dirichlet's th50rsm on primes in an arithmetic progression
TRANSCRIPT
DIRICHLET'S TH50RSM ON PRIMES IN AN
ARITHMETIC PROGRESSION
by
THELM ANI ' GHENAULT, B.S.
A THESIS
IN
MATHEMATICS
Submitted to the Graduate Faculty of Texas Technological College in Partial Fulflllnent of
the RequireTisnts for the Degree of
MASTER OF SCIENCE
Approved
Accepted
May, 1966
PK'CG- \\\\J^
A/o. \2
ACKNOWLEDGMENT
I am deeply indebted to Dr. Anthony Gioia for his
direction of this thesis.
ii
TABLE OF CONTENTS
Page
ACKNOWLEDGMENTS ii
INTRODUCTION 1
PART I. SPECIAL CASES OF DIRICHLET'S THEOREM
Chapter
I. ELEMENTARY RESULTS 3
EUCLID'S PROOF
PROOF OF THE GEOMETRIC PROGRESSION ^n+3
PROOF OF THE GEOMETRIC PROGRESSION 6n+l .
II. ANALYTIC RESULTS 7
EULER»S PROOF
DIRICHLET'S PROOF OF THE PROGRESSION kn+1
Part II. THEOREMS NECESSARY TO THE PROOF OF
DIRICHLET»S THEOREM 21
III. CHARACTERS AND THEIR PROPERTIES 22
IV. THE GENERAL THEORY OF DIRICHLET SERIES 30
Part III. DIRICHLET'S THEOREM AND SOME OF ITS
CONSEQUENCES 51
V. PROOF OF DIRICHLET'S THEOREM 52
VI. CONSEQUENCES OF DIRICHLET'S THEOREM 66
LIST OF REFERENCES 70
iii
INTRODUCTION
This paper contains an extensive and detailed proof
of Dirichlet's theorem via analytic methods. This famous
theorem states that in any arithmetic progression kn+a,
n=0, 1, 2, .,., where k, a, are positive integers, having
no common factors gre -ter than one, there is an infinity
of primes.
Some special cases of Dirichlet's theorem were proven
much earlier by Euclid, Suler, and others. Euclid proved
that there are an infinite number of primes of the form of
2n+l by elementary methods, and two thousand years later
Euler gave an analytic proof of this same theorem.
Dirichlet (I805-I859) patterned his method of proof
after that of Euler. Dirichlet's proof by analysis is in
itself a miracle, for the theorem is concerned with inte
gers, vihereas analysis deals with continuous and non-integral
numbers. For this reason much credit is c iven both Euler
and Dirichlet in the development of modern analytic theory
of numbers. In this branch of m^.thematics, analysis, the
mathematics of continuity is apolied to problems concerning
the discrete domain of the integer.
Many elementary proofs have been given for special
cases of Dirichlet's theorem, but it was not until l^k^
that Salberg gave an elementary proof of Dirichlet's theorem.
PART I
SPECIAL CASES OP
DIRICHLET'S THEOREM
CHAPTER I
ELEMENTARY RESULTS
Almost the first questions anyone might ask about
primes are, "How many primes are there? Are there an infi
nite number?" Euclid was the first to give a proof that
there are an infinite number, and his proof is as signifi
cant as it was when it was discovered over two thousand
years ago. Time has not mired its simplicity or beauty.
Let us look closely at it.
Theorem 1.1. There exist infinitely many primes.
Proof. Suppose on the contrary that there are only
finitely many, say p-., P2. ...t PT^. We form the number
(1) N = n p + 1.
i=l i
Then either N is a prime distinct from p- , ..., p, or is
divisible by a prime p. p cannot be any of p- , ..,, p- ,
for we should have from (1) k
p I N, P I TT p^ , i=l
and hence pll, which is impossible. Therefore p is a
pri.re distinct from j>-^, ..., p, , contradicting our hypothe
sis.
Another way of stating this result is as follows:
There exists infinitely many primes of the arithmetic
progression 2n+l, which makes this result a special case
of Dirichlet's theorem.
Euclid's method of proof has become one of mathema
ticians' finest tools. This method of proof will be used
in the next two theorems.
Theorem 1.2. There exist infinitely many primes of
the form ^n+3.
Proof. Suppose that there exist only finitely many
primes of the form ^n+3f say p, , p p, . We form the
number k
(2) N = 4 n P. - 1.
i=l
Then either N is prime and = 3 (mod 4) or N is composite.
Not all prime factors of N can be of the form 4n+l, for
otherwise N itself would be = 1 (mod 4). This is easily
seen by takineg the example x=4n-. + 1 and y=kn^ + 1 , t hen
xy = l6n-|n^ + 4n^ + 4n2 + 1 , or xy = 1 (mod 4). There
exists, therefore, at least one prime factor p = 3 (mod 4).
Again, p cannot be any one of p... p^. .... PT^, for if it
were, then from (2) k k
p I 4 Tf p. and p » 4 TT p. - 1 , i=l ^ i=l ^
which implies pi 1. This is impossible. Thus p is a prime
distinct from p, , ..., p, .
Now using the quadratic law of reciprocity of Gauss-
Legendre, we will prove the next theorem.
Theorem 1.3* There exist infinitely many primes of
the form 6n+l.
Proof. Suppose on the contrary that there are only
finitely many, say p , P2f ••.! PT * We form the number
(3) N = 12 ( n p.)^ + 1.
i=l
Then N = 1 (mod 6). Either N is a prime other than p]_,
P2» .... Pj or it is a composite. If N is a prime, we have
our contradiction, otherwise assume that a prime p divides
N. If p is two or three, then p I 12 and pi N, which implies
p divides one. This is impossible. Therefore p i 2, 3.
and pL N, so from (3). the congruence
3 x^ + 1 = 0 (mod p)
is solvable; i.e.
3 x^ = -1 (mod p),
3 (3 x2)= -3 (mod p),
(3 x)2 = -3 (mod p).
Thus
but by the quadratic law of reciprocity
Therefore
( ! ) • ^ ••
i.e., p = 1 (mod 3). o^ p=3k+l, where k is an integer, p
is odd since p ^ 2 and may be expressed as p=2m+l where m
is an integer greater than one.
So
P = 2m + 1 and p = 3k + 1
or 2m + 1 = 3k + 1,
2m = 3k.
But 2 l2m, so 2 1 3k. Since 2 t3, 2 I k, or k=2x where x is
an integer. So p is expressible in the form
p = 3(2x) + 1,
= 6x + 1.
Therefore there exists at least one prime divisor p of N
and p = 1 (mod 6). p cannot be one of the primes p^, ...»
p. ; for if it were, then from (3) k 2 ^ ^ 2
p 112 ( It p. ) + 1, p 112 (."n P ) ,
i=l ^ "- ^
and hence p I 1 which is impossible. Therefore there exist
infinitely many primes of the form 6n+l.
3y using this method, it is possible to prove a number
of special cases of Dirichlet's theorem.
CHAPTER II
ANALYTIC RESULTS
Euler gave a fundamentally different proof of Euclid's
theorem. This proof is done by analytical means, which is
the fundamental technique employed by Dirichlet.
In this chapter v:e will turn our attention to Euler's
proof and special cases of Dirichlet's theorem which may be
proven by analytic methods.
Let s be real and consider the series
(4) 2 -K n=l ^ ^
This series converges for s > 1, since
1 1
1 . 1 1 2 ^ •*• ""TS- 1 —oS: + ~Q? = oS-1
" ^ •*• "53" + ~ p " + "ys- 1 -^ **• - T ^ + "Tjjs- + - ^ - ifs-=ir-
etc, where we consider 1, 2, 4, 8, ... terms of the series.
It follows that the sum of any finite number of terms of
the given series is less than the geometric series
ls-1 - - 2 ^ **• " 4 ' ^ * " " 8 ^ + ... = i.(|)y-l
V7hich converges for s > 1. Thus by the comparison test the
series (4) converges for's > 1, and its sum is denoted the
Riemann zeta function ^ (s).
The zeta function, studied extensively by Riemann, is
8
fundamental in prime number theory. Its importance is due
to a remarkable theorem discovered by Euler which expresses
the function as a product of primes only [ ll] •
Theorem 2.1. For s > 1,
(5) ^ (s) = TT (1-p-^)"- . P
where the product is taken over all primes p.
Proof. Let
(6) P(x) = TT (l-p-^)-l, p<x
then since p > 2 and s > 1, we have
(7) • (1-p-^)-^ = 1 + p"^ + p-2^ + ... .
00
k=0 ^ .
Nox'j if each series (7) were absolutely convergent for all
P = P^. P2» P- * •••» ^-r— '^* ^^^ terms could be rearranged
in any order, and the rearranged series would converge to
the same sum [jTl, Each series (?) is absolutely convergent
since by the n root test C3l]»
-ks lim\lp-H)^'^^ = lim ^s k->cx) k->oo "
= P'^ < 1.
since p > 2 and s > 1. So
P(x) = TT d-p"^)"^. p<x
= n 2 p-^^ p<x k=0
= (1 + pi~^ + P-L"^^ + • • • ) • • • (1 + p / ^ + -2s ^ ^
Multiplying these series together, the general term resulting
is of the form,
Pl-^lSp2-^-2Sp^.a3s...p^-ars (^^^Q, n^>0, . . . a^>0).
Letting n = p ^ - p Po ^•••Pp . we have
(8) P(x) = 2 n" .
The number n will occur if and only if it has no prime factor
greater than x, and by the fundamental theorem of arithmetic,
it occurs once only. Hence in (8) the summation extends
over numbers formed from the primes p £ x.
