NATIONAL TECHNICAL UNIVERSITY OF ATHENS
DEPARTMENT OF NAVAL ARCHITECTURE AND MARINE .ENGINEERING
On the solvability of a two-dimensional water-wave radiation problem by G. A. Athanassoulis (REVISED VERSION)
October 1983 Report Number 11-1983
Address: Department of Naval Architecture and Marine Engineering Naval Architecture Laboratory
TABLE OF CONTENTS
Page
ABSTRACT 1
Introduction 1
Notation and terminology 4
Formulation of the radiation problem
and some preliminary results 8
Background theory 15
The weak solvability of the radiation problem 19
Regularity of solution 28
APPENDIX 30
ACKNOWLEDGMENTS 31
ABSTRACT
The existence of a unique weak solution for the two-dimensional water-wave radiation problem arising when a floating rigid body oscillates on
the free surface is established for all but a discrete set of oscillation frequencies. The body-boundary is taken to be o 'class (see §2) and the body-boundary condition is satisfied in L2sense. The proof relies on an expansion theorem (Athanassoulis [1] ) and on the property of the associated
water-wave multipoles to be a Riesz basis of [(-n,o),a fact which is
established in the present paper.
Under stronger geometrical restrictions on the body-boundary it is
proved, using a method due to Ursell [10] , that the weak solution is
actual-ly a classical one, that is the velocity field is continuous up to and including the body-boundary.
1. Introduction
This paper is concerned with the study of the boundary-value problem arising when a two-dimensional body floating on the free surface of an
unbounded, infinitely deep, incompressible and inviscid fluid,performs time-harmonic oscillations of small amplitude about a fixed mean position. This problem is to be studied under the additional assumption that energy is radiated in both directions towards infinity,in which case it is usually called a radiation problem.
The solvability of the water-wave radiation problem for either two or three-dimensional floating obstacles of general shape has been studied by John [2] , Beale [31 , Lenoir and Martin
[4] , [5] , and Lenoir E6]
John [2] reduced the problem to a Fredholm singular integral equation of the second kind and treated the three-dimensional case assuming uniform,
finite depth; however, his results are also valid ir the two-dimensional case for either finite or infinite depth. He established the existence and uniqueness of a classical solution when the wetted surface of the floating
2 body satisfies certain geometrical conditions, namely it is of class
it intersects the free surface perpendicularly and is intersected by every vertical line in at most one point. Beale [3] proved the unique weak
solvability of the same problem for all but a discrete set of oscillation frequencies removing John's geometric restrictions. Lenoir and Martin [4]
[5) treated the case of three-dimensional bodies floating in a fluid of infinite depth. Using the limiting absorption principle they proved the existence of a generallized solution for all oscillation frequencies for which uniqueness holds. They also provided a general uniqueness theorem which, however, as pointed out by Martin and Ursell, was not correct (see
the discussions of a paper by Euvrard et al [27] ). Lenoir [6] treated, by the same method, the two-dimensional case for either finite or infi-nite depth.
For the two-dimensional case with infinite depth Athanassoulis [i) showed, using a method due to Ursell [7] , [8] and the conformal mapping technique,
that,if a solution of the radiation problem exists, it can be expanded in
an infinite series of special functions, called water-wave multipoles.
In the present paper this multipole expansion is to be used to establish the weak solvability of the radiation problem for all but (possibly) a dis-crete set of oscillation frequencies (Theorem 5.3) provided that the body-boundary is sufficiently smooth, i.e. it belongs to the class t (see § 2), and, as a consequence, it intersects the free-surface perpendicularly. It is worth mentioning that in our approach John's convexity condition is no more necessary. Subsequently, under stronger geometrical restrictions on the
-3-body-boundary,it is proved that the unique weak solution is,..in fact, a clas-sical one, i.e. the velocity field is continuous up t and including the body-boundary. Morecvr, it is established that the multipole expansion coef-ficients of the velocity field are of order
O(iI'i?)
(Theo'rerrvi 6.1.).
The idea of using relevant multipole expansions to prove the solvability of water-wave radiation problems goes back to Ursell (see Ursell [9] , [10]
and Yu and Ursell [ii] ) who has treated the case of a semi-circular floating
body. Ursell reduced the question of the solvability of the radiation problem to that of an infinite linear system and studied it using Fredhoim theory for compact operatm-s. In our treatment the proof of the weak solvability theorem,presented in §5, makes use of the basis property of the water-wave multipoles, which is also established in the same paragraph,whilst the proof of the regularity theorem,presented in § 6, is along lines due to Ursell
jio].
In establishing the basis property of the water-wave multipoles we
need some elements of the spectral theory of compact operators analytically dependent on the spectral parameter. The pertinent background material along with the needed elements of the theory of bases in Hilbert spaces are summa-rized in § 4.
