This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
0 i s s u f f i c i e n t l y l a r g e .
.
tends t o i n f i n i t y i n
Then (3.12) implies t h a t
Furthermore,
N -h f.(x)-a.= z c x + ~ ( x - ~ ) J h=l j h
as
x
tends t o i n f i n i t y i n
B
as
x
tends t o i n f i n i t y i n
B
I)
Assume t h a t
.
Therefore
-.
By u t i l i z i n g the d e f i n i t i o n of F ( p l ( x ) , . . . , p k ( x ) ) we can complete the proof o f Theorem 3.1. The following r e s u l t s a r e s p e c i a l cases of Theorem 3.1:
where
8
i s e i t h e r open o r closed.
11) Assume t h a t
Then
6
ASYMPTOTIC EXPANSIONS
6 i s a closeil s e c t o r
where
Let us consider a s e c t o r
Then, i f
cofO
M' i s s u f f i c i e n t l y large, we
and t h e p o s i t i v e number
have m
-N l / f ( x ) ~1," c cNx N=o
*
(x+m
~
j
LB) .
in
W e s h a l l now t u r n t o a j u s t i f i c a t i o n of term-by-term
i n t e g r a t i o n and
d i f f e r e R t i a t i o n of asymptotic expansions. T I I E O M 3.2:
Assume t h a t
3 .i3) where
B
i s e i t h e r open o r c l o s e d .
(3.14) where t h e p a t h of i n t e g r a t i o n should be taken i n Proof:
We assume without l o s s of g e n e r a l i t y t h a t
.
B rD
i s a closed s e c t o r .
Since
we have
hrthermore
sm[
N
x N
ch5-hld5=
h=2
h=2
c
h x-h+l -h+l
This proves Theorem 3 . 2 .
THEOW 3.3:
6
be a closed s e c t o r d e f i n e d by
rD: a L a r g x l b
where &
i;i
gi&
b
,
1x1
are real numbers and
t h e i n t e r i o r of
rD :
LM , M
i s a p o s i t i v e number.
Denote
7
PROPERTIES
fit:
,
a<arg x< b
1x1 > M
.
Assume t h a t f(x)
m
=
c
.I
&
(x+=
"$-A
0)
.
N=o m
-N-1 f l ( x ) 5 I: (-N)cNx N-
where Proof:
fl(x)
(x+m
denotes t n e d e r i v a t i v e of
f(x)
&
fit)
,
with respect t o
x
.
Let us put N
P,(X) =
z
ChX
-h
h=o
and = f ( x ) -pN(x) We must show t h a t
-N-2)
%(x) = O ( x
s u b s e c t o r of t h e open s e c t o r sector
9
.
gl
'
as
x
tends t o i n f i n i t y i n a closed
To do t h i s , l e t us c o n s i d e r a closed
defined by
where a [$ sin($) 1
y(w0)
]Y(t)
1
< ep 0 s i n ( 2 6 ) [ Y ( o )
t
-1,
(3
1
1
-3 s i n ( 2 6 )
sin($))-'[e
2 0
- 111
.
< 3 [ sin($) 1-l
This completes the proof of Lemma 10.1. W e s h a l l now complete t h e proof of Lamma 9.1 by using Lemma 10.1.
do t h i s , consider a s e c t o r 8 i n the
( al,
...,a m )-space
? - p l a n e and a domain
To
;Qr i n the
which a r e defined r e s p e c t i v e l y by 8: l a r g ho+ (mt2)arg
53 ,
51
151 2n
and pr:
lall t
...+(%I
m-1
(i)
cpk
i s a polynomial i n
(al
i s a polynomial i n
..., N - m t l )
( h = 0 , 1 , **.,m-2)
,...,am,Xo, ...,$-1)
are i n t e g e r s ; ( i i ) Qh
( k = 1,
(a1,
,
9
whose c o e f f i c i e n t s
...,am,Xo, ...,sJ-mtl) whose c o e f f i c i e n t s
are integers. I n this manner w e can determine
qn(x)
and
Sn(x) ,
t i o n of (15.8) and the second equations o f (15.9) determine
, respectively.
pn(x)
The q u a n t i t i e s
pn(x)
.
( n = 0,1,. .)
The second equaand p,(x) contain a r b i -
t r a r y constants which we will f i x by t h e conditions pn(0) = 0
(15.16)
16.
( n = 0,1,.
..) . Let u s write
A subdominant s o l u t i o n of equation (15.2).
S ( X , E ) as
m,U s(x,c)=
z
m-h sh(c)x
h=2 where
(16.1)
.
.
S are polynomials i n (a1,. .,a ) The power series (16.1) h,n m a r e formal. However, by using Theorem 5.1 (Chapter 1), w e can f i n d m-1
and t h e
functions
Ph(a,c)
(i) the
8,
where
, cl
Ro
( i t ) the (16.3)
ph
(h=2,
...,m)
such t h a t
a r e holomorphic i n t h e domain:
and
PO
a r e a r b i t r a r y but f i x e d p o s i t i v e numbers;
s a t i s f y t h e asymptotic conditions
A RELATION
60
uniformly f o r
€1 < P o
law
BETWEEN bm AND
+ ...+lamI L R o
lall
-
as
tends t o zero i n t h e s e c t o r
c
We s h a l l consider t h e d i f f e r e n t i a l equation (15.2) w i t h
(16.A) If we s e t
{
=
(16.5)
(h=l)
+:
,
€Bh(a,d
(h=2,
...,m)
,
we obtain a s o l u t i o n
(16.6)
w = f ( x , a , c ) =bm(x,al(a,c),
of (15.2)
.
.. . , a m ( a , c ) )
This s o l u t i o n i s c e r t a i n l y subdominant along t h e p o s i t i v e real
axis i n x-plane.
Since b m ( x , a ) i s e n t i r e i n (x,al function f ( x , a , g ) a s a power s e r i e s i n
,...,am 1 , we can express t h e (cP,, ...,e~,) . S e t
(16.7)
where
(p1 = p 2 + . . . + p m
and t h e
-.
fm (x,a) ,P2’ * ’Pm
are entire i n
(x,a)
.
From (16.7) we d e r i v e
(16.8) If we s e t
al
(h=l)
,
ch(a,B) = { ah + r eieh where
r
...,m) , t h e oh a r e real v a r i a b l e s ,
(h=2,
i s an a r b i t r a r y p o s i t i v e number and
then
(16.9)
f
mYP2’.
..tPm( x , a )
Therefore, the power s e r i e s (16.7) and (16.8) a r e uniformly convergent for O<x 0
i s s u f f i c i e n t l y l a w e , system (19.12) has a solu-
tion -
i n t h e domain defined bx (19.1)
and
( i ) C1(x,A)
(19.15)
as
C (x,h) 2
lal[ t
w ( x , h ) = 1 t G2(x,h)
(19.3) such t h a t
tend t o zero uniformly f o r
...t l a m [I R o
,
O<x 0
positive real axis.
Hence
bm(x,al
, solution y(x,A)
,...,am-1’
(19.17) i s subdominant along t h e
i s a constant m u l t i p l e of am t A )
.
We s h a l l determine this constant m u l t i p l i e r . ( 19.18)
x
bm(x,al
To do this, set
,...,am-1’ am +A)=tm(al ,...,am-1’
a +A)y(x,h)
m
.
ESTIMATES IN
As
x
tends t o
+a0
, we
Im g
5 n-
71
po
have
i -&
where
A(
(m: odd)
,
and -h h=l The q u a n t i t i e s
bh
and
Section 6 , Chapter 2.)
If
r
m
a r e polynomials i n
al
,...,am-1’
a +?,
m
m i s odd,
Theref ore (19.20)
t m ( al , . . . , a m-1’ am + A )
h=l
(m: odd) If
m is even,
Theref ore,
.
.
(Cf.
ASYMPTOTIC BEHAVIOR OF bm(x,a)
72
c)
( 19.21
Cm(al,
...,am-1'
am + h )
(m: even) Set
-
f i P ( ~ , h )2 -
T
h(al, ".,a
1- h 7
- $ n tCt l )
\T
] d ~
h=l
0
(19.22)
7
.
(m: odd)
,
m+A) = h=l
0
(m: even) m=l
In p a r t i c u l a r , i f
.
,
J(m+ 2) = 3/2 2
and hence (19.23)
%(al+A)=J
--1
1.1
$Q
[ P ( T , A ) 2 - ' f 2 - L (2 ~
1+ A ) T
0
If
(19-24)
la1 2Mo
9
larg X I 9 - P o
9
and Mo i s s u f f i c i e n t l y l a r g e , t h e r e are no singular poin i n the s e c t o r
9
of
P(
0 5 a r g T<arg A or 02-g
T2-g
A
.
Therefore, by v i r t u e of t h e behavior of t h e integrand of (19.23) a t and
, we
can change T=ht
t o obtain
T
by (O(_t_arg 7 2 , a r g A Therefore, we can change -1 2 T=A t
T
by (O ~ w -4 ~exp[Kw ~ 3 61 , -1 lo5
f(O,w-'A)
( 2 5 .Y)
a
If (25.16) and (25.24)
tends t o i n f i n i t y i n s e c t o r (25.22).
are i n s e r t e d i n t o (25.15), we g e t
lo2 (1) = o ( l ) e x p [ K ( l + w 3 )X 6 1 012
W
This estimate is not s a t i s f a c t o r y .
.
In order t o get a b e t t e r estimate
we
s h a l l use t h e following f o e a s :
(25- 2 5 )
I
f(0,A)
= 1 7 [ ~ ( o , ~+~wIf ()0 , w 2 uI C(W
f '(0,A) =-
A>
I
[wf l(o,W%)
+ fl(0,w2A) I ,
c(w3A)
which are derived from (25.12) by replacing A
(25.25) i n t o (25.151, we get (25.26) where A(k) =
(25.27)
and B(A) =
(25.28)
It i s n o t d i f f i c u l t t o prove t h a t A ( A ) =wu0,~(w3a) ,
(25.29) To estimate
A(?)
, we
assume t h a t
3
by w X
.
Upon i n s e r t i n g
STOKES MULTIPLIERS
-
A plane
In s e c t o r (25.30) we have l=g[w 3A J (~T I -
(25.31)
Po
1arg[w%11
P,
.
Hence, (25.24) and
-2 -1 52 f(0,w3X) = [ l + o ( l ) ] w % exp[Xw2X6] , 21
(25.32) €[(Q,w3X)= [-l+ o(l) hold as
A
25
1 ~ 5 exp[h2X61 ,~
tends t o i n f i n i t y i n s e c t o r (25.30).
If (25.24) and (25.32)
a r e i n s e r t e d i n t o (25.27), we g e t
g
2 (25.33)
as
A
A(h) = 2w4[l -k o(l) ]exp[X(w tends t o i n f i n i t y i n Sector (25.30).
3 s e c t o r (25.30), the q u a n t i t y w X
- l)h
as
A
2 1
tends t o i n f i n i t y i n s e c t o r (25.30).
1 ( A )= 2 W 4 [ [ l t d
092
or
13
is i n s e c t o r (25.19). Hence
c(w3X) = w 2 [ l t o ( l ) ] e x p [ - X ( l t w
W
1
Wthermore, i f
-1 (25.34)
s6
-6
6 )A ]
Therefore,
lo o(1) ]exp[K(w
5 2
-6
+ w )A
2 1 -6 6 + [ l + o ( l ) ] e x p [ K ( l + w ) A 11
6
1
is i n
CASE WHEN
P(x,A) = x 3 + h
-
i 2 5
1 c(A) = w z { [ l +
(25.35)
107
-6
o(l)]exp[K(w
- w6 ) A 6 1 1 2
-6
+ [ 1 + o ( l ) I e x p [ K ( 1 + ~ )A as
tends t o i n f i n i t y i n s e c t o r (25 .30)
A
arg A
The neighborhood of the d i r e c t i o n s e c t o r s (25.19) and (25.30).
.
6
11
=-$
I n order t o estimate
is n o t covered by i n the neigh-
(X)
W 092
borhood of t h i s d i r e c t i o n we use the following formulas:
(25.36)
I
= 1 2 [ f (0,w2A) + w f ( O,wh)
f(O,w5)
J
,
c(w A ) f ‘(0,w-lA)
=y [wf ‘(0,w2A) + f ‘(0,WA) 1 , 4 w
A)
which a r e derived from (25.25) by replacing
A
by
w-~A
.