These numbers n include all numbers < x, so that
n=l (x) x+1
and the l a s t sum tends t o 0 when x->flo • Hence
n=l x->»(x)
= l im n ( l - p " ^ ) " . X->£» p<x
= TT d - p " ^ ) " . P
Hence T (s ) = Vi (1-p"^) ( s > l ) . P
Theorem 2.1. may be regarded as the analytic analogue
of the fundamental theorem of arithmetic.
Theorem 2.1. is a special case of a more general theo
rem which "Till be useful later. We recall that a function
f(n) is multiplicative if f(mn) = f(m)f(n) when (m, n) = 1,
and comoletely multiplicative if f(mn) = f(m)f(n) without
restrictions.
Theorem 2.2. If f(1) = 1 and f(n) is multiplicative,
then
10
(9) E ^M = TT (l+f(p)+f(p2)+...+f(p )+...) = n S f(p^)
n=l p p k=0
provided that either side is absolutely convergent. If
f(n) is completely multiplicative then (iO) TT Z f'(P ) =TT (l-f(p))-l.
p k=0 p
In both instances the product is taken over all primes p. CO
Proof. Suppose 2 f'(n) is absolutely convergent. n = l •
Let
(11) p(x) = n 2 f*(p^)-p<x k=0
00 00 Since 2 lf(n)l converges and 2 'f(p )• is a subseries of
n=l k=0
the series 2 ff (n)l , 2 If (D )l converges [ 4] . Since n=l k=0 00 ,
the series 2 f(p ) is absolutely convergent, we can rear-k=0
range terms and multiply the series together.
p(x) = n 2 f(p^) = '^ ( i+f (p)+f (p^)+-- - )
p<x k=0 p<x
Letting p^, Pp. .... P^ < x, we have P(x) = TT (1+f(p)+f(p2)+...)
P5.X
= Q + f (Pi)+f*(Pi)+- • -n iii+^(P2)+^(P2^^'*" • -H- • •
[;i+f(p^)+f ( p ^ ^ ) + . . 0
= 2f(Pi''^)f(P2^^)---f(Pr'''*)
where a's may take on any non-necrative integral value. The
summation is over all possible choices of a-^, a^. .... CL^.
Since f(n) is a multiplicative function and any two primes
p. and p. (1 5> j) are relatively prime.
11
P(x) = TT 2 f(P^) = 2 f(n) p<x k=0 (x)
where the last summation is taken over all n having no
prime divisor which exceeds x, and each such n occurs pre
cisely once.
The summation 2 ^(n) includes all functions f(n) (x)
up to f(x), so that 00 CP
• 0<2 f(n) - 2 f(n) < E f(n) ; n=l (x) x+1
and the last sum tends to 0 when x->ao , Hence 00 00 00 , 2 f(n) = lim 2^M = lim TT E f(P ) = TT 2 f ( p ^ ) .
n=l x->oo (x) x->oo p<x k=0 p k=0
Now l e t
(12) TT 2 ^(P ) ® absolutely convergent, and consider p k=0
(13) Q(x) =17 2 f(P^).
p k=0
The absolute convergence of (13) follows as a corollary of
the convergence of (12) C5D« Since the series in (13) are
finite and Q(x) is an absolutely convergent product, we may
multiply the series together and rearrange factors of the
product without affecting the value of the product [yT]. Q(x) =T\ 2 f (P ) .
p k=0
= TT Cn-f(p)+f(p^)+---f(p^)D • P
Multiplying the series together, the general term resulting
is of the form
f(Pl l)f(P2''2)f'(P3'' )---. (0<ai<x; i=l, 2. 3. • • • ) .
Noting that we are dealing with multiplicative functions and
12
(p^tPJ = 1 where i j- j, we have
A number n will occur if and only if it has no prime factor
with an exponent greater than x, and each n occurs pre
cisely once. Hence Px ^
Q(x) = 2 f'(n). F^ = (p-j_P2P3...) .
Thus ^ 00
0 < 2 f(n) - Q(x) < 2 f(n). n=l F +i
The last sum tends to zero as x ->00 . Hence 00 GO V 00 V E f (n) = lim Q(x) = lim IT 2 f (p ) ^ U E f'(P )• n=l x->oo x->oo p k=0 p k=0
Now v:e wish to consider the second part of the theo
rem (10). Since we are dealing with completely multiplicative
functions, 00 00 ]^
-n 2 f(P ) =TT 2 Criv)3\ p k=o p k=0
= TT (1 + f(p) + [If(p)I] + &(P)I] +•••). P
Now we shall show that 2 3 -1
1 + f(p) = &(P)I] + Cf (P)I] + ••• = Cl-f(p)Il when
lf(p)l < 1. 2 n-1
Let S„ = 1 + f (p) + &(p)Il + ••• + Cf (P)I1 2 n-1 n
then f (p)S^ = f(p) +Cf (P)II + ••• + &(P)I] + &(P)I] • n
Subtracting, D-^(p)IlSn = 1 - &(p)Il *
i-[:f(p)J °^ ^n = l-f(p)
n Now we need to show that lim df (p)]] = 0. To prove this we
n->00
13
must show that given € > 0 we can find N such that n
I Q(p)D - 0|< € for n > N.
This result is true for f (p) = 0; hence we consider f (p) 5 0.
Now |Q(P)I] 1= |f(p) < €
when n Inlf (p)| < In £ , or n > (In £ ) / (In I f (p)l ) = N
Qsince if f (p) < 1, In f(p) is negative 3* We have there-n
fore found the required N, and lim Qf(p)] = 0. Thus 2 n->QO
1 + f(p) +Q(p)I] + ••• = lim S^, n->co
^'^ i-f(p) , r^p),
Thus we have completed the proof of Theorem 2.2,
In order to prove that there are an infinite number of
primes, Euler proved that the series 2 1/P diverged. P
Theorem 2.3.
s->l" 0 p pS
wiTEiB the sum is over all primes p. There exist therefore
infinitely many primes.
Proof. From (5). we have
iis) =Tr ( i -p -^ )"^ , s > 1. ^ p
Now taking the logarithms of both sides, we have
log llis) = log TT (1-p-^)"-^. ^ P
= - 2 log (1-p-^). P
But since s > 1 and p > 2, we have -l<p~^ < 1; therefore we
can expand log (l-p*^) as log (1-p-S) = - ^ 1 ^
k= 1 k p ^ .
14
Hence log ^ (s) = 2 S - j ^ ^ p k=l ^P
00
p k=2 ^P
= 2 VP^ +2 ^ 1 ^ . p p k=2 ^
Now we need to look at the behavior of log T (s) as
s->l ^. In order to do this, we will first look at the behavior of ^ (s) as s->l"*"0.
We can write )^ (s) in the form
^ (s) = 2 n - ^ = J"x-^dx + § j''"' (n- -x-S)dx. 1 1 I n
Here \ x~ dx =sn"
since s > 1. Also
n-s - x-s = j' 0 < n-^ - x-^ = \^ st"^'^ dt < nl n
if n < X < n+1, and so
0 < 1"*^ (n-s - x-^) dx < la : ^n
and m r n+1 _ _« 2 I (n'^ - X ) dx is positive i V n
and numerically less than s2 n~ .
Hence
^ (s) = + 0 (1) =(^) (l + 0(s-l)j ,
and log ^ is) = log s^x + log (l + O(s-l))
= los s ^ + O(s-l).
Therefore Log t (s) ->«> as s -> 1, so if we could prove that 2 2 TT^S remains bounded as s -> 1, then we would
p k=2 ^P have proved the theorem.
1 5
I t R(s) =2 £ r p ^=2^^"^
then JR(s)| < 2 2^ Ij^l p k=2
1 2 2 ^pte p k=2 -
1 *2 2 p p k=2 ^
ri
^^t 2 5te = (i-p-2)"i. (s > 0)
k=0 ^
CO ^ 1 1 - , 1 P-^^
-2s or IR(S)I<|2 f- s
P ^
5. 2 - T-s 2 P •
: 1 ^ -2s 2 1-2-2
< r— . 2 n n=l
o. -2s Now looking at the behavior of the series 2^=1 n" as
s -> 1 , we see
lim 2 n" = 2 "H " • -he series on the right-s->l" ^ n=l n=l
hand side of the equation is a hyperharmonic series, and
converges \J?'2 • Hence as s -> 1 , R(s) remains bounded
and
im , . 2 ^P^ -s->l+0
> cx>
Corollary 2.3. As a simple corollary, we infer that
2 /p diverges. P
Proof. Suppose 2 1/P converges to a. P
Then for s > 1,
16
2 l/P < a. P
and this contradicts the statement of the theorem. Thus
2 1/p diverges, P
We now wish to modify Euler's argument to prove a spe
cial case of Dirichlet's theorem. To do this, we introduce
a character X(n) and a series
L(s.X) = f % i n=l ^
We make the following requirements of X(n):
(1) X(n) must be completely multiplicative; i.e.,
X{m)X{n) = X(mn),
(2) X{n) depends on the residue class modulo k to which n
belongs, and
(3) X(n) = 0 if (n,k)> 1,
Using these modifications, we will prove the following
theorem:
Theorem 2.4.
->a?. Ii F - i->i ° pslTmod 4)P^ S'
There exist, therefore, infinitely many primes in the pro
gression 4n+l,
Proof. Let us define our character X'(n) as follows: . ,,(n-l)/2 ^ (-1)' for odd n,
X(n) =
0 for even n.
The second and third conditions for a character are obviously
displayed by our function. We need only to show that our
function is completely multiplicative. If either m or n or
17
both ra and n are even, then X(m)) (n) = /(mn). Now consider
both m and n odd,
X(m) = (-1)^^-^)/^, ^
X(n)= (-l)^^-^^/^
X(m) X(n) = (.l)(-l)/2(.i)^-^)/2^
and X(mn) = (.i)(^-l)/2.
