• Nie Znaleziono Wyników

(1)An additive problem with primes and almost-primes by T

N/A
N/A
Protected

Academic year: 2021

Share "(1)An additive problem with primes and almost-primes by T"

Copied!
15
0
0

Pełen tekst

(1)

An additive problem with primes and almost-primes

by

T. P. Peneva and D. I. Tolev (Plovdiv)

1. Introdu tion. In 1937 I. M. Vinogradov [10℄ proved that forevery

suÆ ientlylargeoddintegerN theequation

p

1 +p

2 +p

3

=N

hasa solutioninprimenumbersp

1 ,p

2 ,p

3 .

Twoyears latervan derCorput [9℄used themethodof Vinogradovand

establishedthatthere existin nitelymanyarithmeti progressions onsist-

ingofthree di erentprimes. A orrespondingresultforprogressionsoffour

ormore primeshasnotbeenproved sofar. In 1981,however, D. R.Heath-

Brown[5℄proved thatthereexistin nitelymanyarithmeti progressionsof

fourdi erentterms,threeofwhi hareprimesandthefourthisP

2

(asusual,

P

r

denotesanintegerwithnomore thanrprimefa tors, ounteda ording

to multipli ity). One of the main points in [5℄ is a result of Bombieri{

Vinogradov'stype forthesum

X

x<p

2

;p

3

2x

p

1 +p

3

=2p

2

p2 2p30(modd) u(p

1 )u(p

2 )u(p

3 );

where d is squarefree, (d;6) = 1; u(n) = (logn)=log3x for n  5 and

u(n)=0otherwise.

Re entlyTolev [8℄foundan analogousresult forthequantity

J

k;l (N)=

X

p1+p2+p3=N

p1l(modk) logp

1 logp

2 logp

3

;

whereN isa suÆ ientlylargeoddinteger and(l;k)=1. In [8℄the Hardy{

Littlewood ir le method and the Bombieri{Vinogradov theorem were ap-

plied,as wellassome arguments belonging to H.Mikawa.

1991 Mathemati sSubje tClassi ation: Primary11N36.

(2)

It would be interesting to prove that there exist in nitely many arith-

meti progressions of three di erent primes su h that for two of them, p

1

and p

2

, say, both the numbers p

1

+2, p

2

+2 are almost-primes. In the

presentpaperwestudythisproblem. OurmaintoolisaresultofBombieri{

Vinogradov'stype whi hweestablish usingthemethoddevelopedin[8℄.

Let x be a suÆ iently large real number and k

1 , k

2

be odd integers.

Denote byD

k1;k2

(x)thenumberof solutionsof theequation

(1) p

1 +p

2

=2p

3

inprimesp

1 , p

2 ,p

3

su h that

(2) x<p

1

;p

2

;p

3

3x

and

p

1

+20 (mod k

1 ); p

2

+20 (mod k

2 ):

Let usalso de ne

(x)=

X

x<m

1

;m

2

;m

3

3x

m1+m2=2m3

1

logm

1 logm

2 logm

3

; 

0

=2 Y

p>2



1

1

(p 1) 2



:

We prove thefollowing

Theorem. For ea h A>0 thereexists B =B(A)>0 su h that

XX

k1;k2

p

x=(logx) B

(k

1 k

2

;2)=1

D

k1;k2 (x)



0 (x)

'(k

1 )'(k

2 )

Y

pj(k1;k2) p 1

p 2

 x

2

(logx) A

:

Forsquarefree odd k we de ne J

k

(x) asthe numberof solutions of the

equation (1)inprimessatisfying(2)and su hthat

(p

1

+2)(p

2

+2)0(modk):

The Theoremstated above implies

Corollary 1.For ea h A>0 thereexists B =B(A)>0 su h that

X

k

p

x =(logx) B

(k;2)=1

 2

(k)

J

k

(x) 

0 (x)

Y

pjk

2p 5

(p 1)(p 2)

 x

2

(logx) A

:

Remark. We shall prove the Theorem and Corollary 1 with B =

16A+100.

(3)

Corollary 2. There exist in nitely many triples p

1 , p

2 , p

3

of distin t

primes su h that p

1 +p

2

=2p

3

and (p

1 +2)(p

2

+2)=P

9 .