2. Notation and terminology
A Cartesian coordinate system Oxz.3 is introduced with xaxis on the mean free surface, x3-axis directed vertically upwards, i.e. in direction opposite to that of the acceleration of gravity, and the center
0
inside the floating body. A point in the (x2X3)-plane is denoted by=xr3) or
in complex notation.The mean fluid domain is defined by D where
D is a compactum in the lower half-plane, intersecting the
x-axis
and representing the floating body. The mean free surface of the fluid,i.e. the part of D-axis lying outside D, is denoted by'D, and
the mean wetted surface of the floating body, i.e. the common boundary of D and D, is denoted by 1DS. The symbols
D, D
denote the pointsets which are symmetric tot, D
with respect to D.-axis; see figure 1. We also defineD= DU
D*IU
DUF
KF-4
('D
K5
K
D
T
-5
The class of quadratically Lebesque-integrable, complex-valued functions, defined almost everywhere in (c,6) is denoted by
L(cI).
The scal:r product and the norm in are denoted by () and($,) ,respectively.
2
The space
L(-nc)
is decomposed into two orthogonal subspaces: the subspace ,spanned byjco9
and the subspacein which forms a complete orthogonal sequence.
Some smoothness requirements are needed for the body-boundary D8.
We shall say that8 belongs to the class
.ifbis a simple,
closed curve, described parametrically by equations
x2=x2W), x3=X3(Q))
E-nn]
and the functions
X(0),X3(e)
have the following properties:They are continuously differentiable and their first derivate sat-isfy a Holder condition with exponent o(o,L)
Either
c>LI2
or the first derivaties of X2(9) and G) are of bounded variation.((9))(d.))2o
for everyThe above conditions are needed to ensure the validity of the Lemma 3.1 below.
More restrictive assumptions on the body-boundary will be introduced in §6, where the regularity of the radiation problem will be studied.
-6
A4=A24
4acncL
v.44
=
= O+jO
V I\f2 =
Lç.= i4O
=(2?)+çw),
ecCj
=
in the definition of the zero and unit elements
(2.2)
(2.3)
(2..4)(.5)
(2.6')
C2.(.8)
I 4Two noninteracting imaginary units 3 and t are used, making neces-sary the distinction between the
j
-complex numbers c3c(i-j ()e 1k))
and the -complex numbers(c1) .Products of
j-
and %-complexnumbers also occur, leading to objects of the form
O(4Je+S
(ijji'
which are called ij-complex numbers. The sets of and j-complex numbers are denoted by
,
and C1. respectively. The basic no-tions and operano-tions in are defined below.
Let be U_complex numbers. They can be represented
in the form A= +jw t
s4jW,5
weTe Z)W,
,v/$ct.
Then, we define
Equality
Addition Zero element Multiplication -Unit element Scalar multiplication Absolute valueThe subscript
and the absolute value will be omitted in the sequel whenever no confusion is likely to occur.
The system
(12+)
'
O 4)
is a comrriutative ring, which) however, is not a field (there exist non-zero, nonivertible elements).Furthermore,
.
is a norm, and the set t.1;
,equipped with the
ope-rations (2.3), (2.5), (2.) and the norm (2.9), becomes a comutative Banach algebra. For a rather complete study of the algebraic and topo-logical structure ofjj see Athanassoulis [1] , appendix i.
-7-Finally, if A=+ +i.'+1,jS ELj
we define
RetA
+3
LA
=
-8
3. Formulation of the radiation problem and some preliminary results
Under the assumptions made in 1, the fluid motion is described by a velocity potential
where
cr),
thea-complex amplitude ofc (x-)
, satisfies the Laplace equation+c)O
x...eTt,
(3d')
and the boundary conditions
KD)_c1$()=O) _/
t)
_LC),
C13
(3.3)
>0 S=3
(3.4)
and4(x)
jKO(x)- QxcD)
x1I'oô.
(3.5)
Here and below, w is the frequency of oscillation, is the accelera-tion of gravity and/'r
denotes differentiation along the normalof directed outwards with respect
to
the fluid. Introducing the j-complex potentialwhere
W(xis
the i-complex amplitude of the stream function.4) (?Fc),
equations (3.1)-(3.5) are tranformed toF (w') b ;,_
a-ki
C
VOC3
I
cLF(w
.
(w)
c)
I
ctF(w)
cund
f(w)
LW
t12"-9
->
CD,-
CO1
(3.io
respectively.Re-Eo-G (3.5) and (3.10) are two fonns of the radiation condition. For a discussion of this condition see, e.g., Stoker [12] , §4.3, 6.7, Athanassoulis[1] , 2,4.