X(mn) = X(m)X(n) if the exponents have the same parity.
To show that this is the case, we need only to show that
the difference of the two numbers is an even number,
mn-1 (m-l) + (n-l) _ mn-m-n-H _. (m-1) (n-1) 2 " 2 •" 2 2
Since m and n are both odd, both (m-1) and (n-l) are even.
Thus (m-lVn-1) is even. Hence X(n) is a completely mul-2
tiplicative function.
Consider now for real s the series
L(s, X) = § iLLnl n=l n^
1 i + 1 1 + = 1 - 3 s + s " s + • • • •
n=0(2n+l)s ^
From the elementary theory of infinite series, vie now
show that this series converges for s > 0, and converges
absolutely for s > 1, Let
^n = (k!n-H)s
The alternating series 2 (-l) a (n=0, 1, 2,... and a^ > 0)
converges if (1) an>an+i (n=0, 1,...), and (2) lim a ^ = OQ 8]],
If s > 0, this is the case since 1 ^ 1
(2n+l) 2 - L2(n+l)+in . and lim 1 = 0.
n->00 (2n+l)^
18
The series 2 Tvn-H)^ converges absolutely if the n=0
CD series 2 (2 +1)^ converges. This is the case for s > 1
since the series 2 —=^ (s>l) converares and each term of the T nS n=l ^
00 series 2 /^ -, \o is less than or equal to the corresponding
^2. T2n+1) s CD
term of the series 2 ~s • Thus by the comparison test, n=l ^
the series 2 , vo converges absolutely for s>l FQ"] . (2n+l)^ ~
Since X(n) is completely multiplicative, and L(s,X) is
absolutely convergent for s>l, we have from Theorem 2.2,,
I-L(s,x) i n Q - ^%I T"^ . odd p P'
Taking logarithms, we get
Log L(s.X) = log n C l - ^ n '^. odd p
= - 2 log.ci - ^ : • odd p P
Since s > 1, p > 2, and 'X(p)l < 1, vie can expand the loga
rithm as
log Ci - Mia = - S | M . pS =2. 'P
Thus
log L(s,X ) = 2 I fM^. odd p k=l '-P*^
Odd p P k=2 ^P
_ K X + ^ ^ X(p^)
Odd pP odd p k=2 P
19
1 - "^ 1 + R, ( s ) . p=l(mod 4) P ps3(mod 4) P
where we c a l l e d the t h i r d sum RAs).
As be fo re we have
Hi(s)l< X 2 4 i # odd p k=2 '^P^^ ,
k' < i 2 2 - ^ ^ odd p k=2 P
ca CO
^3-
but
so
odd p k=2 P
^ - k r = d - p ' ^ ) " - ^ fo r s > 0. k=0 P^^
5* 1 _ ^
or
Now s
Bi(s)| < * 2 odd p
< i 1 - 2 l-2"2
< -1 1
ince the series
p-2s 1-p-s
£ p-2^
n=l
n=l converges as s -> 1 ,
Rl(s) remains bounded. To isolate the primes p=l(mod 4),
vje add log ^ (s) where
log ^ (s) = 2 ^ ^ + R(s). P P
= -Is- + 2 -^ + s(s) . ^ odd p P
= 4^ + 2 4 ^ + 2 - ^ + R ( s ) . ^ p=l(mod 4) P p=3(mod 4) P
20
log L(s.x) + log t (s) =-is- •*• 2 2 ; ^ + Rl(s) ^ ^ p=lTmod 4) P
+R(s).
We know that log t (s) ->oo as s -> 1 and the terms
R-j_(s), R(s), and 2 all remain bounded as s -> 1 . Also,
L(s,X) -> L(1,X) a convergent series. Our theorem would
be proved if we know that L(l,x) ? 0. This can be seen as
follows
LCI, X) = (1 - ) + ( - ) + ••• > 0.
thus proving the theorem.
PART II
THEOREMS NECESSARY TO THE
PROOF OF DIRICHLET'S THEOREM
CHAPTER III
CHARACrERS
In order to prove the general case of Dirichlet's
theorem, we need to develope the general theory of charac
ters. In this chapter we will develope the theory of
characters for an Abelian group G of order h and then apply
it to our particular needs.
Recall we made the following requirements of a char
acter:
(1) X(n) must be completely multiplicative, (2) X(n) de
pends on the residue class modulo k to v:hich n belongs —
the most important being the reduced residue class, and
(3) X(n) = 0 if (n, k) >1,
If X(a) = 0 for every element in our group, then
X(a) satisfies the multiplicative property, but we agree
to exclude this trivial case. Also the multiplicative
property is satisfied if X(a) = 1 for all elements in the
group. V/e shall call this the principal character and de
note it X]_(a). It is not obvious that there exist other
characters, but we shall show for an Abelian group G of
order h, there are h distinct characters.
A few properties can be proved directly from defi
nition,
22
23
Theorem 3.1. If e is the identity of the Abelian
group G, then X(e) = 1.
Proof. We have
X(e) = X(e2) = (e):! .
Hence X(e) = 0 or X(e) = 1. If X(e) = 0, then X(a) =
X(ae) = X(a)X(e) = 0, for each a in the group, and we have
excluded this case.
Theorem 3.2. X(a) is different from 0 for all a in G.
Proof. Suppose on the contrary that for some a in G,
X(a) = 0. Then X(e) = X(aa-l) = X(a)X(a"^) = 0. Thus if
X(a) = 0. then X(e) = 0, which is a contradiction to Theo
rem 3.1.. so X(a) is different from zero for all a in G.
Theorem 3-3. X(a) is an h^^ root of unity; i.e. X(a)
h satisfies the equation X (a) = 1 for all a in G,
Proof. Let a ^ G. Since G is a group of order h,
a = e. Thus
X^(a) = X(a^) = X(e) = 1.
We now state a fundamental theorem of Abelian groups
which is necessary in our next theorem. A proof to this
theorem can be found in Theory of Groups by Marshall Hall CloU'
Theorem. Let G be an Abelian group of order h; then
there exist in G elements a , a^ a^ of order r-,, r^,
..., r respectively so that every element g of G may be
written uniquely in this form
g = a-, la<, 2 ... a^^^ where G < k . <r, - l ( i = ^ 1 2 m — i — i
l,2,...m) , Moreover ^1*^2'**^m ~ *
24
Theorem 3-4. There are precisely h different charac
ters for a group G of order h.
Proof Let an, ,.,, a be a basis from G and let r, J- m j
be the order of a. (;) = 1,2,... ,m), We define
(1) X(aj) = Pj (j = l,...,m),
th where p. is the r. root of unity and may be expressed as ^ 2i£l
(2) p. = exp ^1 * i 0 < n.<rj - 1.
ki kr Also If g = a- .,, a , let
^1 •. ^1 r (3) X(g) = p{^^ ... p_-^ .
Then X so defined Is a character, for if a and b are two
elements of the group G, we have
SI S2 Sm a — a a ) .,, a J. 2 m
and
ti t2 tn b = a-L -Lag ''...ajj ^ .
Hence from (3)
y/ N ^ SI S2 ^ Sm X(a) = pL P2 •• -Pm
X(b) = p-L • P2 •••Pm^ m
and X(ab) = X (a ^ "' ,, .a/^""^^).
Suppose that s^ + t^ = u. (mod r^), 0 < u^ < r^ - 1 si+ti kiri+ui , ,
then a^ = ^i ^^^ ^^^°^ 2)
= p .
= p"i.
Thus X(ab) = Pi Pa"^ • • • Pm' '" !
25
and X(a)X(b) = p^^^^^Sa'^**^. • •p/"'**'".
I n
But pj = pj i f 1 = n(mod r ^ ) , and hence X(ab) = X(a)X(b).
The numbeir of poss ib le choices for n . I s r . and hence the
t o t a l number of poss ib le choices for X(g) = X ( a . ^ ^ . , , a ^'^) ± m
is ^1*^2" *^m " * ^"^^^^OYQT, no two distinct choices lead
to the same character. For if X, and X are two characters u V
defined by
X (a ) = exp 2ni . ^ 0 < u < r, - 1, " J r j J " " j " " j
and
X (a ) = exp 2TTi . 0 < v, < r, - 1,
then for some j, u ^ . Thus there exist an element g
in G for which
X^(g) 5 X^(g) ,
the element being namely g=a.. There are, therefore, at
least h characters. On the other hand if X is a character,
then X(g) = X(a^ • )*''X(a ° ). Each of the factors is an
th r. root of unity, and so X(g) must coincide with one of
the above choices. There are, therefore, exactly h distinct
characters.
Characters may be combined in the following way: Let
X and X be two characters on the group G, We define for g
in G
X X (g) = x(g) X(g),
Theorem 3»5- Under this rule of composition, the
26
characters of a group G form themselves an Abelian group.
Proof. First, because of its usefulness in the rest
of the proof, we will show x X (g) = X x (g). Let
x(a.) = p where p. = exp(2lll . n,) J J J r . J
where (j = 1, ..., m) and 0 < n. < r - 1 ;
and let X (a ) = q. where q. = exp (2ni/r. . m )
were (j = 1, ..., m) and 0 < m. < r - 1.
Also if g = a, - •••a let 1 r
/ \ ^1 ^r x(g) = p^ ... p^ and x/f \ ^1 ^r
x(g) = qi .-. q^ . Then
X X(g) =x(g) X(g),
. ki ^r N / il kp. ' = (p^ - ...P- ) iq-^ - ...qp ^ ) .
= exp 2TTi C ( ^ + ... + !il±r)^(5i^ +...+!^):] • 1 Tp ri r^ -•
= exp 2TTi L ( - ^ + ... + _£_^) +
(21^ + . . . + ^v^) :i
= (q -'l .-. q/^)(Pi^^--.Pp^^^) '
= X(g) x(g) = X x(g)
The set of characters is closed. Let U/= x X, Then
l^(ab) = X X(ab) = x(ab) X(ab)
= x(a)x(b)X(a)X(b).