2. Notations. LetxbeasuÆ ientlylargerealnumberandAapositive

onstant. The onstantsinO-termsand-symbolsareabsoluteordepend

only on A. We shall denote by m, n, d, d

1 , d

2

, a, q, k, k

1 , k

2

, l, r, h, f

integers, byp,p

1 ,p

2 ,p

3

prime numbersand byy,z,t, real numbers. As

usual (n), '(n) denote Mobius's fun tion and Euler's fun tion; 

k (n) is

thenumberof integer solutionsof theequation d

1 :::d

k

=n; (n)=

2 (n).

We denote by (m;n) and [m;n℄ the greatest ommon divisor and the least

ommon multiple of m and n, respe tively. For real y, z, however, (y;z)

denotes the open interval on the real line with endpoints y and z. The

meaning is always lear from the ontext. Instead of m  n (mod k) we

shall writefor simpli itym n(k). We shall also use the notation e(t) =

exp(2it). The letter denotessome positive realnumber, notthe same in

all appearan es. This onvention allowsusto write

(logt)e

p

logt

e

p

logt

;

forexample.

We de ne

(3)

H =

p

x

(logx)

16A+100

; Q=(logx) 4A+20

;  =xQ 1

;

E

1

= [

qQ q 1

[

a=0

(a;q)=1



a

q 1

q

; a

q +

1

q



; E

2

=



1



;1 1







E

1

;

S

k ( )=

X

x<p3x

p 2(k)

e( p); S( )=S

1

( ); V( )= X

x<m3x e( m)

logm

;

E =

XX

k

1

;k

2

H

(k1k2;2)=1

D

k1;k2 (x)



0 (x)

'(k

1 )'(k

2 )

Y

pj(k1;k2) p 1

p 2

:

3. Proof of the orollaries. Supposethat 2

(k)=1and M

1 ,M

2 are

integers. The followingidentityholds:

(4) (k)

X

[k

1

;k

2

℄=k

k1jM1

k

2 jM

2

(k

1 )(k

2 )=



1 ifkjM

1 M

2 ,

0 ifk-M

1 M

2 .

A similar identity has been stated in [1, Lemma 8℄. For onvenien e we

(4)

Ifk-M

1 M

2

theequality(4)isobvious. SupposethatkjM

1 M

2

. Wehave

(k) X

[k1;k2℄=k

k

1 jM

1

k

2 jM

2

(k

1 )(k

2

)=(k) X

k1jk

k

1 jM

1 X

djk1

kd=k

1 jM

2

(k

1 )



kd

k

1



=

X

k

1 j(M

1

;k)

M

2 k

1

0(k)

X

dj(k

1

;M

2 k

1

=k)

(d)=

X

k

1 j(M

1

;k)

M

2 k

1

0(k)

(k1;M2k1=k)=1 1=1;

sin etheonlyintegerwhi hsatis esthe onditionsimposedinthelastsum

isk

1

=k=(k;M

2

): This ompletesthe proof of(4).

Using (4)we get

J

k (x)=

X

x<p

1

;p

2

;p

3

3x

p1+p2=2p3

(k) X

[k

1

;k

2

℄=k

k

1 j(p

1 +2)

k2j(p2+2)

(k

1 )(k

2 ) (5)

=(k) X

[k

1

;k

2

℄=k

(k

1 )(k

2 )D

k

1

;k

2 (x):

Supposethat  2

(k)= 2

(l)=1 and (k;2) =(l;2)=1. We de ne

t(l)= Y

pjl p 1

p 2

; %(l)=

(l)t(l)

'(l)

and

(6) L(k)=(k)

X

[k

1

;k

2

℄=k

(k

1 )(k

2 )

'(k

1 )'(k

2 )

t((k

1

;k

2 )):

It is learthat

L(k)=

(k)

t(k) X

[k1;k2℄=k

(k

1 )(k

2 )t(k

1 )t(k

2 )

'(k

1 )'(k

2 )

=

(k)

t(k) X

[k1;k2℄=k

%(k

1 )%(k

2 )

=

(k)

t(k) X

k

1 jk

X

djk

1

%(k

1 )%



kd

k

1



=

(k)%(k)

t(k) X

k

1 jk

X

djk

1

%(d)

=

(k)%(k)

t(k) X

djk

%(d)



k

d



=

(k)%(k)(k)

t(k)

X

djk

%(d)

(d)

=

(k)%(k)(k)

t(k)