A conformal mapping W=f()is now introduced, transforming the ex-terior of the unit circle [: jSt>i in the s-plane onto the domain
in the W-plane. Such a function always exists, provided that
has at least two boundary points (Riemann's mapping theorem) and may be chosen so that the pointsets
K
SI1O)
Ski-KF-LP1,SO)
Sti
andKIm<O) ii>i1
in the 3-plane to be tranformed to the pointsets
andD
in the w-plane)respectively (see figure 1).
Finally, we defineK=ImO
KJ\K
If D belongs to the class , then:
The function (3) can be extended in the region S:
es-tablishing a one-to-one and bicontinuous correspondance between
the regions [S 3I'iand EcUM (Osgood-Caratheodory's theorem). The function is continously differentiable and its derivative
10
-3. The function is expanded in a Laurent series of the form
2-t
c4>o,
t_-L,
where all CS are real numbers since the domain is symmetric with respect to 2-axis.
4. The equation
Using the change of variable we may transform the equations (3.6)-(3.10) into the equations
F') b. i-ccdy1jc
,Ij
+
LciS
K0
cl$
F0)
36 KF)
Re
J
cLF(S)/d3
1)-
0 E1j ± ) whereF,(3)
circLs
4-cr
F(--(S))
It should be noted that the complex potential
(s)
depends also on the3)
Re±o01
(3J3')
(3A4')
w(-(e')
(3.L2)
realizes a parametric representation of the boundary
I5
such thatthe functions and
satisfy the conditions (i), (ii), (iii) introduced in the definition of
"C
parameter K0=W/. To emphasize this fact we shall often write instead of (3).
The three sets of equations (3.1)-(3.5), (3.6)-(3.1O) and (3.13)-(3.17) are mutually equivalerrL under the assumption In the present paper we shall work with the third formulation and, for the sake of brevity,
the equations (3.13)-(3.17) will be collectively referred to as the radiation xroblem P(<0)..
We shall now state an expansion theorem for the ware potential
1SK0),
which is of fundamental importance for the subsequent treatments. This theorem provides an extension of the multipole expansion introduced by Ursell [26]for body-boundaries symmetric with respect to the vertical axis.
Theorem 3.1 ( The expansion theorem; Athanassoulis [1] , § 4). A function
çi<
satisfies the conditions (3.13), (3.14), (3.16) and (3.17)if
and only if it may be represented in the formF (.K')
=
t\6 (&);
(-?-(S; K0)-t-II_ d'm Mw(
K:))
(3.:L)
')
whereAO)A.)cL.
cE1
-1
(3.i)
tW
6'0(W i<) =4(i-j)
-j-(3.2d.')
fl
(. ao
-
<cW
Gj
(w3
K)=- (iu)
K0 e
+
P (w Ko))
2fl
w
_iK0W(
_.s 1(OU.(W c) e
tte.
cLud)
S
C3. 21
0(WK0)
*i<J(w;c)
cjWo
cLWit follows that and
(wC)/Lv'l
12
-and the path of integration in the last integrals is taken to be in the lower half plane.
Remark 3.1.1. The functions G0(w;)and
E(w;Ko)
represent free surface flows which are, respectively, symmetric and antisymmetric with respect to the axisOx3
Remark 3.1.2. The function
(w
K0 can be written in the following alternative formF(W;
K0)e
cow')
-of KGQ..
and they remain bounded asK0> o
for eachIntroducing now (3.18) into (3.15) and differentiating term-by-term we obtain
IJ \c
2-e-i
where is Euler's constant. With the aid of the above expression we see
that F(W,K)is,for each WQJ an analytic function of K0 in
where Q. is any open,bounded, simply connected region of .containing a part of the positive real axis but not containing the origin. Moreover, the
function K0F(WKo) tends to zero as O
, for each fixed v14D. Since;:
s=iIa)
are also analytic functions
= v(e) eE-[-n,o]
,
'4'e're
s_c.)
13
-The termwise differentiation in (3.22) is justified, in the sense of
pontwise convergence, if the series EJ2-l.Q and
1'ct
converge. The convergence of the first series is a consequence of the body-boundary smoothness assumption (see Lemma 3.1 below). The convergence of the second series will be considered in § 6, where will be established thatd= O(iIw3),
provided that is represented by a finite Laurent series and that is a simple,closed curve with continuous tangent.However,if the left hand side of (3.22) is interpreted in a limiting sense, that
is as the limit when l3)-i+, the weaker condition is sufficient to make (3.22) valid in L2-sense. This point of view will be adopted in § 5, where the weak solvability of the problem will be proved, without restricting to be a finite Laurent series.
Setting now
1\D
cycL
mlt
)2)3J..)
(3.
we can rewrite the equation (3.22) in the form
DC)
GEn)O.
(3.2C)
The central question is now the following : Under what conditions and
in what sense may the function
V(f3)
expanded with respect to the sequence0
It would be noted that the forcing term
V()
may be also depended upon the parameter K0,as, for example, in the diffraction problem. But this fact does not introduce any further complication.14
-We now state a lema on the conformal mapping coefficients of a
boundary, which is often used in this work.
