PREHISTORY OF THE VORONIN UNIVERSALITY THEOREM 1 ...

PREHISTORY OF THE VORONIN UNIVERSALITY THEOREM 1 ... PREHISTORY OF THE VORONIN UNIVERSALITY THEOREM 1 ...

siauliaims.su.lt
from siauliaims.su.lt More from this publisher
17.01.2015 Views

’iauliai Mathematical Seminar 1(9), 2006, 4153 PREHISTORY OF THE VORONIN UNIVERSALITY THEOREM Antanas LAURINƒIKAS Vilnius University, Naugarduko 24, LT03225, Vilnius, ’iauliai University, P. Vi²inskio 19, LT-77156 ’iauliai, Lithuania; e-mail: antanas.laurincikas@maf.vu.lt Abstract. The results of H. Bohr, R. Courant, B. Jessen and S. M. Voronin which inuenced the Voronin universality theorem for the Riemann zetafunction are surveyed and discussed. Key words and phrases: denseness, Kronecker theorem, permutation of terms, Riemann zeta-function, universality. Mathematics Subject Classication: 11M06. 1. Introduction The theory of the Riemann zeta-function ζ(s), s = σ + it, occupies one of central places in analytic number theory. We recall that ζ(s) is dened, for σ > 1, by ζ(s) = ∞∑ m=1 1 m s , and by analytic continuation elsewhere. Many problems of analytic number theory and, in general, of mathematics are inuenced by value distribution of the function ζ(s). In particular, the zero distribution of ζ(s) is closely related to the prime number theorem. For the rst time, this was observed by B. Riemann in 1859 [9]. Ch. de la Vallée Poussin [10] and J. Hadamard [5] successfully developed Riemann's ideas and proved independently the prime number distribution law π(x) = ∑ px 1 ∼ lix = ∫ x 2 du , x → ∞. (1) log u

’iauliai Mathematical Seminar<br />

1(9), 2006, 4153<br />

<strong>PREHISTORY</strong> <strong>OF</strong> <strong>THE</strong> <strong>VORONIN</strong><br />

<strong>UNIVERSALITY</strong> <strong>THE</strong>OREM<br />

Antanas LAURINƒIKAS<br />

Vilnius University, Naugarduko 24, LT03225, Vilnius,<br />

’iauliai University, P. Vi²inskio 19, LT-77156 ’iauliai, Lithuania;<br />

e-mail: antanas.laurincikas@maf.vu.lt<br />

Abstract. The results of H. Bohr, R. Courant, B. Jessen and S. M. Voronin<br />

which inuenced the Voronin universality theorem for the Riemann zetafunction<br />

are surveyed and discussed.<br />

Key words and phrases: denseness, Kronecker theorem, permutation of terms,<br />

Riemann zeta-function, universality.<br />

Mathematics Subject Classication: 11M06.<br />

1. Introduction<br />

The theory of the Riemann zeta-function ζ(s), s = σ + it, occupies one of<br />

central places in analytic number theory. We recall that ζ(s) is dened, for<br />

σ > 1, by<br />

ζ(s) =<br />

∞∑<br />

m=1<br />

1<br />

m s ,<br />

and by analytic continuation elsewhere. Many problems of analytic number<br />

theory and, in general, of mathematics are inuenced by value distribution<br />

of the function ζ(s). In particular, the zero distribution of ζ(s) is closely<br />

related to the prime number theorem. For the rst time, this was observed<br />

by B. Riemann in 1859 [9]. Ch. de la Vallée Poussin [10] and J. Hadamard [5]<br />

successfully developed Riemann's ideas and proved independently the prime<br />

number distribution law<br />

π(x) = ∑ px<br />

1 ∼ lix =<br />

∫ x<br />

2<br />

du<br />

, x → ∞. (1)<br />

log u


42 A. Laurin£ikas<br />

Here and in the sequel, p denotes a prime number, and f(x) ∼ g(x), x → a,<br />

means that lim f(x)/g(x) =<br />

x→a<br />

1. The equality (1) follows from the fact that<br />

ζ(σ + it) ≠ 0 for σ 1. The famous Riemann hypothesis (RH), non-proved<br />

and non-disproved till our days, asserts that all non-trivial zeros of ζ(s) lie<br />

on the critical line σ = 1 2, i. e., ζ(σ + it) ≠ 0 for σ > 2. 1 H. von Koch proved<br />