Y



1+

%(p)

2



:

(5)

Usingthede nitionsof %(l) and t(l) we easily ompute

(7) L(k)=

Y

pjk

2p 5

(p 1)(p 2) :

Nowweapply(3), (5){(7) andthe Theoremto obtain

X

kH

(k;2)=1

 2

(k)

J

k

(x) 

0 (x)

Y

pjk

2p 5

(p 1)(p 2)

= X

kH

(k;2)=1

 2

(k)



(k) X

[k

1

;k

2

℄=k

(k

1 )(k

2 )



D

k1;k2 (x)



0 (x)

'(k

1 )'(k

2 )

Y

pj(k

1

;k

2 )

p 1

p 2





XX

k

1

;k

2

H

(k

1 k

2

;2)=1

D

k

1

;k

2 (x)



0 (x)

'(k

1 )'(k

2 )

Y

pj(k

1

;k

2 )

p 1

p 2

 x

2

(logx) A

:

Corollary1 isproved.

Consider thesequen e

A=f(p

1

+2)(p

2

+2)jx<p

1

;p

2

3x; (p

1 +p

2

)=2 primeg

and letB bethe setof odd primes. De ne

X =

0

(x); !(k)=k Y

pjk

2p 5

(p 1)(p 2) :

We applyTheorem10.3of [3℄ hoosing=2, =1=4, =4:1,  =0:4. It

is lear thatwe may get ridthe extra fa tor 3

(d)

in the onditionR (; )

using,forexample, theCau hyinequality. We obtain

jfP

9 :P

9

2Agj x

2

log 5

x :

Sin e the ontributionof theterms forwhi h p

1

=p

2

is at most O(x),

thelastestimate provesCorollary 2.

4. Proof of the Theorem. Itis learthat

D

k

1

;k

2 (x)=

1 1=

\

S

k

1 ( )S

k

2

( )S( 2 )d =D (1)

k1;k2

(x)+D (2)

k1;k2 (x);

(6)

where

D (i)

k

1

;k

2 (x)=

\

E

i S

k

1 ( )S

k

2

( )S( 2 )d ; i=1;2:

Consequently,

(8) E E

1 +E

2

;

where

E

1

=

XX

k1;k2H

(k

1 k

2

;2)=1

D

(1)

k1;k2 (x)



0 (x)

'(k

1 )'(k

2 )

Y

pj(k

1

;k

2 )

p 1

p 2

; (9)

E

2

=

XX

k

1

;k

2

H

(k

1 k

2

;2)=1 jD

(2)

k1;k2 (x)j:

(10)

TheproofoftheTheoremfollowsfrom(3),(8){(10)andfromtheinequalities

E

1

 x

2

(logx) A

; E

2

 x

2

(logx) A

:

4.1. Theestimate of E

1

. Wehave

(11) D

(1)

k

1

;k

2 (x)=

X

qQ q 1

X

a=0

(a;q)=1

I(a;q);

where

(12) I(a;q)= 1=(q)

\

1=(q) S

k

1



a

q +



S

k

2



a

q +



S



2



a

q +



d :

If

(13) qQ; (a;q)=1; j j

1

q

thenwehave

(14) S



2



a

q +



= h(q)

'(q)

V( 2 )+O(xe

p

logx

);

where

h(q)= q

X

m=1

(m;q)=1 e



2m

q



=





q

(q;2)



'



q

(q;2)



'(q)

(the proofis similarto thatof [6,Lemma 3,X℄).

Consider S

k

(a=q+ )fora, q, satisfying(13). Wearenotableto nd

(7)

some hypotheses whi h have notbeenproved yet). We shall nd,however,

an asymptoti formulawithan errortermwhi his smallon average.