4,C(
Lemma 3.1 If
DG
and{c
are the Laurent expansion coefficientsof
the conformal mapping fw'ction (see equation (3.11)), thenCQ
1 oo.
i
The proof of the above lemma is based on two classical theorems on the absolute convergence of Fourier series (Zygmund [13] , Theorems 3.1 and 3.6,
pp.
240-241) and the Denjoy-Lusin Theorem on the absolute convergence of the series of Fourier coefficients (ibid.p.232).Finally, we collect, in the form of a theorem, some properties of the functions ')(c)S=Oji tV (Lko'))tn 23J...) and
which will be needed in § 5.
Theorem 3.2 Suppose that Then
i) 65((s)) <o))
ci, are continuously differentiable with respect to G."< for any fixed Theyare also analytic functions of K0Q,for each SEKii.. are uniformly bounded with respect to
'w
and continuously differentiable with respect toS{s:
31>.Lj,the infinite series containedbeing terimise differentiable
iii) (G 0')1So,i, are continuous with respect to
OE-fl16]
for any fixed K0EQ. They are also analytic functions of K0E, for eacheE-noT1.
uniformly bounded with respect to ' and continuous with respect to E'
E-n111].
15
-4. Background theory
In this section we present some definitions and theorems underlying the proof of the solvability theorem, given in the next section. The materials presented concern the spectral theory of compact operators analytically depen-dent on the spectral parameter and the theory of bases; they are given for the Hubert space case, although can be properly generalized for Banach or even,
in some cases, for linear topological spaces (See I-Iarazov 122] , Singer [14] ).
Let E be a separable Hubert space and be The set of all bounded linear operators in
E.
By (.,.) and (. II are denoted the inner product and the induced norm inE
, respectively.Definition 4.1.
Let T()(resp.
) be an operator-valued (resp.yector-valued) function defined on a simply connected region G.Cc1,with values in
B(E)
(resp.E ).
We shall say thatT(x)
(resp. ) is analytic inO.
if the scalar-valued function (T(yX)\J) (resp. ((.x'X))is analytic in
Q
for anyXJEE.
This apparently uweakllnotion of analyticity is actually equivalent to the strong" one, based on the existence of a Fréchet derirative with respect to
X.
(See,e.g., Taylor[24],
p. 205, or Rule and Phillips [2.5], p.93).Theorem 4. 1. (Gohberg and Krein [23], p. 21). Let T(.t) be an operator-valued function, analytic in an open, simply connected region
QCC,
and sup-pose that all valuesof T()
are conipact operators. Then,for all pointswhere is a set
of
isolated pointsof Q,
the number t(2)of
linearly independent solutions
of
the equationT())xO,
eE
16
-In particular, if t()Q for at least one point in Q.. , then for all
the operator I-T() has a bounded inverse.
Definition 4.2. Let be two sequences with elements in
E
.
Then, the two sequences X y,,will be said to be quadratically near each other ifo3 (4.L')
the two sequences çxj3l.j,,,3 will be said to be equivalent if there exists an automorphism
T
, i.e. an isomorphism ofE
onto itself, such thatTheorem 4.2. (Higgins [25] , p. 75). Suppose that and are two equivalent sequences in
E
. Then, I is a (Schcruder, Bessel, Hilbert, Riess) basis ofE if
and only if is a (Schcruder, Bessel, Hubert, Riesz) basis ofE.
We shall now state and prove a theorem providing a criterion for the equi-valence of a sequence in a Hubert space
E
to a complete orthonormal system ofE
. This theorem can be considered as an extension of some relative theoremsof Bary (see Bary [16] or Kato {17] , pp. 246-266).
Theorem 4.3. Suppose that is a complete orthonormal system of a
Hilbert space E , and
()3J
eQCIJ is a family of sequences inE
definedby
=
(4)
where
q())y=i,21..J
are analytic rector-valued functions of JEQ,with valuesin
E.
Suppose also that(z1.3)
-
17-where
B
is a positive constant, independent of1)
and thatE.l).kLfor
at least one pointof a.
Then the sequences
{x
fjj))
are equivalent for allQ,where
is a set of isolated points ofG..
Moreover,fy)J
is a Riesz basis ofE
forRemark 4.3.1. The set . may be contable, finite, or even empty.
Remark 4.3.2. Condition (4.3) is equivalent to the quadratic nearness of the sequences
[X.n3
andTo prove Theorem 4.3 the following Lemma is needed.
