an introduction to the use of integral equation for ... · an introduction to the use of integral...
TRANSCRIPT
An introduction to the use of integral equation
for Harmonic Maxwell equation
J.C. NÉDÉLECCMAP, École Polytechnique
Integral for Maxwell
Sommaire1. Introduction
2. The plane waves solutions
3. Fundamental Solution and RadiationConditions
4. Elementary differential geometry
5. The single layer for Helmholtz equation
6. Integral Representations
7. Calderon projectors
8. The perfect conductor
9. EFIE, MFIE and CFIE equations
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
10. Approximation for EFIE
11. Single layer for the Laplace Problem
12. Homogeneous and pseudo-homogeneouskernels
13. Pseudo-homogeneous kernels
14. The Helmholtz decomposition
15. The dielectric case
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
1 – Introduction
Electromagnetic waves are defined by theelectric fieldE and themagneticfield H at each point inR3. An isotropic dielectric mediumis characterizedby the electric permittivityε andthe magnetic permeabilityµ.
The speed of wavesin this dielectric medium is1/√
εµ. We denote byε0,µ0 respectively the permittivity and the permeability of the vacuum, and byc the speed of light in a vacuum, which is
c =1√ε0µ0
. (1)
The relative permittivity and permeability of the medium are defined by
ε = εrε0, εr ≥ 1,
µ = µrµ0, µr ≥ 1.
(2)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
In the absence of electric and magnetic charges and currents, the electric
and magnetic fields in a dielectric medium are governed by theMaxwell
system of equations
−ε∂E∂t
+ curlH = 0,
µ∂H∂t
+ curlE = 0.
(3)
We must complete these equations bythe transmission conditionsat
interfaces that separate different dielectric media.The tangential
components of the fieldsE andH are continuous acrossa surfaceΓ of
discontinuity ofε or µ. We denote byn the unit normal toΓ and then these
jump conditionstake the form
[E ∧ n]Γ = 0,
[H ∧ n]Γ = 0.
(4)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Isotropic conducting media are characterized bythe conductivityσ, whichis a real positive number. Maxwell equations in such a mediumare
−ε∂E∂t
+ curlH − σE = 0, (Ampère−Maxwell law),
µ∂H∂t
+ curlE = 0, (Faraday law).
(5)
The interface conditions(4) on dielectric media are unchanged.
We introduce theelectric inductionD and themagnetic inductionB:
D = εE, (6)
B = µH. (7)
From the Maxwell equations, it follows that
divB = 0. (8)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Theharmonic solutionsof Maxwell equations are complex-valued fieldsE
andH such that the following fields(ω is the pulsation)
E(t, x) = ℜ(E(x)e−iωt
),
H(t, x) = ℜ(H(x)e−iωt
),
(9)
satisfy the Maxwell system. They satisfy the system ofHarmonic Maxwell
equations
iωεE + curlH = 0,
−iωµH + curlE = 0.
(10)
It follows from Maxwell equations that
divD = 0. (11)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
In the case of a conducting media, we can redefinethe new electricpermittivity asε = ε + iσ/ω and the equation (10) is just similar.
Theperfectly conducting mediumis a common model. The conductor is abounded domain withboundaryΓ. We take the limit whenthe conductivityσ and thusε tends to infinityin the conductor. Then for a fixed pulsationω,the two fieldsE andH tend to zero in this medium. The following interior
limits E ∧ n|Γ is zero, but it is not the case forH ∧ n|Γ. A surfacic electriccurrent appears and also a density of electric charges.From the first part ofthe jump condition(4) the following exterior limit is also zero
E ∧ n|Γ = 0, (12)
while the exterior limitsH ∧ n andE · n are non-zero.We thus only have
to find the solution of (10) and (12) in the infinite exterior dielectricmedium. Aclassical exterior problem ( the PEC problem)is that of aperfectly conducting objectlit by an incident plane wave.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
2 – The plane waves solutions
They are specific solutions inR3 of the Harmonic Maxwell equations (10)
for an homogeneous media. Asε andµ are constant inR3, we have
divE = divH = 0; (13)
Using the vectorial identity
∆E = ∇divE − curl curlE (14)
yield thatE andH are also solutions to the wave equations
∆E + k2E = 0,
∆H + k2H = 0, k = ω√
εµ.(15)
The plane waves are associated with a scalar wave equation.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Looking forE, H such that
E(x) = ~e ei(~k.~x),
H(x) = ~h ei(~k.~x), with |~k| = k
(16)
leads to the equations
ωε~e + ~k ∧ ~h = 0,
−ωµ~h + ~k ∧ ~e = 0.(17)
We notice thatthe three vectors~e,~h,~k are mutually orthogonal, and therelations(17) now describe all the plane wave solutions.The vector~k givesthe direction of propagation of the wave. We see thatthe electric and themagnetic fields lie in the plane orthogonal to~k, called thephase plane.These three vectors can be real or complex.Whenk is real, the modulus ofthese solutions is constant.When~k is real and~e is complex of the form~e = ~α1 + i~α2 with (~α1 · ~α2) = 0, we speak of acircular plane wave.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
3 – Fundamental Solution and Radiation Conditions
We denote byG the fundamental solution of the scalar Helmholtz equation
which satisfies the outgoing radiation condition:
G(x) =1
4π
eik |x||x| . (18)
The fundamental solution for a system of equationswith n unknowns and
n equations of the form
au = 0, (19)
wherea is a differential operator with constant coefficients, is defined as a
matrixn × n, denoted byA, such that
aA = δI, (20)
whereδ is the Dirac mass at the originandI is then × n identity matrix.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Theterms ofA are in general distributions, and this will be the case here, in
contrast with the fundamental solution of the Harmonic Helmholtz equation
given by(18) which is an integrable function.
Thus, the fundamental solution of the Maxwell system is a6 × 6 matrix.
Due to the symmetry of the operator, we only need to compute half of this
matrix or equivalently two3 × 3 matrices, solutions of the system
iωεE + curlH = δI, I is the 3 × 3 identity matrix,
−iωµH + curlE = 0.(21)
We obtain the other half of the fundamental solution by exchanging the role
the fieldsE andH, and the parametersε andµ.
Thewave numberk is
k = ω√
εµ. (22)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Theorem 1 The first half of the fundamental solution of the Maxwell
system, solution of(21), is given by
E(x) = iωµG(x)I + iωεD2 G(x),
H(x) = curl(G(x)I).
(23)
The second half of the fundamental solution of the Maxwell system is given
by
E(x) = curl(G(x)I),
H(x) = −iωεG(x)I − iωµD2 G(x).
(24)
The second derivativeD2 should be understood in the sense of
distributions inD′(R3).
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Proof
We introduce thescalar potentialV and thevector potentialA
E = ∇V + A,
divA − k2V = 0 ( Lorentz gauge condition).(25)
In the case of the fundamental solution, given by equation(21), thecorresponding quantities are respectively the scalar potential V , which is avector inR
3, and the vector potentialA which is a3 × 3 matrix. Taking thedivergence in the decomposition(25), it follows that
divE = ∆V + divA, (26)
which is obtained from the divergence in the first part of equation (21),
divE =1
iωεdiv(δI) = − i
ωε∇δ. (27)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
From the gauge condition(25), we can eliminateA and we obtain an
equation forV :
∆V + k2V = − i
ωε∇δ, (28)
which outgoing fundamental solution is
V (x) =i
ωε∇G(x). (29)
Taking the curl in the decomposition(25), it follows that
curlE = curlA. (30)
EliminatingH in (21) yields
curlcurlE − k2E = iωµδI. (31)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
From(25), we have
∇divA − k2(E − A) = 0, (32)
Using(32) in connection with(31) and(30) gives
∇divA − curlcurlA + k2A = −iωµδI, (33)
or equivalently
∆A + k2A = −iωµδI. (34)
The outgoing fundamental solution of(34) is
A(x) = iωµ G(x)I. (35)
Thus, we have obtained(23). The expression(24) follows by exchanging
the quantitiesE andH, andε andµ, without forgetting to changei into−i.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
It is important to notice thatthe coefficients of the fundamental matrix arenot integrable functions, nor usual finite parts. In particular,the trace ofD2 G is ∆G, which has a Dirac mass singularity at the origin.