We have

(15) S

k



a

q +



=

X

1mq

(m;q)=1

m 2((k;q)) e



am

q



T( );

where

T( )= X

x<p3x

p 2(k)

pm(q)

e( p):

Using the elementary theory of ongruen es one may easily prove that

ifthe integersk, m, q satisfy(k;2) =(m;q) =1 and m  2((k;q))then

there existsan integer f =f(k;m;q) su h that(f;[k;q℄)=1 and su hthat

for any integer n the ongruen e n  f([k;q℄) is equivalent to the system

n 2(k),nm(q). Hen e we have

T( )= X

x<p3x

pf([k;q℄) e( p):

We de ne

(t;h)=max

yt max

(l;h)=1

X

py

pl(h) logp

y

'(h)

:

UsingAbel'sformulawe obtain

T( )= 3x

\

x



X

x<pt

pf([k;q℄) logp



d

dt



e( t)

logt



dt+



X

x<p3x

pf([k;q℄) logp



e(3 x)

log3x

= 3x

\

x



t x

'([k;q℄)

+O((3x;[k;q℄))



d

dt



e( t)

logt



dt

+



2x

'([k;q℄)

+O((3x;[k;q℄))



e(3 x)

log3x

= 1

'([k;q℄)



3x

\

x

(t x) d

dt



e( t)

logt



dt+2x

e(3 x)

log3x



+O((1+j jx)(3x;[k;q℄)):

(8)

3x

\

x e( t)

logt

dt=V( )+O(1)

to get

T( )=

V( )

'([k;q℄) +O



Q

q

(3x;[k;q℄)



:

We substitute this expressionfor T( ) in (15) and we ndthat under the

ondition(13)wehave

(16) S

k



a

q +



=

k (a;q)

'([k;q℄)

V( )+O(Q(3x;[k;q℄));

where

(17)

k

(a;q)=

X

1mq

(m;q)=1

m 2((k;q)) e



am

q



:

Anexpli itformulaforthequantity

k

(a;q)isfoundin[7,p. 218℄. Itimplies

that

(18) j

k

(a;q)j1:

Furthermore, weshall usethetrivialestimates

(19)

S

k



a

q +



 x

k

; jV( )j x

logx

; jh(q)j1:

From(14),(16),(18),(19)andthewell-knownestimate'(n)n(loglogn) 1

we get

S

k

1



a

q +



S

k

2



a

q +



S



2



a

q +

 

=S

k1



a

q +



S

k2



a

q +



h(q)

'(q)

V( 2 )+O



x 3

k

1 k

2 e

p

logx



=S

k

1



a

q +



h(q)

'(q)



k

2 (a;q)

'([k

2

;q℄)

V( )V( 2 )

+O



Qx 2

qk

1

(3x;[k

2

;q℄)



+O



x 3

k

1 k

2 e

p

logx



= h(q)

k

1 (a;q)

k

2 (a;q)

'(q)'([k

1

;q℄)'([k

2

;q℄) V

2

( )V( 2 )+O



x 3

k

1 k

2 e

p

logx



+O



Qx 2

qk

(3x;[k

1

;q℄)



+O



Qx 2

qk

(3x;[k

2

;q℄)



:

(9)

Forthe integralI(a;q) de nedby(12), we nd

I(a;q)= h(q)

k1 (a;q)

k2 (a;q)

'(q)'([k

1

;q℄)'([k

2

;q℄) 1=(q)

\

1=(q) V

2

( )V( 2 )d (20)

+O



xQ 2

k

2 q

2

(3x;[k

1

;q℄)



+O



xQ 2

k

1 q

2

(3x;[k

2

;q℄)



+O



x 2

k

1 k

2 e

p

logx



:

We also have

(21)

1=(q)

\

1=(q) V

2

( )V( 2 )d = (x)+O(q 2

 2

)

(theproofisanalogoustothatin[6,Lemma4,X℄).Using(18){(21) we nd

that

I(a;q)= h(q)

k

1 (a;q)

k

2 (a;q)

'(q)'([k

1

;q℄)'([k

2

;q℄)

(x)+O



q 2

 2

'(q)'([k

1

;q℄)'([k

2

;q℄)



(22)

+O



xQ 2

k

2 q

2

(3x;[k

1

;q℄)



+O



xQ 2

k

1 q

2

(3x;[k

2

;q℄)



+O



x 2

k

1 k

2 e

p

logx



:

Set

b

k

1

;k

2 (q)=

q 1

X

a=0

(a;q)=1

k

1 (a;q)

k

2 (a;q);

(23)



k1;k2 (q)=

h(q)b

k1;k2

(q)'((k

1

;q))'((k

2

;q))

' 3

(q)

: (24)

From (11), (22){(24) and thewell-knownformula

'([k;q℄)'((k;q))='(k)'(q)

we get

D (1)

k

1

;k

2 (x)=

(x)

'(k

1 )'(k

2 )

