ) =
_

0
R(x +t, )dt. Then, we have
lim
0
_
R
6
S
0
(x)w

0
(x +y)

([x +
y
2
], y)dxdy =
1
(2)
3
_
R
3

S
0
() w
0
()g(0, )d.
Proof. The proof of this result is written in the appendix.
Using the proposition above together with Proposition 3.2, we get that
lim
0
Q

, g

)
= 1m
_
1
(2)
3
_
R
3

S
0
() w
0
()g(0, )d +
1
(2)
3
_
R
3

S
1
() w
1
()g(q
1
, )d
_
=
1
(4)
2
__
R
3
[

S
0
()[
2
(
2
1)g(0, )d +
_
R
3
[

S
1
()[
2
(
2
1)g(q
1
, )d
_
.
18 ELISE FOUASSIER
Thus, the radiation condition (4.4) is proved.
Appendix A. Proof of Proposition 4.5
In the sequel, we denote
G

= lim
0
_
R
6
S
0
(x)w

0
(x +y)

([x +
y
2
], y)dxdy.
A.1. Bounds on G

. In order to study G

, we need a preliminary result on the

test function g

.
Lemma A.1. Let R (

c
(R
6
= 0). We denote g

the solution to
(A.1)

+.
x
g

= R(x, ).
It is given by the explicit formula
(A.2) g

(x, ) =
_

0
exp(

[[
1
s)
1
[[
R(x

[[
s, )ds.
Then we have the estimate
(A.3) M 0, [

(x, y)[ C
x)
M

M

y)
M
,
where . . denotes the inmum of two numbers, and C is a constant depending on
M and R.
Proof. Let a be a multiindex such that [a[ M. We denote =

||
. We write
y
a

(x, y) = T
y
_
_

0
(i

)
a
_
e
||
1
s
1
[[
R(x s, )
_
ds
_
=
_
R
3
de
i.y
_
+
s=0
ds

b,c,d,e,f
C
(a,b,c,d,e,f)
e
||
1
s
(i

)
b
(

[[
1
s)
(

[[
1
s)
f
(i

)
c
(s)(i
x
)
d
(i

)
e
_
1
[[
R
_
(x s, ). (A.4)
Using that
R (

0
(R
6
[[ = 0) so
there exists r
0
, A, B > 0 such that supp(R) [x[ r
0
A [[ B,
R and its derivatives belong to L
1
(R
6
).
[(i

)
b
(

[[
1
s)[ Cs, [[ A.
[(i

)
c
(s)[ Cs, [[ A.
in the integral above, s [[x[ r
0
, [x[ + r
0
], so for [x[ large enough, we can use
the equivalence s [x[ x) (where we dente for a, b > 0, a b if c
1
, c
2
>
0/ c
1
a < b < c
2
a), we get in (A.4),
[y
a

(x, y)[ Cx)

M
exp(

B
x)).
The desired estimate follows.
HIGH FREQUENCY ANALYSIS OF HELMHOLTZ EQUATIONS: CASE OF TWO POINT SOURCES 19
Using this lemma, we estimate
[G

_
R
6
w

0
(x +y)S
0
(x)

((x +
y
2
, y)dxdy

C
_
R
6
[w

0
(x +y)[
x +y)
1
2
+0
x +y)
1
2
+0
x)
N1
[S
0
(x)[x)
N1
(x +
y
2
))
M

M

y)
M
C|w

0
|
B
|x)
N1
S
0
(x)|
L
2

_
R
3
sup
xR
3
x +y)
1
2
+0
x)
N1
([x[ + [y[))
M

M

y)
M
dy.
Now, we prove that the integral
1

=
_
R
3
sup
xR
3
x +y)
1
2
+0
x)
N1
([x[ + [y[))
M

M

y)
M
dy
is bounded uniformly with respect to .
Let us dene the following three subsets in R
6
(A.5)
A

= (x, y) R
6
/[x[ [y[, B

= [x[ [y[, [
10
y[ 1,
C

= [x[ [y[, [
10
y[ 1,
where
10
means
1
with > 0 suciently small.
If (x, y) A

, then
x +y)
1
2
+0
x)
N1
([x[ + [y[))
M

M

y)
M
Cx)
N1+
1
2
+0
_
x)
M

M

_
y)
M
.
Now, we distinguish the relative size of x) and

:
if x)

, we have
x)
1
2
+0N1
_
x)
M

M

_
x)
1
2
+0N1

M

(
1
2
+0N1+M)
.
if x)

, we have x)
1

x), hence we get

x)
1
2
+0N1
_
x)
M

M

_
x)
1
2
+0N1

M

M

(+1)(
1
2
+0N1)
.
Now, we choose 1/2 + 0 such that N
1
<
1
2
+ 0. Then, we get the following bound
for the contribution of the set A

to 1

C
M(+1)(1/2+0N1)
_
R
3
y)
M
dy.
Since N
1
>
3
+1
+
1
2
, we can choose M > 3 such that this contribution is uniformly
bounded with respect to .
If (x, y) B