= x(a)X(a)x(b)X(b),
= xX(a)xX(b),
= li/(a) "tp'(b). Thus lif is a character. The
associative law nolds for characters. Let a, 3, and y be
27
three characters defined on G such thata-3(g) = x(g) and
PvCg) = X(g). We Wish to show that aX(g) = xyCg).
aX(g) = a(g)X(g) = a(g)aY(g) = a(g) 3(g) yCg).
XY(g) = x(g)Y(g) = a3(g)Y(g) = a(g)3(g)Y(g).
Thus the associative law holds for characters. The princi
pal character X^ acts as the identity since
XjXis) = XX3_(g) = X(g),
If g is an element of G and
X(g) = Vi • .. .Pj. .
let {jjie) = Pi'^'^.-.p/^^ .
Then X(g)i^(g) =i^(g)X(g) = 1 = X^(g).
Thus for a character X(g), there exists an inverse li/(g).
Hence we have shown that the characters of a group G form
an Abelian group.
Theorem 3.6.
h if X = X. ,
> 2 ^(g) = .
g6 G |0 if X 5 X^,
the sum being for a fixed character summed over all elements
of the group.
Proof, If X = X3_, then X^(g) = 1 for all g in G
and so 2 ^1(s) = ^ since the order of the group G is h. g€ G 1
If X 5« X-j_, then let
S = 2 ^(g)-
Suppose that a is an element of G for which X(a) 5 1. Such
an element must exist, for otherwise X viould be the principal
28
character. As g ranges over the elements of G, then so
does a«g once and only once. Then
S = 2 > (g) = 2 5((ag) = 2 X(a)X(g) g€ G g£ G g^ G
= X(a) 2 X(g) = X(a) S, g€G
Since S » X(a) S and X(a) 4 1, then S must be zero. Thus
we have completed our proof, i^
VJe now prove the so-called dual of this theorem.
Theorem 3 . 7 . fh i f g = e ,
2x(g) =. X |0 o t h e r w i s e . \
the sum being taken over all characters X for a fixed
element g of G.
Proof. If g = e then by Theorem 3.1. X(g) = 1 for all
X, and since there are precisely h different characters for
the group G, then
2 5i(e) = h. X
Suppose g 5 e; then we show that there exists a character
X such that X(g) 5 1. To this end we let
g — a-| ...a ^1 ^ im
m since g 5 e; then for some j, k. 0 (mod r^). We define
X(a.) = exp and
X(a^) = X(a2) = ••• X(a 3_) = X(a^^^) = ••. = X(a^) = 1,
Then, as we saw above X defines a character and X(g) =
"9^^^ ? 1 Esince k. 4 O(mod r .) and p. .1 = exp iULi 4 \ \ . J «J J J I* <
Now we let
29
S =2 ^is) •? and since by Theorem 3.5, x X ranges over all characters
whenever X does, we select x(g) 4 1, then^
S = 2 3f(g) =2x(g) X(g) = x(g) 2 ^ ( g ) X X X
= x(g) S
and since x(g) 4 1 and S = x(g) S, then S = 0, Hence we
have completed our proof.
yrl t'r . •r^:r .
CHAPTER IV
THE GENERAL THEORY OP DIRICHLET
SERIES
• = * , -
A Dirichlet series is a series of the form
(1) ? a(n) n=l
Though Dirichlet used only functions of real variables in
his proofs, we shall generalize this by letting s = a + it
and by letting a(n) be a sequence of real or complex num
bers. Here we have followed the standard notation of
denoting the real part of s by the Greek letter a and the
imaginary part by the Latin t.
The theory of Dirichlei series involves many delicate
questions of convergences; here we will establish a few of
these results which will be useful in further development
of our sub.iect.
Theorem 4.1. If the series (1) converges for
S- = a +it , then it converges for all s for which o>a_, o o o o
Moreover, the convergence is uniform inside any angle for
which larg (s-s^)! < n/2-6 for O<6<TT/2.
Proof, The proof is based on a lemma for partial
integration of Stieltjes integrals which we state in the
follovring form.
30
imMiN.
31
Lemma 4.1. Let x > 1 and let 1 (x) be a function vxith
continuous derivative for x > 1. Let Six) = 2 G(n) n<x
where C(n) are r e a l or complex numbers. Then
(2) 2 C(n) ^^(n) = Six) XJJix) -• j ^ S ( t ) ' l / ; ' ( t ) d t .
Proof of Lemma 4.1. Let k be an integer such that
k < X < k+1; then in this interval vre have
tx] k (3) S(x) = 2 C(n) = 2 C(n) = 2 C(n) = S(k).
n<x n=l n=l
where QT] denotes the g r e a t e s t in tege r in x .
We define S(0) = 0., and we get
2 C(n) \ljin) = 2 C(n) ihin) n<x ^ n=l ^
= 2 C(n) -[bin) n=l ^
But s ince
S(n) = 2 C(x) x<n
and S (n - l ) = 2 G(x), x<n-l
we have
S(n) - S (n - l ) = 2 C(x) - 2 C(x).
x<n x<n-l
= C(n).
So r ep lac ing C(n) by S(n) - S(n- l ) and proceeding x\*e have
2 C(n) l//(n) = 2 Cs(n) - S(n- l ) 2 ll/(n) Kx ^ n=l ^
32
2 [IS(n) ^ ( n ) - S ( n - l ) xfjin) J . n=l
k k = 2 Sin) xjjin) - 2 S ( n - l ) i/;(n) ,
n=l ^ n=l ^
k k - 1 = 2 S(n) \ i ;(n) - 2 S(n) i / /(n+l) ,
n=l ^ n=0 ^
on changing the index i n the second sum. Now proceed ing
we have
k k -1 2 C(n) z//(n) = 2 S(n) t/;(n) - S(0) xj/il) - 2 S (n ) i / / (n+ l ) n<x n=l n=l
= 2 " S(n) W;(n) + S(k) W;(k) - 2 S (n ) \ l / (n+ l ) n=l ^ ^ n=l ^
= k - 1 2 S(n) [ : \^ (n) - \pin+l)2 + S(k) i / ; ( k ) .
n=l
By adding and s u b t r a c t i n g S(k) xjjix), we have
k-1 2 Oin) xhin) = 2 C l/ (n) -l/;(n+l) S(n)n + n<x n=l
s(k) H \/;(k) - i/;(x):] + s(k) \ j j i x ) . , I C n+1 , "I
We now replace \p in) - XjJin+l) ^y - - n r ^^^^^ ^^
Jx "? ^1/; (t)dt and obtain
k-1 r n+1 » 2 G(n) W;(n) = - 2 S(n) ] W/ (t)dt n<x ^ n=l ^ n r "" (X •
- S(k) ) ijj (t)dt + S(k)'l//(x).
Since S(x) = S(k) from (3) and S(n) is constant on the
intervals n to n+1 and k to x, we may treat S(n) as a
33
constant and let S(x) = S(k) to obtain
2 C(n) \hin) n<x ^
^-1 /n+1 , t (y. I - 2 ) S(t)i/; (t)dt - )l Sit)xlf (t)dt n=l n ^ k r
+ S(x)V^(x),
= s(x) \/;(x) - 2 i"""- s(t) \//'(t)dt n=l ^
- J s(t)i/;*(t)dt,
= S(x) l/;(x) - C J S(t);/;*(t)dt
+ $2 S(t)i^*(t)dt +••• +
ik-1 s(t)i/;'(t)dt:] - i^ s(t)i/;'(t)dt
= S(x)i|;(x) - ^^ S(t)l/;'(t)dt,
Thus we have completed the proof of the lemma and now re
turn to the proof of Theorem 4.1,
We put 00 GO ^ a(n) = ^ a(n) ^-^ --«=! ^ - n o
a(n)/_s,^ ././„\ _ ^-(S-SQ)
^^3_ ns JTn n'=*o ns-So n=l
= X and C(n) = " ' " V n ^ o , •^^(x) =
then
S(x) = £ C(n) = E ||^ n<x n<x
and xj/'ix) = -(S-SQ) X-^-'^'-^O' .
Now
34
M , . M N Z aiSi = £ aM _ ^ a(ni
= 2 a M . -1 . 2 a(ni . 1
Now using lemma 4.1. we obtain
M z ^ n=N+l
I|i = S(M) ;//(M) - ij S(t) ;/;'(t)dt
- S(N) l/;(N) + J' S(t) l/;'(t)dt.
But since ^(M) = M-(S-SO) ^ r^.j -(S-SQ)
and, l//*(t) = -(S-SQ) t"^-(s-So) ^ ^ ^^^
2 § ^ = S(M)M-^^-^O) + (s.s ) (M s(t) t-l-(s.So)^t n=N+l n " 1
-S(N) N-(s-So) . (3_3^) jN g(^)^-l-(s-So)^^^
Adding and subtracting S(N) M"^^"^o) and combining the
integrals, we have
M 2 ^igi = CS(M) - S(N);] M"^^'^o^
n=N+l n / ^ ^ ^ •+ S(N) Qr^^'^o^ - N- ' o-'I]
+ (S-SQ) i,J S(t) t-l-(^-^o)dt. o N
But S(N) QM'^^'^O) « N'^^'^o^]
= -(S-SQ) ijj 3(N) t"^"^^-^o^dt,
so 2 ^ ^ = CS(M) - S(N)I| iC^^'^o) n=N+l ^
- (s-So) \l S(N) t-l-(^-^o)^t.
fw.r.,v--' 35
+ (s-s^) iJJ S(t) t-^-(^-^o)^t, " N
= CS(M) - S(N)3 M"^^'^O^
+ (S-SQ) iVs(t) - S(N):] t-^-^^-^o^dt. N
00 Since we have assumed the 2 a(n)/n ° converges, then for
n l € > 0 there exists N. such that for all x > N > N-, we
IS(x) - S(N)| < £ ; thus
+ (s-s^) i: Cs(t) - S(N)J t-^-^^"^o^dtl. N
<|S(M) - S(N)| • I M-^s-So) I
/Ml + S-SQ )j^|s(t) - S(N)
<iS(M) - S(N)| I n'^^'^o)
.-l-(s-So) dt ,
by using S(x) - S(N) N
< £ . •
dt .