We have used thefundamental solution of the Harmonic Helmholtzequationgiven by(18).
This function satisfies the followingSommerfeld conditionor outgoingwave condition
∣∣∣∣∂u
∂r− iku
∣∣∣∣ ≤c
r2at infinity, (r = |x|). (36)
We twice used thisSommerfeld conditionto select a unique fundamentalsolution for the scalar potential and the vector potential.
We then need to describe precisely the behaviour of the solutions at infinity,which is equivalent todefining the radiation conditions. They appear in theexpression of the fundamental solution.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Theorem 2 The fundamental solution(23) and(24) (Theorem1) satisfies
the usual radiation conditions (r = |x|)∣∣∣∣∂E
∂r− ikE
∣∣∣∣ ≤c
r2, for large r, (37)
∣∣∣∣∂H
∂r− ikH
∣∣∣∣ ≤c
r2, for large r, (38)
and the followingradiation conditions of Silver-Müller : (n = −→r /r)
∣∣√ε E −√µ H ∧ n
∣∣ ≤ c
r2, for large r, (39)
∣∣√ε E ∧ n +√
µ H∣∣ ≤ c
r2, for large r. (40)
Any of these four radiation conditions selects a unique fundamental
solution. Moreover, each component ofE andH behaves as1/r at infinity
The proof is based on the expansion at infinity of the functionG and itsderivatives.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
A more concise expression of theSilver-Müller radiation conditionsfor the
Maxwell system is (c denotes a generic constant )
|E(x)| ≤ cr , for large r,
|H(x)| ≤ cr , for large r,
∣∣∣√
εE −√µ H ∧
−→rr
∣∣∣ ≤ cr2 , for large r.
(41)
Remark 1 : The plane wave solution do not satisfy the Silver-Müller
radiation conditions (as for example they do not tend to zeroat infinity).
This condition is characteristic of thescattered part of the fieldcreated by
the bounded obstacle.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
In the case of an object lit by anincident plane wave, the total fieldtakes theform
E = Einc + Es H = Hinc + Hs (42)
wherethe scattered fieldEs, Hs satifies the Silver-Müller conditions.
Now a formulation of thePEC problemfor a general excitation associatedto a boundary datag, is: FindE andH such that
iωεE + curlH = 0, x ∈ Ωe,
−iωµH + curlE = 0, x ∈ Ωe,
E ∧ n|Γ = g, x ∈ Γ, g a given vector tangent to Γ,
E and H satisfy (41).
(43)
and in the case aboveg = −Einc ∧ n|Γ . (44)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Figure 1: Notations
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
The interior harmonic Maxwell problemis
iωεE + curlH = 0, x ∈ Ωi,
−iωµE + curlE = 0, x ∈ Ωi,
E ∧ n|Γ = g, x ∈ Γ, g a given vector tangent to Γ.
(45)
Remark 2 : Wheng = −Einc ∧ n|Γ, its solution is just minus the
incoming plane waveE = −Einc, H = −Hinc.
It is thus known in that case.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
4 – Elementary differential geometry
A surfaceΓ in R3 is usually defined by a system of chartswhich maps apart of this surface on a bounded squared domain ofR2. We introduce somedifferential operators on such a surfaceΓ. These operators are needed in thestudy of the Maxwell system and its integral representation.
For every pointy in R3, we denote byδ(y) the distance ofy to the surfaceΓ
δ(y) = infx∈Γ |y − x| . (46)
A collection of points whose distance to the surface is less thanε is called atubular neighbourhood of the surfaceΓ. Forε small enough and when thesurface is regular and oriented, any pointy in such a neighbourhoodΓε, hasa unique projectionP(y) on the surface which satisfies
|y − P(y)| = δ(y). (47)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
It is easy to check that for a regular surfaceΓ which admitsa tangent plane
at the pointP(y), the liney − P(y) is directed along the normal to the
surfaceΓ at this point. Notice thatthe unit normaln is the gradient of the
functionδ(y) oriented towards the interior or the exterior depending on the
position of the pointy. It holds that
n(P(y)) = ∇δ(y), when y ∈ Ωe
n(P(y)) = −∇δ(y), when y ∈ Ωi.(48)
Any pointy in the tubular neighbourhoodΓε is located usingP(y) and
δ(y), or more precisely we write:
y = P(y) + s n(P(y)), −ε ≤ s ≤ ε,
s = δ(y), when y ∈ Ωe, s = −δ(y), when y ∈ Ωi.(49)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
To any functionu defined on the surfaceΓ, we associate thelifting u
defined on the tubular neighbourhoodΓε by
u(y) = u(P(y)). (50)
We continue todenote byn(y) the vector field∇s which is defined onΓε
and has the valuen on the surfaceΓ.
We introduce the familyΓs of parallel surfaces
Γs = y; y = x + sn(x); x ∈ Γ . (51)
The vector fieldn is the field of normals to the surfacesΓs.
We introduce the first family ofdifferential operatorswhich actsonfunctions defined on the surfacesΓ or Γs. Thetangential gradient∇Γu isdefined as
∇Γu = ∇u|Γ . (52)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Thetangential rotational of a functionis defined as
−−→curlΓu = curl(un)|Γ . (53)
The field of normals is a gradient, which implies
curln = 0. (54)
We use the vectorial calculus formula
curl(u−→v ) = ∇u ∧ −→v + ucurl−→v (55)
which yields−−→curlΓu = ∇Γu ∧ n. (56)
Thecurvature operatorRs is
Rs = ∇n. (57)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
When we are on the surfaceΓ (i.e., whens = 0), we omit the indexs.
Theorem 3 The curvature operatorRs is a symmetric operator acting in
the tangent plane. We denote by1/R1(s) and1/R2(s) its two eigenvalues,
calledprincipal curvatures. We define itsmean curvatureHs to be half its
trace
Hs =1
2divn =
1
2
(1
R1(s)+
1
R2(s)
). (58)
We define itsGauss curvatureGs to be its determinant
Gs =1
R1R2= det
(Rs|Γ
). (59)
The area element on the surfaceΓs at the pointy = x + sn(x) is related
to the area element on the surfaceΓ at the pointx through the relation
dγs(y) =(1 + 2sH(x) + s2G(x)
)dγ(x). (60)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
The normal derivative of the curvature tensorRs is
∂
∂sRs = −R2
s. (61)
Moreover, it holds that
∂
∂sHs = −1
2trace
(R2
s
)= −1
2
(1
(R1(s))2 +
1
(R2(s))2
), (62)
∂
∂sGs = −2HsGs, (63)
[Rs(v ∧ n) − 2Hs(v ∧ n)] ∧ n = Rsv, (for every vector v). (64)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We introduce now afamily of operators acting on tangent vector fieldsv.
We need to define a liftingv of this field in the neigbourhoodΓε. It is
natural to ask it to be tangent to the surfacesΓs.
We transport the tangential vector fieldsin the neighbourhoodΓε using the
following lifting
v(y) = v(x) − sRs(y)v(x). (65)
We introduce the followingfirst order differential operators:
Thesurfacic divergenceof a tangent vector fieldv is
divΓv = divv|Γ. (66)
Thesurfacic rotational of a vector fieldv is
curlΓv = (curlv · n(x))|Γ. (67)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We introduce the followingsecond order differential operators:
Thescalar Laplacianor Laplace-Beltrami operator acting on a functionu is
∆Γu = divΓ∇Γu = −curlΓ−−→curlΓu. (68)
Thevectorial Laplacian or Hodge operator acting on a tangent vectorfield v is
∆Γv = ∇ΓdivΓv −−−→curlΓcurlΓv. (69)
Theorem 4 Letu ∈ C1(Γ) be a function andv ∈(C1(Γ)
)2a tangent
vector field defined on the surfaceΓ. The followingStokes identitieshold:∫
Γ
(∇Γu · v) dγ +
∫
Γ
u divΓvdγ = 0, (70)
∫
Γ
(−−→curlΓu · v
)dγ −
∫
Γ
u curlΓvdγ = 0, (71)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
divΓ−−→curlΓu = 0, (72)
curlΓ ∇Γu = 0, (73)
divΓ(v ∧ n) = curlΓv. (74)
Moreover, ifw is a functionC2(Γ), we have:
−∫
Γ
∆Γw udγ =
∫
Γ
(∇Γw · ∇Γu) dγ =
∫
Γ
(−−→curlΓw · −−→curlΓu
)dγ. (75)
If w is a tangent vector field(C2(Γ)
)2, it holds that
−∫
Γ
(∆Γw · v) dγ =
∫
Γ
divΓw divΓvdγ +
∫
Γ
curlΓw curlΓvdγ. (76)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
The surfacic operators in the domainΓε and the usual three dimensional
operators inR3 are linked.
