J
OURNAL DE
T
HÉORIE DES
N
OMBRES DE
B
ORDEAUX
A
NTANAS
L
AURIN ˇ
CIKAS
Limit theorem in the space of continuous functions for the
Dirichlet polynomial related with the Riemann zeta-funtion
Journal de Théorie des Nombres de Bordeaux, tome
8, n
o2 (1996),
p. 315-329
<http://www.numdam.org/item?id=JTNB_1996__8_2_315_0>
© Université Bordeaux 1, 1996, tous droits réservés.
L’accès aux archives de la revue « Journal de Théorie des Nombres de Bordeaux » (http://jtnb.cedram.org/) implique l’accord avec les condi-tions générales d’utilisation (http://www.numdam.org/conditions). Toute uti-lisation commerciale ou impression systématique est constitutive d’une infraction pénale. Toute copie ou impression de ce fichier doit conte-nir la présente mention de copyright.
Article numérisé dans le cadre du programme Numérisation de documents anciens mathématiques
Limit Theorem in the
space
of continuous
functions for the Dirichlet
polynomial
related with the Riemann zeta-funtion
par ANTANAS
LAURIN010CIKAS
RÉSUMÉ. Dans cet article on prouve un théorème limite dans l’espace des
fonctions continues pour le polynôme de Dirichlet
$$
où
d03BAT (m)
sont les coefficients du développement en série de Dirichlet dela fonction
03B603BAT (s)
dans le demi-plan 03C3 > 1, 03BAT =(2-1
loglT) ½,
03C3T =½ +
log2 lT/lT,
lT > 0, lT ~ log T et lT ~ ~ lorsque T ~ ~.ABSTRACT. A limit theorem in the space of continuous functions for the Dirichlet polynomial
$$
where
d03BAT (m)
denote the coefficients of the Dirichlet series expansion of the function 03B603BAT(s)
in the half-plane 03C3 > 1,03BAT=(2-1 lnlT)-½,
03C3T =½ + ln2/
lT
and lT > 0, lT~ ln T and lT ~ oo as T ~ ~, is proved.Let s be a
complex
variable and(8),
asusual,
denote the Riemannzeta-function. To
study
the distribution of values of the Riemann zeta-functionthe
probabilistic
methods can beused,
and the obtained resultsusually
arepresented
as the limit theorems ofprobability
theory.
The first theoremsof this
type
were obtained in~1~,~2~,
andthey
wereproved
in[3]-[5]
using
other methods. In modernterminology
we can formulate it as follows. LetManuscrit regu le 20 octobre 1993
*This paper has been written during the author’s stay at the University Bordeaux I in the framework of the program "S~T Cooperation with Central and Eastern European Countries" .
C be the
complex
space and letB(S)
denote the class of Borel sets of the space S. Letmeas{A}
be theLebesgue
measure of the set A and...
where in
place
of dots we write the conditions which are satisfiedby
t. We define theprobability
measureTHEOREM A. For u
> 2
there exists aprobability
measure P on(C,
13(C))
such that
PT
convergesweakly
to P as T ~ oo.More
general
results were obtained in[6~.
Let M denote the space offunctions
meromorphic
in thehalf-plane
Q> 2 ,
equipped
with thetopology
of uniform convergence on
compacta.
Define theprobability
measureTHEOREM B. There exists a
probability
measureQ
on(M,
B(M))
such thatQT
convergesweakly
toQ
as T -~ oo.Note that the
explicit
form of the measureQ
can beindicated, and,
obviously,
Theorem A is acorollary
of Theorem B.The situation is more
complicated
when adepends
on T and tendsto 2
as T - oo, or cr= 2.
It turns out that in this case some powernorming
is necessary.Let lT
> 0 and letlT
tend toinfinity
as T 2013~ oo,or lT
= oo.We take
The case
iT
= oocorresponds
to8T
= 2 ~
The function
is called the characteristic transform of the
probability
measure P on thespace
(C,13(C))
[7].