Lemma 4.1. Suppose that {x
is
a coirrplete orthonormal systemof
E
andis
a sequence inE
such that IfoOThen
the operator T E
E
defined on {xJ by TDc,, can be linearly extended on the whole
E
as a compact operator.Proof. To start with, let us extend the operator
T
on the whole E.If only a finite number of in the expansionc(X
of an elementzeE
are different from zero, then we defineT=Io
,Tx,=.c*',.The
same definition remains valid when the seiesoçX
has infinite terms, provided that the series'o( converges inE
. This is indeed the case since (N>rJ)IIc4,
II1c1)). >IIfl
'T\T1
and the series LQ(T) and converge. The compactness of the operator
T
follows from the condition<
See, e.g. Smirnov [18] , § 138.Proof of the Theorem 4.3. We introduce the operator
T(>)
defined on byT)x=q()
and extend it on the wholeE
by means of the definition18
-According to Lemma 4.1,T(A) is a compact operator for each G) . Furthermore,
it is analytic with respect to
Consider now the operator
A()I+LT()
. Obviously,thus the equivalence of the sequences {x and is equivalent to the invertibility of operator
AG).
Since,LIT()IJ = i:;1
II TO)
IItIB,
it follows that there exists a point
X=X,
in which (j.&0T(X)Hi; this ensures the invertibility of A(>).Accordingly, applying Theorem 4.1, we conclude thatA)
is invertible for all 2kQ\'.Finally, invokingTheorem 4.2, we see that is a Riesz basis of
E
for allXQ\.
This completes the proof of the theorem.04
2.
2
Remark 4.3.3. Since
A(o(.J (),where
1l4.)xand
jo(.)=JJxi)od,
e=f
the invertibility of the operator
A(X)
is equivalent to the condition2
4ct=O
or
e
yefl\(
(A)
i.e. the e-iinear independence of the sequence $)),1.In the application to the radiation problem condition (4.4) is, in fact, a uniqueness theorem.
Then, )
2.
ii) Il2e)c
i(+i-a)Oi
=-
cx3sa)
whereby it follows that
19
-5. The weak solvability of the radiation problem
The main tool in establishing
the solvability of the radiation problem '(K0)is the basis property of the sequence
(K) }
in the spaceL(_n,o).
Let us note that the functions(;K)
depend analytically upon the Parameter k0Cj,which effects as the parameter in the general theory of the preceding section.
1,
Lenvna 5.1. Suppose that
and consider the multipoles
The proof of the above results is staightforward,
so it will be omitted.
According to (5.3) the sequence
is quadratically near to the system which is an orthogonal
basis of the Hubert space
L28. However, we cannot deduce any completeness or basis property
of from this fact, since, in general
(&K0')
To proceed, we shall modify the multipoles
'(& ç)
in such a manner that the modified multipoles, say (4)(. <) :
j) belong .o2
ii) satisfy a condition similar to (5.3).
20
-t)(O; K0')
3(O '(s)
'ifl=,3and try to determine the constants by means of relations
(KA(K)f\
(K0)=
0)
00 01
(KoA(Ko)=O)
where
=
(CO5(5))
(O;
K0) ) . Relations (5.5) areequivalent to the condition (G; K0)
L
The system (5.5) is uniquely solvable for each
rn'),
provided that' (
°<°
(), (s.G)
(In any case, since
i(K0)
is an analytic function of 1it may have onlyisolated zeros in
G..
Consequently, we can modify, if necessary, the region Q,,exc1udinq from it the zeros of (K0')) . Then, the constants can beexpressed in the form
o4 <Q 0)4
A=A (Ko)=
,(Ko)) M2J)
where
(-©
(rf)
() 1L\(<))
Om 44 11-Joj
C<o) =(-O
O'Yflo)+4(%c)')
/ LØK0),
34O
e.4(j) __tt
(e2
(2-2)
2
and-S .(5.4)
(5.5'
(*)
The geometrical meaning of this condition is that the projections ofthe functions (Q;Ko) and (e;o) in the subspace
i
The infinite series in (5.9) converges, provided that . From
the equations (5.8) and (5.9) it is easily seen that
I 0)4 I o4
'ni=2,3...,
We can now state and prove the following
Lenna 5.2. Suppose
DBE4, Mo)*O)
and consider the modified multipo leewhere 21
-40q (;Ko),
cJn
K0)j2Q)ceSiY
-
J-.
(5.{o)
(5j{%)
;;;ç 1(K0)G;K0).
(5.ia
Then,1) 9(;Ko)cC([-n,o) arnL
K0)1Zoo
,et-n1o1&,
(5J)
2.ii) 41(;K0)
£LB
I I(;
K0)B
(5.i4)
where a positive constant independent of K0
Proof. The proofs of i) and ii) are straightforward, hence will be omitted. To prove iii) we observe that
22
-U
c0) fl £.'- II
(2-QSY) (4'-2)O II
-O) I
I k°')11'
eUsing now (5.2), (5.10) and Theorem 3.2-(iii) we obtain
II(;K)ft
whereby it follows the desired result (5.14), and the proof of the lemma is completed.