Theorem 5 Letu be a function defined onΓε andv a field of vectors inR3
defined onΓε. The following decompositions hold:
∇u = ∇Γsu +∂u
∂sn, (77)
v = (v · n)n + vΓs , vΓs = n ∧ (v ∧ n), (78)
divv = divΓsvΓs + 2Hs(v · n) +∂
∂s(v · n), (79)
curlv = (curlΓsvΓs) n +−−→curlΓs(v · n) + Rs(v ∧ n)
−2Hs(v ∧ n) − ∂∂s
(v ∧ n),
(80)
∆u = ∆Γsu + 2Hs∂u
∂s+
∂2
∂s2u, (81)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
∆v = ∆ΓsvΓs + 2Hs∂
∂svΓs +
∂2
∂s2 vΓs
+
[∆Γs(v · n) + 2Hs
∂
∂s(v · n) +
∂2
∂s2 (v · n)
]n
+
[2 (v · ∇ΓsHs) − 2divΓs(Rsv) + 2(v · n)
∂
∂sHs
]n
+ (2HsI − 2Rs)Rsv + 2Rs∇Γ(v · n) + 2(v · n)∇ΓHs.
(82)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
5 – The single layer for Helmholtz equation
The Helmholtz equation appears in acoustics problems. It described the
equation satisfy by the pressure in the case of harmonic solution.
The two classical boundary conditions for the interior Helmholtz Equation
are theDirichlet and the Neumann problems
∆u + k2u = 0, x ∈ Ωi,
u|Γ = ud.(83)
∆u + k2u = 0, x ∈ Ωi,
∂u∂n
∣∣∣Γ= un.
(84)
In the case ofexterior problems, we seek complex-valued solutions. We
need to specify the behavior at infinity and in particular to impose the
radiation condition.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
The two model equations are the Dirichlet and the Neumann problems
∆u + k2u = 0, x ∈ Ωe,
u|Γ = ud.(85)
∆u + k2u = 0, x ∈ Ωe,
∂u∂n
∣∣∣Γ= un.
(86)
In both cases, the conditions at infinity are (c is a fixed constant) |u| ≤ c
r , r large,
|∇u| ≤ cr ,
(87)
and theSommerfeld condition∣∣∣∣∂u
∂r− iku
∣∣∣∣ ≤c
r2. (88)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Notice that this last condition implies that the solutions are complex-valued,
althoughk2 is real. The associated operator is not symmetric.
This condition is satisfies by the scattered part of the wave and it means that
this wave is going to infinity (outgoing).
Thefundamental solution of the Helmholtz equation which satisfies the
outgoing radiation conditionis
G(x) =1
4π
eikr
r, r2 = x2
1 + x22 + x2
3. (89)
This function is singular at the origin. It is possible to check that it is
integrable and that its trace on a regular surface crossing the origin is still
integrable on this surface.
We can used this fundamental solution to obtain integral representation of
the solutions of the problems (83), (84), (85) and (86).
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Theorem 6 (Representation theorem)Letu be a function such that (with
possiblyk = 0)
∆u + k2u = 0 in Ωi, (90)
∆u + k2u = 0 in Ωe, (91)
u satisfies the outgoing radiation condition (88) and its tracesuint and
uext, (∂u/∂n)|int and(∂u/∂n)|ext belong toC0(Γ). We define
[u] = uint − uext, (92)[∂u
∂n
]=
∂u
∂n
∣∣∣int
−∂u
∂n
∣∣∣ext
. (93)
For y /∈ Γ, the followingrepresentation formula holds:
u(y) =
∫
Γ
G(x − y)
[∂u
∂n(x)
]dγ(x) −
∫
Γ
∂
∂nx(G(x − y))[u(x)]dγ(x),
(94)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
and fory ∈ Γ
uint(y) + uext(y)
2=
∫
Γ
G(x − y)
[∂u
∂n(x)
]dγ(x)
−∫
Γ
∂
∂nx(G(x − y))[u(x)]dγ(x).
(95)
Proof We use the Green formula with on one sideu, and on the other sidethe functionG(· − y). For any domainΩ, whose boundary is∂Ω, and foranyy /∈ Ω, we have
0 =
∫
Ω
(∆u(x) + k2u(x)
)G(x − y)dx
−∫
Ω
(∆G(x − y) + k2G(x − y)
)udx
=
∫
∂Ω
[∂u
∂n(x)G(x − y) − ∂
∂nxG(x − y)u(x)
]dγ(x).
(96)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Let y be in the domainΩi andBε be a ball whose center isy, radiusε and
boundarySε. We also introduce a ballBR whose center isy, radiusR,
chosen large enough to contain the domainΩi, and whose boundary isSR.
We use the above Green formula in the two domains:Ωi − Bε, Ωe ∩ BR.
Adding the two corresponding contributions yields
∫
Γ
G(x − y)
[∂u
∂n(x)
]dγ(x) −
∫
Γ
∂
∂nxG(x − y)[u(x)]dγ(x)
−∫
Sε
G(x − y)∂u
∂n(x)dγ(x) +
∫
Sε
∂
∂nxG(x − y)u(x)dγ(x)
+
∫
SR
G(x − y)∂u
∂n(x)dγ(x) −
∫
SR
∂
∂nxG(x − y)u(x)dγ(x)
= 0.
(97)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
The functionu is regular in the ballBε. Thus, whenε tends to zero, the firstintegral onSε is bounded by
∣∣∣∣∫
Sε
G(x − y)∂u
∂n(x)dγ(x)
∣∣∣∣ ≤ εsupx∈Bε
∣∣∣∣∂u
∂n(x)
∣∣∣∣ (98)
and tends to zero. The second integral has the value
∫
Sε
∂
∂nxG(x − y)u(x)dγ(x) = u(y)
∫
Sε
∂
∂nxG(x − y)dγ(x)
+
∫
Sε
∂
∂nx(G(x − y))(u(x) − u(y))dγ(x),
(99)
∂
∂nxG(x − y) =
(ik
4πr− 1
4πr2
)eikr, (100)
∣∣∣∣∫
Sε
∂
∂nx(G(x − y))(u(x) − u(y))dγ(x)
∣∣∣∣
≤ (kε + 1)supx∈Bε|u(x) − u(y)| .
(101)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
This last term tends to zero. The first term in the right hand side of(99)
admits the limit−u(y).
Let us now examine the terms on the sphereSR. They take the form
−∫
SR
(∂
∂nxG(x − y) − ikG(x − y)
)u(x)dγ(x)
+
∫
SR
G(x − y)
(∂u
∂n− iku(x)
)dγ(x).
(102)
The first of these terms is bounded by∣∣∣∣∫
SR
(∂
∂nxG(x − y) − ikG(x − y)
)u(x)dγ(x)
∣∣∣∣ ≤ supx∈SR|u(x)| .
(103)
ForR large enough, the condition(87) implies that this term tends to zero.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
The condition(88) shows that the second term is bounded by∣∣∣∣∫
SR
G(x − y)
(∂u
∂n(x) − iku(x)
)dγ(x)
∣∣∣∣ ≤C
R, (104)
and thus tends to zero whenR tends to infinity.