, then [y[ 1 so we obtain

x +y)
1
2
+0
x)
N1
([x[ + [y[))
M

M

y)
M
Cy)
M+
1
2
+0
.
Thus, the corresponding contribution to 1

is bounded by C
_
R
3
y)
M+
1
2
+0
dy which
is convergent.
If (x, y) C

, then
x +y)
1
2
+0
x)
N1
([x[ + [y[))
M

M

y)
M
Cy)
1
2
+0
y)
M

M

y)
M
.
20 ELISE FOUASSIER
Since [y[
1+0
, if we denote z = y, we have [z[

, for some > 0, which

implies that

2
z).
Thus, we have
1

C
3
_
R
3

)
M+
1
2
+0
(z)
M

M
)dz
C
M(1)4
_
R
3
z)
M+
1
2
+0
(z)
M

M
)dz
where we used the hypothesis (H1)

.
We distinguish two cases, according to the relative size of z) and

. We have
_
z

z)
1
2
+0
z)
M

M
z)
M
dz
M
_
z

z)
1
2
+0M
dz
C
(
3
2
+0)
,
and
_
z

z)
1
2
+0
z)
M

M
z)
M
dz
_
z

z)
1
2
+0
dz
C
(
3
2
+0)
.
Hence, we get
1

C
M(1)4(
3
2
+0)
.
To conclude, we choose M >
4+
3
2

1
, which gives that 1

is uniformly bounded with

respect to .
A.2. Convergence of G

. We decompose G

in the following way

G

=
_
R
6
w

0
(x +y)S
0
(x)
_

((x +
y
2
), y)

g

(0, y)
_
dxdy
+
_
R
6
w

0
(x +y)S
0
(x)
_

(0, y) g(0, y)
_
dxdy
+
_
R
6
w

0
(x +y)S
0
(x) g(0, y)dxdy
= I

+II

+III

Using the same method as in Section A.1, we prove that I

, II

0. Then, we
may write
III

=
_
R
6
w

0
(x)
_
S
0
g(0, .)
_
(x)dx.
Moreover, we established in the proof of Proposition 4.3 that if is rapidly decreas-
ing at innity, then S
0
B. Hence, since (w

0
) converges weakly- in B

to w
0
,
we get
lim
0
III

=
_
R
6
w
0
(x)
_
S
0
g(0, .)
_
(x)dx,
which ends the proof.
Acknowledgement. I would like to thank my advisor Fran cois Castella for having
guided this work.
HIGH FREQUENCY ANALYSIS OF HELMHOLTZ EQUATIONS: CASE OF TWO POINT SOURCES 21
References
[1] S. Agmon, L. H ormander, Asymptotic properties of solutions of dierential equations with
simple characteristics, J. Analyse Math., 30 (1976), 1-38.
[2] J.D. Benamou, F. Castella, T. Katsaounis, B. Perthame, High frequency limit of the
Helmholtz equation, Rev. Math. Iberoamericana 18 (2002), 187-209.
[3] N. Burq, Mesures semi-classiques et mesures de defaut, Seminaire Bourbaki, Vol. 1996/97.
Asterisque No. 245 (1997).
[4] F. Castella, B. Perthame, O. Runborg, High frequency limit of the Helmholtz equation.
Source on a general manifold, Comm. P.D.E 3-4 (2002), 607-651.
[5] L. Erd os, H.T. Yau, Linear Boltzmann equation as scaling limit of the quantum Lorentz
gas, Advances in di. eq. and math. physics (Atlanta, GA, 1997) 137-155 Contemp. Math.,
217, Amer. Math. Soc., Providence, RI, 1998.
[6] P. Gerard, Mesures semi-classiques et ondes de Bloch, In Seminaire Equations aux derivees
partielles 1988-1989, exp XVI, Ecole Polytechnique, Palaiseau (1988).
[7] P. Gerard, P.A. Markowich, N.J. Mauser, F. Poupaud, Homogeneisation limits and
Wigner transforms, Comm. pure and Appl. Math., 50 (1997), 321-357.
[8] L. H ormander,The Analysis of Linear Partial Dierential Operators I, Springer-Verlag,
1983.
[9] P.-L. Lions, T. Paul, Sur les mesures de Wigner, Revista Matem atica Iberoamericana, 9
(3) (1993), 553-618.
[10] L. Miller, Refraction of high-frequency waves density by sharp interfaces and semiclassical
measures at the boundary, J. Math. Pures Appl. (9) 79 (2000), 227269.
[11] G. Papanicolaou, L. Ryzhik, Waves and Transport. Hyperbolic equations and frequency
interactions, IAS/Park City Math. Ser., 5 (1997).
[12] B. Perthame, L. Vega, Morrey-campanato estimates for the Helmholtz equation, J. Funct.
Anal. 164(2) (1999), 340-355.
[13] L. Tartar, H-measures, a new approach for studying homogeneisation, oscillations and
concentration eects in partial dierential equations, Proc. Roy. Soc. Ed., 115 A (1990),
193-230.
[14] E. Wigner, On the quantum correction for thermodynamic equilibrium, Phys. Rev., 40
(1932).