For a > JQ, vie have
j«|,-l-(s-So)|,, . S« t-^-(°-o)dt = N-(' -°o)-H-'°-°o) N'" "• -N
Now substituting this result in, we have
M 2
n=N+l
a(n) ns
<ls(m) - S(N)
+ 6 Is-s,
. -(s-so)
^vj-(g-go),M"'^^"^Q^
Q-Co
36
We know (S(M) - S ( N ) | < € s i n c e M > N, a n d t h a t | M " ^ ^ " ^ O ^ I =
M-<°-°o) , 0 .
2 ^ < € M-^^-^o) +|s-So)r-N-(cJ-ao).j^-(a-ao)-,\ ln=N+l n " ^ cf-oo '- . - ' /
Now s i n c e j s - s j >/Re ( S - S Q ) | = a-a^ ,
M-^^-^o) <lfl!o! M-^''-°°^ . and - a-Oo
^ inS+1 "^1 ' ( S ' ' ' '^CN-^-o) -M-^^-^'
|s-sc| N--( - o) < £ 1° q N
M O-OQ
s Since 6 is arbitrary, (3) implies that 2 a(n)/n may be n+1
made arbitrarily small for a fixed C!>OQ. Hence the series
converges. Now we need only to show that the series con
verges uniformly inside any angle for which
larg (S-SQ)I <• n/2 - § for 0 <§< n/2.
We know that
arg |s-So|= arc cos jg^s^ ' .
or
COS arg |fe-So) = -j- - j .
so
/s-s I- ^-^o '^ ol - COS arg IS-S-QI .
Hence
37
I S-Sr>l |cf-^ol "
a^a a. COS arg )s-s
1
I • T ^
= cos arg |s-s-| ,
But
cos arg 1S-SQ) — cos |arg s-s )
since arg J(S-SQ)|<^
and js-s jis real.
so
Is-s-s a i^^-^oi — cos Iarg S-SQI
1 = cos(n/2-6) .
1 sin6
Hence from (3) we have
M a(n) 2—^ n+1 ^ ^
(z
slnS
and the convergence is thus uniform inside the angle.
Corollary, If
00
2 n=l
a(n) n
converges, and
f(s) = 2 §l§i ntl ^ ^
then
lim f(s) = 2 ^ ^ S->So n ^
where s->So through values in the region of uniform con-
vergence described above.
38
Proof. Since we are dealing with a uniform convergent
series, under our given hypothesis
lim f(s) = lim 2 ^ ^ s->s, s->s, n-
= 2 lim i ^
= 2^^ A Dirichlet series may converge for all values of s,
or for no values of s, or may converge for some values and
diverge for others. For example, the series
00
1 n In^ converges
for every s, for by the ratio test, vre have for every s
s lim n->a) (n+1):(n+1)" nln^ = lim n
n->oo (n+1)(n+1) = 0 ;
on the other hand, the series
? n'
is always divergent, for by the ratio test
lim n-x»
(n+1 Tn+T
.' . n.' = lim n->aD
n TE+n =± = 00
for every s.
Theorem 4.2 The series (1) either converges for all
values of s or diverges for all values of s, or there exists
a line in the complex plane, a = a, such that (1) con
verges at all points to the right of this line and diverges
at all points to the left, VJe call a the abscissa of con
vergence •
T>
39
Proof, Prom Theorem 4.1, we see that if the series
(1) converges for SQ^a^+itQ, it converges for all s for
which a > OQ, We now divide the real numbers into two
classes R and L. Y belongs to R if series. (1) converges
for all values of o > Y. Otherwise, Y is placed in L, By
Theorem 4,1, every member of R is to the right of every mem
ber of L, and there are members in both classes. Therefore
a Dedekind cut is defined, and the real number defined by
that cut is a, the abscissa of convergence. Thus the re
gion of convergence is a half-plane to the right of a = a.
On the line a = a series (1) may converge or diverge.
Theorem 4.3. The series (1) whose abscissa of conver
gence is a, represents an analytic function in the half
plane a > a.
Proof. Let s_ = a^ + it^ and suppose that a_ > a. o o o o
Then there exists a neighborhood of s^ lying in a region
of uniform convergence. Since each term of the series is
analytic, it follovjs that the sum of the series is an ana
lytic function, for the series converges uniformly in the
prescribed neighborhood of s^ CllH* Theorem 4.4. If a(n) > 0, and f(s) is the function
00 determined by 2 a(n) n , the f(s) has a singularity at
n=l s=a, the abscissa of convergence.
Proof, . The proof is sinilar to its power series
analogue. Let a-, > a and expand f(s) in a Taylor series
about a . We get
40
f(s) = 2 tillio^L. (s-ai)^ k=0 ^! ^
If a is not a singular, then the interior of the circle with
centre a^ and radius a^ - a lies in the half plane a > a,
and has only regular points on its boundary. Hence the
radius of convergence r is greater than a^ - a since the
circle of convergents of the power series must have at
least one singularity on its boundary. The idea of this
proof is that the circle of convergence goes to the left
of a, and so there must be a point to the left of a = a,
at which the Dirichlet series converges. This contradicts
the choice of a. Here are the details: choose € > 0 so
that the radius of convergence of the power series
r > a - L - a + 6 > a - j _ - a . Then the Taylor expansion of
f(a - € ) is;
k=0
This series converges. Since o-^ is in the circle of con
vergence of the power series, the series f(a-j ) =
00 2 ^^^' ^ay be differentiated term by term, and the
n=l n 1 derived series converges and represents the derivative of
the sum of the original power series Cl2]] . Thus
f(ai) = 2 ^ n=l
CO
(4) f(a-6) = I f^^Uo^) ^i=^iS^
f '(ai) = 2 afai (-logon) n=l 1 n ^
4-1
f"(o^) = f a(n) [:-log (n)i2 n=l nc l
f(^)(aO = f a(n) C-logenH^ n ^ nc l
Now substituting this result into (4) we have
ria^) = g Ia=£-lii I (-10 n)^ a(n) k=0 ^' n=l " -
= 5" 5" C(a-£-ai) (-log n)2 a(n) k=0 n=l kJ n^l • .
= 2 2 C(gl-a- €)(log n)n^ a(n) k=0 n=l 2 n l •
Since by hypothesis a(n) > 0 and o-^ - CL +€ > 0, all terms
involved in the double sum are non-negative; we therefore
have an absolutely convergent series and may interchange
the order of summation Cl3ll.
f(a-£) = 2 2 C(gl- - )(log n)J a(n) n=l k=0 k! n l
_ ^ a(n) f Cdos n)(ai-a+£)n^ •" 4^ y,al ^ TTj
n=l n k=0 ^. V/e now have, s ince CO r- . , ^j^\"[^
S* L d o g n)(a2^-a+€)J k=0 k]
ex £ ( l o g n)(ai-a-i^)>n' '^"^-"'^ ,
0 0 , X ^
f (a-e) = j ; ^ 1 : ; ^ . n-cJi+a-^ ,
„^j,<'
42
CO = 2 jg] nei H ^ .
Thus the series with s=a-(£ converges contradicting the
fact that a is the abscissa of convergences. n
Theorem 4.5. Let S(n) = 2 a(k). If 3> 0 and S(n) = k=l
0 (n^), then the abscissa of convergence a is < 3. Thus
if Y is the lower bound of numbers 3 for which S(n) = 0(nP),
then a < Y«
Proof. Since
n n-1 S(n) - S(n-l) = 2 a(k) - 2 a(k) = a(n).
k=l k=l
we have for M > N
M M 5 a(n) = 5; S(n) - S(n-l) n=l\[ nS n=N n^
M M = y ^M ^ S(n-l)
i N ""n^ ^N n^
M . . M S S(n) 5* S(n-l) S(N-l) rfeN""? n=N nS - N^
M _, . M ^ S(n) S S(n) S(N-;1) ^ n^N "^I^ ^ N (nvl)^ ' " " N ^ ,
< S(n) 4 S(^). . S(M) S(N-l) - ^ N " ^ ^ '"n=N Ti TiTys •" Ti+rys - NS ^
4 ^rn^ r 1 1 \ . S(M) S(N-l)
( n+1 ^ ^ •} 1
^ ° 2 • > n 3E^ = -i^ - T ^ s . so
43
M 2 ^ n=N n""
M 2 n=N
S(n) n+1
n dx S(N-l) ^ S(M)
(M+1)S
and since S(n) is constant over the interval n to n+1.
M , . M , n+1 2. 4 ^ = s 2 J n^N n n=N n
S(x)dx - S(N-l) , S(M) xs+x NS "*• (M+1)
= s N+1 ( ^"^ S(x)dx +. . (M+1 S(x)dx N M sTT
S(N-l) S(M) NS IM+TT^
= sf M+1
• N
S(x)dx S(N-l) . S(M) (M+l)S
= 0 , ( " 1 IS(x)l dx ] ^ O(N^) ^ O(M^)
0( isl i H^^ ^ ^ ^ ) ^ 0(N^-^) + 0.(M - )
I M+1 = 0 I (si N x-^-l+^ dxy+ 0(N^-^) + 0(M^-°) ,
= 0 ( |sl^C(M-*-l)^-^-N^-^: )+ 0(N^-^)+0(M^-^).