Let us recall that the set QCcCis a bounded, open, simply connected
region containing a part of the positive real axis but not containing the origin. Since,however, inequality (5.14) remains valid
as K0*O (see Remark 3.1.2),
we can take as a region containing a pointK,
such thattKc,=lZ
in which case . tKO
<
j
. Definingccj
in this way and using Lemma5.2 and Theorem 4.3 we arrive at the following
Theorem 5.2. Supppose that
*j (0)KOEQC.Then, for
allk'L,
where is a setof
isolated pointsof a)
the sequenceis
a (nonorthogonal) Riesz basisof
the Hilbert spaceB'
equi-valent to the orthogonal basisRemark 5.1.1.
KE
if and only if there exists a sequence of scalars{c(ynJ such that
(5.i5)
(See Remark 4.3.3). Clearly (5.15) is a nonuniqueness condition for the ra-diation problem
V(K0)
In the remaining part of this paper we shall simplify the notation by negle-ting K0 from the arguments of various functions. Accordingly, we shall
oW
We cave the basis property of the sequence
Theorem 5.2. Suppose that
D6
to
,cod-
K0
Then, the sequence { is a (nonorthogonal) Riesz basis
of
theHubert space
0
Proof. Let
V()be
an arbitrary element oft.(-n16).
In the first place we seek two scalarsCLO) oh.,
such thatCondition (5.16) is equivalent to the system
cL
+
+ L4
-(V5= (ccSsG),\'(G))), which is always solvable defining uniquely c[c, and
cLL)
since/x z/: 0
Since 1J(e)
L B
and the sequence2
is
d.Riesz basis of
LB
(Theorem 5.1), there exists a unique sequence of scalars such thatU)
=1()
i-B,
&cL
-
Q3.
(5.i
Using (5.16) and (5.18) we obtain the following expansion of terms of where 04
V(Y_
co)
(-n,°);>
lb
c1
-
2324
-The series in the right hand side of (5.t)is absolutely convergent.
For, according to (5.10), (5.18) and Schwarz inequality, it follows
(M>)
\Thus,the proof of the theorem has been completed.
We shall now state and prove the main result of this work. Theorem
5.3.
(The weak solvability theorem). Suppose thato, U(e)
cLKOEQ\. (In the physical
problem
K0(o
oo) ) Then, there exists a wiique weak solutionof
the radiation problem'(K0)
;
that is, there exists a unique conrplexpotential J() such that:
it is continuous throughout
K={S: Im3c S1
L
,
it satisfies conditions (3.23), (3.14), (3.16),(3.27)
andif
is
the limiti
F(re.)
'c>
L+(4)
exists for almost all
&E-n,oJ
and defines a functionF(o) 1(-n,o),
satisfying the boundary condition(3.25)
inProof. belongs to
2
L-sense.
V()LD(
PiO
It should be noted that the values of the functions tX(-) and
oO.
(5.2o)
Remark 5.3.1. In fact, the limiting behaviour described in iii) remains valid as approches along any path lying in
K
which is not tangent to the unit circle.Under the stated hypotheses the function
a
L (-n,o)
and the sequence is a Riesz basis of La(fl,O) . Therefore, there exists a unique sequence of scJoxs {'D°such that25
-as well -as the values of the scalars
D
are 3-complex numbers. Accordingly,H
(5.222)
Consider now the function
(3)
=
4c))+ D1(s))
(5.2
i)1(3)is continuous on
K
. To establish this fact it suffices to prove the continuity of the functionM(5),
3K.
C5.a')
Since the sequence
MS)
is uniformly bounded onK,
the series (5.24) is dominated by the series which is absolutely convergent (see equation (5.20)). Accordingly, the series (5.24) converges uniformly onK
defining a continuous function there.1()
satisfies the conditions (3.13), (3.14), (3.16) and (3.17) because of the Expansion Theorem 3.1.(4)
It remains to prove that the function
f.
(S)
zd()/ctS
converges for almost allGEfl,cJ
as S)-'i-I-, defining an-function which satis4ies the boundary condition (3.15). Difecentiat-ing (5.23) we obtain (1)
-'i-)
(.25)
where (a)cL)
{1)
D
kM(S)
(5.26)
H
(3)
can be also written in the form(.2)
where
First, we study the function
4 (e )
96\flifl]
--9
According to (5.21)
> oe
6
L (_n,n')
whilst(DI)1L,y(eie)
represents a continuous function, as
be proved in a manner similar to that used for proving the continuity
(i') iO 2
of -1() in i). Consequently,
\ (
) is an L -function on the unitcircle.
(41
Now we study the limiting behaviour of the function i-t (') as
(S1-+i+. Let us note that this function is univalent and i-analytic
throughout the open domain
si>i3.