[7] that RH is equivalent to the estimate<br />

π(x) = lix + O(x 1 2 log x).<br />

Similarly ζ(σ + it) ≠ 0, for σ > α, 1 2<br />

< α < 1, if and only if, for any ε > 0,<br />

π(x) = lix + O(x α+ε ).<br />

All these results show the importance of the value distribution of the function<br />

ζ(s).<br />

In 1975, a Russian mathematician S. M. Voronin (19471997) discovered<br />

that the set of values of ζ(s) in the strip {s ∈ C : 1 2<br />

< σ < 1} is very rich,<br />

every analytic function can be approximated by shifts ζ(s + iτ). He called<br />

the latter property of ζ(s) the universality. More precisely, the Voronin<br />

universality theorem is the following statement [6], [14].<br />

Theorem V. Let 0 < r < 4. 1 Let f(s) be a continuous non-vanishing function<br />

on the disc |s| r which is analytic in the interior of this disc. Then,<br />

for every ε > 0, there exists a real number τ = τ(ε) such that<br />

(<br />

max<br />

∣ ζ s + 3 ) 4 + iτ − f(s)<br />

∣ < ε.<br />

|s|r<br />

Theorem V is a result of development of investigations of H. Bohr, R. Courant,<br />

B. Jessen and S. M. Voronin on the denseness of some sets of values of<br />

the function ζ(s). The aim of this note is to recall the mentioned results.<br />

2. Bohr's denseness theorem<br />

H. Bohr, probably, was the rst who began to study the value distribution of<br />

the Riemann zeta-function. Among his other results, the following theorem<br />

obtained jointly with R. Courant occupies a particular place, see [1], [2], [3].<br />

Theorem B 1 . Let σ, 1 2<br />

< σ 1, be xed. Then the set {ζ(σ + it) : t ∈<br />

(−∞, +∞)} is dense in C.<br />

First H. Bohr proved Theorem B 1 under RH.


Prehistory of the Voronin universality theorem 43<br />

We will give a sketch of the proof. Consider the set of values of the<br />

function<br />

F x (σ; θ 2 , θ 3 , ...) = − ∑ px<br />

log<br />

( )<br />

1 − e−2πiθp<br />

p σ ,<br />

where θ 2 , θ 3 , θ 5 , ... are real variables indexed by prime numbers. It turns<br />

out that this set contains a disc which radius tends to innity as x → ∞.<br />

Therefore, for every a ∈ C, there exist x and a vector (θ2 0, θ0 3 , θ0 5 , ...) such that<br />

F x (σ; θ2, 0 θ3, 0 θ5, 0 ...) = a. (2)<br />

It is well known that the set {log p : p is prime} is linearly independent<br />

over the eld of rational numbers. Hence, by the Kronecker theorem, for<br />

every ε > 0, there exists a dense set {τ j } of real numbers such that<br />

∥<br />

τ j log p<br />

2π<br />

− θp<br />

0 ∥ < ε.<br />

Here ‖x‖ is the distance of x from the closest integer. This together with<br />

(2) shows that, for every ε > 0, there exists a dense set {t j } of real numbers<br />