If a>P, the right-hand side -> 0 as N ->oo , thus proving
that the series converges.
Since the Riemann zeta function is an important exam
ple of a Dirichlet series, v:e will now develop one of its
important properties.
Theorem 4.6. The function
oo (4) ^ (s) = 2
n=l n' where s = a+it is analytic in the
half plane a > 0 except for a simple pole at s = 1 with
residue 1.
44
Proof.
S(n) = 2 a(k) = 2 1 = n, k=l k=l
>
By Theorem 4.5. If 3 > 0 and S(n) = O(n^), then
a is < 3. If vje choose 3 = 1, then these conditions are
siatisfied and a < 1, but (4) converges for a > 1 and di
verges for a = 1 so the abscissa of convergence a must
equal one. V/e now wish to effect a continuation of the
zeta function by the use of the Euler-Mac Laurin formula
which follows.
Lemma 4,2, Let \h{x) be defined and continuous for
n < X < m, where m, n are positive integers; then
^0,) = i : i | i . % ) , $^^u),
+ ')l ^'(x)(x-Cxn-l) dx.
Here CTQ denotes the greatest Integer in x.
Proof of Lemma 4.2. Starting viith the last term on
the right, we have
i^ XJj'ix) x-CxH-t <ix = in^*(x) . X dx - i^"^'(x) K dx
- h )^IJJ (x)dx.
(m I ( n + 1 I (n+2 , i Now ) ' \jj ( x ) x d x = ) ^ \y ( x ) xdx + J^^^*^ (x ) xdx
+ • • •+ ; -, l l ; ( x ) d x , m-1 T
45
"'- k+1 li=n '^ r
s imularly
-i C;i;'(x)dx = -i "I" 'il'^^xh' (x)dx , •m , ., ,. , ^-1 rk+1
k=n
and /m m-1 ,1 +1
- )nDG^'(x)dx = I )^ Zx^2•^|J<U)ix. m-1 /Ir+I
= 2 k r^^ T//'(x)dx ,
k=n ^ ^
by ignoring the discontinuity in the Riemann-Stieltjes
integral at x = k+1. Substituting these results in, we
have frn m—1 /V+"l
(5) )^\}J'M (x-CxH -4)dx = 2 )^ i/;'(x) xdx -I 2 C^^i/;'(x)dx
k=n - ^ m-1 "' • (k+1 , ., ,,
- E ^ )i T^'(x)dx. k=n
If v:e integrate the first integral on the right of (5)
by parts, we get
m-1 f]£+l i ^ - l , ; 2 K it/'(x) xdx = 2 i/;(ic+i) • ( -1) - '^WM k=n ^ ^ k=n^
^^^1 (k+1 I f , ^
k=n ^ ^
Substituting this into (5) and integrating the second and
46
third terms of (5), we have
(m I I m-1 )nV^ (x) ( x - H - i )c ix = 2 C(1^+1);(;(k+1) -k i / ; (k ) J
k=n ^ ^
m-1 - 2 I
k
^ /k+1 m.-l > ). iiy(x)dx - * £ C Ti;(k+l) --d/Ck)] =n ' " !s:=n '
m-1
K.—n
2 i//(k+i) - 2 i r v ^ dx k=n k=n
m - i 2 cv^(k+i) -w/(k):i,
k=n ^ ^
m-1 2 i / / (k+i)
k=n
m-1 m-1 i 2 l / / (k+l) + 1 2 ^ ( i ^ )
m-1
k=n
(k+1
k=n
\ ? . n V ( ) ' k=n
m-1 m-1 /Q = J: 2 \Lii'^+i) + i 2 i i ;(k) - L'u!/(x)dx,
k=n ^ k=n ^ ^ ^
m m-l i 2 \^(^) + - ^2"il;( ) - r^//(x)dx,
k=n+l K=n
1
k=n
- i \i;(x)dx, n -
m (m = 2 V^^^) " 4V^(n) - Sl//(m) - )^ l / ; (x)dx,
k=n Hence we have proved the lemma.
We return to the proof of one continuation of L (s). In the above lemma we put \hix) = x~s x"or s 9« 1 and invest Israte the remainder of the series for the zeta function.
47
E x-^ = *n^ + im^ + \l x-^ax - s "^ x-"-\x-CxD-*)dx. k=n ^ n ' -
.Now investigating the last term on the right, we see
0 < X -CxH < 1 for all x, orfx - C^J - ||< 1,
so it follows that
I s 'll x-^-1 ix - CxH - J)|dx <lsl i 'lx- -l I I x - H -ildx. n (ni _a-l
~ n
<'f a/n" - l/m°).
•" a or
sjl X-"-l (X - Cx: - 4)dX = 0 ( i | . ^f .
Hence if a > 0, the second integral (6) is bounded and
converges absolutely and uniformly as m ->oo. We now let
in ->oo in (6),
m / li2i 2 | s = l ini l /2n^+l /2m^+ ) ^ x"^dx m->a>k=n m->a>^ "
- s )^ x"^" (x -Cx] - i )dx y ,
(?) ^ 2 ^ i s = ^ + )n ^ ^ ^ ^ - s )^ x - " - \ x - C x > * ) d x ,
= 2E^ - \ l i - - S )^ X"^"-^ (x-[IxI]-^r)dX.
CD Now L (s) = 2 ^s ^^^ "'® w r i t t e n ,
^ k=l-^
(8) ^ ( s ) = " 2 i s + 2^^ - ^ 3 - - s C x - ^ " ^ x - C x > ^ ) d x ,
48
This is true for any n. If n=l, then (8) becomes
(9) H s ) = - -1- - s f^ x-^-1 (x- X -*)dx. *-* i-s -
The integral in (9) represents an analytic function for
o > 0, and hence t (s) is analytic for a > 0 except for
the' simple pole at s=l where the residue is 1. Hence we
have proved Theorem 4,6.
Theorem 4,7, If
f(s) = § ainl n=l nS
and g(s) = 1 £%I rfrl n^
and the series converge forOa^, and if f(s) = g(s) for
a > a^, then a(n) = b(n) , n=l,2,3.... .
Proof. Let
GO , .
do) F(s) = 2 ^ T ^ = 0 forvQ > OQ n=l n^
then c(n) = 0 for all n. Indeed, by Theorem 4.1, there
exists a region enclosing the real axis for s > a , in which
the series (10) converges uniformly. To prove that c(n) = 0
for all n, we suppose that c(m) is the first non-zero coef
ficient. Then
(11) 0 = F(s ) = c(m) m -s i J c l m + l i f m + l ] - s U c ( m + 2 i [ m + 2 H . , 1 L V c(m) \ m J J c(m) V m / J ,
= c(m)m"'^ ( l + G(s)) ,
49
where G(s) = S (?- ) (E^] k=1 C(m) V m y
-s
If aQ< a3L< a, then
X r . - ( C - C T ) -ai
m+kj - < m+1j (m+k]
and
JG(S)| < 1 ( m+l\" °'" l m°l^ [c(m+k)l - lc(m)l \ m I 1 1 (m+k)oi .
which tends to 0 when s ->00 through real values of s.
Hence
11 + G(s)l > -I
for sufficiently large s; and (11) implies c(m) = 0, a
contradiction. It follows that if f(s) = g(s) for a > a^,
then a(n) = b(n) for all n. V/e shall occasionally refer
to this theorem as the identity theorem.
Theorem 4.8. If
00
^ ' = nil ^
CO ^ / V
and g(s) = 2 ^
converge absolutely for a > OQ , then for o > o^ ,
f(s)g(s) = I ai&i . i ^ n=l ^ n=l n
00 / N ^ c(n)
— ^.^ y^S n=l ^
where XtAAS TECHNOLOGlCAt COUtUfe LUBBOCK. TEXAS ^ LIBRARY
50
C(n) = 2 a(j) b(k) = 2 a(k) b(§) jk=n k m
= 2 a(§) b(k), kfn ^
the summation in the first expression being over all terms
a(j) b(k) for which jk = n and in the second and third ex
pressions over all factors k of n.
Proof, Since for a > a^, the series f(s) and g(s)
are absolutely convergent, we may multiply them together
and rearrange terms.
CO . . 00 ^,, .
f(s) g(s) = 2 ^ ^ • 2 ^
j=l J k=l ^
When we multiply these two series together in the sense of
forming all possible products with one factor selected from
each series, the general term resulting is a(.l) b(k) _ si( O b(k)
(jk)s - 'i^
where n = jk. If now we add together all terms for which
n has a given value, we obtain a term c(n) n~^, where
c(n) = 2 a(j) b(k) , but jk=n
there is a term of the form c(n) n"^ for each value of n
as n ranges from 1 to oo .
Thus
f(s) g(s) = 2 c(n) n-s n=l
as required.
PART I I I
DIRICHLET'S THEOREM AND SOME
OF ITS CONSEQUExNfCES,
I'
,• '• n
CHAPTER V
PROOF OF DIRICHLET'S. TiiSOREM
V/e now prove Dirichlet's theorem which states that in
any arithmetic progression kn+a, n = 0, 1, 2, .,., where k
and a are relatively prime, there is an infinity of primes.
We let X be a character defined on the set of reduced
residue classes modulo k. Recall in Chapter III we devel
oped the theory of characters for an Abelian group of order
h, hence it is necessary for us to show that the set of
reduced residue classes modulo k form an Abelian group.
Let S be the set of reduced residue classes modulo k,
together with the binary operation multiplication modulo k.