Upon i n s e r t i n g
(25.36) i n t o (25.15) we g e t
I n the same way a s we derived ( 2 5 . 2 9 ) , w e g e t
..
2
(25.39)
= L 4,
In order t o estimate
@A)
( 25 -40)
B(A) = W W ~ , ~ ( A )
, we
-TT+p o < m g
-7
..
.
assume t h a t
n
X i - 5-
po
.
( C f . Fig. 25.5.)
d(25.40)
~
c -
A- Plane Fig. 25.5.
STOKES MULTIPLIERS
108
In s e c t o r (25.4O), w e have (25 -41)
lmg
Ln-
P,
IardwAIl
9
5.- P,
-
Hence, (25.16) and
52. 11 % 4 exp[& 6x 6 I ,
I
f(0,wX) = [I+o ( 1 ) 1w
(25-42)
11
fl(o,wa) = [-i+ o ( i ) 1w%4
hold a s
tends t o i n f i n i t y i n (25.40)
X
.
Ii 6 6 exp[Kw a I , Thus
i.z
2
B(A)
(25.43)
6 6
3
= 2 W 4 [ l t o(l)]exp[K(ltw ) X
as
tends t o i n f i n i t y i n s e c t o r (25.40). Furthermore, i f 2 (25.4O), the q u a n t i t y w X i s i n (25.19). Hence c
1
as
5
tends t o i n f i n i t y i n (25.40)
.
z
2 5 (1tw6)w3=-(1+w-6)
(25.45)
A
is in
c c
Observe t h a t
.
P u t t i n g everything together, we g e t again (25.35) a s
X
tends t o i n f i n i t y
.
i n (25.40)
Thus we obtain (25.21) as
1
-
1 2 1 c(X) =w2[1+o(l)]exp[K(1+w-6 ) A 6 ] tends t o i n f i n i t y i n the s e c t o r
c
-2
+ as
A
c
6 6 [ 1 t o ( l ) ] e x p [ K ( 1 t w ) h 13
tends t o i n f i n i t y i n the s e c t o r
(25.46) (See Fig. 25.6.)
-n-$+Po<arg
1s:-
p,
.
Sectors (25.19) and (25.46) cover a-plane completely.
3
CASE WHEN P ( x , h ) = x + h
log
Mg. 25.6. Now, we s h a l l investigate t h e r e l a t i o n between the two asymptotic formulas (25.21) and (25.35) i n the common p a r t of the two s e c t o r s (25.19) and
(25.46).
To do this, l e t us f i n d the term i n formula (25.35) which domiSuch a term can be found i f w e examine t h e real p a r t of
nates t h e o t h e r . the quantity
2 2
(25.47)
2 2
2 2 2
‘6 6 6 K ( l + w 6 ) A 6 = K ( ~ + w - ~ ) A ~ - K (- ww ) A
i n s e c t o r (25.46).
Observe t h a t
I 6
(25-48)
arg(l+w )
= 6n
.
Hence, i n (25.46), (25.49) This implies that
F
STOKES MULTIPLIERS
110
i n s e c t o r (25 .46).
Therefore, if b 0 > 0
i s g i v e n , we have
-1 (25.51)
c(h) =w
2
2
5 2
-6 6 6 [ l + o ( l ) ] e x p [ K ( w - w )A ]
and
1 2 5 -6 6 c(A) = w 2 [ 1 t o(l)JexpCK(1 t w ) A 3
(25.52)
as
tends t o i n f i n i t y i n t h e s e c t o r
A
(25.53)
-r-+n+
p 0 < a r g 11-9tio
and i n the s e c t o r
-4+ 5 t i o ~ a r g1515- p, ,
(25.54) respectively.
(See F i g . 25.7.)
The common p a r t o f s e c t o r s (25.19) and
(25.46) i s contained i n s e c t o r ( 2 5 . 5 4 ) .
Thus w e found t h e r e l a t i o n between
t h e twc asymptotic formulas (25.21) and (25.35) i n t h e common p a r t o f
(25.19) and ( 2 5 . 4 6 ) . T
Mg. 25.7. Now, l e t u s make t h e f o l l o w i n g o b s e r v a t i o n .
(25.55)
-+ n
p,<arg
A52n-$-
6,
Assume t h a t
.
5
Then - n - $ n t p o 5 a r g [ w - 5 ~ ~ < - $ - 6, and hence t h e q u a n t i t y
wV5A
-1
,
i s i n s e c t o r (25.53).
5
Therefore
2-32
c ( h ) =w2[1+o(l)]exp[K(w -6 - w 6 ) w 6 k 6 ]
as
1
t e n d s t o i n f i n i t y i n sector (25.55).
Since
CASE WHEN
and
2-3
6 6 =w w w we g e t again (25.56) as
A
-? =w
P(x,X) = x 3 + A
111
-:-;2 , =-w
w
-6
-
1 5 2 c(A) = w 2 [ l + o ( l ) ] e x p [ K ( 1 + w-6 ) A 6 1 tends t o i n f i n i t y i n (25.55)
.
Thus we have proved the following
result: The q u a n t i t p
admits an asymptotic representation
c(A)
-1 2 2 c(A) = w 2 [ l + o ( l ) ] e x p [ K ( l + w-6 ) A 6 1 as
A
tends to i n f i n i t y i n t h e s e c t o r
-?+b 0 5 a r g A ( 2 n - y -
b0
.
Furthermore, asymptotic formula (25.35) i s v a l i d a s
?,
tends t o
i n f i n i t y i n the s e c t o r (25.57)
IargA+?l = O
,
Fh(c) = O and
Fh+l(c) = O
.
L e t us d e r i v e a contradiction from the assumption t h a t
#'
Fh-
(37 -4)
Fh(c) f
9
o
9
Fh+l(c) #
'
Set
(37.5 Then (37.6
k
where
C (a)
(37.7)
i s defined by (21.4). Furthermore, -r (Cf. (21.211.) Wo(a) =2w m,l
.
Lst us set
Assumption (37.4) implies t h a t (37.11)
Y,(C)#O
>
YZ(C)#O
Hence t h e r e exists a neighborhood (37.12)
v,(a)#o
f o r every a E V
.
Denote by
;O
,
9
V
of
v,(a)#o
,
q(c)#O c
@#O
*
such t h a t
~ , ( a ) # o, k2(a)#o
Assume t h a t the closure o f
6 ,M
9
V
i s compact.
a s e c t o r i n x-plane which is defined by
( 3 7 -13)
6 i s a small p o s i t i v e number, and M i s a l a r g e p o s i t i v e number. The s e c t o r B i s contained i n ah Therefore, by u t i l i z i n g the asymp 6 ,M t o t i c representations (36.6) o f ( x , a ) , we obtain
where
.
m,k
DISTRIBUTION OF ZEROS
158
a€ V
uniformly f o r if M
as
x
tends t o i n f i n i t y i n
'6,M
This means t h a t ,
*
is sufficiently large,
(37.15)
for large
z(a(n') E
%
.
( n = 1 , 2 , . .)
n ,
l i m z( a ( n ) ) =a n++ z(a(n):! a r e contained e i t h e r i n t h e s e c t o r
Since
2h-1 l a r g x - ~ n 56 l
(37.17) f3r l a r g e
n
.
and
,
1x1 L M
,
the points
,
(See Fig. 37.1.)
0 2h-1
m+2
Fig. 37.1. W e s h a l l prove t h a t , i f
cannot contain any p o i n t s
M i s s u f f i c i e n t l y l a r g e , s e c t o r (37.16)
z(a(n))
.
S i m i l a r l y , we can a l s o prove t h a t
s e c t o r (37.17) cannot contain any p o i n t s large.
z(a("))
, if
M
i s sufficiently
This will l e a d u s t o a c o n t r a d i c t i o n .
L e t us f i n d t h e zeros of t h e s o l u t i o n
cp(x,a)
i n s e c t o r (37.16).
do t h i s , we s h a l l use the asymptotic r e p r e s e n t a t i o n s (36.6) o f
for
k=h
equatior.
and h + 1
, and
formula (37.8).
To
Qmlk(x,a)
From (37.8) w e d e r i v e an
PROOF OF THEOREM 36.1
159
(37.18) and R1
There e x i s t two p o s i t i v e numbers Ro
such t h a t
(37 -19)
, s i n c e (37.12) holds f o r ( x , a ) imply t h a t of U m,k
aEV
for
tations
uniformly f o r
a€V
as
x
.
Also, the asymptotic represen-
tends t o i n f i n i t y i n s e c t o r (37.16), where
r=r m,h+l-rm,h
(37.21)
aEV
r ” = h r + rm,h+l
’
*
Set
and
Then equation (37.18) i s equivalent t o (37.22) where
,
g(x,a) =(2nN)i+ $(a) N
is an a r b i t r a r y i n t e g e r .
Let u s set
Then from (37.20)
we derive
l i m H(x,a) = O
uniformly f o r
a€V
as
x
Now i t i s easy t o prove t h a t t h e
tends t o i n f i n i t y i n s e c t o r (37.16).
image o f s e c t o r (37.16) by the mapping (37.23)
u=g(x,a)
(for every fixed
a
in
V)
contains a s e c t o r (37
.a)
i n u-plane, where
Img
-6
~ - 17 1 5-69
1.
i s a s u f f i c i e n t l y small p o s i t i v e number,
s u f f i c i e n t l y l a r g e p o s i t i v e number, and in V
2z and
% is a
a r e independent o f
.
Denote by U(N)
the set i n u-plane which c o n s i s t s of p o i n t s
a
160
DISTRIBUTION OF ZEROS
(2nN)i+ $ ( a )
, where
(aEV)
i s a positive integer.
N
s u f f i c i e n t l y small.
UCN)
Observe t h a t , i f
i s sufficiently large.
(37.24).
N
and
l i m g ( x , a ) ==
U(N)
U(N) rp(x,a)
as
i n (37.24), i f
.
NfNI
i n s e c t o r (37.16) a r e tends t o i n f i n i t y i n
x
g ( z ( a ( n ) ) ,a‘n) )
i s i n s e c t o r (37.16) and
z(a(”))
is
V
i s contained i n s e c t o r
aEV , i t follows t h a t
s e c t o r (37.16)uniformly f o r f o r some
a r e mutually d i s j o i n t i f
U(N1)
Pathermore, s i n c e the zeros of
s o l u t i o n s o f (37.22), and
Then. i f
E U(N)
].(a
(4 )I
i s sufficiently large.
Assume t h a t t h e r e a r e i n f i n i t e l y many Then (37.2) implies t h a t t h e r e exist two positive integers
and
N
z(a(”))
i n s e c t o r (37.16).
z ( a ( n ) ) and
, and
z ( a( n ’ ) )
two
such t h a t
NI
(i) N # N ~ ;
,
(ii) g ( z ( a ( ” ) ) , a ( ” ) )E U ( N )
U(N’)
c
to
.
a=a(t) a(n’)
(OLtLl)
, where
W e assume t h a t
long
€18
.
t
,
a ( 1 ) = a( n ’ )
, where
U(N(t))
Note t h a t t h e curve a
and t h a t
rl
joins
tends t o
C
terminates a t along C
c
Iz(a(t))(
a(tl)
(ii) z ( a ( t ) ) i s (iii) N ( t l ) # N ( t 2 )
Let
r2
i p
to
Consider
.
z(a(t))
As
g ( z ( a ( t ) ) , a ( t ) ) of
.
c
, and
rl
that
z(a)
tends
Hence we can assume t h a t
is s u f f i c i e n t l y l a r g e .
a(t,)
of
To
Assumptions
( i ) and
such t h a t
;
s e c t o r (37.16) f o r
t l < t < t 2;
.
be the image of
rl
joins g ( z ( a ( t l ) ) , a ( t l ) ) t o
by
tradicts the f a c t that
u=g(z(a(t)),a(t))
g(z(a,(t)),a(t,))
dz(a(tl)),a(tl)) i s i n U(N(tl)) Since a l l U(N) a r e d i s j o i n t , I’2
37.2.)