Thelognormal probability
measure on(C,B(C))
is definedby
the characteristic transformTHEOREM C . The
probability
measureconverges
weakly
to thelognormal probability
measure as T 2013~ oo.Here if
~(s) ~
0, a
ER,
then(a(s)
is understood asexp{alog~(s)}
where
log ~(s)
is definedby
continuousdisplacement
from thepoint s
= 2along
thepath joining
thepoints 2,
2 + it and (1 + it.ivhen %T
= 1
Theorem C wasproved by
A.Selberg
(unpublished),
seealso
(8],
and for different form oflT,
it was obtained in[8]-(10], [5].
Now it arises the
problem
to obtain some results of the kind of TheoremC in the space of continuous functions.
Let
Coo
= C U{oo}
be the Riemannsphere
and letd(si, S2)
be a metricon
Coo
given
by
the formulaeHere s, sl, s2 E
C. This metric iscompatible
with thetopology
ofCoo.
LetC(R)
=C(R, Coo)
denote thespace of continuous functions
f :
R - Cooequipped
with thetopology
of uniform convergence oncompacta.
In thistopology,
sequenceC(1f8)}
converges to the functionf
EC(R)
ifas n -~ oo
uniformly
in t oncompact
subsets of R.The functional
analogue
of theprobability
measure in Theorem C is themeasure
Does this measure converge
weakly
as T ~ oo to someprobability
measureon
(C(R),
,ri(C(II8)))?
At this moment thisquestion
is open and it seems to be very difficult.In the
proof
of Theorem C aninportant
role isplayed by
the Dirichletwhere
d,~ (m)
denote the coefficients of the Dirichlet seriesexpansion
of thefunction in the
half-plane
u > 1(see
[11], [12] ) .
Therefore the aim of this paper is to prove the limit theorem in the space of continuous functionsfor
Su (s) .
This theorem will be the firststep
tostudy
the weak convergenceof the
probability
measure(1).
Now let 2
log T,
aT= 1/2
+ and let lTMoreover we suppose that
for all U > 0 as T --j oo. Here B denotes a number
(not
always
thesame)
which is bounded
by
a constant.THEOREM There exists a
probability
measure P on(C(lI8), t3(C(It8)))
suchthat
PT,ST
convergesweakly
to P as T - oo.Proof of the theorem is based on the
following
probability
result. LetSl
and be two metric spaces, and let h :Si
-~S2
be a measurablefunction. Then every
probability
measure P on(81, B(8I))
induces on(S2, Zi(S2))
theunique
probability
measure definedby
theequality
Ph-’(A)
=P(h-’A),
A EB(82).
Now let h and
hn
be the measurable functions fromSl
intoS2
andLEMMAl. Let P and
Pn
be theprobability
measures on(81,8(81)).
Sup-pose that
Pn
convergesweakly
to P as n - oo and thatP(E)
= 0. Thenthe measure
P,,hn 1
convergesweakly
toPh-’
as n - oo.Proof. This lemma is Theorem 5.5 from
[13].
Let ~y denote the unit circle on
complex
plane,
thatis 1
=Is
E C:1
s1=
1 }.We
putwhere 7P = 1 for each
prime
p. With theproduct topology
andpoint-wise
multiplication
the infinite-dimentional torus 52 is acompact
AbelianThe Fourier transform
g(k)
of the measure P is definedby
the formulaHere k =
(k2,
k3,
...)
whereonly
a finite number ofintegers
kp
are distinctof zero, and xp E -y.
LEMMA 2. Let be a sequence of
probability
measures on(Q,
Ii(S2))
and let
(k) }
be a sequence ofcorresponding
Fourier transforms.Suppose
that for every vector k the limit
g(k)
= limgn(k)
Then there exists -n-oo
a
probability
measure P on(S2, j3(S2))
such thatPn
convergesweakly
to Pas n ~ oo.
Moreover,
g(k)
is the Fourier transform of P.Proof. The lemma is the
special
case of thecontinuity
theorem forcompact
Abelian group, see, forexample,
[14].