Furthermore, we have?2fl
(2
2
LD
=
lJ DOs-13Tn
' 2. whererry
-
26-2
±
Using the above relations we find
1) (t) e oO 2
a
S t)9(D\
I_(+i)
1. 00 oQ 'I-(ti#2)
_)&
=>-,>_ (+i)
'=a
1=2(5.2)
C5.3o)
c)(\2\)2Qo (5.M)
it can
27
-whereby, in conjunction with Abel's theorem (see, e.g., Goldberg [19] § 9.6) we obtain
1 ti)
f
)
(j
.5. 32)
t')
The last condition characterizes the function as a Hardy function of class
62;
see Hoffman [20], p.39 and Walsh [21], § 6.10, 6.11.
U)
Since
H ('t.°)
is harmonic and satisfies (5.32) and'tj
(e )
is anI-function of the unit circlewe conclude, applying Fatou's theorem "
(Hoffman [20] , ch.3) , that the limits ..irni
) exist for almost all
OE-flfl]
and the equality(i)
ry4 (tre )
.'.t- 4_+
holds almost everywhere in . In fact, (5.33) is valid as
the point approaches along any path in the open set
{:
>t3
which is not tangent to the unit circle.Using now (5.25), (5.26) and (5.33) we see that the limits
(i)
[ (;,') exist for almost all
ee-n, o]
andQ;
S)
=
for almost all
E[-nO]
asS
approaches -nontangentially.(.1)
G
The function ( )satisfies the boundary condition (3.15) in
L2-sense since, by construction,
ot
ee (e)_D)
and, by the definition of
vc9)=>
06. Regularity of solution.
Let us, introduce the following geometrical assumption
(A) The body-boundary is such that the Laurent series of the
con-formal mapping function has only a finite number of terms.
In the present section we shall prove,using a method due to Ursell
[io]
the following regularity resultTheorem 6.1. (The regularity theorem). Suppose that is a simple, closed curve with continuous tangent, satisfying the assumption ('i). Suppose
also that and that
#o
K0eQ\L.Then,
the uniqueweak solution
of
the problem (K0) given by the expansion (5.23), is actually a classical one, i.e. the velocity field is continuous thoughoutThe coefficients
D
of
the expansion (5.23) are of order0
(5.jrw) Remark 6.1.1. Under the assumptions stated the curve is actually an analytic one.Proof. According to Theorem 5.2 (see equation (5.18))
2
where
O() V(LO9()
1(G))
V(G) = U(G)
cke
/cLS
and &to,ckj are given by (5.17). Equation (6.1) is equivalent to
where LJ5=(co5(seU(0)) and q . Multiplying (6.2) by
Os %1
S we obtain
*sUs=ss+
(6.3)
It can be proved, with the aid of the assumption (A), that
n
-
28-a
D
S-23)...)
dswi
I
oOq
2
1 I29
-(see Appendix). Moreover, since is analytic and
e)c(C-n1oJ), u(e)
kas continuous second derivative, from which it followsthat UO(I!?). Thus
E
2
Consider now (6.3) as a functional equation in P. . Because of (6.4),
2
the solution of equation (6.3) must be in ; that is
(6.5)
Using (6.3), (6.4), (6.5) and Schwarz's inequality we obtain
IsUs-st)5I
£4Ko>1Ijs1a
00
52.'
from which it is concluded that
O(i/S2 )
. The termwisedifferentia-bility of the expansion (5.23) and the continuity of
c-j.,f(5)JL3
throughout¼YOFUB
now at once. Thus, the proof of the theorem is completed.Remark 6.1.1. Ursell
[ioj
bas studied the existence, uniqueness and regularity of solution of the radiation (heaving) problem for the semicircular boundary based directly on equations (6.2) and (6.3),i.e. withour resorting to the basis properties of the water-wave multipoles,developed in § 5. This more effective approach can be easily extended for body-boundaries satisfying the assumption (A), and this was essentially done in the present section. However, the author was not succeded to prove inequality (6.4) when the nonzero conformal mapping coefficients are infinite.Here we shall sketch the proof of inequality (6.4). Since, in the present case, it follows that we obtain 5's APPENDIX 30 -FJ
+#','4
0(-i)
c45-2)(e+-s-2) -
___
ss1
(A
-
pyy2Using (5.10) and the inequality
rn=i
a
22
2
a-eI .ICej
2(+i-a) 5
tT tn(.is-a)t(.Pi_s_a)
a4
where A(N),K and A are positive constants. Moreover,
() and
(.) have continuous second derivatives ; hence =O(L/.S2)ooI
iSo(
consequently
11sj2Ac'c
z
J52
Each term in the sum appearing in the right hand side of (AZ) dominated by the quantity
S2.