When the pointy is on the surfaceΓ, the ballBε is split intoΩi ∩ Bε and
Ωe ∩ Bε, which yields two pieces in the corresponding integrals. Whenε
tends to zero, these two pieces are asymptotically separated by the tangent
plane. Thus the associated integrals onSε give rise to the term
(1/2) (uint(y) + uext(y)). We must notice that in this case, the integrands
associated with the surfaceΓ admit a singularity at the pointy. The one
associated with the kernelG(x − y) is singular as1/|x − y| and thus
belongs to the spaceL1(Γ) for x onΓ.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
In the expression of the other integrand appears the kernel
∂
∂nxG(x − y) =
eik|x − y|4π|x − y|2
(ik − 1
|x − y|
)((x − y) · nx), (105)
which singularity is equivalent to1/|x − y|, since its numerator
((x − y) · nx) vanishes as|x − y|2 in a neighbourhood ofy, and thus this
kernel belongs to the spaceL1(Γ) for y ∈ Γ. This is no longer true wheny
is not onΓ.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
The following integral expression is calleda single layer potential:
u(y) =
∫
Γ
G(x − y)q(x)dγ(x); (106)
The following integral expression is calleda double layer potential:
u(y) =
∫
Γ
∂G
∂nx(x − y)ϕ(x)dγ(x). (107)
These two types of potentials satisfy the Helmholtz equation in Ωi andΩe
and the outgoing radiation condition at infinity.
The singularity contained in the kernel implies that these potentials admits
some jumps in their values when the current pointy cross te surfaceΓ. In
the case of the single layer these discontinuity are described in the
following theorem:
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Theorem 7 The single layer potential(106) is continuous with respect to
the variabley in R3 and especially when crossing the surfaceΓ where its
value is
u(y) =
∫
Γ
G(x − y)q(x)dγ(x), y ∈ Γ, when q(x) ∈ C0(Γ). (108)
Its tangential derivatives are also continuous. Its normalderivative has a
discontinuity when crossingΓ. Its limit value on both side ofΓ are
∂u
∂n
∣∣∣int
(y) =q(y)
2+
∫
Γ
∂
∂nyG(x − y)q(x)dγ(x), (109)
∂u
∂n
∣∣∣ext
(y) = −q(y)
2+
∫
Γ
∂
∂nyG(x − y)q(x)dγ(x). (110)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Proof
The property(108) results from the integrability of the singularity ofE in a
tubular neighbourhood ofΓ. This singularity is uniformly integrable with
respect to the variabley and thus the classical Lebesgue theorem implies
continuity with respect toy and(108). The theorem6 can be used when the
functionu is represented as a single layer potential. It follows that
[u] = 0 and q =
[∂u
∂n
]. (111)
The gradient of a single layer potential is
∇u(y) =
∫
Γ
∇yG(x − y)q(x)dγ(x). (112)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We use the specific moving system of coordinates introduced above. A
pointy+ in a tubular neighbourhood of the surfaceΓ is determined by its
projectiony0 and the distanceρ to Γ. We introduce the “symmetric” point
y− : y+ = y0 + ρny0
,
y− = y0 − ρny0.
(113)
We compute the following sum:
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
((∇u (y+) + ∇u (y−)) · ny0)
=
∫
Γ
((∇y+
G (x − y+) + ∇y−G (x − y−)
)· ny0
)q(x)dγ(x)
=
∫
Γ
[eik |x − y+|
4π|x − y+|2(
ik − 1
|x − y+|
)
+ eik |x−y−|4π |x−y−|2
(ik− 1
|x−y−|
)]((y0−x) · ny0
) q(x)dγ(x)
+
∫
Γ
[eik |x − y+|
4π |x − y+|2(
ik − 1
|x − y+|
)
− eik |x − y−|4π |x − y−|2
(ik − 1
|x − y−|
)]ρq(x)dγ(x).
(114)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
The classical Lebesgue theorem applies to the first part of the right-handside of(114) since((x − y0) · ny0
) is equivalent to|x − y0|2. Its limit is
2
∫
Γ
∂
∂nyG(x − y)q(x)dγ(x), when ρ tends to 0.
It is more difficult to show that the second part tends to zero with ρ. Let usstudy the more delicate term which is
∫
Γ
(1
|x − y+|3− 1
|x − y−|3
)ρq(x)dγ(x)
=
∫
Γ
(|x−y−|2−|x−y+|2
)(|x−y+|2+|x−y−|2+|x−y+||x−y−|
)
|x − y+|3 |x − y−|3 (|x − y+| + |x − y−|)
× ρq(x)dγ(x).(115)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We can then write
|x − y−|2 − |x − y+|2 = 4ρ ((x − y0) · ny0) . (116)
This integral is bounded by
∣∣∣∣∣
∫
Γ
(1
|x − y+|3− 1
|x − y−|3
)ρq(x)dγ(x)
∣∣∣∣∣
≤ C
∫
Γ
ρ2 |x − y0|2(|x − y0|2 + ρ2
)5/2dγ(x).
(117)
We split this last expression into two parts:
– when|x − y0| ≤ ρ
∫
|x−y0|≤ρ
ρ2 |x − y0|2(|x − y0|2 + ρ2
)5/2dγ ≤ Cρ,
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
– when|x − y0| ≥ ρ
∫
|x−y0|≥ρ
ρ2 |x − y0|2(|x − y0|2 + ρ2
)5/2dγ
≤ ρε
∫
Γ
|x − y0|4−ε
(|x − y0|2 + ρ2
)5/2dγ ≤ ρε
∫
Γ
dγ
|x − y0|1+ε,
from which it follows that the limit is zero. The other terms can be treated
similarly. Equalities(109) and(110) follow from (111) and the limit(114).
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
6 – Integral Representations
We exhibit the integral representations of the Maxwell equation and give
their principal properties. There are obtain when the values of the constants
ε andµ are the same for the interior domainΩi and the exterior domainΩe.
Theorem 8 (Representation Theorem)LetΩi be a bounded regular
interior domain, which boundary is the regular surfaceΓ and denote byΩe
the associated exterior domain. The exterior unit normal toΓ is denoted by
n. LetE andH be regular solutions of the Maxwell equations
curlE − iωµH = 0,
curlH + iωεE = 0,inΩi, (118)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
curlE − iωµH = 0,
curlH + iωεE = 0,inΩe,
E andHsatisfy the radiation
conditions of Silver-Müller.
(119)
Let’s denotej andm the electric and magnetic tangent currentsto the
surfaceΓ
j = Hi ∧ n − He ∧ n
m = Ei ∧ n − Ee ∧ n;(120)
whereHi andHe are respectively the interior and exterior limits of the
fieldH, whileEi andEe are the interior and exterior limits of the fieldE.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Then, the fieldsE and H admit the integral representation
E(y) = iωµ
∫
Γ
G(x − y)j(x)dγ(x)
+ iωε∇
∫
Γ
G(x − y)divΓj(x)dγ(x)
+curl
∫
Γ
G(x − y)m(x)dγ(x), y /∈ Γ.
(121)
H(y) = −iωε
∫
Γ
G(x − y)m(x)dγ(x)
− iωµ∇
∫
Γ
G(x − y)divΓm(x)dγ(x)
+curl
∫
Γ
G(x − y)j(x)dγ(x), y /∈ Γ.
(122)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Threeintegral operatorsappear that are :
Sj(y) =
∫
Γ
G(x − y)j(x)dγ(x), (123)
Tj(y) = ∇∫
Γ
G(x − y)divΓj(x)dγ(x) (124)
Rj(y) = curl
∫
Γ
G(x − y)j(x)dγ(x). (125)
We will need the value on the surfaceG of the fields represent by theintegral representation(121) and(122). These values are associated to therespective jumps of the above operatorsS, T andR.
The operatorsS andT have continuous values acrossG. So is theirtangential derivatives. ButT exhibit a discontinuous normal component andS a discontinuous normal derivative acrossG.
The operatorR have discontinuous tangential values acrossG. But itexhibit a continuous normal component acrossG.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Proof
It is based on the known results for the integral representation of the scalarHelmholtz equation given in Theorems6 and7, and the use of someproperties of distributions.