such that ∣ (<br />

)<br />

∣∣∣ log 2<br />

F x σ; t j<br />

2π , t log 3<br />

j<br />

2π , ... − a<br />

∣ < ε 1,<br />

or, by the denition of the function F x , for s j = σ + it j ,<br />

∣ log ∏ px<br />

(<br />

1 − 1<br />

p s j<br />

) −1 − a∣<br />

∣ < ε 1. (3)<br />

Next, it is proved that there exists x such that<br />

∣<br />

) −1 ∣ ∣∣∣<br />

< ε 1 . (4)<br />

∣ log ζ(s j) − log ∏ px<br />

(<br />

1 − 1<br />

p s j<br />

Now by (3) and (4), for suciently large x, we can nd a subset A ⊂ {t j }<br />

such that<br />

| log ζ(σ + it j ) − a| < 2ε 1<br />

if t j ∈ A. This proves the theorem.<br />

Using similar arguments, H. Bohr and B. Jessen obtained [4] the following<br />

important theorem.<br />

Theorem B 2 . Let a ∈ C \ {0}. Then, for every σ 1 , σ 2 , 1 2 < σ 1 < σ 2 1,<br />

there exists a constant c = c(a, σ 1 , σ 2 ) > 0 such that the number of solutions<br />

of the equation ζ(s) = a in the rectangle σ 1 < σ < σ 2 , |t| < T is ∼ cT as<br />

T → ∞.


44 A. Laurin£ikas<br />

3. Generalizations of Theorem B 1<br />

In 1972, S. M. Voronin obtained [11], [12] a multidimensional analogue of<br />

Theorem B 1 . He considered the value distribution of the vectors<br />

and<br />

(ζ(s 1 + iτ), ζ(s 2 + iτ), ..., ζ(s n + iτ))<br />

(ζ(s + iτ), ζ ′ (s + iτ), ..., ζ (n−1) (s + iτ)),<br />

where s, s 1 , ..., s n are xed complex numbers, and τ ∈ (−∞, ∞).<br />

Let n > 1, C n = C × ... × C, and let C denote the extended complex<br />

plane, C n = C × ... × C.<br />

} {{ }<br />

n<br />

} {{ }<br />

n<br />

Theorem V 1 . Let h > 0, (s 1 , s 2 , ..., s n ) ∈ C n , 1 2 < Res j 1, j = 1, ..., n,<br />

and s j ≠ s k for j ≠ k. Then the sequence of points from C n<br />

η m = (ζ(s 1 + imh), ζ(s 2 + imh), ..., ζ(s n + imh)), m ∈ N,<br />

is dense in C n .<br />

The proof of Theorem V 1 is based on the following statement (main<br />

Lemma).<br />

Lemma V 1 . Suppose that (s 1 , s 2 , ..., s n ) ∈ C n , s j ≠ s k , k ≠ j, 1 2 < Re s j 1<br />

and Im s j > 2, j = 1, . . . , n. Then, for every square matrix (a kl ) ∈ C and<br />

n2<br />

arbitrary y > 0, ε > 0, there exists a nite set of prime numbers M such<br />

that:<br />

1 0 M contains all prime numbers less than y;<br />

2 0 For k, l = 1, ..., n,<br />

∣ (log ζM (s k , θ 0 )) (l−1) ∣<br />

− a k,l < ε,<br />

where, for θ = (θ 2 , θ 3 , ...),<br />

ζ M (s, θ) = ∏<br />

p∈M<br />

(<br />

1 − 1 ) −1<br />

p s e−2πiθ p<br />

and θ 0 = ( 1<br />

2 , 2 2 , 3 2 , 4 2 , ...) .<br />

Proof. S. M. Voronin, dierently than H. Bohr, applies a theorem of<br />

E. Steinitz on the permutation of the terms of series in R n . Suppose that


Prehistory of the Voronin universality theorem 45<br />

u m ∈ R n , and the series<br />

∞∑<br />

m=1<br />

converges. If, for every unit vector e ∈ R n , the series<br />

∞∑<br />

m=1<br />

u m (5)<br />

(u m , e) (6)<br />

converges relatively, then for each vector a ∈ R n , there exists a permutation<br />