Ba
(+(5+)a
+
cx=c,L
is(A3)
+ K
A
51 2.(Aa)
31
-where . Following Ursell (rio\
, p. 295) we can easily show that
4
4i1
d
Using now
(/z) (A3)
and(t.'1)
we obtainA(t'J)
(M
'= a
which is the required inequality. Since AL(1I))oQ as 'J> o3 this proof breaks down if the number of nonzero coefficients is infinite.
ACKNOWLEDGMENT
The author is grateful to Professor F. Ursell and to the reviewer for their suggestions concerning the possibility of obtaining regularity results applying Ursell's method. The author would also like to express his deepest graditute to his friend and colleague P.D. Kaklis who gave much of his time on reading a preliminary draft of this paper pointing out errors and deficien-cies and suggesting improvements.
32 -REFERENCES
[i] G.A.Athanassoulis,
An expansion theorem for water-wave potentials,
Department of Naval Architecture and Marine Engineering, National Technical University of Athens, Report No.3, 1982f2] F.John,
On the motion of floating bodies II,
Comm.Pure Appi. Math. 3, 45-101 (1950)J.T. Beale
Eigenfunction expansions for objects floating in
an open seas
Comm. Pure Appl.ath. 30, 283-313 (1977)M.Lenoir and D.Martin,
An application of the principle of
limiting absorption to the motion
of floating bodies,J.
Math.Anal. Appis. 79,370-383 (1981)
5] M.Lenoir and D.Martin,
Etude thorjque et nwithrique
du
probl'eme linaris
dv. mouvernent sur la houle tridimensionnel,
Ecole National Superieure de Techniques Avances, Paris, Rep. No. 124, 1980
M. Lenoir,
Method de couplage en hydrodynainiae naval et
application
la resistance de vagues bidirnentionnelle,
Ecole National Superieure de Techniques Avance'es, Paris, Rep. No. 164, 1982, chapter III
[7] F.Ursell,
Surface waves on deep water in the presence of a
submerged circular cylinder II,
Proc.Camb.Phil.Soc. 46,153-158 (1950)
[8
F. Ursel 1,The expansion of water-wave potentials at great
33
-{ 9] F.Ursell, Short surface waves due to an osillating zrnmersed body, Proc. Roy. Soc., Ser.A, 220, 90-103 (1953)
[io]
F. Ursell, A probtem in the theoryof
water waves, in Numerical solutionof
integral equations, Clatendon Press, Oxford, 1974ii Y.S.Yu and F.Ursell, Surface waves generated by an oscillating circular cylinder on water
of
finite depth: theory and experiment, J. Fluid Mech. 11, 4, 529-551 (1961)J.J.Stoker, Water waves. The mathematical theory with applications, Interscience, New York, 1957
A.Zygmund, Trigonometric series Vol.1, Cambridge University Press, London, 1959
I.Singer, Bases in Banach spaces, Springer-Verlag, Berlin,1970
is]
J.R. Higgins, Completeness znd basis propertiesof
setsof
special funetions, Cambridge University Press, London, 1977 1i6] N.Bary, Sur les systmes complets desfonctions orthogonales,Mat. Sbornik 14 (56), 51-108 (1944)
17] T.Kato, Perturbation theory for linear operators, Springer-Verlag, Berlin, 1976
[18] V.I.Smirriov, A course
of
higher mathematics, vol.V.Integration and functional analysis, Pergamon Press, Oxford, 1964[ ig] R.R.Goldberg, Methods of real analysis, Blaisdell Publishing Co., New York, 1964
[20]
K.Hoffman, Banach spaces of analytic functions, Prentice Hall Inc., Englewood Cliffs, N.J., 1962S
34-[21] J.L. Walsh, Interpolation and approximation by rational functions in the complex domain, Fifth edition, Colloquium Pub]. vol.XX, Amer. Math. Soc., Providence, 1969.
D.F. Harazov, On the spectrum
of
completely continuous operators depending analytically on a parameter,in linear topological spaces, (in Russian), Acta Sci. Math. (Szeged), 23, 38-45 (1962).I.C. Gohberg and M.G. Krein, Introduction to the theory
of
linear nonselfajoin-t operators, Transl. Math. Monographs, Vol. 18, Amer. Math. Soc., Providence, 1969.A.E. Taylor, Introduction to functional analysis, J. Wiley & Sons, New York, 1957.
E. Hille and R.S. Phillips, Functional analysis and semi-groups, Revised edition, Colloquium Publ. Vol. XXXI, Amer. Math. Soc., Providence, 1957.
F. Ursell, On the rolling motion of cylinders in the surface of a fluid, Quart. J. Mech. App]. Math. 2, 335-353 (1949).
D. Eurrard, A. Jami, M. Lenoir and D. Martin, Recent progress towards an optimal coupling between finite elements and singularity distri-bution procedures, Proc. Third. Intern. Conf. on Numerical Ship