X

qQ



k1;k2

(q)+O



 2

(logx) X

qQ q

2

[k

1

;q℄[k

2

;q℄



(25)

+O



xQ 2

X

qQ

(3x;[k

1

;q℄)

k

2 q



+O



xQ 2

X

(3x;[k

2

;q℄)

k

1 q



+O



x 2

k

1 k

2 e

p

logx



:

(10)

Considerthefun tion b

k

1

;k

2

(q). From(18) and (23)wehave

(26) jb

k

1

;k

2

(q)j'(q):

ItisnotdiÆ ulttosee thatb

k1;k2

(q) ismultipli ativewithrespe tto q and

thatforprime p we have

(27) b

k

1

;k

2 (p)=

8

>

>

>

<

>

>

>

:

p 1 ifp-k

1 ,p-k

2 ,

1 ifpjk

1 ,p-k

2 ,

1 ifp-k

1 ,pjk

2 ,

1 ifpjk

1 ,pjk

2 .

We alsohaveb

k1;k2

(4)=0. Thereforethefun tion

k1;k2

(q)de nedby(24)

is multipli ative with respe t to q and 

k

1

;k

2 (p

l

) = 0 if l  2. We apply

Euler's identity (see [4, Theorem 286℄) and also (19), (26), (27) and the

de nitionof

0

. After some al ulationswe get

(28)

X

qQ



k1;k2

(q)=

0 Y

pj(k1;k2) p 1

p 2 +O



X

q>Q (k

1

;q)(k

2

;q)

' 2

(q)



:

From(25), (28)and thetrivialestimate

(x) x

2

log 3

x

we obtain

D (1)

k

1

;k

2 (x)=



0 (x)

'(k

1 )'(k

2 )

Y

pj(k

1

;k

2 )

p 1

p 2 +O



x 2

X

q>Q (k

1

;q)(k

2

;q)logq

k

1 k

2 q

2



(29)

+O



xQ 2

X

qQ

(3x;[k

1

;q℄)

k

2 q



+O



xQ 2

X

qQ

(3x;[k

2

;q℄)

k

1 q



+O



 2

(logx) X

qQ q

2

[k

1

;q℄[k

2

;q℄



+O



x 2

k

1 k

2 e

p

logx



:

Using(9)and (29) we nd

(30) E

1

xQ 2



1 +

2

(logx)

2 +x

2



3 +x

2

e

p

logx

;

where



1

= X

k

1

;k

2

H X

qQ

(3x;[k

2

;q℄)

k

1 q

; 

2

= X

k

1

;k

2

H X

qQ q

2

[k

1

;q℄[k

2

;q℄

;



3

= X

k1;k2H X

q>Q (k

1

;q)(k

2

;q)logq

k

1 k

2 q

2 :

(11)

Consider 

1

. We have



1

(logx) X

kH X

qQ

(3x;[k;q℄)

q

=(logx) X

hHQ

(3x;h)(h);

where

(h)= X

kH X

qQ

[k;q℄=h 1

q

= X

dQ X

kH X

qQ

[k;q℄=h

(k;q)=d 1

q

 X

dQ X

qQ

q0(d) 1

q

log 2

x:

Hen e



1

(log 3

x) X

hHQ

(3x;h):

Nowweusethe de nitionsofH,Qand theBombieri{Vinogradovtheorem

(see [2, Chapter28℄, forexample) andwe nd

(31) 

1



x

(logx) 12A+72

:

We now treat 

2

. We have



2

= X

d

1

;d

2

Q d

1 d

2 X

k

1

;k

2

H X

qQ

(k

1

;q)=d

1

(k2;q)=d2 1

k

1 k

2

Q X

d

1

;d

2

Q d

1 d

2

[d

1

;d

2

℄ X

k

1

;k

2

H

k10(d1)

k

2

0(d

2 )

1

k

1 k

2

Q(log 2

x)



;

where





= X

d1;d2Q 1

[d

1

;d

2

= X

dQ X

d1;d2Q

(d1;d2)=d d

d

1 d

2 (32)

 X

dQ 1

d X

d1Q=d 1

d

1 X

d2Q=d 1

d

2

log 3

x:

Hen e

(33) 

2

Qlog 5

x:

To ompletetheestimateof E

1

we have to onsider

3

. Obviously

(34) 