The fieldsE andH, solutions of(118) and(119), satisfy in the sense ofdistributions inR3 :
curlE − iωµH = mδΓ,
curlH + iωεE = jδΓ.(126)
Thus,E andH are the sum of two contributions, one associated withj, theother associated withm. Let us compute the contribution associated withj,which solves the equation
curlE − iωµH = 0,
curlH + iωεE = jδΓ.(127)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
It takes the form ofthe sum of a vector potentialA and a scalar potentialV :
E = A + ∇V. (128)
Exactly as in the computation of the fundamental solution, these potentials
must be linked by a gauge condition, which we choose to be theLorentz
gauge
divA − k2V = 0. (129)
The potentialsA andV are continuous across the surfaceΓ.
We have
iωεdivE = div (jδΓ) = (divΓj) δΓ, (130)
which, in combination with the gauge condition yields
∆V + k2V = − i
ωεdivΓjδΓ. (131)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Thus, the potentialV is given by a single layer potential for the scalar
Helmholtz equation, with density(idivΓj)/(ωε) :
V (y) =i
ωε
∫
Γ
G(x − y)divΓj(x)dγ(x). (132)
From equation(127), it follows that
∆A + k2A = ∇divA − curlcurlA + k2A
= k2∇V − k2∇V + k2E − curlcurlE
= −iωµjδΓ.
(133)
Thus, the vector potentialA is given by a single layer potential for the
scalar Helmholtz equation , with densityiωµj:
A(y) = iωµ
∫
Γ
G(x − y)j(x)dγ(x).
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
This give the expression of this part of the fieldE and the similar part ofH
E(y) = iωµ
∫
Γ
G(x − y) j(x) dγ(x)
+i
ωε∇∫
Γ
G(x − y)divΓj(x)dγ(x),
(134)
H(y) = curl
∫
Γ
G(x − y)j(x)dγ(x). (135)
By an argument of symmetry, the part associated withm is
E(y) = curl
∫
Γ
G(x − y) m(x)dγ(x), (136)
H(y) = −iωε
∫
Γ
G(x − y) m(x)dγ(x)
− i
ωµ∇∫
Γ
G(x − y)divΓm(x)dγ(x).
(137)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Theorem 9 Theinterior projectoris the operatorCint which maps the
couple(m, j) to theinterior limit values of the integral representation. It is
given by
Ei ∧ n =m
2− ny ∧ (Rm +
i
ωεT j + iωµSj), (138)
Hi ∧ n =j
2+ ny ∧ (
i
ωµTm + iωεSm − Rj). (139)
Theexterior projectoris the operatorCext which maps the couple(m, j) to
the to theexterior limit values of the integral representation. It is given by
−Ee ∧ n =m
2+ ny ∧ (Rm +
i
ωεT j + iωµSj), (140)
−He ∧ n =j
2− ny ∧ (
i
ωµTm + iωεSm − Rj). (141)
The operatorsCint andCext are projectors.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
The expression ofthe interior and exterior valuesare deduced from the
properties of single layer potentials given in Theorems6 and7. (continuity
across the surfaceΓ of the potential and the tangential part of its gradient,
discontinuity of the normal derivative). Besides, using(80), we have
curlu ∧ n =∂
∂nuT −∇T (u · n) + RuT . (142)
Thus, for a single layer potentialu, the jump ofcurlu∧ n is the same as the
jump of∂uT /∂n, whereuT is the tangential component. It follows that
limy→Γ±
(curly
∫
Γ
G(x − y)j(x)dγ(x)
)∧ ny
= ±j(y)
2+
∫
Γ
(curly(G(x − y)j(x)) ∧ ny) dγ(x).
(143)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We have
curly(G(x − y)j(x)) ∧ ny = (∇yG(x − y) ∧ j(x)) ∧ ny
= ∂G∂ny
(x − y)j(x) −∇yG(x − y) (j(x) · (ny − nx)) .(144)
In order to exhibit the limit of the gradient term, we use
ny ∧∇∫
Γ
G(x − y)ρ(x)dγ(x)
=
∫
Γ
ny ∧∇yG(x − y)ρ(x)dγ(x)
(145)
and thus
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
ny ∧∇∫
Γ
G(x − y)ρ(x)dγ(x)
=
∫
Γ
((ny − nx) ∧∇yG(x − y)) ρ(x)dγ(x)
+
∫
Γ
−−→curlΓxG(x − y)ρ(x)dγ(x)
=
∫
Γ
((ny − nx) ∧∇yG(x − y)) ρ(x)dγ(x)
−∫
Γ
G(x − y)−−→curlΓρ(x)dγ(x),
(146)
from which we obtain formulas(138), (139), (140) and(141).
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
7 – Calderon projectors
We introduce theimpedancez =
õ
ǫand thefrequencyk = ω
√εµ.
We introduce also theoperatorΠΓ which is the projector on the tangent
plane at a point ofΓ. We define theoperatorA which is called the EFIE
operator.
A = k ΠΓS + 1/k ΠΓT (147)
We associated to any operatorL thetwist operatordefined as
Lt = −ny ∧ L (148)
So we consider the following twist operatorsAt, St, Tt andRt.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
The Calderon projectorsintroduced in theorem9 are now (in matricial
form)
Cint =
I2 + Rt − i
z At
izAtI2 + Rt
(149)
Cext =
I2 − Rt
iz At
−izAtI2 − Rt
(150)
We thus have the identity (decomposition of the identity)
Cint + Cext = I. (151)
Then, both projectors satisfy
C C = C (152)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
which is also equivalent to the two identities
(I
2− C) (
I
2− C) =
I
4. (153)
Cint Cext = Cext Cint = 0 (154)
The two identities (152) (or (153)) are also equivalent to (151) and the
following one
(Cext − Cint) (Cext − Cint) = I. (155)
These identities are equivalent to the two following ones:
At At + Rt Rt =I
4,
At Rt + Rt At = 0.(156)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Two other expressions are the following
(At + Rt) (At + Rt) =I
4
(At − Rt) (At − Rt) =I
4.
(157)
Let B be the rotation or twist operator(which operates in the tangent planeto the surfaceΓ)
Bj = n ∧ j. (158)
Express in term of this operator, the Calderon identities are also
B (A + R) B (A + R) =I
4.
B (A − R) B (A − R) =I
4.
(159)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
8 – The perfect conductor
We examine in this section the specific properties andthe variational
formulation of the integral equationassociated withthe perfect conductor
problem. We have seen that the scattering of a plane wave, by a perfect
conducting object immersed in vacuum, leads to the exteriorMaxwell
problem
curlE − iωµ0H = 0, inΩe,
curlH + iωε0E = 0, inΩe,
E ∧ n|Γ = −Einc ∧ n, on Γ,∣∣√ε0E −√
µ0H ∧ n∣∣ ≤ c
r2 .
(160)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We associate with it, the interior Maxwell problem
curlE − iωµ0H = 0, inΩi,
curlH + iωε0E = 0, inΩi,
E ∧ n|Γ = −Einc ∧ n, on Γ,
(161)
which always admits the solution
E = −Einc,
H = −H inc.(162)
This solution is not unique when the interior problem is not invertible, i.e.,
whenk2 is an eigenvalue of the interior Maxwell problem.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We applythe representation Theorem8 to the solutions of the problems
(160) and(161). We obtain
E(y) = iωµ0
∫
Γ
G(x − y)j(x)dγ(x)
+i
ωε0∇∫
Γ
G(x − y)divΓj(x)dγ(x),
(163)
H(y) = curl
∫
Γ
G(x − y)j(x)dγ(x), (164)
j = −H inc ∧ n − H ∧ n. (165)
Remark 3 :The tangent fieldj is determined by the boundary condition on
the surfaceΓ, which expression is given by(140).
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We have obtained theintegral equation
(Einc ∧ n
)(y) = iωµ0ny ∧
[∫
Γ
G(x − y)j(x)]dγ(x)
+i
ωε0ny ∧
[∇y
∫
Γ
G(x − y)divΓj(x)]dγ(x).
(166)
In this first expression of the integral equation, only integrable kernels
appear. But,derivatives of the integral operator appear, which expression
will lead to a non integrable kernel if we derivate under the integral symbol.
Instead we introducea variational formulationfor this equation, where
appears only first derivatives of the unknownj, and yet no finite part or
non-integrable kernels.
