Roger'Webstor 10 REVUE D'ÄNALYSE NUMÉRIQUE ET DE THÉORIE DE L'APPROXIMATION
258 Tome XXVI, Nd 1-2, 1997, PP' 259-268
RE FERENCES
1.E.Artin,EínfihrungindieTheoríe¿¿¡'Qammanktio-n,.Teubnar'-Laipng'1931'
2. H. Bohr and J. Moll-erup,
iirr"øog
í Mathematik AnalyseIII,
Koponhagen, 1922, pp. 149-164.¡. N. drlurtaf<i, Élëments de Mathématique,BooklV, Chaptor
YIl
Lafonctíon gamma,Pafis, 1951.¿. p. ¡. bavis , Leonhar¿ Euler's íntegral. A hístorícal profile of the gamma functíon, Amar' Math. MonthlY 66 (1959), 849-869'
5.
F. \---'"
7r (1939)' 175-189'6.
R.
theorem' Amor' Math'a ^
(1938), x f(x)' Acta Math' 705',7-62.
s.A.v/.RobertsandD.E.VarbergConvøtFunctions,Acadgrlichess,NowYork'1973.
9. R. W"brtur, Convøcíty, Oxford Ùniversity Pross, Oxford' 7994'
APPROXIMATIONoFCOMPLEXVARIABLEFI-INCTIONS
AND APPLICATIONS FOR SOLVING SINGULAR
INTEGRAL EQUATIONS
V.
ZOLOTAREVSKI,'V. SEICHIUKRecoivod MaY 15, 1996 Pure Mathematics Section School of Mathematiu and Statístics
UniversitY of Sheffield Sheffield 53 7RH
England
E-maíl: R.J.WEBSTER@SHEFFIELD'AC'UK
The direct methods of solving singular integral equations (SIE) use results concerning the approximation of functions of complex variables by polynomials' These functions are defined on closed smooth contours' some results have been giveninourpapefs[1_3]withoutploof.InthisNotewewillprovesomeassertions about the approximationuy polynomials of functions defined on arbitrary closed smooth contours. These resrilts havebeenused in numerical anaþsis (collocation'
-q*O*t
.,
reduction and spline methods) for solving SIE'l.Furtheru'eneedthefollowingdefinitionsandnotations:Byf'wewill
denote an arbitrary closed smooth contour bounding a simply-connected region F+
containing the point
i=
o. By F_ wewill
denote the complementof 4 uf
in theentire complex plane. Let
c(i¡
be the space of continuous functions onf
. Assume,* ;;(ai
is a set of tuncìions defined onf
and satisffing a Hölder condition with the exponent B, 0 < B <l.
The nolm on¡¡u(r)
is defined [4,p'
173] byllqllp
Tgl'( )+
q)
-
.p(t'= llell" +
ã(q;Þ)
t sup
lÉ12
lþt2el
l', -'rl'
(1)
From [4,
p' ll3lwe
knowthat Í/u(f) with
the norm(l) is
a Banach space' According to [5,p.
109],if
0 <p
<ø (
1' thenH*(f) c Au(f)'
In this caseit
also frol<ls llellu. r',llql[,
where tp €/r-(f)'
Thisis
a consequenceof
the relation1991 AMS Subject Classification: 30E10' 65R20'
260
(s)
V. Zolotarevski, V. Soichiuk 2
Proof. Letus first
assumethat lrr-rrlr-!Q,,,r, ef).
Then,fron
(4),Solving Singular Integral Equations
t1)
-,pl(tt) -
q(t2) + q*n(tz4) -
q(tz26t
lq(t,)-
q(,,)ll',
-
',1' sup
lrxt2 tr,lTel
< max{ I sup
,ll,'-øl=t
3
we have
Let
xn = X,,(f)
be the set of allpoþomials
of the typeØ q'(r) = f'o'o'/e r'
k=-n where roare arbitrary real or complex numbers,
. For the function
q
the value 8,,(q)=
infenexnllq- q,ll"
is called the best uniform approximation bypoþomials
of the form (2), The polynomiat9l
forwhich
E,(q) = ll* - *; ll. *
called the best polynomial uniform approximation of g. Recall 16, p. 4321that for each flrnctionq
cC(f) in X,,
there always exists such auniquepoþomial
rpi,The module of continuity [5, p. 107] of function
g
isr(ô, e) =
sup ^le(r,)-
.p(¿r)l (at
o; ty,t2 et)
,lr,-trl<ô'
If
<p e¡/"(f) (o . o <
1), then(3)
c,)(¡,q)<
ä(<o;c,)ô".Itisknown l7,p.3lll thatif e
ec(l),
ttrenn,(q)sþ2,'[;,-),
rvherep,
are constants not depending on
r
and on the functionq.