.
i s a p o s i t i v e i n t e g e r depending
N(t)
( i i ) imply t h a t there e x i s t s a subarc
(i)
which j o i n s
C
i s the curve which defines the s i n g u l a r i t y a t
z ( a ( t ) ) i s i n s e c t o r (3’7.16), t h e p o i n t
t o i n f i n i t y as a(t)E V
C
be a subarc of t h e curve
a(0) = a ( n )
u-plane must be i n on
and
U(N)
are i n (37.24).
ro:
Let a(n)
, where
g ( z ( a ( ” ’ ) ) , a ( ” ’ ) )E U ( N 1 )
.
.
The curve
T,
Note t h a t
and g ( z ( a ( t , ) ) , a ( t , ) ) is i n U(N(t2)). must go o u t s i d e of U ( N ) mis con-
g ( z ( a ( t ) ) , a ( t ) )E U ( N ( t ) )
Hence s e c t o r (37.16) cannot contain any ficiently large.
.
(t, 2
In the case
(40.24)
or L =y
, we
..
k = 0,1,. ,p-1,
Theref ore,
-
So,
, we
...,SPm1
L =y + L =y - 1
.
have
.
2x
or
g(x) = e
have as
and
gk, a r e
( 4 0- 2 5 )
=m
-2x
g(x)+ck
and the s e c t o r s
2
g(x) = e
either
{
L =y
y1 = 0 and y
In p a r t i c u l a r , i f
where
{
either
(40.22)
%,
x-*m i n co,
..., cP-1
are p o i n t s on t h e Riemann sphere.
t h e same a s w e defined i n Section 7 with cover x-plane completely.
ck E I y l ,
and t h e r e e x i s t , f o r each
-.,Y,I j
-
(k = 0 ,
, exactly
n
-
*
Pp-1)
.
Furthermore, 9
distinct indices
j
m=p- 2
k
such t h a t
k' - Y j ' W e s h a l l prove t h e s u f f i c i e n c y of conditions (i)and (ii)of Theorem
39.1 by constructing a s u i t a b l e covering s u r f a c e 62 over t h e Riemann sphere. W e s h a l l explain such a method by u t i l i z i n g a very simple case. The proof f o r the g e n e r a l case i s similar, and will be l e f t t o t h e readers.*
a. Construction
.
of a coverinp s u r f a c e R
In t h i s s e c t i o n and s e c t i o n s
&, 45, and 46, we s h a l l consider the case when (41J)
m=3
and
(41.2)
i
c0 =
o,
cl=l, c2=1+i, c3=i, c,+ =1+i
,
and w e s h a l l prove t h e e x i s t e n c e o f a s o l u t i o n of system (39.11). t h a t the q u a n t i t i e s
co, cl,
c2, c3
(ii)of Theorem 39.1.
*
see, for example, G. a f v i n g [ 81.
and
c
4
Note
s a t i s f y conditions (i)and
177
A COVERING SURFACE
Let us consider a pentagon whose v e r t i c e s a r e denoted by A3
and
A
4 '
{Ao,A1,%,A3,A4] (41.3) where
(See Fig. 4 1 . 1 . )
A o , A1,
A2,
from t h e s e t
=%
(k=0,1,2,3,4)
a r e given by ( 4 1 . 2 ) . In this s e c t i o n , we s h a l l construct a
ck
covering s u r f a c e 62,
over t h e Riemann sphere such t h a t
i s a simply connected and open Riemann surface;
(ii) Ro A.
c(A)
i n t o t h e Riemann sphere by c(%)
(i) Ro
Define a mapping
contains a domain which is homeomorphic t o t h e pentagon
%%SA4
i n such a way t h a t each v e r t e x
mic branch-point of
Ro
over t h e base p o i n t
%
corresponds t o a l o g a r i t h c ( %)
, and
t h a t t h e orien-
t a t i o n i s preserved; (iii) Ro admits no s i n a u l a r p o i n t s o t h e r than the f i v e loearithmic branch-points described i n (ii)
.
Flg.
41.1.
There a r e f o u r d i s t i n c t p o i n t s (41.4)
0 ,
1 , l+i,i
.
Consider t h e square on the FEemann sphere {co,c1,c2,c3,c4] whose v e r t i c e s are a t t h e f o u r p o i n t s (41.4). (See Fig. 4 1 . 2 . )
i n the set
Fig. 0.2.
A G m R A L BOUNDARY VALUE PROBLEM
178
This square d i v i d e s t h e Riemann sphere i n t o two domains D1 assume t h a t t h e p o i n t a t i n f i n i t y belongs t o
D2
.
triangle
AoA3A4
'I"nis s u b d i v i s i o n y i e l d s a t r a p e z o i d
.
The f o u r v e r t i c e s
Ao,
A3
A,,
A3
and
,
3'
of t h e t r a p e -
Similarly
of t h e t r i a n g l e a r e l o c a t e d on t h e
A
4
boundary i n t h e counter-clockwise s e n s e .
, c(%)
A
and a
zoid a r e l o c a t e d on t h e boundary i n t h e counter-clockwise s e n s e . the three vertices
to
A,
AoA1A2A3
and
A1,
We
(See F i g . 41.2.)
Let us subdivide t h e pentagon by t h e line-segment j o i n i n g (See n g . 41.1.)
.
D2
and
On t h e o t h e r hand, t h e f o u r
,
c(A ) = i a r e t h e ver3 t i c e s of t h e square on t h e Riemann s p h e r e which i s given by F i g . 41.2.
points
c(A1) = O
=1
c(AZ) = l + i and
These f o u r p o i n t s are l o c a t e d on t h e boundary o f clockwise s e n s e . c ( k ) = 1t i
4
Similarly, the three points
a r e l o c a t e d on t h e boundary of
D1 i n t h e counter-
c(Ao) = O
,
c(A3) = i
, and
i n t h e counter-clockwise
D2
This o b s e r v a t i o n shows t h a t mapping (41.3) can be extended t o a
senee.
homeomorphism from t h e pentagon onto t h e Riemann sphere c u t a l o n g t h e a r c c o n s i s t i n g o f t h r e e line-segments 1
,
lfi
SO
and
lti
to
i
L1
, L2
L3
which j o i n
Since
LyLyL3
and
, respectively.
0
to
1
,
i s a c u t on
t h e Riernann sphere, each of t h e s e t h r e e line-segments r e p r e s e n t s two l i n e segments. where
D2
i
For example,
1 9 1
.
L1
r e p r e s e n t s two line-segments
, and , L2,2 , and
i s on t h e boundary o f
Similarly, we define
41.3.)
.
L2,1
D1
L
192 L 311
L 191 and L 1 , 2 ' i s on t h e boundary o f
'
L3,2
.
(See F i g .
l+i
2 . 3 '
I.
4.2
1
F i g . 41.3. Let u s denote by the vertex
A. to J The homeomorphism from t h e pentagon onto t h e Riemann
4, .
s p h e r e , c u t along
(41.5)
L"L'L
Aj-%
t h e line-segment j o i n i n g t h e v e r t e x
, maps
t h e l i n e segments
1 2 3 - - $'L1 ' 4%' A2A3 ' A3A4'
and
ALAo
A COVERING SURFACE
179
e i t h e r onto
L1,l’ L2,1
(41.6)
’
U
L3,1
L3,2
9
9
a d
L2 , 2
’
L1,2
r e s p e c t i v e l y , o r onto
(41.7)
’
5’2
’ L3,2 ’
L2,2
L3,1
’ and
’
L2,1‘%,1
In order t o preserve t h e o r i e n t a t i o n , we must choose ( i J . 6 )
respectively.
a s t h e image of the f i v e line-segments (41.5).
The end-points of these
five arcs ( u . 6 ) are
(41 -8)
0
The p o i n t
D1
, 1 , lti , i , and
l+i appears twice.
, while
l+i
.
The first p o i n t
lii i s on t h e boundary of l+i i s on the o t h e r s i d e of t h e c u t . Thus
t h e second p o i n t
these two p o i n t s w i l l eventually represent two d i s t i n c t p o i n t s over t h e
same base p o i n t
l+i
.
The next s t e p i s t o a t t a c h a s u i t a b l e domain along each of the f i v e a r c s (41.6) so t h a t t h e f i v e p o i n t s (iJ.8) points.
become logarithmic branch-
For example,
Such s u i t a b l e domains a r e c a l l e d logarithmic-ends.
consider t h e a r c
.
L2,yL1,2
I n order t o c o n s t r u c t a logarithmic-end
along this a r c , we s h a l l u s e countably many copies of the Riemann sphere which is c u t along
.
L2v5
We denote by
S
( j =1,2,.
j of t h e Riemann sphere each of which i s c u t along an a r c
is a copy of
LyL1
which a r e copies of two a r c s
‘j ’2
4
j
and
.
S
The surface
, where
4,
j ’2 i s on t h e boundary of
D
j
j’2
4,
j
‘
these copies
.
Lj has two domains D
, respectively.
D1 and D2
..) j
The c u t
C
j i s on t h e boundary of
(See Fig. 41.4.)
The a r c and
D
&
j
j,2
represents D
j
.
, and
ir-----
0
I , Fig. 41.4.
S1 t o L2,yL1,2 by i d e n t i f y i n g C 191 with L z , y a l s o a t t a c h Sj+l t o S by i d e n t i f y i n g Cj+l,l with &tJ,2 9
W e attach
j
4,2. We
A GENERAL BOUNDARY VALUE PROBLESI
180
where
j =1,2,
L2,2 "L 1,2
.
....
I n this manner, w e can a t t a c h a logarithmic-end along
This process i s t o p o l o g i c a l l y equivalent t o a t t a c h i n g an open
semi-disc t o t h e pentagon
-
along
Ao4%A3A4
A4Ao
.
(See Fig. 41.5.)
Fig. 41.5. I n order t o c o n s t r u c t a s u i t a b l e L1,l * w e s h a l l again consider countably many logarithmic-end along this a r c , Let us next consider t h e arc
- ....
.
We denote by 3 ( j =1,2, ) L, j these copies of the Riemann sphere, each o f which i s c u t along an a r c .t
copies o f the Riemann sphere, c u t along The a r c
-
l:
and
i s a copy of t h e a r c
J-
, where
4
j' ,1 j ,2 i s on t h e lower s i d e o f
L, I
.
,j ,1
t,
j
.
The c u t
.c",
j
r e p r e s e n t s two a r c s
J
i s on t h e upper siae of
-
j
(See Mg. a . 6 . )
, while
j ,2
N
Fig. 0 . 6 .
we j o i n
9,
to
by i d e n t i f y i n g
by i d e n t i f y i n g i to S j+l,2 'j+l j w e can a t t a c h a logarithmic-end along
Ao% .
.
d
with L,l 1,2_ and 4, ,1 , where
%,1
.
We a l s o a t t a c h
j = 1,2,
. ...
Thus
This process i s again topo-
l o g i c a l l y equivalent t o a t t a c h i n g an open semi-disc t o t h e pentagon
Ao%%A3A4
along
(See Fig. 41.5.) S i m i l a r l y , w e can c o n s t r u c t a
logarithmic-end along each o f t h r e e o t h e r a r c s of ( a . 6 ) .
Thus w e
REMARKS construct a covering surface Ro
181
which s a t i s f i e s conditions ( i ) ,(ii) and
(iii).
42.
Remarks on non-uniaueness and s i n w l a r i t i e s of s o l u t i o n s .
I n the
preceding s e c t i o n , we constructed a covering surface sphere s o t h a t
Ro
over the Riemann Ro s a t i s f i e s the t h r e e conditions ( i ) ,(ii) and (iii)
.
I n this s e c t i o n , we s h a l l present some examples which will show t h a t such covering surfaces are not unique.
Example I:
which is c u t along
(l)"L(l)u L1,l 2,l
S(l) and S ( 2 ) o f the Riemann sphere The c u t on S(') represents two a r c s
Consider two copies
eLyL3
L3,1 and
s e n t s two a r c s
.
(')"
( 2 ) WL(2) +2) 2,1 3 , 1
%,1
L ( l ) and L(2) are copies o f j ,k j ,k (See Fig. 42.1.)