Let
LEMMA 3. The
probability
measureQT
convergesweakly
to the Haarmeasure Tn on
(S2,
B(S2))
as T 2013~ oo.Prooi The Fourier transform
9T(k)
of the measureQT
isgiven
by
Here
xp E
q, k
=(ki ,
k2,
...) .
By
definition of the Fourier transform ofprobability
measure on~5~,13(S~~),
only
a finite numberof kj
are distinctfrom zero. Since the
logarithms
ofprime
numbers arelinearly
independent
over the field of rationalnumbers,
we find thatWe define the function
hT :
S2 ~C(R)
by
the formulaHere
pa
11
J~ means thatpO:
h
butpc’+’
Then,
clearly,
Let,
forbrevity,
and let
Let K be a
compact
subset of R. For every 6 > 0 we define the setA~~
by
and we
put
LEMMA 4.
m(Ak(K))
= 0 for every E >0, K,
and k G N.Proof.
By
theChebyshev
inequality
Using
theCauchy formula,
we have thatwhere L denotes the
restangle,
enclosing
the set iK =lia, a
EKI,
withto the real axis.
Moreover,
we suppose that the distance of L from the setiK is > From this
equality
it follows thatHence, having
in mind theinequality
(5),
we obtain thatwhere !
I L
is
thelength
of L. From the definitions of andSn ( Un +
z +
iT)
we havethat,
for z = u +iv,
Since
hence we find that
By
a similar manner we find thatFrom the definition of the contour L it follows that
for z = u + iv E L.Then
(8)
together
with(2)
and the well-known estimatewhere /0 is the Euler
constant,
yields
for n > no. Here we have used the
inequality
01, n >
no,which follows
trivially
from themultiplicativity
of(rrt)
and from theinequality
0l, n >
no,implied by
the formula[11], [12]
Thus,
Consequently,
in view of(
Hence,
for mn,
Repeating
theproof
of Lemma 3 from[10)
andtaking
into account(10),
1we see that , , ,
Consequently,
this and(15)
give
the estimateFrom
this,
(6),
(7)
and(11)
we find thatfor every E > 0 and k G N. Thus it follows from the definition of the set
that
The lemma is
proved.
Proof of Theorem. We will deduce the theorem from lemmas
1,
3 and 4.Let
converges to
as T 2013~ oo, and let E denote the
e"2,
...) I
of elements of Q suchthat
-does not converge to some function
h(t; eiTl, eiT2 ,
...)
as T --+ 00. In order to prove the theorem we must show thatm(E)
= O-Since Q iscompact,
itis
separable. Consequently
[13],
E EB(Q).
does not converge to some function
h(t; ei’Tl, ei’T2, ...)
as T 2013~ oc. We willprove that
m(EI)
= 0. First we consider thesequence
hn(t; ei’Tl, ei’T2, ...).
Note that there exists a sequence of
compact
subsets of R suchthat
--Kj
C and if .K is ascompact
of R then KC Kj
forsome j .
LetThen
is a metric in
C(R) .
Since
C(R)
is acomplete
metric space, we have that every fundamentalsequence is
convergent.
Thus it follows from the definition of the funda-mental sequence thatThus, by
Lemma4,
From the definition of the function
hT, using
the estimates oftypes
(13)
and(14),
we find thatuniformly
E R and inei’T2 , ...)
E 9.Therefore,
in view of( 16) ,
We have shown that there exists a function h such that for almost all
. , . ,
uniformly
in t oncompact
subsets of R.Similarly
as above in the case ofthe variable t it can be
proved
using
theCauchy
formula that for almostall
(ei7"l,
... )
E S2 the relation(17)
is validuniformly
in Ti oncompact
subsets of
R,
uniformly
in 72 oncompact
subsets ofI~, ....