Fròm tÉese two last inequalities we deduce thatif
<p e.H.(I'),
then(4) E,(q) < p,#H(q;o).
TrTEoREM
L Let
<p uã"(f)
andq)
be the polynomial of the best uniform approxímation of p. Thenltt
1,"-ø
lq(q)
- qi(,') - q(,,) * qi(r,)l
< 2*qðlq(,) - el.)l,p
=#H(q;o).
l', -',1'
'erIf
lr,- ,rl .I,
according to [3,p.
42],wereceive(6) r(ô,q;) 3 þt'
o(õ,q),
since
s . H"(r),
trren le(4)-
,pþr)l.
lr'-
hl"H(qta).
From (3) and (6)we deduce
lei(,') -
,p;(/r)l<
p¿lr'-
trl"H(rp;a)'
Therefore,sup
o<14-trl<t
),
suPI rtl2 lt,l2el
lq(t')- q(t')l l\ -
"1"
1,,
- ,,
1,,- ,,
lle
-
eillu. H(q;") (0. p
<o <
1)=F$H(q;o) . +#H(q;o) .
ts#H(q;o)
Lelr¡tøe
L
For each polynomial rpnfrom X,
we havellqill" <
pu',llq,ll".
Proof
qrþ) -- i*, = {'(r-n r r-n¡{*'..+,nt2') = t-nP2,,(t), '
k=-n
q', (t) = -nt-n-rP2,(r) * t-n(r2,Q))'= -u!rqr(r)
+t-'(r2,Q))'
'Assume that the
function , = \f(w) =
cw + cs + ctw-t -+ '.'
performs aconformal mapping of the exterior of the unit circle {l
,
It l}
onto F.- so thatV(*) =
co andV'(*) =
c>
0.According to the Cauchy theorem about derivatives o f the analytic functions, we have
(7) P',zn(t)=*!,#0"=*I,ffi*,
262 V. Zolotarovski, V. Soichiuk 4 5 Solving Singular Integral Equations 263
where
fo =
{/;t = V(pw),lr,l = t,p t l}
is the level line for some p>
1, We may construct acircle lr - tl =
¿(p- t)
b>
0,with
the center on an arbitrarypoint r
ef
such thatit
does not intersect the levelline Ç.
Let us chooset = y(rv),r = V(pr),Thenweobtain
l' -
rl=
l.r,(or)- v(,)l = lv'(6)llo", - *l= lv'(E)(p -
r),From[7,p,
1s1]0<m <lv'(E)l <M <-,lEl>1.
Then*(p-t) <l'-4<u(p-t).
In this case, the circle
lr -
4= î(p - f)
does not intersectI,
. gV the Cauchytheorem and from (7) it follows that
P'r,(t)=* J -.1'(') .,o,
It-ll=ó(P-l)
r-n(r -
t)"Further,
(x + å(p
-
l)cosv)' +(t *
b(p -l)sin<p)2x'+ y'
f
dq=
:u+m5i
x2 + y2* b'(p - t)'*
zb(p-
1.xcosg+ysin
t222
dg=
x +y
ó(p
-
t)2
l2
+ (.r cos <p + -1, sin q)
1+2b(P
I
) 22x +y
d<pLeto=t+a 'nb
Then2Í
J t*
0
ålo
-
1)T+(.rcos9+Ysinq)
tC +1t22
2
2b(P Ð
dq=
l,-'
r',,(,)l = l,-'l * n_uJ^,_,,ffi ø
=',ft_-rj,l*, ol+*{,, lËl# =
B+¡*ir,[' lil'ø,
where
' =:flËhlq'(")l
Let t -
t = b(p-t)eio (o <,p<2n). Then r =t +b(p-r)e¡v.
dr =å(p-t)iejetfq;
lo"l =
(p-t)aa9,Let t =
x+ry.