First, using the projectorΠΓ, we can rewrite equation(166) as
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
−ΠΓEinc(y) = iωµ0ΠΓ
[∫
Γ
G(x − y)j(x)]dγ(x)
+i
ωε0∇Γ
[∫
Γ
G(x − y)divΓj(x)]dγ(x).
(167)
We multiply equation(167), written onΓ, by a test vectorjt. Using the
Stokes formula(75) on the surfaceΓ, we obtain
iωµ0
∫
Γ
∫
Γ
G(x − y)(j(x) · jt(y)
)dγ(x)dγ(y)
− iωε0
∫
Γ
∫
Γ
G(x − y)divΓj(x)divΓjt(y)dγ(x)dγ(y)
= −∫
Γ
(Einc · jt
)dγ, for any jt tangent toΓ.
(168)
This variational formulation is calledthe Rumsey principle.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
9 – EFIE, MFIE and CFIE equations
There aretwo integral techniques currently useto solve the equations (43)
and (45). The first one is the one introduced above in equation(167). It is
called theElectric Field Integral Equation or EFIE. It has also the following
form using the notation(147) with g = −ΠΓEinc.
Aj =i
zg. (169)
The second one (calledthe MFIE equation) usesthe representation ofH
given by(160)
H(y) = curl
∫
Γ
G(x − y)j(x)dγ(x), y /∈ Γ. (170)
In the case where theinterior conductor domain is not flat, theinterior
value ofH ∧ n (cf remark 2) and then using(139), we can write an
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
equation for this interior value:
1
2j(y) − n ∧ curl
∫
Γ
G(x − y)j(x)dγ(x) = −H inc ∧ n, y ∈ Γ. (171)
Thus,the MFIE equationis
(I
2+ Rt)j = −H inc ∧ n (172)
The CFIE equationis any convex combination of EFIE and MFIE
((1 − α)A + α(I
2+ Rt))j = (1 − α)
g
z− αHg ∧ n. (173)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
10 – Approximation for EFIE
Thevariational formulation for EFIE (called the Rumsey’s principle) leadsto a natural approximation. It is (here we chooseg = −Einc)
k
∫
Γ
∫
Γ
G(x − y)(j(x) · jt(y)
)dγ(x)dγ(y)
−1k
∫
Γ
∫
Γ
G(x − y)divΓj(x)divΓjt(y)dγ(x)dγ(y)
= − i
z
∫
Γ
(Einc · jt
)dγ; for any jt tangent toΓ.
(174)The operatordivΓ is the tangential divergencedefined on the surfaceΓ.
We solve this equation in the Hilbert space
H−1/2div (Γ) =
v ∈ TH−1/2(Γ); divΓv ∈ H−1/2(Γ)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
A finite element approximation, which is well adapted to this space,
consists of the well-known 2DRaviart-Thomas mixed finite element(which
leads to stable and convergent approximations).
Let (Th) be a family of triangulations ofΓ and let us denoteS1(Th) the
associated finite element space. Then, there is one unknown for each edge
of (Th) which is the electric flux crossing this edge.This approximation
correspond exactly to the so-called RWG basis for the EFIE. It is
k
∫
Γh
∫
Γh
G(x − y)(jh(x) · jt
h(y))dγ(x)dγ(y)
−1k
∫
Γh
∫
Γh
G(x − y)divΓhjh(x)divΓh
jth(y)dγ(x)dγ(y)
= − i
z
∫
Γh
(Einc · jt
h
)dγ; for any jt
h ∈ S1(Th).
(175)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
11 – Single layer for the Laplace Problem
We give and prove in this section some specific properties ofthe single
layer potential and the first kind integral equation associated with the
Laplace equation. The associated representation theorem was introduced in
theorem6 in the specific casek = 0. We introduce a variational
formulation. We study and prove a coercivity result for these variational
formulation. They will use in the Helmholtz (and also for Maxwell) case,
since the kernel associated with the Laplace operator is theprincipal part of
the kernel of the associated integral equations. We will give also a more
precise meaning of the notion of principal part.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Theorem 10 The integral equation associated with thesingle layerpotential
u(y) =1
4π
∫
Γ
1
|x − y|q(x)dγ(x) (176)
for the Dirichlet Laplace problem is
1
4π
∫
Γ
1
|x − y|q(x)dγ(x) = ud(y), y ∈ Γ. (177)
A variational formulation for this integral equation is
1
4π
∫
Γ
∫
Γ
q(x)qt(y)
|x − y| dγ(x)dγ(y)=
∫
Γ
ud(y)qt(y)dγ(y), ∀qt∈H−1/2(Γ). (178)
The corresponding operator, denoted byS, is an isomorphism ofH−1/2(Γ) onto H1/2(Γ) which satisfies the coercivity property
∫
Γ
(Sq)qdγ ≥ α ‖q‖2H−1/2(Γ) , α > 0, ∀ q ∈ H−1/2(Γ). (179)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Proof For q ∈ C0(Γ), the potentialu satisfies the system of equations
∆u = 0, inΩi and Ωe,
[u]Γ = 0,[∂u∂n
]
Γ= q.
(180)
Let W 1(Ωe) denote the Hilbert spaces:
W 1(Ωe) =
u;
u
(1 + r2)1/2∈ L2(Ωe),∇u ∈ L2(Ωe)
.
When the domainΩe is the whole spaceR3, we have
Theorem 11
||v||W 1(R3) ≤ c||∇v||L2(R3), ∀ v ∈ W 1(R3) (181)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
A variational formulation for the problem (180) is∫
R3
(∇u · ∇v)dx =
∫
Γ
qvdγ, ∀ v ∈ W 1(R
3). (182)
Theorem11 shows that the bilinear form in the left-hand side of(182), is
coercive on the spaceW 1(R
3). Thus,the Lax-Milgram theorem proves the
existence of a unique solution to this problem whenq is in the space
H−1/2(Γ), as this space is the dual of the trace spaceH1/2(Γ).
Notice that theintegral in(177) is a Lebesgue integralwhenq ∈ L∞(Γ),
but ceases to be so for a less regularq. It is not even a duality when
q ∈ H−1/2(Γ).
From the trace theorem, we know that the trace on the surfaceΓ of the
solution of the problem(180) belongs to the spaceH1/2(Γ). Thus, the
operatorS is continuous fromH−1/2(Γ) into H1/2(Γ). The inverse
operator associated withud ∈ H1/2(Γ), the value ofq.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Both interior and exterior Dirichlet problems admit a unique solution inH1 (Ωi) andW 1 (Ωe). The interior normal derivative is then defined as
∫
Γ
∂u
∂n
∣∣∣int
vdγ =
∫
Ωi
(∇u · ∇v) dx, ∀ v ∈ H1 (Ωi) , (183)
while the exterior normal derivative is defined as∫
Γ
∂u
∂n
∣∣∣ext
vdγ = −∫
Ωe
(∇u · ∇v) dx, ∀ v ∈ W 1 (Ωi) . (184)
The trace theorem proves that both of them are inH−1/2(Γ) and we have
q =∂u
∂n
∣∣∣int
−∂u
∂n
∣∣∣ext
(185)
and satisfies, using the Representation Theorem6 (if q is regular)
Sq = ud. (186)
Thus, the operatorS is a bijective mapping fromH−1/2(Γ) ontoH1/2(Γ).
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Moreover, subtracting equalities(183) and(184), we obtain∫
Γ
(Sq)qdγ =
∫
Γ
qudγ =
∫
R3
|∇u|2dx. (187)
The continuity of the trace implies that∫
R3
|∇u|2dγ ≥ β ‖u‖2H1/2(Γ) , β > 0, (188)
while the continuity ofS−1 is equivalent to
‖u‖2H1/2(Γ) ≥ γ ‖q‖2
H−1/2(Γ) , γ > 0. (189)
Combining(188) and(189) yields(179).
Remark 4 :Another meaning of the property(179) is to assert that∫
Γ
∫
Γ
q(x)q(y)
|x − y| dγ(x)dγ(y) (190)
is a scalar product on the spaceH−1/2(Γ).