Since thefamily
of sets of m-measure one is closed under countable
intersection,
hence wehave that
(17)
is true for almost all(ei7"l, eiT2, ... )
E Quniformly
in t oncompact
subsets ofR,
the convergencebeing
uniform in 7j oncompact
subsets of R, j = 1, 2, ....
Since,
for every M >0,
in view of the estimate
we have that
h(t;
ei’T2,
...) ~
oc for almost allei’T2,
...)
E 9. The relation(17)
and the uniform convergenceimply
that for almost all(ei’TI,
,.. . )
E 9uniformly
in t oncompact
subsets of R. Thisyields
rri(E)
= O.The latterequality together
with Lemmas 1 and 3 proves the theorem.!!:I..
COROLLARY. There exists a
probability
measure P on(C(R),
13(C(R)))
such that
PT,SnT
convergesweakly
to P 00.Proof Let K be a
compact
subset of R. Denoteby
ZT(it+iT)
the diferenceLet CT =
(log
1T) -’.Then
In view of the
Cauchy
formulawhere L is the contour similar to that in the
proof
of Lemma 4. Hence wefind
by
theMontgomery-Vaughan
theorem fortrigonometrical polynomials
[15], [12]
thatFrom this and from
(18)
we deduce thatBy
(19)
the secondintegral
in the latter formula iso(T)
as T 2013~ oo, andthe first
integral trivially
isBET
T. Hence and from(20)
as oo for every E > 0.
Thus,
thecorollary
follows from Theorem and Theorem 4.1 from[13]:
Let(S, p)
be aseparable
space andXn
andD P
Yn
he S-valued random elements. If andp(Xn,
Yn) ---+
0,
thenD
Acknowledgement.
For the warmhospitality
during
ourstay
at theUniversity
Bordeaux I we express ourdeep
gratitude
to Prof. M.MENDTS
FRANCE,
Prof. M. BALAZARD and to theircolleagues.
We also would like to thank Prof. J.-P. ALLOUCHE and the referee for severalhelpful
remarks.
REFERENCES
[1]
H. Bohr and B. Jessen, Über die Wertverteilung der Riemannschen Zeta funktion, Ernste Mitteilung, Acta Math. 51(1930),
1-35.[2]
H. Bohr and B. Jessen, Über die Wertverteilung der Riemannschen Zeta funktion,Zweite Mitteilung, Acta Math. 58
(1932),1-55.
[3]
B. Jessen and A. Wintner, Distribution functions and the Riemann zeta-function,Trans.Amer.Math.Soc. 38
(1935), 48-88.
[4]
V. Borchsenius and B. Jessen, Mean motions and values of the Riemannzeta-function, Acta Math. 80
(1948),
97-166.[5]
A. Laurin010Dikas, Limit theorems for the Riemann zeta-function on the complexspace, Prob. Theory and Math. Stat., 2, Proceedings of the Fifth Vilnius
Confer-ence, VSP/Mokslas
(1990),
59-69.[7]
A.P.Laurincikas, Distribution of values of complex-valued functions, Litovsk. Math. Sb. 15 Nr.2(1975),
25-39, (inRussian);
English transl. in Lithuanian Math. J.,15, 1975.
[8]
D. Joyner, Distribution Theorems for L-functions, John Wiley (986).[9]
A.P. Laurincikas, A limit theorem for the Riemann zeta-function close to the criti-cal line. II, Mat. Sb., 180, 6(1989),
733-+749, (in Russian); English transl. in Math. USSR Sbornik, 67, 1990.[10]
A. Laurincikas, A limit theorem for the Riemann zeta-function in the complexspace, Acta Arith. 53 (1990), 421-432.
[11]
D. R. Heath-Brown, Fractional moments of the Riemann zeta-function, J.London Math. Soc. 24(2) (1981), 65-78.[12]
A.Ivic, The Riemann zeta-function John Wiley, 1985.[13]
P. Billingsley, Convergence of Probability Measures, John Wiley, 1968.[14]
H.Heyer, Probability measures on locally compact groups, Springer-Verlag,Berlin-Heidelberg-New York