Thenl,-'r,,,,(,)l =, +dry'Ï)
n oro-,)<re == sL, I, 2flx+¿y+ å(p-
rXcos<p+¡sin<p)l'IrrT tl ;,
-l
o'=
=T'-,1
I
2n**cosq+/smq
1
x
22+y ur={['.1 o
_--an1, t2^
I
¿.p<I
exp
2tt
I
=J
0^ i
xcos9+/sm9
¿n
x z2 +y
d9rcosg+ysin<p
,'*y' dg,
where
^ I
)r+*cosq+/smq
*'*y'
However,
rcosq+ysing
cosq+ -l ,sing=cos(ï-q) <l
x'+ y'
*'*y'
6 ,| Solving Singular Intograt Equations 265
,t el
264
Hence
V. Zolotarevski, V. Soicbiuk
l,
-.
r,,.(,
)l.
B+, *r(^¡r)r- =*n
['
. ffi)"t þ,
I tcft
-, | = ¿ ( p - r) )Let us denote
by rp
the inverseof
the frrnction ry. Then we perfonn a conformal mapping of the exteriorof f
onto {lwl >t}
so thatg-(æ) =
oo. Recallthat
q-
has the forme-(r) = I, *? . þ .
Let,l?
>
1 andf"
be a level line of mappingV for
lwl=l?.
Consider the Q,function
(_ I
. This function is analytic onF_,inc
ling thepoint co. Therefore, by the maiinium principle, we have*lffil .***ffi=#xl#
Since D,*
c F*,
then23+le,(r)le.t,)]'l = -sl*,(')[e.(')]"1 = T.irle,(')[*.(')]'l = 1e*le'(r)l
In this relation we consider that
for r
eF, lq.(t)l = I
. Therefore,(e) llq"ll"r,;1"
<llq,ll.t.l'
The function <p, is analytic in the ring Fbounded
by f,
andfo
' Hence, bythe maximum principie and by inequalities (8) and (9), it follows that
i.Ele,
(r)l= -u*{ll*,
ll"1ç¡ ; lle, 11",,^,1='.*{il*,
11"1.¡},
ilr, 11",.,^' }'
Let r =t-!
and R= l+!.tn.n
Rn<
eandrn > 1 nn+
so
llq,ll.rot= allç"ll.rrl
and, rinally,llq,ll"r.l <
t urllq,,ll.rr¡: rüh€reznltl'z
[,.**,[,.+)),n
=_
maxr.r;l ,p"(t)l.R,
We note
that ¡t; is
apart of F-, which is the exterior for the levelline fo
We also note that
for I
el^
we ftuut l,n-(r)l=
.R, SinceDl c f
, then-,{ffil = -r,lffil = **]# =
mqxle,(')
INow we use that
fo¡ r
er,
le-(r)l-
1. Hence(8) 1i3¡le,(r)l = +g.l,p,(,)li?"
Further, let
w =
rp* (r) be the mapping which is the interior off
on{lrl . t}
suchthat,p.(0)
= 0.Itisknownthat
q.(¿)= t +yrtz rTzt3 +...
.Then[*.]'*,
is an analytic function onF'n.
Letr
< 1,f,be
the level line of themapping9*(r) for
lwl= r.
Note tnat[e.(r)]=t i,
un inverse functionof
<p* andD"*
is a part ofF*,
which is in the interior o1f,
, Then, by the maximum principle,ry+1e,,(')[e.(,)]'l = *f{|q,,(')llq.(')1"} = llq,ll.r,,r' /'.
Tmoneu 2. Let g,
€X,.
Then(10) llq,llp<p,rnpllq,llc, o(p<1'
Proof.
First assum ethatlt,- bl> j,{,r,t, el)'
Then1
lÌ.
< znpllq,,ll.
Now consider that lr,
- rrl.L .mthis
case, fromle,(r,)
-,p,,(,r)l
=|
I *',,,"ru"
= llq',11" 'leng'thtrt,3
constllc',ll"l,t- trl
and Lemma 1, we have
266
(q) - q,(t, < cotst
¡rø
.rllqrll
,lrt - rrl'-u =
const'þs.nß llqrll, l', -'rlu
From these cases we deduce (10), where p7
=
max(l + const' po;3),Further, we establish the estimation relation of the interpolating Lagrange polynomial for the function
q
defined onf
.LEtvfl\l,t 2.
Let
,p ۋ"(f)
andq,,
e X,, be a poþnomial such thatllq- q,ll"
< (p',*¡rnhn)ø,(rp):
Then
(11) llq-q"llu .@j#'¡aã(q;o) (o.p < o <
r)Proof.