of the terms of series (5) such that the obtained series converges to a.<br />

Let p m be the mth prime number, e ∈ R be an arbitrary unit vector,<br />

2n2<br />

and<br />

u m =<br />

(<br />

p −s 1<br />

m e mπi , − log p m p −s 1<br />

m e mπi , ..., (− log p m ) n−1 p −s 1<br />

m e mπi , ...,<br />

p −s n<br />

m<br />

e mπi , − log p m p −s n<br />

m e mπi , ..., (− log p m ) n−1 p −s n<br />

m e mπi) .<br />

S. M. Voronin proves that there exists a partial series of (6) in R 2n2<br />

∑′<br />

(u m , e)<br />

m<br />

which diverges to +∞. In this case, m is even. Similarly, he constructs a<br />

partial series<br />

∑<br />

m<br />

′′<br />

(u m , e)<br />

which converges to −∞. Hence, it follows that the series (6) converges relatively<br />

in R 2n2 . This implies the relative convergence of the series<br />

where<br />

û m =<br />

(<br />

∞∑<br />

(û m , e),<br />

m=1<br />

log ( 1 − p −s 1<br />

m e mπi) −1<br />

,<br />

(<br />

log ( 1 − p −s 1<br />

m e mπi) −1 )′ , ...,<br />

log ( 1 − p −s n<br />

m e mπi) −1<br />

,<br />

(<br />

log ( 1 − p −sn<br />

m e mπi) −1 )′ , ...,<br />

(<br />

log ( 1 − p −s 1<br />

m e mπi) −1 ) (n−1)<br />

, ...,<br />

(<br />

log ( 1 − p −sn<br />

m e mπi) −1 ) (n−1) ).


46 A. Laurin£ikas<br />

Therefore, by Steinitz's theorem, for every a ∈ C n2 , there exists a permutation<br />

of the series<br />

∑<br />

∞ û m<br />

m=1<br />

in C such that the obtained series converges to a. Now to prove the lemma<br />

n2<br />

it suces to take a rather long partial sum of the obtained series.<br />

Proof of Theorem V 1 . By Lemma V 1 , for every y > 0, the set of<br />

vectors<br />

(<br />

ζM (s 1 , θ 0 ), ..., ζ M (s n , θ 0 ) ) ,<br />

where M runs over all nite sets of prime numbers containing all prime<br />

numbers less than y, is dense in C n . Let θ 1 = (θ p (1)<br />

1<br />

, θ (1)<br />

Then the set of vectors<br />

θ (1)<br />

p m<br />

=<br />

{<br />

0 if p m y,<br />

m<br />

2<br />

if p m > y.<br />

p 2<br />

, ...), where<br />

(<br />

ζM (s 1 , θ 1 ), ..., ζ M (s n , θ 1 ) ) (7)<br />

is also dense in C n . Let (a 1 , ..., a n ) ∈ C r be arbitrary. In view of the denseness<br />

of the set (7), we can nd a set M such that, for k = 1, ..., n,<br />

|ζ M (s k , θ 1 ) − a k | < ε.<br />

Since ζ M (s k , θ) is continuous with respect to θ p , p ∈ M, hence there exists<br />

δ > 0 such that, for k = 1, ..., n,<br />

|ζ M (s k , θ) − a k | < ε (8)<br />

if, for all p ∈ M,<br />

|θ p − θ p (1) | < δ. (9)<br />

In the sequel, S. M. Voronin separates two cases. Let, for prime p,<br />

q p = h log p<br />

2π .<br />

Since the system {log p : p is prime} is linearly independent over the eld of<br />

rational numbers, the system {q p } also has the same property. Therefore, if<br />

the equality<br />

∑<br />

α p q p = 1 (10)<br />

p


Prehistory of the Voronin universality theorem 47<br />

with rational coecients α p , which are non-zeros only for a nite number of<br />

primes p, holds, it is unique. S. M. Voronin considers separately the cases:<br />

1 0 Equality (10) is true.<br />

2 0 Equality (10) is not true.<br />

Denote by P M the space of the variables θ p , p ∈ M, and denote by K a<br />

cube of points from P M satisfying (9). Let P 1 = {p : α p ≠ 0 in (10)}. If (10)<br />

is true, then we take some p ′ ∈ P 1 and nd from (10) that<br />

q p ′ = 1 c 0<br />

(<br />

b 0 +<br />

∑<br />

p∈P 1 \{p ′ }<br />

b p q p<br />

)<br />

with integers c 0 > 0 and b 0 , b p . Then, in this case, in place of {η m } the<br />

sequence η ′ m = η c0 m is considered, and this corresponds h ′ = c 0 h and<br />