, while
(')"
4,2 L2,2
L
Lg,2
( 2 ) " ~ ( ~( 2) ) ~ 4,2 2,2 L3,2
and j,k
the c u t on
'
where
i
j=1,2,3
I
.
and
S(2)
repre-
The two a r c s k=1,2
.
l+i
Fig. 42.1.
.
(2)uL(2)u (2) Thus we have constructed a covering s u r f a c e Q over 2 , l L3,1 the Riemann sphere. There e x i s t s a homeomorphism from t h e pentagon
'1,l
Ao%A2A
A
34
onto
(42.1)
such t h a t t h e images of f i v e line-segments
- - - ' 0 %
' %% ' 2'3'
' 3'4'
and
Eo 4
are respectively
(42 *2)
Ey a t t a c h i n g s u i t a b l e logarithmic-ends t o Q along t h e f i v e arcs (42.2), we can construct a covering surface 61 which s a t i s f i e s the same conditions
A GENERAL BOUNDARY VALUE PROBLEM
182
(I), (ii)and (iii)a s t h e s u r f a c e Ro
.
from Ro EkamDle A ~ .A
11:
does.
L e t us subdivide t h e pentagon
However,
Ao4A2A3A4
R
is d i f f e r e n t
by the line-segment
(See ~ F i g . 42.2.) A0
Fig. 42.2. Consider two copies
S"(l)and
of t h e Riemann sphere, where
s(2)
i s c u t along an a r c which is a copy o f
an a r c which i s a copy of
of
L
j ,k
,
respectively.
q L y L 3
.
LyL2
, while s"(2)
The c u t on
(See F i g . 42.3.)
+)
i s cut along
s"(l)r e p r e s e n t s two
REMARKS
183
There e x i s t s a homeomorphism from t h e t r i a n g l e
onto
A,%%
such
t h a t the images of the t h r e e line-segments
-
- -
(42.3)
’
AoAl
!LA2
and
%Ao
are, respectively, (42.4)
There also e x i s t s a homeomorphism from t h e trapezoid
onto
AoA2A3A4
4 2 ) S
such t h a t t h e images o f t h e f o u r line-segments
, A2A3 , A3A4
AoA2
(42.5)
and
A A
40
are, respectively, (42.6)
Let us a t t a c h -(2)“-(2) L2,1
4,l
-S( 2 )
to
by i d e n t i f y i n g
-(l) 5 .2
with 2.2
I n this way, we construct a covering s u r f a c e
B“
over the
There e x i s t s a homeomorphism from t h e pentagon
Riemann sphere.
Ao%A2A 3A4
such t h a t the images of t h e f i v e line-segments ( 4 2 . 1 ) a r e , respec-
onto tively, (42.7)
.-
By a t t a c h i n g s u i t a b l e logarithmic-ends t o
B along t h e f i v e a r c s ( 4 2 . 7 ) ’
we can c o n s t r u c t a covering s u r f a c e which s a t i s f i e s the same conditions (i), (ii)and (iii)a s t h e s u r f a c e Ro
, but
i s d i f f e r e n t from R o I n Section 39, we i n v e s t i g a t e d the cases m = l and m = 2
39.1.
I n p a r t i c u l a r , i n case
m= 2
equation (39.27)
If we regard
i
a
1- i t a n [1 - n ( l - az) ] 1
rxc3. , equations
.
o f Theorem
(39.11) were reduced t o the
1+ i tan[-n(l- a2) 1 4
the points = O and c = a a r e 3 ’ c3 3 logarithmic branch-points of this f u n c t i o n . This means t h a t equations 2
a s a function of
c
(39.11) may have i n f i n i t e l y many s o l u t i o n s f o r f i x e d values of The non-uniqueness o f s o l u t i o n s of (3.11) i s c l e a r l y
C0’C1, * .,c mtl r e l a t e d with the non-uniqueness of covering s u r f a c e s t h a t w e have shown
by Examples I and 11.
The function a2 of t h e q u a n t i t y However, the condition
c
3
is holomorphic a t
c =I 3
.
A
12%
GENERAL BOUNDARY VALUE PROBLEM
y12- T2Tllf 0
(39.9) i s violated i f
Z ( 1 - t a2) =0,-1,-2
I
,... ,
1
,... .
OI’
- ( 3 + a2) =O,-1,-2
4
Hence t h e r e a r e c e r t a i n s i n g u l a r i t i e s of s o l u t i o n s o f (39.11)a t Let us consider the case when m = 3
,
c0 = o
(42 .e;j
cl=l
,
, and
c2=l+i
,
c3=€i
,
c 3 =1
.
c4=l+i
uhere
(42.9)
.
O 0
m M lA 5LL:
1 1 .1
. a * +
bI,
IRo
9
i s s u f f i c i e n t l y small.
a,b,c,a,p,y,t,T,
none o f them i s zero.
&, if
g&
c
be complex numbers.
Assume t h a t
PROBLEM 49.1 AT a = 0
205
i t follows t h a t
(51.5)
a = a ,
Proof:
b=!,
.
c=y
A straight-forward computation shows t h a t
q, and c a r e a r b i t r a r y complex q u a n t i t i e s . Therefore, (51.5) follows from (51.4). This completes the proof of Lemma 51.1. a , b, c , 5 ,
where
LEMMA 51.2: according a s
(49.9).
rk
Let the matrices
m i s odd o r even.
Assume t h a t the
rk
be defined by e i t h e r (49.13) Let the matrices
Sk(a)
or
be defined by
m, if
s a t i s f y condition (49.14).
(49.17) a
i n domain ( 5 1 . 3 ) , the system of equations
(51.6)
%(a)
=rk
(k=O,l,
S,(a)
=rk
(k=O,
...,m+l)
i s equivalent t o (51.7)
...,m-2) .
It i s s u f f i c i e n t t o show t h a t (51.6) follows from (51.7).
Proof:
t h e matrices
%(a)
Since
s a t i s f y r e l a t i o n (49 .lo), we derive
rmtlrmrm-l = s,,(a)sm(a)sm-,(a)
(51.8)
from (49.10), (49.14) and ( 5 1 . 7 ) .
( 51 -9)
By v i r t u e of Lemma 51.1, i t follows t h a t r m = S m ( a ) , rmtl=Smt,(a)
rm-l =sm-l(a)
.
This completes the proof of Lemma 51.2. LEUMA
Assume t h a t
complex v a r i a b l e s . tions
a,(y)
(i)
a,(y)
each
(51.11)
i s odd.
&&
yo
such t h a t
i s holomorRhic f o r Iyj-(l+w)J= U ( W
-k
x,Ck ( a ) )
.
( C f . (21.26).)
Hence
k -k k I \ ( u ( x , a ) ) = C ( G ( u ( x , a ) ) ) = C(U(W x,G ( a ) ) )
(52.20)
= exp{ZEm(w-kx,Gk(a)) }C(Gk(a)) k = e x p { ( - l ) 2Em(x,a)}Ck(a)
.
In t h i s computation, we used formulas ( 2 1 . 7 ) , ( 2 1 . 2 3 ) , and (ii)of Theorem 21.1.
Therefore, we o b t a i n (52.17) f o r
(52.20).
As t o
fo(a)
bl
(52.21)
Zm+l if C
,
from (52.6) and
observe t h a t
m
1
( a ) = - J1 [xm+ 2 %x m-k 2ni C k=l
15 dx
i s a simple closed curve such t h a t a l l r o o t s o f t h e polynomial a r e
contained i n i t s i n t e r i o r . ( 5 2 -22)
bl
v(a) = b
(a)
Hence
(u) =bl
F+l Since
...,m-2
j =1,
(a)
P1
.
( c f . ( 2 1 . 1 2 ) ) , w e o b t a i n (52.17) f o r
j =O
$1 completes t h e proof of Lemma 52.2. Let u s d e r i v e (52.1) from ( 5 2 . 2 ) .
To do this, assume t h a t
.
This
PROBLEM 49.1: THE GENERAL CASE
Yo,...,Ym,2
yj # 0
.
Then, i f
u
and
a r e i n domain (52.4), and t h a t
...,um(x,a)
Let u l ( x , a ) , a"
211
s > 0 i s so small t h a t
denote the q u a n t i t i e s defined by (52.16).
s a t i s f i e s (52.2), i t follows from Lemma 52.2 t h a t
5 i s an a r b i t r a r y complex number.
a l s o s a t i s f i e s ( 5 2 . 2 ) , where
u(5,E) Observe
that
Co(u(5,s)) =exp{2Em(5,E)}Co(g)
(52.23) Since Em(Z,8)
i s a polynomial i n
5
from zero, we can f i n d
5
, and
.
Co(g)
and yo a r e d i f f e r e n t
so that
(52.24.)
c o ( u ( 5 , a )=yo
This means t h a t
a=u(s,g)
.
satisfies (52.1).
Thus we proved the following theorem.
THEOREM 52.1:
Assume t h a t
p o s i t i v e number
E
m24 and m i s even.
such t h a t ,
m, t h e r e e x i s t s
a
if
and matrices (49.17) s a t i s f y condition (49.l4), Problem L9.2 has a s o l u t i m . 53.
Problem L9.1:
I n this s e c t i o n , we s h a l l show t h a t
t h e general case.
condition (49.14) i s s u f f i c i e n t f o r the existence of a s o l u t i o n of Problem 49.1. LEMMA 53.1:
Assume t h a t
s a t i s f y condition (49.14).
m
m23
i s odd, and t h a t matrices (49.13)
Set
,
c 0 =m
cl=o, (53.2) U
k-2
(k=2,
C k = z
Then
co,
...,c
...,m t l ) .
a r e well-defined points on t h e Riemann sphere, and they
s a t i s f y the following two conditions.
(i) c ~ + ~ # (ck =~O ,
...,mtl;
cm t 2-- c 0 ) i
A PRESCRIBED
212
...,cI U t l 3
(ii) t h e s e t (co,
STOKES PHENOMENON
c o n t a i n s a t l e a s t t h r e e d i s t i n c t p o i n t s on t h e
Riemanr, s p h e r e . Proof:
Note t h a t
L_
.
d e t H,#O
This means t h a t
defined p o i n t s on t h e Riemann s p h e r e . follows from p r o p e r t y
(53.3)
'ktl
-2
#O
.
For
.
(k=2,
.
Hence
are well-
m i s odd, p r o p e r t y
(ii)
( i ) , w e s h a l l prove t h a t
To prove
# 'k
Ho = T o
Observe f i r s t t h a t
-
(i)
Since
. . .,c &l
co,
T~
...,m) ,
=yo
and
k>_O , we have
a =1 0
.
This shows t h a t
%+l = r k t l % ' m d heme
' k t l ='k+l7kt
',"herePore,
" k t l ''ktlak
-t % '
-' k t l
-
7
k
0
0
ktl
'%is shows t h a t
'
'k
'k
b
k
c k t l # ~ for
= -det H , # O
,
'k
k=2,
...,m .
Condition (49.14) can be
vrifter: as
Hence
This shows t h a t
T
~
=-- w -~l #
0
.
Therefore,
c f mtl
a
.
This completes t h e
proof of Lemma 53.1.
moRn*!5 3 . 1 :
Assume t h a t
.
s a t i s f y c o n d i t i o n (49 .I&) Proof:
m i s odd, and t h a t m a t r i c e s (49.13) Then Problem L9.1 h a s a s o l u t i o n .
m>_3
By v i r t u e of Lemma 53.1, we can u s e Theorem 39.1 t o f i n d
p l e x numbers
al,
...,am
such t h a t t h e d i f f e r e n t i a l e q u a t i o n
m y " - [ x mt
z "kx
k=l
m-k ly=O
m
com-
PROBLEM 49.1:
THE GENERAL CASE
has two linearly independent solutions yl(x) the boundary conditions
213
and y,(x)
which satisfy
(53.5) Since c0 = a and c1 = O
, we must have
(53.6) where 1 is a non-zero complex number. By utilizing the matrices %(a) defined by (@.lo), we can write the connection formulas (49.1) as
[ld m,k (X,a),bm,k+l(x,a) 1 = [km,k+l(x’a)9bm,k+2(x9a) lsk(a) (k=O,
...,m + l ) .