Let
<pi be the polynomial of the best approximation for the func- tion <p. From (5) and (10) it follows thatlle
-
e,llu= ll* -
eillu* llol -
e,llu< p,#ã(q;o) * p,,ullel -
r,,11,.Using (a) for the second term of the last inequality, we deduce that
llel -e,ll. = ll*i -*ll. *llq" -qll. , *,#n(q;o)+(p,,
+pnrnn)^E',,(rp) << (t * pr *
pslnn)¡:"r# r@to)
= (p,,, + ¡r,,,lnlþ u@;").
Then
for
Fro=
F¡* þtl\z
ifrrd prrr=
pzlqs we obtain inequality (11), Let {t r}'"-_rbe a consequ ence of2n+
1 distinct points fromf
and2n
Solving Singular Integral Equations 26',7
u,(s, t) = iî,1,¡gç,,¡
.t=0
It is evident that
U,X, =
Xn. As in [5, p. 539], we establish thatlr,(t,¿)-
g(¿)l<
(t* )',)n^(e), g(r)
ec(r),
where
2t
L, = TS,.IlLt'l¡. -
j=oLet us
consider thefunction
V,V(w) =
cw + "orr-' +. .., which
mapsconformally the exterior of the unit circle with the center 0 onto
Fl
so thatV(*) = -
,v(-) = c>
o'Let tu, =
eXp#htt -n) (i'=:1,i =
0,1,...,2n) be a system of
equidistance points on
lo
={lrl : l}
, and(12) t,=v(w,), i=0,1,...,2n.
TuBonRv 3.
If
tj,i :0, l,
...,2n
are defined by (12), thenI, I[rr++¡r*lnn,
The proof of this theorem is very long, We
will
omit it.2. Now, using our previous results, we propose a substantiation
of
the collocation method.We consider SIE with the Cauchy kernel
(13)
c(r)e(r).+{9 a,+};[*r',t)<p(t}1t = f(t),t et,
in the Banach space
Hp(f)
, 0 < p < 1. Here c, d,K
andf are known functions inøu(f) '
According andq
is unknown,to the collocation method, we seek an approximate solution of (13) in the form ofapoþomial
n ,,
.p"(r)
= luf){, ter
k=-n
{"f'}-=_, = {c,r}'r=-,
are found from the followingV. Zolotarovski, V. Seichiuk 8 9
I¡(')
=fl('-'o)
k=0; k+
j
fl (,, -,0)
k=0;k+
j
tj
2n t
n
= >^?* (re r, i=o2n)
k=-n
By
Unwe denote the operator which maps any function g continuous onI
onto its interpolatin
g
Lagrange poþnomial defined by using the nodes{t,}',',=o
This is apolynomial of the form
(14)
The unlarown c oeffrcienfs system of linear eqttations
268 V. Zolotarevski, V. Soichiuk 10
(1s) "(',)ä.'otj
+r(t,)/,ootl *
þ,*r*[*r,,,,),e.rr = Í'(,,),
where a(t)
= c(t)+d(t), b(t)= {t)-a(t)andt,i:0,1,...,2nform
aset ofdistinctpoints
onf.
. Tmoneu
4. Supposethefunclions a, b and K(uniformwithrespectto both variables) belongto
the spdce¡1"(I.),
0<
Pa a Í l,
andlet
thefollowing condítions hold:t)
a(t)b(t) É 0,/
e F,2)
intd a(t)U-'(t)=
0,,
€ I-,3) the lrernel of the operator cowesponding to the left side of
(I3)
is empty.In addition, let t¡,
i =
0,2n, be calcalated according to (I2).Then,
þr
stfficiently large n, the system(I5)
has a uníquesolutø"
{op]J'o=
,,
The ap¡troximatesolutions (14) convergein thenorm of
Hr(f)
as n-+
æ to the exact solution <pof (I3),
whatever the functionf
eH"(f). For
the rrtte oJ convergence, the following estimate holdsllq
-
q"llu3
(P,u + P,r lnn)nþ-"n(e;ø)'
For the proofsee [1-3].
REFERENCES
RDVTIE D'ANALYSE NT]MÉRIQUE ET DE THÉONTN DE L'APPROXIMATION Tome XXVI, N* 1-2, 1997, pp. 269-275
BOOK REVIEV/S
Partial Differential Equatíons and Functional Anølysis,lnl. Memory of Píerre Grisvard, edíleó,
by J. CEA, D. CHENAIS, G. GEYMONAT and J.-L. LIONS, Progress ín Nonlînear
Dffirential Equations and Their Applícatíons, Yol. 22, Birkhäuser, Boston-Basel- Berlin, 1996, 264 pp., ISBN 0-8176-3839-3.