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
12 – Homogeneous and pseudo-homogeneous kernels
Let Γ be a bounded regular surface contained inR3. Pseudo homogeneous
operatorsare integral operators of the form
g(y) =
∫
Γ
K(y, x − y)ϕ(x)dγ(x) ; y ∈ Γ. (191)
We establishcontinuity properties of these operators fromHm(Γ) tosimilar Sobolev spacesand any positive integerm. The associatedkernelsK are regular with respect to the variabley and quasi-homogeneous withrespect to the variablex − y.
We establish similar continuity properties forthe family of adjoint operatorswith respect to the scalar product inL2(Γ) which are of the form
g(y) =
∫
Γ
K(x, y − x)ϕ(x)dγ(x), y ∈ Γ. (192)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We examinehomogeneous kernelsK which are the restriction to thesurfaceΓ of kernels defined inR3.
Definition A homogeneous kernelK(y, z) defined inR3 is of class−m,
for an integerm such thatm ≥ 0, when
supy∈R3sup|z|=1
∣∣∣∣∂|α|
∂yα
∂|β|
∂zβK(y, z)
∣∣∣∣ ≤ Cα,β, ∀α and ∀ β. (193)
∂|β|
∂zβK(y, z) is homogeneous of degree − 2
with respect to the variable z, for |β| = m.
(194)
For any planeH whose equation is(h, z) = 0, and anym-uple of vectorsz1, . . . , zm in H,
∫
S1
Dmz K(y, z′) (z1, . . . , zm) dz′ = 0, (195)
whereS1 is the intersection of the sphereS2 with the hyperplaneH.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We give some examples of these operators that we will encounter.
Example 1: Let m be a positive integer. We consider the kernel
K(x) = |x|2m−3. (196)
This operator is of class1 − 2m. Condition(194) is satisfied, since the
partial derivatives of order2m − 1 are odd.
Example 2 : Let α be a multi-index andℓ an integer. A more general
example, which includes the previous one, is
K(x) = (x)α|x|ℓ, (x)α = xα1
1 xα2
2 xα3
3 , |α| = α1 + α2 + α3. (197)
It is of classm = −(|α| + ℓ + 2), for a positivem. Condition(195) is
satisfied whenm is odd and|α| evenor whenm is even and|α| odd, since
in both cases, the derivatives of order−m are odd.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
The following theorem gives the continuity properties for these examples:
Theorem 12 The operator associated with the kernelK by either(191) or
(192) is continuous fromHr(Γ) into Hm+r(Γ), for any real positiver,
when the kernelK is of class−m and the surfaceΓ regular enough.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
13 – Pseudo-homogeneous kernels
For an integerm such thatm ≥ 0, akernel K(y, z) ispseudo-homogeneous of class−m when it admits for any positive integers an expansion of the form
K(y, z) = Km(y, z) +
ℓ−1∑
j=1
Km+j(y, z) + Km+ℓ(y, z), (198)
whereKm+j is of class−(m + j) for j = 0, ℓ − 1, andl is chosen suchthatKm+ℓ is s times differentiable.
Theorem 13 LetK be a pseudo-homogeneous kernel chosen of class−m.
The associated operators given by(191) or by (192), are continuous from
Hr(Γ) into Hr+m(Γ) for any integerr.
This theorem is a direct consequence of Theorem12 using the expression(198) for s big enough.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We describe now some examples of this situation:
Example 3
Let nx be the unit normal to the surfaceΓ at the pointx. The kernel
K(x, z) =(nx · z)
|z|3 (199)
looks homogeneous of degree−2, but is rather homogeneous of degree−1.
The vectorz which appears in the integral operators(191) and(192) is
eitherx − y or y − z. Whenx andy are close, this vector is asymptotically
tangent. LetP be the projection operator on the surfaceΓ, which is defined
in a tubular neighbourhood of this surface. Then, for az of the formy − x
with y andx on the surfaceΓ, the kernelK takes the form
K∗(x, z) =(nx · (P(x + z) − P(x)))
|z|3 . (200)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
A Taylor expansion of the numerator is
P(x + z) = P(x) + DP(x) · z +1
2D2P(x)(z, z) + · · · . (201)
Using the moving frame on the surface, (x = P(x) + sn(P(x))), the
differential ofDP(x) is (I + sR)−1 which acts in the tangent plane to the
surface at the pointP(x), and thus,
(nx · (DP(x) · z)) = 0. (202)
This shows that the operator associated with the kernelK∗ is
pseudo-homogeneous of class−1, the first non-zero term of the Taylor
expansion being homogeneous of degree−1.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Example 4
The single layer potential kernel associated with the Helmholtz equationis
K(z) =eik|z||z| . (203)
A Taylor expansion of the exponential in a neighbourhood of zero is
eik|z| = 1 + ik|z| − k2|z|2 + · · · , (204)
from which it is clear thatthis kernel is pseudo-homogeneous of class−1.
Example 5
The double layer potential kernel associated with the Helmholtz equationis
K(x, z) =∂
∂nx
(eik|x − y||x − y|
). (205)
It is pseudo-homogeneous of class−1 (combining the two previous cases).
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
14 – The Helmholtz decomposition
We introduce a saddle-point formulation equivalent to the variational
formulation(168). It will be more suitable to prove existence and
uniqueness, using the following abstract theorem.
Theorem 14 (Fredholm alternative) LetV andW be two Hilbert spaces.
Leta(u, v) be a bilinear form continuous onV × V which satisfies
ℜ [a(u, u)] ≥ α ‖u‖2V − c ‖u‖2
H , α > 0, ∀u ∈ V, (206)
whereH is a Hilbert space containingV . Letb(q, v) be a bilinear form
continuous onW × V which satisfies:
sup‖u‖V =1
|b(q, u)| ≥ β ‖q‖W − c ‖q‖L , β > 0, ∀ q ∈ W, (207)
whereL is a Hilbert space containingW . Consider the following
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
variational problem, withg1 ∈ V ∗ andg2 ∈ W ∗:
a(u, v) + b(p, v) = (g1, v), ∀ v ∈ V,
b(q, u) = (g2, q), ∀ q ∈ W.(208)
Denote byV0 the kernel of the bilinear form b in V , i.e.,
V0 = u ∈ V, b(q, u) = 0 , ∀ q ∈ W .
Suppose thatthe injection from V0 into H is compactand thattheinjection from W into L is compact. Suppose that there exists an element
ug2∈ V such that
b(q, ug2) = (g2, q), ∀ q ∈ V.
Then, the variational problem(208) satisfiesthe Fredholm alternative,i.e., - either it admits a unique solution inV × W , - or it admits a finite
dimension kernel, and a solution defined up to any element in this kernel,
when the right-hand side(g1, g2) vanishes on any element in this kernel.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Proof Let us first prove that this kernel has finite dimension. From(206)
and(207), any element in this kernel satisfies
‖u‖2V ≤ 1
αℜ [a(u, u)] + c ‖u‖2
H ≤ c ‖u‖2H + c
1
β‖p‖L ‖u‖V ,
or else
‖u‖V + ‖p‖W ≤ c ‖u‖H + c ‖p‖L . (209)
Moreover,u ∈ V0. From the compact injection hypothesis, it follows thatthis linear space has a compact unit ball and thus has finite dimension. Fromthe property ofg2, it is equivalent to consider the case whereg2 is zero.We then introduce, using the Galerkin technique, an approximated solutionfor our problem, denoted(uε, pε), in the quotient space(V × W )/N . Wecontinue to denote this new space byV × W . Notice thatuε belongs toV0.Estimates of this solution are
‖uε‖V + ‖pε‖W ≤ c ‖uε‖H + c ‖pε‖L + c ‖g1‖V ∗ + c ‖g2‖W∗ . (210)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We proceed by contradiction to show the convergence of this sequence:
– either the sequence‖uε‖H + ‖pε‖L is bounded, and then from(210), wededuce that(uε, pε) is bounded inV × W . We then have a weakconvergence inV × W .
– or the sequence‖uε‖H + ‖pε‖L is not bounded, andthenuε = uε/(‖uε‖H + ‖pε‖L) is bounded inV , whilepε = pε/(‖uε‖H + ‖pε‖L) is bounded inW , and their weak limits(u, p)
are such that a(u, v) + b(p, v) = 0, ∀ v ∈ V,
b(q, u) = 0, ∀ q ∈ W.(211)
Thus(u, p) are in the kernel of this problem.