3

= X

d

1

;d

2

H d

1 d

2 X

k

1

;k

2

H X

q>Q

(q;k1)=d1

(q;k )=d

logq

k

1 k

2 q

2

=

4 +

5

;

(12)

where



4

= X

d

1

;d

2

H

[d1;d2℄>Q d

1 d

2 X

k

1

;k

2

H X

q>Q

(q;k

1 )=d

1

(q;k2)=d2

logq

k

1 k

2 q

2

;



5

= X

[d

1

;d

2

℄Q d

1 d

2 X

k1;k2H X

q>Q

(q;k

1 )=d

1

(q;k

2 )=d

2

logq

k

1 k

2 q

2 :

We have



4

 X

d

1

;d

2

H

[d

1

;d

2

℄>Q d

1 d

2 X

k

1

;k

2

H

k

1

0(d

1 )

k

2

0(d

2 )

1

k

1 k

2

X

q>Q=[d1;d2℄

log (q[d

1

;d

2

℄)

q 2

[d

1

;d

2

℄ 2 (35)

(logx) X

d

1

;d

2

H

[d

1

;d

2

℄>Q 1

[d

1

;d

2

℄ 2

X

k

1

H =d

1 1

k

1 X

k

2

H =d

2 1

k

2 1

X

q=1

(1+logq)

q 2

(log 3

x) X

h>Q 1

h 2

X

[d1;d2℄=h

1(log 3

x) X

h>Q



3 (h)

h 2

 log

4

x

Q :

Forthe sum

5

we nd



5

(log 3

x) X

[d

1

;d

2

℄Q 1

[d

1

;d

2

℄ 2

X

q>Q=[d

1

;d

2

℄ logq

q 2

 log

4

x

Q





where



is de nedby(32). Consequently,

(36) 

5

 log

7

x

Q :

Finally, ombining(30), (31), (33){(36) and usingthe de nitionsof Q and

 we get

E

1

 x

2

(logx) A

:

4.2. Theestimate of E

2

. It is lear that

E

2

 X

k1;k2H

\

E2 S

k1 ( )S

k2

( )S( 2 )d

:

(13)

Usingthede nitionof S

k

2

( ) weget

E

2

 X

k1;k2H X

x<p3x

p 2(k

2 )

\

E2 S

k

1

( )S( 2 )e( p)d

 X

k1;k2H

X

(x+2)=k

2

<r(3x+2)=k

2

\

E2 S

k1

( )S( 2 )e( 2 )e( rk

2 )d

 X

k

1

H X

n3x+2



X

k

2

H

X

(x+2)=k2<r(3x+2)=k2

rk2=n

1





\

E

2 S

k1

( )S( 2 )e( 2 )e( n)d

 X

kH X

n3x+2

(n)

\

E

2 S

k

( )S( 2 )e( 2 )e( n)d

:

By Cau hy'sinequalityweget

E

2





X

kH X

n3x+2

 2

(n)

k



1=2



X

kH k

X

n3x+2

1 1=

\

1=

f( )e( n)d

2



1=2

;

where

f( )=



S

k

( )S( 2 )e( 2 ) if 2E

2 ,

0 if 2E

1 .

We nowapplyBessel'sinequalityto obtain

E

2

x 1=2

(log 2

x)



X

kH k

\

E

2 jS

k

( )S(2 )j 2

d



1=2

(37)

x 1=2

(log 2

x)



X

kH k

X

x<p

1

;p

2

3x

p

1

p

2

 2(k)

\

E

2

jS(2 )j 2

e((p

1 p

2 ) )d



1=2

=x 1=2

(log 2

x)

1=2

; say :

We have

 = X

kH k

X

jrj2x

r0(k)



XX

x<p

1

;p

2

3x

p

1

p

2

 2(k)

p

1 p

2

=r 1



\

E

2

jS(2 )j 2

e( r)d

(38)

x X

kH X

jrj2x

r0(k)

\

E

2

jS(2 )j 2

e( r)d

0 00

(14)

where

 0

= X

kH



\

E2

jS(2 )j 2

d



;

 00

= X

kH X

1jrj2x

r0(k)

\

E2

jS(2 )j 2

e( r)d

:

Obviously

(39) 

0

H 1

\

0

jS(2 )j 2

d  Hx

logx :

By theCau hyinequalitywe nd

 00

 X

kH X

1r2x

r0(k)

\

E

2

jS(2 )j 2

e(r )d

= X

1r2x



X

kH

kjr 1



\

E

2

jS(2 )j 2

e(r )d

 X

1r2x

(r)

\

E2

jS(2 )j 2

e(r )d





X

1r2x

 2

(r)



1=2



X

1r2x

1 1=

\

1=

g( )e(r )d

2



1=2

;

where

g( )=



jS(2 )j 2

if 2E

2 ,

0 if 2E

1 .