Piorro Grisvard, one of tho most distinguished contomporary French mathomaticians, died on April 22, 1994. A conforonco was hold in November 1994 out of which grew tho invited articlos contained in this votumo. All papers are rolated to functional analysis applied to partial differential equations, which was Grisvard's speciality. Indood, his knowledge of this ar€a was extremely broad. Functional analysis applied to partial differontial equations is an a¡oa enjoying rebirth and reexamination in many parts of tho world, particularþ with rospect to its recent thrusts into mathomatical physics. Griwa¡d also becamo a specialist in tho study of opti- mal rogularity for partial difforontial equations with bounda¡y conditions. Ho oxaminod singularitios coming ftom coefhcients, boundary conditions, and mainly non-smooth domains.
Griward left a logacy of preciso results vory usoful in applied mathematics, which wero published in journals and books. This volume contains a bibliography of his works as woll as his last paper, which is published horo posthumously.
Tho contents aro as follows: P. Griwmd, Problèmes aux limitets dans des domainec avec points de rebroussemenf; T. Apel and S. Nicaisa, Ellíptic problems in domains with edges:
anisotropic regularity and anisotropicfníte elernent meshes; M. S. Baouendi and L. Preiss Rothschild, (Inique continuation of harmoníc functíons at boundary poinß and applications to problems in complex analysis; C. Ba¡dos and M. Bolyshov, The wave shapíng problem; P. G.
Ciarlet, Modélisation mathématique des coques linéøirement élastiques; G. Da Prato, Fully non- Iinear equ,ations by linearization and maximal regularity and applicatiorls; M. Dauge, Strongly ellíptic problems near cuspidal poinß and edges; L. Boutot do Monvol, Star produît associë à un crochet de Poisson de.rang constant; P. Destuynder and F. Santi, La méthode de,s lâchéø Qe' tourbíllons pour lecalcul des efforß aérodynamíques; A. Favini, Sum of operators' method in abstract equations; G. Goymonat ancl O. Tcha-Kondor, Constructive methods
þr
abstract dif- þrential equations and applícatiozs,' V. A. Kondratiov and O. A. Oleinik, On asymptotics oJsolutíotts of second order elþtic equations in cylindrical domains; V. A. Kozlov ancl V. G.
M*'yq
Singularities to solutions in mathetnatical physics problems in non-smooth domaíns; J.-L.
Lions and E. Sanchoz-Palancia, Problèmes sensítifs et coques ëlastiques minces; M. T. Niano, Contrôlabílité eractefrontière de l'ëquøtion dec ondes en présence de singularités; E. Sinestrari, Interpolation and extrapolation spaces in evolution equations; M. Zamar, Localísatíon des
singularit* sur la frontière et partitions de I'unitë.
Ths rosoarch or oxpository articles included in this volumo, many of them with diroct roforences to Grisvard's contributions, will bo of groat intorost for all spocialists in nonlinear partial difforontial oquations and thoir applications.
R Precup 1991 AMS Subject Classification: 65899
l.Y.
Zolotarevski and V. Seichiuk, The collocation methodfor solving singular integral equations gíven on a Lyapunov contour of íntegration, Soviot. Math. Dokl.23 (1981),468470.
2. V. Seichiuk, On the rate of convergence of collocation and mechanícal quadratures method
ral
equ
apunov contout', Chislonnoe roshonie zadach ,$tiints
108-118 (in Russian).3.
V. riimens
lving singulør integral equations on closed contours of íntegration, Çtiintsa, Chiçinãu (1991) (in Russian).4. N. Mushelisvili, Siigular Integral Equations, Nauka, Moscow, 1968 (in Russian); English transl., Noordhoff, 1972.
5. I. Natanson, Constructívè Functíon Theory, GITTL, Moscow, 1949 (in Russian); English transl., Vol. III, Ungar, Now York, 1965.
6. J. Walsh, Interpolating and Approximation by Rational Functions ín the Complex Domain, Moscow, 1961 (in Russian).
7. P. Suetin, The series on Faber's Polynomíals, Nauka, Moscow, 1984 (in Russian).
Received May 15, 1996 State Uníversíly of Moldova ChiSínãu
Moldova