Set
...,m-1) .
(k=O, Then
and hence
Therefore, (53.8) Set
(53.9) We shall Drove that if k is even, if k is odd, where (53.11)
k=O,
...,m-1 .
214
A
For
k =0
PRESCRIBED STOKES PHENOMENON
, (53.10)
becomes
Also, i t f o l l o w s from (53.1) and (53.7) t h a t
,
t o ( a ) =Co!a) 'To-yo
,
Hence, (53.8) i m p l i e s t h a t
Cl
0
s o ( a ) =1
=1
.
Assume t h a t ( 5 3 .lo) i s true f o r
(53.13)
i
$(a) =
*-'%A
A%A
Let us consider t h e case when k=h+l
, we
.
Co(a) = X Y o
h
This proves ( 5 3 . 1 2 ) .
k = 0,.
. ., h .
Then
if
h
i s even,
if
h
i s odd.
i s even.
If we s p e c i a l i z e (53.10) f o r
get
W e s h a l l prove t h a t (53.14) i s t r u e .
To do this, observe t h a t (53.13)
Since
'h+l =Yh+lThs 'h
'
'htl ='h+laht
'
Therefore, w e d e r i v e
'h
ar++l(a) =yhtl
This proves ( 5 3 . y ) .
from (53.8).
S i m i l a r l y , we g e t
-1 X Chtl(a)
Note t h a t ='yh+l
if
det Hh#O h
.
i s odd.
This completes t h e proof o f (53.10) f o r (53.11). The m a t r i c e s
Sk(a)
f o r (53.11), and s i n c e
satisfy condition
(49.10). S i n c e (53.10) i s true
m - 1 i s even, i d e n t i t y (49.10) becomes
PROBLEM 49.2:
THE GENERAL
215
CASE
.
S&1(a)Sm(a)AHm-1A=12
On the other hand, (49.14) can be written as
=2'
'm+l'mHm-l Hence
Aswl(a)Sm(a)A
*
=relrm .
This implies that
(53.16)
=1
,
c,Ca) =ym
,
.
~ ~ + ~=ydl ( a )
Theorem 53.1 follows from (53.10) and (53.16).
54.
Problem L9.2:
the general case,
method o f solving Problem 49.2.
For
m 2 4 , we have not
found any
Instead, w e s h a l l prove the following
theorem.
THEOFEM 5L.1:
(i)
m24
(ii) yo,
Assume t h a t
m is even; 'ywl , g& p are given complex numbers;
...
(iii) p f ~; (iv) yo#o ; ( v ) matrices (49.17) s a t i s f y condition (49.14)
Then, -
m complex numbers a l , .
there e x l s t
(54 -1)
If
p =0
..,a m
and an i n t e g e r
if
k
is even,
if
k
is odd,
p
(k=O,
f o r some
j
p=O
can be reduced t o the case
.
such
...,mtl) .
then Theorem 54.1 would give a solution o f Problem 49.2.
ever, we s t i l l do not know how t o prove y #O j
.
How-
The case when yo=O but y,#O
To prove Theorem 54.1, l e t us define matrices
Po,
easily.
...,r&l r-
by
A PRESCRIBED STOKES PHENOMENON
216
where
(54.3)
if
k
i s even,
if
k
i s odd,
(k=O,
?k =
...,m) ,
MRtrices ( 5 4 . 2 ) s a t i s f y t h e condition
(54.5)
-
--
rmlfm...rlro = I . ~.
In f a c t , i f we set
’
OP 1
if
k
i s even,
if
k
i s odd,
.
(k = 0 , . ,,m-1)
and
7
I d e n t i t y ( 5 4 . 5 ) follows from (49.L$),(54.7) and ( 5 4 . 8 ) .
Note t h a t
(54.9)
0
=1
.
Hence
THE GENERAL CASE
PROBLEM 49.2:
c Then
0
= a ,
cl=O,
Ok-2 c ='k-2
217
(k = 2 , .
..,m+l)
are well-defined p o i n t s on t h e Riemann sphere, and they
c0,...,cmtl
s a t i s f y t h e condition (54.12)
%+l#Ck
(k=O,
...,W l ;
Cw2=C0)
( C f . the proof of Lemma 53.1.) Furthermore, s i n c e c2 = 1, t h e set {co c ] contains t h r e e d i s t i n c t p o i n t s , 0 , and 1 Hence mtl by v i r t u e of Theorem 39.1, t h e r e e x i s t m complex numbers a l , . .,a and
,...,
. .
two l i n e a r l y Independent s o l u t i o n s
y,(x)
and
y2(x)
m
of the d i f f e r e n t i a l
equation
m m-k yrt- [xm+ c EkX ]y=o k=l
(54.13) such t h a t
Since
c == 0
,
cl=O
where t h e matrices
and
c2=1
, we
must have
Sk a r e defined by (49.9).
o f subdominant s o l u t i o n s by
Let us define another s e t
A PRESCRIBED STOKES PHENOMENON
218
where
f 5.4 .la)
p( a) = m-zv'a)
.
Let
-
(54.19) be
(k=O,
[~k(x),yki~+l(~)]=[~k+l(~) ,?k+2(x)lsk(')
t h e 2onnection formulas f o r t h e subdominant s o l u t i o n s
where
Fmi2 ( x ) =y,(x)
and
7e
3
(x) = y l ( x )
.
?,(x),
-. ...,Y~+~(X),
Then
-
(54.20)
sm (a") =
uhere
(k =0,.
..,m)
and
Then
(54.25) Hence
q-- as
F(x) + ' - 2 -2
x+-
in
ak
(k=2,
...,m + l ) .
219
PROBLEM 49.2: THE GENERAL CASE
Bk = % tk 'k
(54.26 1 Note t h a t %
(54.27)
.
(k=O,
...,m-1) .
- .
.
so(;) =To
Therefore, from (54.27) and (54.26), we can derive
..
,.
s,(a) =rk
(54.28)
(k = 0 ,
i n t h e same way a s i n Section 53.
..,,m-1)
Since t h e matrices
fk
s a t i s f y condi-
sk(Z) s a t i s f y t h e condition .smtl . ( z ) ~ m ( z... ) So(a) =12 ,
t i o n ( 5 4 . 5 ) , and the matrices
we have
A straight-forward computation shows t h a t
-
-
Sm(Z) =rm
.-. smtl(:)
and
..
.
Hence
I 7+1
1 2"+1 B(Z)
=p
(54.29) Q Let
p
9
.
=yk
.
(k = 0 , . ,,m+l)
be an unknown quantity, and define
s
m m-I. x +alx
(54 -30)
+
...+ am-lx
al
,...,am
by
+ am
...+ grn-l ( x + s ) + Ern . a r e polynomials i n s . By v i r t u e o f
= ( x + s p + ZJx+ s ) m - l + The q u a n t i t i e s
al,
...,a m
Lemma 52.2,
we have
(54.31) (k=O,
...,m)
,
and 1
7n
2
(54.32)
Q ~ ( E ) = B(a)
Co(a)Cmtl(a)
.
Then Theorem 54.1 follows from (54.3) , ( 5 4 . 4 ) , (54.29), (54.31), and (54.32). This completes t h e Proof of Tfmxw
Determine
54 .I.
s
so that
Co(a) = y o .
A PRESCRIBED
220
55. of
STOKES PHENOMENON A set
Subdominant s o l u t i o n s a d m i t t i n g a p r e s c r i b e d Stokes phenomenon.
m + 2 n o n t r i v i a l s o l u t i o n s yo,
...,ymtl
of d i f f e r e n t i a l e q u a t i o n (6 .l)
i s c a l l e d a complete s e t o f subdominant s o l u t i o n s of (6.1) ? i f a r e sub:iominant i n
Note ?hat
m+ 2
-
So,
- go,. ..,Sm C l , r e s p e c t i v e l y .
. ..,Swl
solutions
.
y o , . .,ymtl
( C f . D e f i n i t i o n 7.1.)
cover x-plane completely.
For example, t h e s e t o f
i s a complete s e t o f subdominant solu'dm,o, * * ' " b , m t l Fbrthermore, {yo, .,yrtttl} i s a complete set o f subdomi-
..
tions of (6.1).
nant s o l u t i o n s of (6.1) i f and o n l y i f
.
Y,(x) = a , ~ + ~ , ~ ( x , a )( k = 0 , . .,el)
(55.1) khere t h e
4(
are non-zero c o n s t a n t s .
,
If a complete s e t o f subdominant
s o l u t i o n s of (6.1)i s given by ( 5 5 . 1 ) , t h e s e
m+ 2
subdominant s o l u t i o n s
admit t h e connection formulas (55.2)
( k = O , ...,mtl) where i55.3) and
(55.4) Set
and
(55.5)
$=[;
'01
(k=O,
...,ni-tl) .
Then t h e connection formulas ( 5 5 . 2 ) become
Fhthermore, observe t h a t
where (55.8) and t h e m a t r i c e s
(k=O, Sk(a)
...,rtttl)
a r e given by (49.9).
,
S i n c e m a t r i c e s (49.9)
?
221
A PRESCRIBED STOKES PHENOMENON
s a t i s f y i d e n t i t y (49.10), and
( c f . ( 5 5 . 4 ) ) , matrices (55.5)
Tk+2= T0
s a t i s f y t h e condition
... s
SmflSm
(55.9)
0
.
=12
Conversely, l e t us consider m + 2 matrices
;3
4=[;
(k=O,
...,mtl)
such t h a t
(55.11) In this s e c t i o n w e s h a l l prove the following theorem.
mom
5 5 . 1:
(k = 0 , .
qk r
'k
..,m+1)
be d v e n complex numbers s a t i s -
fying t h e conditions t h a t
(i) matrices (55 .lo) s a t i s f y condition (55 .u); f o r some k if. m i s even. (ii) q k # O
,
Then we can f i n d m
complex numbers
esuation (6 .l)has a comDlete set
al,
.
...,am
so t h a t t h e d i f f e r e n t i a l
{yo,. . , Y & ~ }
of subdominant s o l u t i o n s
which admit the connection formulas
(55.12)
[Yk(x) ,yk+l(x) 1 = [yk+l(x) ,yk+2(x)
where
and
Ymt2=Yo
Proof:
-$(a)
= --w
9
*
*
,*l)
0
W e s h a l l f i r s t consider t h e case when m i s odd.
(49.3) .)
(Cf.
Ye3=Y1
(k = '
(k=O,
I n this case,
...,m f l ) .
Therefore,
Also, condition (55.11) implies t h a t mtl (55.13) II r k = - l , k=O Let u s d e f i n e m + 2
(55
complex numbers
ao,
.u)
where
(55.15)
amf2=ao
,
amt3=a1
.
...,a m t l
by the condition
9
A PRESCRIBED STOKES PHENOMENON
222
'70 do t h i s , s e t a I
0
=1
, ( h = 1,
2(h-1)
0
.
in
, the
solution
Hence, we c a l l
f
k
tends t o zero as
fk
x
tends t o i n f i n i t y
a subdominant s o l u t i o n i n t h e s e c t o r
%'
Note t h a t -k -k fk(X,A)=bm(Pwx , a ( p Y ) )
(57.22)
*
In this c h a p t e r , w e shall explain how t o f i n d asymptotic behavior of t h e s e subdominant s o l u t i o n s as variable
58.
x
1 tends t o i n f i n i t y i n s t r i p (57.2).
The
will be r e s t r i c t e d t o a s u i t a b l e domain.
Formal s o l u t i o n s of the a s s o c i a t e d R i c c a t i equation.
L e t us consider
a d i f ' f e r e n t i a l equation 2 y"-x p(x)y=O
,
i s a polynomial i n
x
58.1) where
p(x)
, and
1 is a complex parameter.
In
Sections 58, 59, and 60 we s h a l l d e r i v e asymptotic r e p r e s e n t a t i o n s of solut i o n s of (58.1) with r e s p e c t t o
a
and
x
.
I n such a problem i t i s help-
-I
.
f u l t o r e p r e s e n t s o l u t i o n s i n terms of I P ( x ) ~ Furthermore, a represent a t i o n o f y l / y is derived more e a s i l y than t h a t o f y i t s e l f . Set
(58.2)
u=y'/y
.
