1 t:.
r /7' \ \ ;i1 '
V
:RO TYPE
[IX]
. A~ )
, ell.cn ) "
E'log A2
'
.. m
•. _[
- C27 q,'(N) e·(N).
1
Am...
22 downloads
451 Views
13MB Size
Report
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!
Report copyright / DMCA form
.1 t:.
r /7' \ \ ;i1 '
V
:RO TYPE
[IX]
. A~ )
, ell.cn ) "
E'log A2
'
.. m
•. _[
- C27 q,'(N) e·(N).
1
Am2
- eU(n)
)
, e2.(N+l») 1 -eU(n) -
\
177
PROOF OF THE MAIN THEOREM
[31 ]
11 - (~(N~~ k)21 ~ (~ + ~) logj ~
(1
n1
26q,'(N)e·(N)(q,(N)
+ 1)
2
+ [n log udu _ 2aq,'(N)q,(N)e'HN) _ 26q,'(N)e",(N)
~ nl log nl + n2log ~ - nl - n2 - 2aq,'(N)q,(N)e·(N) - 2aq,'(N)e...(N) ~ nl log n1 + n2 log n2 - 26q,'(N)q,(N)e"' CN ) - 46q,'(N)i(N). For fixed N, nl + n2 is fixed. By differentiating x log x + (c - x) log (c - x), 1 ~ x ~ c - 1, )
it follows easily that it takes its least value for x = 1.
1 - e-2(n-N-1).'(n»)
nl log nl
+ n2 log n2 ~
Thus using (31.13),
(2&P'(N)l(N) - 2) log (2&p'(N)e"'(N) - 2).
Therefore for large N
1 - e-n.'(n»). ,",' log £.oJ
11 -
A~+ k)2 ~ 26q,'(N)e...(N) log (26q,'(N)e",(N) 1
(ewn
- 2)
- 26q,'(N)q,(N)e.(N) - 6q,'(N)e·(N)
: (1 - e·... (n»)
~
2aq,'(N)e...(N) {log q,'(N)e... CN ) + log (26 - 2e-",(N)/q,'(N») I - 26q,'(N)q,(N)e"'(N) - 6q,'(N)e·(N)
~
26q,'(N)e·(N) log q,'(N) - lOq,'(N)e·(N).
Using this and (31.12), (31.11) becomes log I F'(A m )
(31.14)
I ~ 26
(E + f) q,'(n)e·(n) 1
+
.'(N)e",(N) , :~
I
< q,(u). indicated to cover
1
N)e·(N) log (2e·(N»). ~er
values from - n1
11 - e~) I n
26q,'(N)e"' CN ) log q,'(N) - C30 q,'(N)e",(N).
By the same proof as used in obtaining Lemma 30.2, in somewhat simpler form since here a(u) = 1, it follows that
(E + f:) q,'(n)e·(n) og (2e,HN»)
log
N+l
(31.15)
N+l
log)
1- eUA~ I= !q,'(N Cn )
+ !q,'(N + l)e",(N+l) log /1
+ (f
1
N1 - +
-
l)e", cN - 1) log
A~_
eU(N 1)
I
eU~;+1) I
1"" ) q,'(u)e"'Cu ) log 11 - eA~ N+l
11 _
.
2",(u)
I du
+ O(q,'(N)e· CN»).
If the right side of (31.15) is handled by the same method used in the proof of . Lemma 29.1 with the very great simplification that here a(u) is not involved, (31.15) becomes
>,
",
...
"1 ,
,\
"
.~~
'l~ .'
\
[XI
[331
A series
theorems on the gamma and zeta functions.
(T), for large u
..,
Ea 1
,nd it is this that !.O7) is essential. tly much weaker ~(x), (32.07) will all for the usual B the much more
J.l"e-""''''
"
1 - e-""''''
is known as a Lambert series. Thus a corollary of Theorem LUI is that the higher indices theorem holds for Lambert series. This resu'lt was unknown before the proof of Theorem LU in 1937. In fact the Hardy-Littlewood higher indices theorem remained an essentially isolated theorem until the theorem for K(x) was proved. We shall also prove the following result. THEOREM LIV. 5 Let the hypothesis of Theorem LI be satisfied. let xK(x) f L( - 00, 00). Then (32.14)
It can be shown -plane v ~ 0, so lere for the case
lim inf {f(x) - sJ
"'... ..,
=-
lim sup {f(x) - s}
11,
= 11
implies that
(32.15)
lim sup
I
t
s
ak -
1
I~
Clll
where Cl is an absolute constant depending only on L and K(x). if An+l - An ~ 00 tJwn Cl = 1.
>
O.
om well-known
In particular,
33. Reduction to a lemma on biorthogonal functions. The proof of Theorem LI can be at once reduced to the proof of the following lemma. C2 , Cs , ... denote positive constants depending only on Land K(x). LEMMA 33.1. If the hypothesis of Theorem LI is satisfied, there exists a sequence of functions IRn(x, B) I, (n ~ 1), such that
1:
), 00) and N'(x)
for which
In addition
","'00
n-+oo
L
191
REDUCTION TO LEMMA ON BIORTHOGONAL FUNCTIONS
(33.01) (33.02)
Rn(x, B)
1 1:
Xn/2
(33.03)
I Rn(x, B) I dx < C2, I Rn(x, B) I dx
-00
(33.04)
Rn(x, B) dx
and
(33.05)
;i~
1:
=
0,
x
C ~ ~,
A"
1:", K(y) dy = 0,
Rn(x, B) dx
>
CsB,
n
~
1,
k ¢ n,
1:-", K(y) dy = 1.