η ′ m = (ζ(s 1 + imh ′ ), ..., ζ(s n + imh ′ )).<br />

Furthermore, it is proved that on the hyperplane θ p ′ = 0 there exists a<br />

cube K 1 of positive Jordan measure such that<br />

implies<br />

with<br />

and<br />

θ ′ p =<br />

(θ p1 , θ p2 , ...) ∈ K 1<br />

(θ ′ p 1<br />

, θ ′ p 2<br />

, ...) ∈ K<br />

θ ′ p ′ =<br />

∑<br />

p∈P 1 \{p ′ }<br />

b p θ p<br />

{<br />

c 0 θ p if p ∈ P 1 \ {p ′ },<br />

θ p if p ∉ P 1 .<br />

In the case 2 0 , denote by N ∑ ′<br />

m=1<br />

(mq p1 , mq p2 , ...) ∈ K( mod 1). Let<br />

⎧<br />

⎪⎨ 0 if p = p ′ ,<br />

V p = m qp<br />

c ⎪⎩<br />

0<br />

if p ∈ P 1 \ {p ′ },<br />

mq p if p ∉ P 1 , p ∈ M.<br />

In the case 1 0 ,<br />

N<br />

∑ ′<br />

m=1<br />

the sum over m ∈ [1, N] for which<br />

means the sum over m ∈ [1, N] satisfying<br />

(11)<br />

(V p1 , V p2 , ...) ∈ K 1 ( mod 1). (12)


48 A. Laurin£ikas<br />

Now let<br />

A =<br />

N<br />

∑ ′<br />

m=1<br />

N∑<br />

k=1<br />

|ζ(s k + imh) − ζ M (s k + imh, 0)| 2 , (13)<br />

where h = h ′ in the case 1 0 . Then it can be proved that there exists N 0 such<br />

that, for N > N 0 ,<br />

A ε 2 Nmeas{K} (or meas{K 1 }). (14)<br />

Case 2 0 . By the modied Kronecker theorem<br />

lim<br />

N→∞<br />

N<br />

∑ ′<br />

1<br />

N<br />

= meas{K}.<br />

m=1<br />

Therefore, there exists N 1 such that, for N > N 1 ,<br />

N<br />

∑ ′<br />

1 1 2 Nmeas{K}.<br />

m=1<br />

Hence and from (14) there exist m satisfying<br />

N∑<br />

|ζ(s k + imh) − ζ M (s k + imh, 0)| 2 < 4ε 2 .<br />

k=1<br />

Thus, for k = 1, ..., n,<br />

|ζ(s k + imh) − ζ M (s k + imh, 0)| < 2ε. (15)<br />

∑<br />

On the other hand, if m is an index in N ′<br />

, then<br />

m=1<br />

(mq p1 , mq p2 , ...) ∈ K(mod1).<br />

Therefore, by the denition of K, (8) and (9), for k = 1, ..., n, yield<br />

However,<br />

|ζ M (s k , mq 1 , mq 2 , ...) − a k | < 2ε. (16)<br />

(<br />

ζ M (s k , mq 1 , mq 2 , ...) = ζ M s k , m h log p 1<br />

, m h log p )<br />

2<br />

, ...<br />

2π 2π


Prehistory of the Voronin universality theorem 49<br />

and thus (15) and (16) give<br />

= ∏ (1 −<br />

p∈M<br />

= ∏<br />

p∈M<br />

(<br />

1 −<br />

log p<br />

e−2πimh 2π<br />

1<br />

p s k+imh<br />

= ζ M (s k + imh, 0),<br />

p s k<br />

) −1<br />

) −1<br />

|ζ(s k + imh) − a k | < 4ε,<br />

and in the case 2 0 the theorem is proved.<br />

∑<br />

Now suppose that (10) is true. Then m in N ′<br />

satises (14). Since the<br />

m=1<br />

set {1, q p1 , q p2 , ...} \ {q p ′} is linearly independent over the eld of rational<br />