Then
(58.3)
U'+U
2
- a 2p ( x ) = o .
This i s a M c c a t i equation. The m .
M c c a t i equation (58.3) has two formal s o l u t i o n s
FORMAL SOLUTIONS
237
(58.6) -n-1
(58.7)
Pn(x, = p ( x )
and t h e a u a n t i t x coefficients.
Proof:
and
g(x)
%(XI
9
is a polynomial i n p , p l , ...,p (n+l) u i t h r a t i o n a l
Let us consider the formal series (58.4). Observe t h a t
238
SUBDOMINANT SOLUTIONS
This means t h a t i f we d e f i n e
Pn(x)
by
-
Qj(x)Qh(x)I
(n>_z)
j th=n
:'his completes the proof of Lemma 58.1. respect t o
where
deg
g(x) denotes
"heref'ore, the q u a n t i t i e s
the degree of Pn(x)
g(x)
with r e s p e c t t o
a r e bounded f o r l a r g e values of -1
So f a r , we have n o t s p e c i f i e d a branch of
(58.4) and ( 5 8 . 5 ) .
function of ( i ) ;Q
Now, l e t us consider
x
in
B
.
(58 .lo)
is a simply connected domain i n x-plane;
Ip(x)(>,B>o
for
8 such t h a t X E O
.
-1 Let us choose a branch of
p(x)'
i n the domain, and assume t h a t
-1 (58.11)
l a r g ~ x I M) ~ ~for~
XE
B
. x
.
~ ( x as ) ~a holonorphic
P r e c i s e l y speaking, w e assume t h a t
( i i ) t h e r e e x i s t s a p o s i t i v e number
x
P ( x ) ~ i n t h e formal s e r i e s i n a simply connected -1
p(x)
s o t h a t we can s p e c i f y a branch of
domain
with
,
deg g ( x ) I (m-1) (n+l)
(58.9)
p(x)
, then
m
is
x
If the degree of
.
,
239
FORMAL SOLUTIONS
M is
where
a p o s i t i v e number.
p o s i t i v e numbers (58.12)
Kn
.
Under assumption (58.10), t h e r e e x i s t
( n = 0,1,. .)
such t h a t
~ p ( x ) - ~ p ~ ( Lx ) K l
~for
x~
5Kn
xEie
and
a
and
(58.13)
IPn(x)l
for
n=1,2,
I n d e r i v i n g (58.13) we used i n e q u a l i t i e s (58.9).
... .
By u t i l i z i n g Theorems
3.3, 4.1 and 5.1 o f Chapter I, we can c o n s t r u c t a f u n c t i o n two complex v a r i a b l e s , (w,x) i n such a way t h a t (i)
g(w,x)
wo
of the
i s holomorphic f o r
(58.14) where
g(w,x)
larg
WI
52M
,
IwI
2wo>
0
,
xE B
,
i s a p o s i t i v e c o n s t a n t , and
(ii) g(w,x)
s a t i s f i e s t h e conditions:
f o r ( 5 8 . u ) and Note t h a t i f
N=1,2, po
...,
where t h e
EN
a r e positive constants.
i s a s u f f i c i e n t l y l a r g e p o s i t i v e number, then
(58.16) for
( 5 8 -17) where
xEB 6,
,
(Im[kl(l6,
,
i s a given p o s i t i v e c o n s t a n t .
R ~ [ X I > _ F J,~ Therefore, i f we s e t
-1
(58 -18)
2 h(x,X) =g(Ap(x) , x )
the f u n c t i o n h(x,A)
(58.17)
.
F'urthermore,
,
i s holomorphic with r e s p e c t t o h(x,X)
(x,k)
s a t i s f i e s t h e conditions:
i n domain
SUBDOMINANT S O L U T I O N S
240
f o r (58.17) and
...,
N=1,2,
where t h e
Notice t h a t
N a r e positive constants. 1 -
-1
-1
hl(x,A) = & ( x ) - l p ~ ( x ) ~ [ a p ( x ) ~ ~ d g ( ~ p ( x ) ~ , x ) /ag(Ap(x)2,x)/ax aw+ and hence the second and t h i r d i n e q u a l i t i e s of (58.15) imply t h e second i n e q u a l i t y of (58.19)
.
Let us s e t
(58.20)
-1 u = X p ( x )2 - p1( x ) - ' p ! ( x )
th(x,A)+u"
.
Then, t h e R i c c a t i equation (58.3) i s transformed t o 1
where (58.22)
-1 H(X,A)
= [hp(x)'- ~ ( x ) - ' p I ( x ) + h ( x , A ) 1 '
+
-
1 [Ap(xI2 - 5 ( x ) - ' p t ( x )
+ h ( x , h ) 12 - A 2P ( x )
-
Since s e r i e s (58.4) is a formal s o l u t i o n o f ( 5 8 . 3 ) , i n e q u a l i t i e s (58.19) imply t h a t
-1 -N
(58.23)
I H(x,A)l I$/ I P ( X ) ~ ~
f o r (58.17) and Set
N=1,2,
(58.uJ
iT=hp(x) v
to obtain (58.25) where
...,
where the
$ a r e c e r t a i n p o s i t i v e constants.
-1 2
-1
-12
v f t 2[Ap(x)'+ h ( x , ) , ) ] v + [ h p ( x )
]V
2
+ L(x,A) = 0
FORMAL SOLUTIONS
-1 -1 ( 58.26 )
.
~ ( x , h=) [xp(x)23 ~ ( x , h )
I n e q u a l i t i e s (58.23) imply (58.27) f o r (58.17).
-1 -N
1 L(x,A) I L CN,ll
.
( N = 2,3,. .)
W(d2(
We s h a l l prove the following lemma.
-2"; v1
F
l t w1t w 2
2w
t
(w1
1 1
t w;)
( 1 t w t w2)
2
1
1
(58.31) i s a s o l u t i o n of t h e R i c c a t i esuation (58.3).
Consequently,
SUBDOMINANT SOLUTIONS
59.
As.mptotic s o l u t i o n s i n a canonical domain as
Iri t h i s s e c t i o n , r e s t r i c t i n g
system (58.28) f o r
t o a s u i t a b l e domain
,
) I m [ h ] )< a 0
tends t o i n f i n i t x .
, where
Re[h]lO
Ig
6,
, we
simply connected, and t h a t
should be
should s a t i s f y an i n e q u a l i t y
p(x)
)p(x)l> _ B > o for
( 58 .lo)
s h a l l study
i s a given posi-
I n S e c t i o n 58, we r e q u i r e d t h a t t h e domain
t i v e number.
xhere
x
A
XELI
,
p i s a p o s i t i v e number.
Furthermore, w e assumed t h a t
M
The domain
{ j8.1; srne r e
i s a p o s i t i v e number.
J3 must be r e s t r i c t e d f u r t h e r
by t h e behavior o f t h e q u a n t i t y
-1
X
Re[ A p ( t ) 2 d t
as
A
tends t o i n f i n i t y .
x = 50 i s c a l l e d a t r a n s i t i o n p o i n t of t h e x = s i s a zero of p ( x ) d i f f e r e n t i a l equation (58.1) of o r d e r k , 0 of m u l t i p l i c i t y k .
m
: A point
j
For example, t h e d i f f e r e n t i a l e q u a t i o n 2 2 y"-x x ( x - l ) y = O
has two t r a n s i t i o n p o i n t s x = O and x = l i s of o r d e r two, wNle t h e t r a n s i t i o n p o i n t
.
The t r a n s i t i o n p o i n t is of o r d e r one.
x=l
x=O A
t r a n s i t i o n p o i n t o f o r d e r one i s c a l l e d a simple t r a n s i t i o n p o i n t .
DEFTNITION 59.2:
A curve
x = S(S)
( 0 5 s < so)
i s c a l l e d a S t o k e s curve
f o r t h e d i f f e r e n t i a l e q u a t i o n (58.1), when (i)
5(s)
or -
t= ;
i s continuous f o r
(ii) s ( 0 ) (iii) FJ(s)
O < s < so
, where
s
0
i s a p o s i t i v e number
i s a t r a n s i t i o n p o i n t of (58.1);
for
{ i v ) we have
s>O
i s n o t a t r a n s i t i o n p o i n t of (58.1);
ASYMPTOTIC S O L U T I O N S
where t h e i n t e g r a t i o n i s taken along t h e curve
213
x = 5 (s)
.
Some examples o f Stokes curves will be given i n Section 61.
D E F I N I T I O N 59 .l: A simply connected domain
0
& I x-plane i s c a l l e d a
canonical domain f o r t h e d i f f e r e n t i a l eauation (58.1), (i) t h e boundam of (ii) t h e i n t e r i o r of
(iii)
n
n c o n s i s t s of Stokes curves; n does not contain t r a n s i t i o n
p o i n t s of (58.1);
i s c o n f o d l g mapped bx
(59.1)
z=I(x)
=sxp ( t )-d t 2
X
0
onto t h e whole z-plane c u t by a finite number of v e r t i c a l s , each of which
i s unbounded, where
xo i s a Doint on t h e boundary o f
n
.
Examples of canonical domains will a l s o be given i n Section 61.
n
B of a canonical domain
c o n s t r u c t a subdomain
so t h a t
B
We s h a l l satisfies
conditions (58.10) and (58.11).
n
Mapping (59.1) take5 t h e domain
conformally onto t h e whole z-plane
c u t by a f i n i t e number of unbounded v e r t i c a l s . c a l c u t s by V
j
L1,
...,4,.
of c e n t r a l angle
(iii) (h# j )
For each c u t
j
i s on t h e b i s e c t o r o f
t h e c e n t r a l angle
6
of
a r e not contained i n
W e assume t h a t
6
V
V V
j ’ j
domain derived by removing
VluV2w.
V
s h a l l construct a s e c t o r
j ’
i s so small t h a t t h e o t h e r c u t s
j -
i s independent of
, we
L j
such t h a t
i s contained i n t h e i n t e r i o r of
(i) L.
J (ii) L
6
Let us denote these v e r t i -
j
.“Vk
.
Let
U
4,
be the simply connected
from z-plane.
(See Fig.
59.1.)
SUBDOMINANT SOLUTIONS
U
Fig. 59.1.
Denote by onto
&
.
IJ
t h e subdomain o f
The domain
0
(58.11) a r e s a t i s f i e d i n I n t h e domain
which i s mapped by (59.1) conformally
i s simply connected and c o n d i t i o n s (58.10) and
B
.
A w e c o n s i d e r t h e two formal s o l u t i o n s ( 5 8 . 4 ) and
(58.5) of t h e R i c c a t i e q u a t i o n ( 5 8 . 3 ) .
We c o n s t r u c t a f u n c t i o n
g(w,x)
which s a t i s f i e s c o n d i t i o n s (58.15) f o r xE B
(59.2) where
w
,
larg[w]l 5 2 M + 2 n
i s a positive constant.
0
, 1.1
Then, i f
Lwo>O po
,
i s a s u f f i c i e n t l y large
-1 p o s i t i v e c o n s t a n t , we can f i x a branch of
(59.3)
for
I
~ ( x )such ~ that
ASYMP"IC
and
SOLUTIONS
245
-12
(58.18')
h(x9-A) =g(-XP(X> , x >
t h e functions
h(x,X)
and
h(x,-X)
9
are holomorphic with r e s p e c t t o
i n domain (58.17). -Finthemore,
h(x,A)
(58.17) and N = 1 , 2 ,
h(x,-X)
I
( 58.19 ' )
f o r (58.17) and
... ,
lh(x,-h)
while
-
s a t i s f i e s conditions (58.19) f o r s a t i s f i e s the conditions
-1 -n
N
X [-XP(X)~]
(x,h)
Pn(x)l < % l h p ( x )
-12 -N-1 1
n=l
9
1
N
1
n=l
N=1,2,.
.. , where the %
a r e p o s i t i v e constants.