(1) S is a closed set under this operation; i.e. let s^
and Sp^S, then s-, Sp 6S; because, if we let s- = s-j (mod k) I » t
and So = Sp (mod k) where 0 < s-j_ < k and 0 < S2 < k.
Then s- Sp = s- S2 (mod k) , and since (s-j_,k) = 1 and
(sp,k) = 1 it follows that (s]_S2,k) = 1, hence 3 is closed, t
(2) Our operation is associative; i.e., let SQ_ = s-^ (mod k) Sp = Sp (mod k), and s^ = s^ (mod k) S, then
' ' * \ \
s-j_Sp = S-, S2 i'^^orl k) and S2S^ = S2 s^ (mod k)
^SiS2)s3 = s^ S2 s^ (i.od \) s-j_(s2S ) = 3- S2 s^ (mod k).
Hence (s-__"' )s3 = s^is^^s^) {•::io^ k). (3) There is an identity element e =i('-nod 'Oc 3.
52
53
t Let a = a (mod k)€S, then
ae = a (nod k),
ea = a (mod k),
(4) For every element a = a* (mod k) S, there is an in
verse element which may be found by solving the congruence
ax = 1 (mod k).
(5) Our operation is commutative. This follows from the
commutativity of ordinary multiplication and the definition
of congruence modulo a fixed number. Hence we have shown
that S the set of reduced residue classes modulo k with our
binary operation is an Abelian group. The order of this
group is Cp(k), where <i)(k) is Euler's function.
V/e define the Dirichlet L-series as
00 . . L(s,X) = 2-^^-^ where s = a + it.
Theorem 5.1. If X 7 X-, the principal character,
then L(s, X) converges for a > 0.
Proof. In order to show th .t L(s, X) converges, we
need to find the abscissa of convergence a. By Theorem 4.5,
v:e can do this by determining the magnitude of n
S(n) = 2 Xi\n). We wish to divide the numbers m=l
m = 1 to n into coaplete residue systems mciulo k. To do
this we let n = qk + r where q is an integer and r is an
integer snrh that 0 < r < k. Then we have
S(n) = 2 (m) , m=l
54
X 2k 2 ^M + 2 ^M + m=l m=k+l
n=qk+r ••• + 2 ^(°^) •
m=qk+l
Now we wish to divide all except the last of these sums into
two parts, for example the first sum,
2 X(m) = 2 X(m) + 2 (in). m=l m=l m=l
(m,k)=l (m,k)>l
The first sum on the right extends over a reduced residue
system modulo k, and thus by Theorem 3.6. its sum i§ zero.
The second sum on the right is also zero because if (m, k)
> 1, then X(m) = 0 by definition of our character. In a
similar manner, we can show that the next q-1 sums are
each zero, so n=qk+r
S(n) = 2 ^(^). m=qk+l
and n=qk+r
S(n) < 2 IX(m)l, m=qk+l
By Theorem 3.3. X(m), m being an element of the reduced
residue class modulo k, is an ( (k) root of unity, and
hence X(m) can be written as
0 1 P £ Si'^) - 1.
Thus
2
X(m) = expL^I^^ • P
X(m) = 1 if m is an element of the reduced residue
or X(m)
system,
2 = 0 if m is not an element of the reduced
residue system,
55
So n=qk+r
S(n) < 2 1 < r . m=qk+l (m,k)=l
Thus S(n) = 0(1) = O(n^) for every 3 > 0, and by Theorem
4.5. a < 0 making L(s,x) converge fora > 0.
Corollary. The abscissa of convergence is actually
th 0 since for a< 0, the n term does not tend to zero.
Theorem 5*2. For a > 1
L(s,x) =Tr (1 - ^^i|i)-^ . •rx p S P ^
Proof. By definition we have
L(s.X) = 2 ^ n=l
This series is absolutely convergent since by the compari
son test
»X(n)l ^ 1
and the series 2 | - ^ is convergent for a > 1. hence the
L(s,X) series is absolutely convergent. By Theorem 2.2.
we have X(n)v-1 S? X(n) _ TT fi .JLiili)
n=±
the desired result.
Theorem 5.3. L(s,/0 for the principal character X-j_
is analytic for a > 0 except for a simole pole at s = 1.
The residue at s = 1 is
5^
TT (i-i/p) = .dUii = h Plk ^ ^ .
Proof, Examining the principal character of primes,
we see
X^(p) = 1 if (p,k) = 1,
= 0 if (p.k) > 1. i.e. if p k.
From xheorem 5.2., we have
L(s.X) = (1 , Xi(p))-1 ^ p pS
= TT (1 . Xi(p)^~l TT ^ Xi(p))-1 (p.k)=l ps p|k P^
= TT (1 - ^ ) " ^ . (p.k)=l P^
But since
IT (1 - 1/p^)"^ = T[ (1 - i/ps)-l n (1 . l/pS)-l. P (p.k)=l p|k
we have
n (1 - i/pS)-i = n (1 - i/p^)"-^ n (i - i/pS) . (p,k)=l p p|k
so L(s,X) = n (1 - l/p^)~^ V (1 - /P^^ ' P v\^
= ^ (s ) • g (s ) ,
where g(s) = Ti (1 - 1/p^). Since g(s) contains only plk
finitely many terms and each of these terms is analytic,
the product is analytic. By Theorem 4.6 L (s) is analytic
except for a simple pole at s = 1; hence the L-se -ies is
analytic except for a simple pole at s = 1.
51
Residue at s = 1 : lim (s-1) lis) g(s) s->l ^
= lim (s-l)[i - -.1^ - si^ (x-[lxl|-fr)dx s->l
n (1-1/pS) Plk
lim s->l
[ ^ + 1 -s(s-l) i^ (x-dxH-Mx)^
TT (l-l/ps) ] . p|k
= n (1-1/p) , pjk
= Cfe(k)
_ h
We will now prove Dirichlet's theorem by a method
simalar to that used in Euler's proof in Chapter II.
L(s,X) is a complex valued function and has a multiple
valued logarithm. We will choose the branch of log L(s,X)
which is real when s is real and a > 1. For a > 1 we have
log L(s,X) = logTT (1 - ^ - i ) ' ^ P ^
= - 2 log (1 -p
X(p) TDS
)
For a > 1 x\re have X(p) TD'
1 < 1 so
los (1 - ) = - 1 ! ^ n=l ^P
00 (6) log L(s.X) = 2 E § #
p n=l ^
58
^ ps n=2 ^P ^ )
= 2 ^ - ^ E X(P^) p n=:2 ^P
• = 2 % ^ + R(s) . p P
Now we w i s h t o show R ( s ) i s bounded,
R(s)|< 2 P
00 X ( p )
± E 2
n=2 ^P
00
sns
no p n=2nP
00 , 1 ? E 2 rna
p n=2P
00
But 2 - ^ nto P^^
= d - p ' ^ ) " - * - f o r ( a > 0 ) , hence
CO 1 P - 2 ^
„ t 2 F ^ - i-p-° so
-2a lR(s)l l^2f9 0
< 4 . 1 1-2-0
-2a
P
00
2 n -2a 1-2-^ n= l
S i n c e ^ (2 )
a s s - > l ' ^ 0 .
CO = 2 n"^^ i s bounded a s s -> l ' ^^ , R ( s ) r e m a i n s
m=l
59
We now wish to isolate the first sum on the right-hand
side of (6) into two parts, nam.ely p = a (mod k) and p = a
(mod k). Before continuing directly with the general case,
we will look at the case a = 1. V/e now sum (6) over all
characters mod k, and by Theorem 3.4 there are cp(k) = h
of these characters.
Slog L(s.X) =£ 2 ^ +S R(s) . X X p P^ X
The series 2 ps is absolutely convergent since P ^
E p
X(p) 1 E h
which converges for a > 1. Since X is finite and 2 5-
is absolutely convergent, we can rearrange the order of
the terms in the series and obtain
2 log L(s,X) =2 2 ^ + 2 H(s) , X p X P X
2 - L 2 x(p) = 2 H(s) , p P X X
^ - ^ 2 x ( p ) p=l(mod k) X
2 - ^ 2 x ( p ) + 2 H(s). p^Kmod k) P X X
By Theorem 3.7, 2 X(p) can be expressed as X
60
^ h i f p = Kmod k) E X(p) = X 0 otherwTise
Thus
2 log L(s.x) = h 2 -4^+2 R(s) . X p=l(mod k) X
For the more general .case, we find a* such that
aa* = 1 (mod k) .
Multiplying both sides of (6) by X(a*) and summing over
all characters X modulo k, we have E X(a*) log L(s,X) =2X(a*) 2 ^ + 2 X(a*)R(s). X X p ^ X
Since X is finite and 2 ^ E^ is absolutely convergent P P"
2 X(a*) log L(s,X) = 2 - ^ 2 X(a*) X(p) + R*(s) X P ^ X
= 2 - ^ 2 X(a*p) + R*(s) p ^ X
2 ^ ^ 2x(a-p) a «-p=l(mod k) P X
+ 2 1 2 X(a*p)+R*(s), a*PEl(inod k) P ~" X
But by Theorem 3.7,
h if a'--p = Kmod k) 2 X(a- -p) = X 0 otherwise
Also observe that if a*'p=l(mod k). then p = aa^p
=a(mod k), so
61
(7) 2 X(a*) log L(s.X) = h 2 p=a(mod k) ps
+ R*(s),
Now looking at R*(s) = 2 X(a*) R(s), X
we see
1R*(S)1< 21 X(a*) R(s) X
< 2 |H(S)
X
and since R(s) is bound by Theorem 3,7."
R*(s)|< h.
Thus R*(s) remains bounded as s->l'*" , L(s,X) for X 4 Xj^
is analytic at s = 1 by Theorem 5.I. and remains bounded as
s->l**" ; thus log L(s,X) for X 5 X- rem.ains bounded unless
L(1,X) = 0. Also by Theorem 5,3. we see as s->l'*" , L(s,X-j_)
->00 . Hence 2
1 -1+0 . , .