We again applytheBesselinequalityto obtain

 00

x 1=2

(logx) 3=2



\

E

2

jS(2 )j 4

d



1=2

(40)

x 1=2

(logx) 3=2

sup

2E

2

jS(2 )j

 1

\

0

jS(2 )j 2

d



1=2

x(logx) sup

2E

2

jS(2 )j:

Usingthede nitionsofQ, and E

2

we an proveinthesamewayasin

[6, Theorem3,X℄that

(41) sup

2E

jS(2 )j

x

(logx) 2A+7

:

(15)

From (37){(41) we obtain

E

2

 x

2

(logx) A

:

The Theoremisproved.

Finally, the authors would like to thank the Ministry of S ien e and

Edu ationof Bulgariafor nan ial supportunder grant MM{430.

Referen es

[1℄ J.BrudernandE.Fouvry,Lagrange'sFourSquaresTheoremwithalmostprime

variables,J.ReineAngew.Math.454(1994),59{96.

[2℄ H. Davenport, Multipli ative Number Theory (revisedbyH. Montgomery),2nd

ed.,Springer,1980.

[3℄ H. Halberstam and H.-E. Ri hert, Sieve Methods, A ademi Press, London,

1974.

[4℄ G.H.HardyandE.M.Wright,AnIntrodu tiontotheTheory ofNumbers,5th

ed.,OxfordUniv.Press,1979.

[5℄ D.R.Heath-Brown,Threeprimesandanalmost-primeinarithmeti progression,

J.LondonMath.So .(2)23(1981),396{414.

[6℄ A. A.Karatsuba,Prin iples of Analyti Number Theory,Nauka,Mos ow, 1983

(inRussian).

[7℄ H.Maier andC.Pomeran e,Unusually largegaps between onse utive primes,

Trans.Amer.Math.So .322(1990),201{237.

[8℄ D.I.Tolev,Onthenumberofrepresentationsofanoddinteger asasumofthree

primes, one of whi h belongs to an arithmeti progression, Pro . Steklov Math.

Inst.,toappear.

[9℄ J. G.van der Corput,



Uber Summenvon Primzahlen und Primzahlquadraten,

Math.Ann.116(1939),1{50.

[10℄ I. M. Vinogradov, Representation of an odd number as a sum of three primes,

Dokl.Akad.NaukSSSR15(1937),169{172(inRussian).

DepartmentofMathemati s

PlovdivUniversity\P.Hilendarski"

\TsarAsen"24

Plovdiv4000,Bulgaria

E-mail:tpenevaul .uni-plovdiv.bg

dtolevul .uni-plovdiv.bg

Re eivedon24.3.1997 (3150)

Cytaty

Powiązane dokumenty

Mr Smith buys one randomly chosen lottery ticket every day (we assume that tickets from the two lotteries are equally numerous on all days); the choices on different days

Let Z, N, Q be the sets of integers, positive integers and rational numbers respectively, and let P be the set of primes and prime powers. In this note we prove the following

For any fixed 20/21 &lt; γ ≤ 1, every sufficiently large odd integer N can be written as a sum of three Piatetski-Shapiro primes of type γ..

In the present paper we show that if instead of primes one asks for almost primes of some fixed order r (that is, numbers with at most r prime factors, counted with multiplicity, or P

Heath- Brown [6] proved that there exist infinitely many arithmetic progressions of four different terms, three of which are primes and the fourth is P 2 (as usual, P r denotes

However, it is necessary that P(x) and Q(x) assume simultaneously arbitrarily large positive values prime to any fixed natural number.. This condition will be

Besides the main Theorem 1 we shall establish several easier results on norms of ideals in abelian fields in place of primes (see Section 5).. We conclude this introduction

The last three results (one for each discriminant congruent to 5, 4 or 0 modulo 8) together with examples give a complete description of the el- ementary abelian 2-subgroups of