As this kernel is reduced to zero in the quotient space, the quantity‖uε‖H + ‖pε‖L, whose value is1, cannot tend to zero. But, from thecompact injections ofV0 into H and ofW into L, there exists asubsequence which converges strongly. This is contradictory.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
TheDirichlet problem for the exterior Helmholtz problem(85) can be
solved using thesingle layer potential(106). This is the following equation
ud(y) =
∫
Γ
G(x − y)q(x)dγ(x); (212)
Using the previous theorem, we obtain thatthis equation has a unique
solution in the spaceH−1/2(Γ) if ud is in the spaceH1/2(Γ), except when
k2 is an eigenvalue of the interior Dirichlet problem for the Laplace
equation.
The proof is an application of the above results:
– First,the operator(212) is pseudo-homogeneous of order−1. Thus, it is
continuous fromL2(Γ) into H1(Γ). Its dual operator is also continuous
from L2(Γ) into H1(Γ) and thus by duality it is continuousH−1(Γ) into
L2(Γ). Then by interpolation,this operator is continuousH−1/2(Γ) into
H1/2(Γ).
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
– Secondly we can expand the kernel using:
eik|z| = 1 + ik|z| − k2|z|2 + · · · , (213)
which gives that
G(x) = 1/|x| + ik − k2|x| + · · · , (214)
The first term is the kernel of the single layer potential for the Laplaceequation. The second one is a constant kernel which result does not dependson the variabley. The rest of the expansion is a pseudo-homogeneouskernel of order−3 and mapsH−1/2(Γ) into H5/2(Γ).
– The integral kernel(212) is the sum of a kernel coercive in the spaceH−1/2(Γ) and a kernel which mapsH−1/2(Γ) into H5/2(Γ). But theinjection ofH5/2(Γ) into H−1/2(Γ) is compact and thus we can used thetheorem(14). The inverse operator is not defined only whenk2 is aneigenvalue of the interior Dirichlet problem for the Laplace equation.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We now return tothe EFIE equation. We look for the vectorj in the form ofits Helmholtz decomposition. If we suppose that the surfaceΓ is simplyconnected, it has the expression
j =−−→curlΓp + ∇Γq. (215)
It follows from the variational equation(168) that
∫
Γ
∫
Γ
G(x − y)∆Γq(x)∆Γqt(y)dγ(x)dγ(y)
− k2
∫
Γ
∫
Γ
G(x − y)(∇Γq(x) · ∇Γqt(y)
)dγ(x)dγ(y)
− k2
∫
Γ
∫
Γ
G(x − y)(−−→curlΓp(x) · ∇Γqt(y)
)dγ(x)dγ(y)
= iωε
∫
Γ
divΓEincT qtdγ,
for any qt inH3/2(Γ).
(216)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
−k2
∫
Γ
∫
Γ
G(x − y)(−−→curlΓp(x) · −−→curlΓpt(y)
)dγ(x)dγ(y)
− k2
∫
Γ
∫
Γ
G(x − y)(∇Γq(x) · −−→curlΓpt(y)
)dγ(x)dγ(y)
= −iωε
∫
Γ
curlΓEincT ptdγ,
for any pt inH1/2(Γ).
(217)
These two equations are coupled. The principal part of the first equation is∫
Γ
∫
Γ
G(x − y)∆Γq(x)∆Γqt(y)dγ(x)dγ(y),
which, up to a compact operator, is coercive inH3/2(Γ), due to the
properties of the single layer potential for the Laplace operator.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
The principal part of the second equation is∫
Γ
∫
Γ
G(x − y)(−−→curlΓp(x) · −−→curlΓpt(y)
)dγ(x)dγ(y),
which, up to a compact operator, is coercive inH1/2(Γ), due to the
properties of the single layer potential for the Laplace operator.
We prove now that the coupling terms are also compact operators. We
change the integrals using integration by parts on the surfaceΓ
∫
Γ
∫
Γ
G(x − y)(∇Γq(x) · −−→curlΓp(y)
)dγ(x)dγ(y)
=
∫
Γ
p(y)curlΓ
∫
Γ
G(x − y)∇Γq(x)dγ(x)dγ(y).
(218)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
From identity(145), we obtain
∫
Γ
∫
Γ
G(x − y)(∇Γq(x) · −−→curlΓp(y)
)dγ(x)dγ(y)
=
∫
Γ
∫
Γ
(((ny−nx) ∧∇yG(x−y)) · ∇Γq(x)) p(y)dγ(x)dγ(y),
(219)which, using the results on pseudo-homogeneous kernels, shows that thisbilinear form is bounded by
∣∣∣∣∫
Γ
∫
Γ
G(x − y)(∇Γq(x) · −−→curlΓp(y)
)dγ(x)dγ(y)
∣∣∣∣
≤ c ‖q‖H1/2(Γ) ‖p‖H−1/2(Γ) .
(220)
It follows that the system of equations(216) (217) is of Fredholm type. Itis also equivalent to equation(168), and thus it satisfies the existence anduniqueness properties, whenk2 is not an eigenvalue of the interior problem(161) for the Maxwell equation.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We have prove that the system of equation(216) (217) has a unique
solution except whenk2 is an eigenvalue of the interior problem(161) for
the Maxwell equation. If we plug this solution in the EFIE equation(168),
the solutionj is such that
divΓj = ∆Γq
curlΓj = −∆Γp.(221)
As we know thatp is in H1/2(Γ) andq is in H3/2(Γ), it results that the
vectorj belongs to the space
H−1/2div (Γ) =
v ∈ TH−1/2(Γ); divΓv ∈ H−1/2(Γ)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
15 – The dielectric case
Let Ωd be the dielectric, andΩe the exterior complement, with boundaryΓ.
We introduce the following real coefficients (ω is the angular frequency)
ǫd : the electric permittivity ofΩd,
ǫe : the electric permittivityΩe
µd : the magnetic permeabilityΩd
µe : the magnetic permeability ofΩe
kd = ω(µdǫd)12 : the wave number inΩd
ke = ω(µeǫe)12 : the wave number inΩe
zd = (µd
ǫd)
12 : the impedance inΩd
ze = (µe
ǫe)
12 : the impedance inΩe.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
The unknows are the electric fieldsEd, Ee and the magnetic fieldsHd, He.
The harmonic Maxwell system is
(Ped)
curlEd − iωµdHd = 0 in Ωd,
curlHd + iωǫdEd = 0 in Ωd,
curlEe − iωµeHe = 0 in Ωe,
curlHe + iωǫeEe = 0 in Ωe,
Ed × n− (Ee + Einc) × n = 0 on Γ,
Hd × n− (He + Hinc) × n = 0 on Γ,
|√ǫeEe −√
µeHe × n| ≤ C|x|2 when |x| → ∞
whereEinc, Hinc is a known plane wave.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
We introduce the electricj and the magneticm currents:
jd = Hd × n, je = −He × n
md = Ed × n, me = −Ee × n.(222)
Now, two Green’s functions appear denote byGd andGe.
Separately, the fields inΩd and inΩe admits an integral representation
given by (118) and (119) with their respective currents. Theunknowns are
the currentsje andme. The jump conditions on the boundaryΓ in the
equationPed give the currentsjd andmd.
The Rumsey variational formulation consists in equating the two integral
expressions of the currents on the surfaceΓ.
We denoteby Ad = kd Sd + 1/kd Td, Ae = ke Se + 1/ke Te andRd, Re,
the respectiveEFIE and MFIE operatorsassociated to each domain.
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
Then, in matrix notation, the system to solve takes the form
A =
izdAkd+ izeAke Rkd
+ Rke
Rkd+ Rke −i(
1
zdAkd
+1
zeAke)
(223)
CMAP, École PolytechniqueHongKong 2007
Integral for Maxwell
References
[1] J.C. Nédélec“Acoustics and Electromagnetics Equations, Integral
Representations for Harmonics Problems”; Springer-Verlag, 2001.
CMAP, École PolytechniqueHongKong 2007