We
s h a l l prove t h e following theorem. 59 .k*
If
po
i s a s u f f i c i e n t l y l a r m Dositive number,
the
d i f f e r e n t i a l equation 2 y"-A p ( x ) y = O
(58.1)
admits, i n domain (58.17), two s o l u t i o n s of t h e forms:
-1
{
(59.4) where
F+(x,X)
--1
x
Y =xk,u=P(x)' [1+ F-(x,X)lexp[-AJ
and
-l
x
x
$k,h)lexp[AJ p ( t I 2 d t + J h ( t , A ) d t l
y = y + ( x , A T = ~ ( x'[l )+
,
-l x p(tI2dt+ h(t,-A)dt]
are holomorphic i n domain (58.17)
F-(x,A)
Y
uniformly f o r x E &
Proof:
h
We s h a l l construct
tends t o i n f i n i t y i n the s t r i D
y+(x,A)
constructed i n a similar manner.
.
The o t h e r s o l u t i o n
In order t o construct
-
y (x,A)
F+(x,X)
, we
consider t h e system
*
See M.A.
Evgrafov and M.V.
can be
Fedoryuk [11;$4,Remark 4 . 3 on p . 231.
shall
SUBDOMINANT SOLUTIONS
where
( 58.26 )
1
J.
H(x,h) = [ h p ( ~ ) & ~ (-x ) - l p ' ( x ) t h ( x , h ) ] '
(58.22)
4
-1 2
+ [hp(x)
- a 2P(X)
2
- p1 ( x ) - ' p r ( x )
t h(x,X) 1
9
and hence
-12 -N
I L ( X , X ) 1 IcN-J a p h ) I N = 2 , 3 , . .. , where the
( 5 82 7 ) f o r (58.17) and
are c e r t a i n p o s i t i v e con-
CN-l
stants. We s h a l l reduce system (58.28) t o a system of i n t e g r a l equations by choosing the path of i n t e g r a t i o n i n t h e following manner: point i n in
U
&
.
Then
z = I ( x ) is a point i n
.
U
Let
W e shall join
x
be a
z to
by a curve
(59.7)
c
c(z):
along which
Re[XC(s)]
more p r e c i s e l y , choose
=c(s)
,
c(O)=z
,
O R , t h e
path
C( z)
i s e i t h e r a r a y i n t h e e x t e r i o r of d i s c (59.8), o r a curve c o n s i s t i n g of
247
ASYMPTOTIC SOLUTIONS
three parts: (i) a line-segment joining z t o the boundary of d i s c (59.8), (ii) an a r c contained i n d i s c (59.8) and (iii) a r a y i n t h e e x t e r i o r of d i s c (59.8). (See Mg. 59.3.) I n any case, t h e length of the arc i n disc (59.8) can be bounded by a constant which depends only on t h e domain U , and i s independent of z If C(z) i s a ray i n the e x t e r i o r of (59.8), then C(z) i s given by
.
~(z): c = z + s e ie
(59.7 ' 1
,
,
(O(s - R
.
i s a p o s i t i v e number. Therefore by using the 7 f a c t t h a t t h e length of the p a r t of C ( z ) i n d i s c (59.8) i s bounded by a
where
and
c
constant independent o f
f o r a p o s i t i v e number
z
in
, we
a
U
, and
can prove (59.14).
By u t i l i z i n g Lemma 59.1, w e can e a s i l y con-
proved i n a s i m i l a r manner.
s t r u c t a bounded s o l u t i o n w,(x,X) (58.17) i f
po
and
i s sufficiently large.
p o s i t i v e constant
c
w2(x,A)
of (59.12) i n domain
Furthermore, we can a l s o find a
such t h a t
8
Then (59.15) implies t h a t
We s h a l l leave the d e t a i l s t o t h e r e a d e r s . wl(x,h) s o uniformly f o r
Estimate (59.15) can be
x E p as A
,
w,(x,x)
so
tends t o i n f i n i t y i n s t r i p (59.6).
complete the proof of Theorem 59.1 f o r
F+(x,A)= w,(x,x)
,
y+(x,X)
I n order t o
set
+ w2(x,i)
and use Lemma 58.3.
60.
As.wlDtotic s o l u t i o n s i n a canonical domain a s
I n Section 59, w e constructed a subdomain t h a t t h e r e exist two s o l u t i o n s y+(x,h)
&I
and
s a t i s f y t h e conditions given i n Theorem 59.1. i n v e s t i g a t e the asymptotic behavior o f tends t o i n f i n i t y i n behavior of
F+(x,A.)
DEFINITION 60J:*
B
.
and
x
tends t o i n f i n i t x .
n
of a canonical domain y- (x.A)
so
of (58.1) which
I n this s e c t i o n , w e s h a l l
y+(x,l)
and
y-(x.X)
as
x
I n o t h e r words, we s h a l l study the asymptotic
F- ( x , l )
as
A canonical domain
x
tends t o i n f i n i t y i n
n is
B
.
s a i d t o be c o n s i s t e n t ,
or
i n c o n s i s t e n t , according a s t h e v e r t i c a l c u t s i n i t s i m a g e i n the z-plane do, o r do n o t , a l l tend t o i n f i n i t y i n t h e same d i r e c t i o n .
*
This d e f i n i t i o n i s due t o W. Wasow [49].
SUBDOMINAN T SOLUTIONS
250
Set
where
r
i s a p o s i t i v e number.
Then, i f
i s a n c n s i s t e n t canonical domain,
O,S(r)
ari I n c o n s i s t e n t canonical domain,
( S , 8
ea,,h of uhich i s simply connected.
i s s u f f i c i e n t l y l a r g e , and
r
i s simply connected.
If
n
c1 i s
r ) has two connected components
W e s h a l l t r e a t t h e s e two c a s e s separ-
ately.
'I'HEORDl 60.1:"
rf n
i s a c o n s i s t e n t canonical domain,
f3I' (58.i7) N = 1 , 2 ,... , where t h e Praof: bJe s h a l l t r e a t F+(x,A) only.
s t u d i e d i n a s i m i l a r manner.
tim ticjn
$
then
a r e c e r t a i n p o s i t i v e numbers.
The f u n c t i o n
F ( x , l ) can be
I n S e c t i o n 59, we c o n s t r u c t e d a bounded solu-
and w of t h e system o f i n t e g r a l e q u a t i o n s (59.12). 1 2 $ + ( x , A ) i s defined by
The h c -
x,
.
F+:x,X) =w,(x,A) + w 2 ( x , h )
;%eref'ore, we s h a l l s t u d y t h e asymptotic behavior o f tends t o i n f i n i t y i n
&I
.
w1
and
w2
To do this, l e t u s change t h e v a r i a b l e
as
5
in
systen! !59.12) by
c
(60.2)
=:(S)
*
Then,
wl(x,A =
i(
[-2h(S,A)wl(I,h)
+ ~1L ( S , h ) ( w l ( S , h )+ wZ(S,h))
z)
1 -
+ $LCS,X) M S )
2
e x p [ - a ( z - 6 ) 3%
,
(60.33 w2(x,h)
=J-
[ 2h( S ,A )w,( 5 ,A )
c( a )
-$5,
A ) ( ~ ~ (A 5+, w2( I, h )
1 _-
1
- $ 5 , h ) l ~ ( I ) 'dc where
z =i(x)
.
Assilme t h a t a l l t h e v e r t i c a l c u t s
' See
M . A . Evgrafov and M.V.
L1, .. .,$ are d i r e c t e d downwards.
Fedoryuk [ll; Theorems 3 . 3 and 4 . 1 , Remark 4.31
251
ASYMPTOTIC SOLUTIONS
Consider a half-plane
I m [ C ] < r i n C-plane, where r i s a real v a r i a b l e . I n this half-plane, we s h a l l construct a closed s e t O ( r ) which s a t i s f i e s the following conditions: (See Fig. 60.1,)
& ( r ) i s simply connected,
( i ) the i n t e r i o r of
&(r) c o n s i s t s of t h r e e parts: = (-1+i r ) - (is)exp[-ig6] 2 (OlSo
for
Then, mapping (62.9) t a k e s
(62.11)
Re[z]>O
So
as
1 ’
F i g . 61.6-a and Fig. 6 1 -6-b r e s p e c t i v e l y .
(62.8)
j
fo(x,h)
are real numbers, and t h a t
x> x
1 ‘
onto t h e h a l f - p l a n e
.
H
M g . 62.1.
P ( x ) ~s o t h a t
26 3
ASYMPTOTIC REPRESENTATIONS
Let us construct a subdomain
of t h e canonical domain
61+
n+
same way a s we constructed a subdomain 61 o f the canonical domain Section 59.
I n Section 59, w e covered the v e r t i c a l cuts
the sectors
V1,
...,Vk
t i v e number
6
a s the c e n t r a l angle.
of c e n t r a l angle
t h e r e i s only,one v e r t i c a l c u t . o f the canonical domain
n-
.
6
.
L1,.
..,$
i n the
n
in
by
W e use the preassigned posi-
Note t h a t , i n the present case,
Similarly, we construct a subdomain The domain
0
-
- i s shown by Ffg. 62.2.
I n the domain
-
(62.12)
IIm[ll1
w e construct a holomorphic function
5b0 , ho( x , X )
Re[X12po
,
s a t i s f y i n g t h e asymptotic
conditions
f o r (62.12) and
N=1,2,
..., where the
thermore,
is a formal s o l u t i o n o f the R i c c a t i equation 2 u ' t u 2 - X p(x) = o
.
( C f . Lemma 58.1.)
Fig. 62.2.
a r e p o s i t i v e constants.
Fur-
SUBDOMIN AN T SOLUTIONS
264
-1
1 -
Note t h a t w e f i x e d t h e branch of The f u n c t i o n
p(x)
2
so t h a t
p(x)
2
>0
for
X>
x1-
.
can be c o n s t r u c t e d i n t h e same way as we c o n s t r u c t e d
ho(x.h)
i n S e c t i o n 58.
h(x,X)
Ey v i r t u e o f Theorems 59.1 and 6 0 . 1 , t h e r e e x i s t s a s o l u t i o n of (57.1)
o f t h e form:
+s
X
y = p ( x ) '[lt F(x,X)]exp{-XI(x)
(62.14)
ho(t,h)dt]
,
co
xhere
--..
(i: p(xl (iii
--1
1
'
i s t h e branch such t h a t
p(x)
'>
for
0
x > x1
,
F(x,X) i s holomorphic i n t h e domain x€A+
( 6 2 .lj)
,
,
IIm[XlI160
,
Re[h12~,
and s a r ; i s f i e s the c o n d i t i o n s
ioz.:o! f c r !62.15) and
... , t h e
N=1,2,
quantities
$
being p o s i t i v e c o n s t a n t s ,
F i n ii
(iii) ',he path of i n t e g r a t i o n i n t h e right-hand member of (62.14) i s a
ciirve j o i n i n g
$03
to
in
x
at
.
(See F i g . 62.2.)
S i m i l a r l y , w e can
c o n s t r x t a s o l u t i o n of (57.1) o f t h e form:
-1 (62.141 ,
+s
X
y = p ( x ) 4[1+?(x,X)]exp[-hI(x)
ho(t,h)dt)
,
m
where
--1
( i f )p ( x )
i s t h e branch such t h a t
--1 p ( x ) '>
0
for
x>xl
,
(iil) T ( x , h ) i s holomorphic i n t h e domain (62.15 I:
xE9-
)Im[h]l16,
9
Re[X12Po
ani s a t i s f i e s t h e c o n d i t i o n s
(62.16' ;. for (62.15')and (iiiI )
N=1,2
,..., and
t h e p a t h o f i n t e g r a t i o n i n t h e right-hand member o f ( 6 2 . 1 4 ' ) is a
curve j o i n i n g
in
B-
'Theorem 59.1, we must choose
po
+rr,
to
x
.
(See M g . 62.2.)
In order t o use
sufficiently large. The two s o l u t i o n s ( 6 2 . 1 4 ) and (62.141) of t h e d i f f e r e n t i a l equation
( 5 7 . 1 ) a r e subdominant i n t h e s e c t o r
265
ASYMPTOTIC REPRJEXNTATIONS
So: l a r g
s i n c e t h e image of (62.11).
so
0
for
i n t e g r a t i o n i s a curve joining
+-
x
in
to
The two s o l u t i o n s y o ( x , h ) and tor
So
.
fo(x,X)
Hence t h e r e e x t s t s a function
(62.21)
f o ( x , O =c(x)Yo(x,a)
c(h)
, and
x > x1 &+“
.