Before proving the lemma we shall use it to obtain Theorem LI. • For K(x) = e-<"'e'" this theorem was proved by Ingham, On the higher indices theorem oj Hardy and Littlewood, Quarterly Journal of Mathematics, vol. 8 (1937), p. 1.
Iii
ill iii iii iii iii
,
?""
-{
...~
~,-:JP"1 ,
\:'.
•
/
". ,{}
il i !I I'! III III Pi
:Ii
,
~II!
...
I', :1
II'
ii
ii II' ii ',I,il\
[XI
lim sup I n-UO
L: I
yK(y) Idy.
t
ak - s I
1
~
lim sup If(lI,,) - s I fl.-GO
+ lim I t
1
n-'CO
34.1.
For every n
>
0 there exists an H,,(s)
H,,(A,,) = 1;
(34.01)
H,,(Ak) = 0,
where {A,,} are as in Theorem LI. (34.02) then h,,(u)
h,,(u)
= 0, I u I > c,
E
ak - f(lI,,) I =
n.
L( -
00,
00),
sw:,h that
k rf n,
=
h,,(u)
L:
'Iii
ill'I'
Ii!
iii'II
i!i
ii II: ii,
g"(u)e-iu~,,,
111 ill
ii!
I gn(U) I <
0, let
(Jk
= =
(Jk
=
(Jk
Ak).
C6
I g~(u) I <
,
I:i
C6
•
This defines
IUk}
+ An I > }-L for all m > 0, tkL if tkL - Am + An = tL for some m, Am - An if -}-L ~ tkL - Am + An < tL.
lkL if I lkL - Am
uniquely and
I.
I Uk
-
tkL
I~
}-L.
~
}-L.
Also Uk+1 -
I ud
Uk
includes IAm - A,,} for all m rf
T(s) =
n.
IT (1 - ~)(1 - ~). Uk
1
U-k
Then T(s) I sin 21rs/L
Clearly
ill
!I!
i!IIi
H,,(s)e-iOUds,
Proof of Lemma 34.1. Throughout this proof n is some fixed positive integer. For - 00 < k < 00 let
Thus
il
il
(34.04)
Moreover Let
Iiilli
'I'
where
r>
II,
:Ii
I" ii,
and
(34.03)
iii
Iii
If
= (2:)1 /2
il!
II;
I!I
This proves C 1 = 1 when A,,+1 - A" ~ 00. 34. Proof of the lemma. We now turn to the proof of Lemma 33.1. We will require several auxiliary results. The first of these is an interpolation result. LEMMA
ihat s = 0 since, mes
195
PROOF OF THE LEMMA
[ 34]
I=
-~
21r1 s j
IT 111 - - 2s/kL 1111 +- 2s/kL I. S/Uk
1
S/U_k
111!
Ill! il ; jill"
Ii
.~.
"
~1(
"
.,
i i!
!
ii
"on ,,",~
iil i:
il!'
\/ I 1•••••••••••••••••••••••••••••••.il~••••••••••••• \
l-'
\,
,~/
~
:!II
11 ,Iii
illl
lil
,II!
iili
r ilil iiil :111
,ii
[XI t
. ~ + ~i(-v-+l) dJ;. 'nic function in the d~
t 341
I r,,(u, B) I ~ Thus for 1 u
For 1 u \
I
to X(u, v).
Let
CoBI
~
U
Il(u)
u+C
1
u-c
I cjJ(By) Idy
jB(U+C)
1
~ Col U Il(u) B(u--c)
1
I cjJ(y) Idy,
B)
~
I
L:
Col u le6 (2C)
> 2c and B > 2 it follows from 1
r,,(u, B)
6
Thus
L:
(34.19)
I
I cjJ(y) I dy
~
C16 1 u I·
.
_
I lill 'III !,!I!I
cjJ(y) dy ~ 1
(ul
r,,(u, B) 1 du
<
'I'
C17
+u
2'
il
i~
i'l
~I
I!I
ClB •
Ilil
i'l'1l i!I'
il!
illl
iuB
= -,-
ifII
1
From (34.16) we also have .!:..-le'ul.,.r,,(u, B)J du
'1
III
(34.08) that
f
~ C5 1 u le (u)
I
:,i'1 ,
'ii
2c it follows from (34.08) that
I r,,(u,
(34.18) ~te
IIII :1 1
(34.17)
+r
199
PROOF OF THE LEMMA
"'n'"
fU+C v-c
k(y)cjJ(By)g~(u - y)iA""dy
iB ( u/c1(U)) (211") 1/2 k(u) , 1 - k(u)
1 u
+c
v-c
I i!i!
k(y)cjJ(By)g..(u - y)e'>''''' dy.
l!il
il,1
Treating each of the terms on the right in a manner similar to that used on (34.16), it follows easily that of e-g(u).
Thus for
len
I:u {eM"r..(u, B)} I ~ 1 ~u 2 '
(34.20)
u
:'lli
'ill
i!
j!~
iii
If we define R..(x, B)
= (2:)1/2
L:
r,,(u, B)e'u," dU,
i'l r; l!i
it follows from (34.19) that (34.21)
I
R,,(x, B)
1
<
C18
• ,.'
follows. Let <1>(8) d. (34.10) follows
Also
;
,.
R,.(x
+ A", B)
= (2:)1/ 2
L:
= (2:)1/ 2
L. d~
r,.(u, B)c'V>"e'U'" du,
ments of Theorem
..
and integrating by parts,
,
iA""dy.
~
!!~
I:! u
-ixR,,(x
+ An, B)
{r,,(u, B)e'''>''JeiV:< du.
U
I 1;1
By the theorem for Fourier transforms, Theorem E, and (34.20)
:f,j [i!
:i!l lj~
ii
n
;:'~