•^ ->oo as s->l , provided p=a(mod k)
L(1,X) 5 0 for X 5 X-L .
Theorem 5.4. If X 5 X^, then L(s,X) 4 0.
Proof. From (6) we have
(8) logTT L(s,X) = 2 l o s L(s,X) . X X
. ^ S f X(pn) X p n=l ^P
CO
n=l p "P X
62
2 ^ 2 , ; ^ 2 x ( p ^ ) n = l p^^sKmod k) ^P X
CO ^
\ i l p lfmod k)^^F^ f ^^P''^' 00
= h 2 2 ^ n ^ l p nP^^
by Theorem 3 . 7 , Now l e t
h I 2 - ^ = 2 ^ = f(s) , n = l p np^^ m=l ^^
where a(m) d e n o t e s t h e c o e f f i c i e n t s of t h e D i r i c h l e t s e r i e s
and f ( s ) i s t h e sum of t h i s s e r i e s . Note t h a t 1 > a(m)
> 0 . Now l e t a be t h e a b s c i s s a of c o n v e r g e n c e , t h e n
F L ( S , X ) = e^(^) = 1 + f(s) + I ^ + . . . I tAsl X 2 n=0 ^ - .
• C D , X
Now f ( s ) = 2 ^ ? and so f o r a > cc, m=l ^
f ^ ( £ ) ^ ^ a ( - ^ ) ( k )
^ - k = l 1 ^
where we have t a k e n t h e p r o d u c t of n D i r i c h l e t s e r i e s of
t h e form 2 ^ ^ t)y Theorem 4 . 8 . Moreover s i n c e 1 > mti ^^
(n) a(m) > 0 i t f o l l o w s t h a t a ^ ^ ( k ) > 0 . Thus by (8) and
( 9 ) . i f a > a 00
(10) TT L(s,X) = 1 + 2 2 § a(^)(k) m=l k = l ' ^
CO T CQ (rr)
1 -f 2 1 2 a^^^^k) , k = l ^ m=l
63
.g b(k) k"s"
the interchange of summation being justified by absolutely
convergent series (the terms are all non-negative).
The idea of the proof that L(1,X) 4 Q, is as follows.
The left-hand side of (10) has a simple pole at s = 1
arising from L(s,X^), the other L series being regular at
S = 1. If L(1,X) = 0, then this zero would cancel the
pole making the left-hand side regular. But we shall show
that the ri>2-ht-hand side has a singularity in the interval
0 < s < 1. This is a contradiction."
From (9) for a > 1, CO S" a m ) < ^ a(m)^ ^ 1 m=l m=l m=l
which converges for a > 1. Thus f(s) converges for a > 1,
This means a< 1, V/e now wish to show that 1/h < a by
showing f(s) diverges at 1/h where h = (i>(k).
f(l/h) = h f 2 l/m P*"" -2 , 2 -THp m=l p° =l(mod p) m=jh p^=l(mod p)
Now by the Euler-Fermat theorem, if pfk, then p sp* =
Kmod k) C15I]. Hence if we sum over all primes such that
p = l(mod p), x\re sum over all primes such that p f k and
n all primes p which divide k and p = 1 (mod k) so
f (1/h) > h ^ E THHi- > h 2 2 -k^
64
00
= h2 E-^= E E P+k m=jh JhpJ p+ic j = i pJ
> E Y- E ^' E I . ptk P p P pfk P
3y Theorem 2,3, E 1/P diverges, and the second sum has P
only finitely many terms. Thus the series f(l/h) diverges
and a > 1/h, CO
By Theorem 4.4., f(s) = E ^^%^ has a singularity at
m=l °^
s = a since a(m) > 0. Let 3 be the abscissa of convergence
of the Dirichlet series for exp (f(s)). If exp (f(s)) con
verges, then f(s) must converge and vice versa, so a = p
where a is the abscissa of convergence for the Dirichlet of
f(s). Since the coefficients of exp (f(s)) are positive,
exp (f(s)) has a singularity at P = a. This m.eans that TT L(s,X) has a singularity at 3 and 1/h < 3 < 1. On the X
other hand, since L(s,X) for X 5 X-j is regular for a > 0,
it follows that TT L(s,X) is regular for 0 < o < 1 except
X
possibly for a simple pole at s = 1, since L(s,X-|_) is regu
lar except for a simnle pole at s = 1.
If now L(1,X) = 0 for some X (say X = X2) then
lim L(s,X2) = A s->oo s-1
some f i n i t e c o n s t a n t . T h e r e f o r e , u s i n g Theorem 5 . 3 ,
l i m TF L ( s , X ) = l i m ( s - 1 ) L(s,X-,) ^ ^ ^ ' f ^ ) TT L( s .X) , s->oo X s->ao ^'^ l-Ai 2
Xy^x;
e5
= § . A TT L(s,X) = finite limit.
X^X2
That is, TT L(s,X) is regular for 0 < a < 1. This contra
dicts the fact that it has a singularity at 3.
The proof of the nonvanishing of L(1,X) and hence of
Dirichlec's theorem is complete.
CHAPTER VI
SOME CONSEQUENCES OF DIRICHLET'S
• THEOREM
Dirichlet's theorem is related to the general theory
of the "qualitatives" distribution of prime ideals. Its
significance is seen in the following way.
Let A. be any set of prime numbers. V/e wish to con
struct a measure of that proportion of all primes which are
contained in the set A. One such measure can be defined
as follows:
s->1^0 P6A
2 1/P^ = ci. P
The upper sum is taken over all Drim.is in the set A, while
the lower sum is taken over all primes. This lim.it is
called the Dirichlet density and is abbreviated as D,D.
We note that the existence of a positive Dirichlet
density for a set A indicates that the set contains infi
nitely many primes. If for d > 0 there vjere only finitely
many primes in A, then as s -> 1" ^ the nur.era.3r would re-
m.a n finite as the denominator ->oo ; the ratio of the
numerator to denominator ', ould -> 0 contradicT:inr- the fact
that d > 0.
66
61
Theorem 6 . 1 . If ( a , k ) = 1 . t hen
lim^ 2 - ^ s~>K^p=a(mod k) P ^ 1
-nS P P
The primes are thus "equidistributed" in all reduced resi
due classes modulo k.
Proof. From equation (8) in Chapter V we have,
2 X(a*) log L(s,X) = h 2 4 - + R*(s) . X p=a(mod k) P
or
2 log L(s.X) + 2 X(a*) log L(s,X) = h 2 ' "Ts + R*(s). X-L X?X-L p=a(mod k)P
where we recall R * ( s ) . and log L(s,X) for X 5 X^ remain
bounded as s->l .
Also in the proof of Theorem 5.3. we saw
L(s,X. ) = tis) TT (1 - i ) . ^ ^ plk P
Now taking logarithms we have
log L(S,XT ) = log r (s) + log TT (1 - is) . ^ ^ plk ^
but
log fis) = - 2 log (1 - P'^) » P
00 ^
= 2 EA p k=l kP*^
68
= 2^^2 P^ p k=2'"kP^ .
= 2 - ^ + R(s) , p P^
where R(s) remains bounded as s-> l"*" .
So
log L(s.Xn ) = log i is) + log TT (1 - -1-.)
p l k p ^
= 2 - ^ + R(s) + log n (1 -p P plk
and
log L(S,X-L) + 2 X(a*) log L(s,X) = h X5^X
+ R- (s)
p=a(mod k) P'
becomes
2^3 +^(s) + log TT (1 - 1 ) + 2 X(a*) log L(s,X) plk
: h
X? X 1 1
p=aTmod k) P + R-(s).
We now divide each term by 2 -^^ and take the limit as
+0
p
1 +0 s->l' . We recall that 2 — ^ ~ > ^ ^s s->l and also
P ^
2 p=a(mod k) ^
1 -. +0 , -, +0 , 4. . 4. •^ ->00 as s->l . As s->l eacn quotient
+0 -> 0 as s->l except for the quotient
h E psa(modk)
2 ^ p
^ ~ ^
which approaches 1, thus
as s -> 1
p=a(mod k) P'
2 fs P ^
+0
- > 1 "^TkT
LIST OF REFERENCES
1. G.H. Hardy and E.M. V/right, The Theory of Numbers, Oxford University Press, Clarendon, I938, 245,
2. E. Hills, Analytic Function Theory, vol. I. Blaisdell, 1963. 108.
3. L.L, Small, Calculus. ApDleton-Century-Crofts. New York, 1949, 384-386.
4. T. Port, Infinite Series. Oxford University Press. Clarendon, 1930, 26.
5. L.L. Small, Elements Of The Theory of Infinite Processes, McGraw-Hill. New York. 1923. 224.
6. T.J, Bromwick, An Introduction To The Theory Of Infinite Series, Macmillan and Company, London, I926, 107-108,
7. L. L. Smail, Calculus. Appleton-Century-Crofts, New York, 1949, 378.
8. I.I. Hirschman, Jr., Infinite Series. Holt, Rinehart and Winston, New York, I962. 28.
9. L.L. Smail, Calculus, Appleton-Gentury-Crofts, New York, 1949. 381.
10. M. Hall, Theory of Group's. Macmillian Company, New York. 1959. 285-291.
11. E. Hille. Analytic Function Theory, vol. I, Blaisdel, 1963, 192.
12. E. Hille. Analytic Function Theory, vol. I. Blaisdel. 1963. 126.
13. S. Hille, Analytic Function Theory, vol. I. Blaisdel. 1963, 116.
14. W.J. Le Veque, Topics In Number Theory, vol. I, Addison Wesley, Reading, 1956, 42.
70