B-
the path of
(See Fig. 62.2.)
a r e subdominant i n the secof
such t h a t
Z
.
, then
(62.19), (62.20) and (62.21) Will provide an asymptotic representation of fo(x,X) f o r x € & + ~ & a s h tends t o
If we can find
c(h)
i n f i n i t y i n s t r i p (62.1). in
So
, and
-
We can f i n d
c(X)
by l e t t i n g
u t i l i z i n g the asymptotic p r o p e r t i e s of
+.
tend t o
yo(x,X) and
+W
fo(x,l)
The d e t a i l s a r e l e f t t o t h e readers.* In order t o summarize t h e method, again by using case (62.6), we conThis s o l u t i o n i s subdominant i n t h e domain s i d e r t h e s o l u t i o n f2(x,X)
as
x
tends t o
x
.
S2
?+
which i s shown by Fig. 62.3.
See, f o r example, Y. Sibuya
[ 361.
SUBDOMINAN T SOLUTIONS
266
Fig. 62.3.
Ng.
Fig. 62.3 i s a p a r t o f Fig. 61.5. taining domains.
S2
.
The domain
on F i g . 61.5 .) from
no
Let us denote by
no
62.4.
There a r e f i v e canonical domains con-
no
t h e union of these f i v e canonical
i s shown by Fig. 62.4.
( F i g . 62.4 i s also based
a s u i t a b l e neighborhood of the boundary of
-
no no .
W e c o n s t r u c t a simply connected domain
by removing Then, by u t i -
l i z i n g Theorems 59.1, 60.1 and 60.2, we can f i n d an asymptotic representaI
t i o n of
fZ(x,X) which i s v a l i d uniformly f o r
xE
no
as
X
tends t o
i n f i n i t y i n s t r i p (62.1).
63. Simple t r a n s i t i o n p o i n t problems i n unbounded domains. we constructed two s o l u t i o n s
I n Section 59,
26 7
SIMPLE TRANSITION POINT PROBLEN3
of t h e d i f f e r e n t i a l equation (58.1) i n a s u i t a b l e subdomain i c a l domain and
(59.5)
, where
0
F+(x,A)
F+(x,A) 5 0
,
F (x,X) 5 0
,
{ -
uniformly f o r
X
xE 61 a s
The functions
h(x,h)
F- (x,X)
of a canon-
a r e holomorphic i n (58.17),
tends t o i n f i n i t y i n the s t r i p
pmll 18,
(59.6)
and
61
and
Re[XILpo
9
h(x,-A)
a r e holomorphic and s a t i s f y conditions
(58.19) and (58.19’),r e s p e c t i v e l y , f o r (58.17). ( C f . Theorem 59.1.) Furi s a c o n s i s t e n t canonical domain, we have
thermore, i f
I F+(x,X) - I 141AI(x) I -2N
(60.1) f o r (58.17) and
... , where the 4 are c e r t a i n p o s i t i v e numbers,
N=1,2,
and
=s
-
X
(59.1)
I(x)
p(tI2dt
.
xO
( C f . Theorem 60.1.)
n
On the other hand, i f
i s an i n c o n s i s t e n t canonical
domain, then
N = 1 , 2 , . . ., where A and B a r e two a r b i t r a r y r e a l num4 a r e p o s i t i v e numbers depending on A , while t h e 4 are p o s i t i v e numbers depending on B . ( C f . Theorem 60.2,) The domain &
f o r (58.17) and
bers, and the
i s constructed i n t h e following manner:
...,$ .
2;
j
that ( i ) Lj
(ii) Lj
0 i s mapped
= I(x) onto the whole z-plane c u t by unbounded v e r t i c a l s W e covered each L by a s e c t o r V of c e n t r a l angle 6 s o
conformally by
L1,
?he canonical domain
j
i s contained i n the i n t e r i o r of i s on t h e b i s e c t o r of
(iii) the c e n t r a l angle
6
V
j ’
V
j ’
i s so small t h a t the other
(hf j )
are
SUBDOMINANT SOLUTIONS
26 8
n o t contained i n
V
We assumed t h a t
j .
i s independent o f
6
j
.
Then we
...
a s a simply connected domain d e r i v e d by removing v1 Vk -1 from t h e z-plane, and & = I ( U ) , where x = I-1( a ) i s t h e i n v e r s e of
defined
U
z = I(x;
.
'ihe domain Ip(x)l
(58.10;
p
uhere
i s simply connected.
&I
LB>O
for
X
Furthermore,
,~
E
( C f . S e c t i o n 59.)
i s a c e r t a i n p o s i t i v e number.
In deriving the
asymptotic estimates (59.5), (60.1) and (60.6) a s well as i n c o n s t r u c t i n g t h e twc f u n c t i o n s
h(x,i)
and
, condition
h(x,-A)
(58.10) was indispene-
able.
Therefore, t h e asymptotic r e p r e s e n t a t i o n s (59.4) of t h e two solu-
tions
y+(x,h)
and
y-(x,h)
c f any t r a n s i t i o n p o i n t . that
cannot be used i n an immediate neighborhood
i s on t h e boundary of
xo
n
.
, we
z = I(x)
When we d e f i n e d t h e mapping
assumed
H e r e a f t e r , assuming f u r t h e r t h a t
xo
i s a simple t r a n s i t i o n p o i n t , w e s h a l l c o n s t r u c t asymptotic r e p r e s e n t a t i o n s of
y+(x,k)
borhood of'
and
y-(x,X)
.
xo
which can be used even i n t h e immediate neigh-
As i n S e c t i o n s 58, 59 and 60, we assume t h a t
(63.1)
P(X0)
.
x
but f i x e d polynomial i n
=o
#0
-
xo i s a simple t r a n s i t i o n p o i n t of t h e d i f f e r e n t i a l equa-
This means t h a t
tiori (53.1). A neighborhood of t h e t r a n s i t i o n p o i n t
61.1 .)
n
Let
,
-12
=s
X
(63.2;
z=I(x)
X
p(t) dt 0
onto t h e whole z-plane c u t by unbounded v e r t i c a l s assume t h a t t h e c u t three sectors R
.
i s divided i n t o
61 and 62 by Stokes c u r v e s .
conformally by
in
xo
(Cf. S e c t i o n 61, M g . be a c a n o n i c a l domain f o r e q u a t i o n (58.1) which i s mapped
E0
three ssctors
i s an a r b i t r a r y
We f u r t h e r assume t h a t
P'(X0)
9
p(x)
Lo
, 51
6,
starts from and
z=O
.
Lo,L1,...,4( . We
also
This means t h a t two of t h e
i n t h e neighborhood of
As we d i d i n S e c t i o n 59, l e t us cover
xo
a r e contained
L1, ...,$ by s e c t o r s
. ..
V1, ,Vk of c e n t r a l a n g l e 6 s o t h a t c o n d i t i o n s ( i ) , ( i i ) and (iii)o f S e c t i o n 59 a r e s a t i s f i e d . Furthermore, ue denote by Vo a s e c t o r of cent r a l angle
6
such t h a t
( i f )t h e v e r t e x of (ii') (iii1 )
the cut
Lo
Vo
is at the point
bisects the sector
the c e n t r a l angle
tained i n
Vo
.
6
Vo
z=O
,
,
is so small that
L1,
...,\
a r e n o t con-
We s h a l l then d e f i n e a simply connected domain
Uo by
A METHOD DUE TO T.M.
removing
Vo"Vlw.
kV'.
n
(63.3)
Bo=I-L(uo)
.
8,
Finally, we
by
Our main concern i s the asymptotic behavior o f uniformly f o r
269
(See Fig. 63.1 .)
from z-plane. of
define a subdomain
CHERRY
y+(x,A)
and
y- ( x , a )
as 1 tends t o i n f i n i t y i n s t r i p (59.6).
x E Bo
LD
b
Fig. 63.1. I n Section 66, we s h a l l a l s o i n v e s t i g a t e t h e case when the v e r t i c a l cut
Lo
.
This
contains only one o f the three s e c t o r s 60
, 5
, but
means t h a t t h e domain
n
z = O i s not t h e s t a r t i n g point of
and 62 i n the neighborhood of the t r a n s i t i o n point
64.
i s on
z=O
A method due t o T.M. Cherrx.
xo
Lo
.
I n Section 58, w e constructed the two
formal s o l u t i o n s (58.4) and (58.5) o f the R i c c a t i equation (58.3). construction, we used the condition:
lp(x)I 2 p> 0
for
xE
&
.
I n this
Since the
is on t h e boundary of t h e domain Po , w e can not use this method i n go Instead we s h a l l follow t h e scheme due t o T.M. Cherry.* To begin with, we consider a d i f f e r e n t i a l equation transition point
(64.1)
xo
.
2 d w/d5'-
Changing the v a r i a b l e
*
sw= O
.
5 by
See A . P d 6 l y i , M. Kennedy and J . L . McGregor [lo; $4. pp. 465-471], W. Wasow [47; Chapter V I I I , pp. 157-1941, and T.M. Cherry [6].
SUBDOMINAN T SOLUTIONS
270
we d e r i v e from (64.1) t h e d i f f e r e n t i a l e q u a t i o n
Eu'ote t h a t
(64.4) Next, changing t h e unknown q u a n t i t y
w
by
--
1
(54.5)
w=J(x)
'h ,
we d e r i v e from ( 6 44 ) t h e d i f f e r e n t i a l e q u a t i o n
(64.6) On t h e o t h e r hand, i f we change t h e unknown q u a n t i t y
y
by
_(64.7)
Y=P(x)
4v
,
the d i f f e r e n t i a l equation
(58.1)
y"
- i2p( x ) y = 0
is reduced t o
-
This equation can be w r i t t e n a s
(64.8)
We s h a l l transform t h e d i f f e r e n t i a l e q u a t i o n (64.8) i n t o an i n t e g r a l equation i n the next s e c t i o n . Let us prove t h e f o l l o w i n g lemmas.
LEMMA 6C.1:
The f u n c t i o n
(64.9) i s holomorphic i n a neighborhood of t h e t r a n s i t i o n p o i n t x = x 0 Proof:
Put a=pl(xo)
,
b=pll(xo)
Assumption ( 6 3 .l) i m p l i e s t h a t
.
.
A METHOD DUE
TO T.M. CHERRY
271
.
afo Since
1 2 p(x) = a(x-xo) + zb(x-xo) i n t h e neighborhood o f
-1
, 11.
+ O[ (x-x,)
3
]
x=xo
1
2 1-2 2 ~[ a) +-a 4 b(x-xo)+O[(x-xo) 11
2 ~ ( X ) ~ = ( X - X
,
and
1 I ( x ) = ( x - x o ) 3 ~ 2 p a ~ + ~ ( x -+x o[ o )(x-x,) 2 31 3 1oa2
-
.
Hence b 0
Thus, 5a x-xo i n the neighborhood of
x=xo
.
+ o(1)
On t h e o t h e r hand,
0
and
i n t h e neighborhood o f
i s holomorphic a t
x=x
x=xo
.
0
.
The n o t a t i o n
O(1) i n d i c a t e s a term which
Since
we g e t q(x) = O(1) i n t h e neighborhood of
6~.2:
If
in
9,
Proof:
, where
.
This proves Lemma 64.1.
R i s a s u f f i c i e n t l y large p o s i t i v e number, w e have
I$$
(64.10)
x=xo
K
IKJI(X)I-~
for
I I W J2~
i s a c e r t a i n p o s i t i v e number.
It is easy t o s e e t h a t q(x)
= o(x-2)
SUBDOMIN AN T SOLU TIONS
272
as
x
tends t o i n f i n i t y i n
Bo
.
as
x
tends t o i n f i n i t y i n
&lo
, where
respect to
x
.
Hence
n: i s t h e degree o f
p(x)
with
On t h e o t h e r hand,
I(X) = o ( x as
tends t o i n f i n i t y i n
x
If
LEMMA 6L.3:
to
This proves Lemma 64.2.
ro is a s u f f i c i e n t l y small p o s i t i v e number, w e have
for