numbers, by the modied Kronecker theorem we have again that<br />

Hence, for N > N 1 ,<br />

lim<br />

N→∞<br />

N<br />

∑ ′<br />

m=1<br />

N<br />

∑ ′<br />

m=1<br />

1<br />

N = meas{K 1}.<br />

1 > N 2 meas{K 1}.<br />

This and (14) show that, for N > max(N 1 , N 0 ), there exist m such that<br />

N∑<br />

|ζ(s k + imh) − ζ M (s k + imh, 0)| 2 < 4ε,<br />

k=1<br />

and, for all k = 1, ..., n,<br />

Dene<br />

|ζ(s k + imh) − ζ M (s k + imh, 0)| < 2ε.<br />

⎧ ∑<br />

b p V p if p = p ⎪⎨<br />

′ ,<br />

p∈P 1 \p ′<br />

u p =<br />

⎪⎩ u p = V p if p ∈ M.<br />

u p = c 0 V p if p ∈ G \ p ′ ,


50 A. Laurin£ikas<br />

Since (V p1 , V p2 , ...) ∈ K 1 ( mod 1), by the denition of K 1 we have that<br />

(u p1 , u p2 , ...) ∈ K( mod 1). Then<br />

(<br />

m h log p 1<br />

2π<br />

, m h log p )<br />

2<br />

, ... ∈ K( mod 1),<br />

2π<br />

and by the denition of K, for all k = 1, ..., n,<br />

∣<br />

∣ζ M<br />

(<br />

s k , m h log p 1<br />

2π<br />

Thus, similarly to the case 2 0 ,<br />

, m h log p ) ∣<br />

2 ∣∣<br />

, ... − a k < 2ε.<br />

2π<br />

|ζ(s k + imh) − a k | < ε.<br />

The theorem is proved.<br />

We mention also the following Voronin's results.<br />

Theorem V 2 ([11]). Let 1 2 < Re s 0 1. Then the sequence of points from<br />

C n (<br />

ζ(s 0 + imh), ζ ′ (s 0 + imh), ..., ζ (n−1) (s 0 + imh)<br />

is dense in C n .<br />

)<br />

, m ∈ N,<br />

S. M. Voronin in his talk given at the Vilnius conference presented [13] a<br />

generalization of Theorems V 1 and V 2 in the form of one theorem.<br />

Theorem V 3 . Suppose that s 1 , ..., s n are pairwise dierent complex numbers<br />

such that 1 2 < Re s j 1, j = 1, ..., n. Then the curve<br />

t ∈ R, is dense in C n2 .<br />

(<br />

ζ(s1 + it), ζ ′ (s 1 + it), ..., ζ (n−1) (s 1 + it), ...,<br />

ζ(s n + it), ζ ′ (s n + it), ..., ζ (n−1) (s n + it) ) ,<br />

4. Generalizations of Theorem B 2<br />

S. M. Voronin presented the following multidimensional versions of Theorem<br />

B 2 .


Prehistory of the Voronin universality theorem 51<br />

Theorem V 4 ([11], [12]). Let a region D = {(s 1 , ..., s n ) ∈ C n : 1 2 < Re s k <br />

1, k = 1, ..., n} have the positive Jordan measure. Then, for every a 1 , ..., a n ∈<br />

C \ {0}, the number of solutions (s 1 , ..., s n ) of the system<br />

satisfying<br />

(s 1 , ..., s n ) ∈<br />

⎧<br />

⎪⎨ ζ(s 1 ) = a 1 ,<br />

. . .<br />

⎪⎩<br />

ζ(s n ) = a n ,<br />

N⋃<br />

(D + im(h, ..., h)), h > 0,<br />

m=1<br />

for suciently, large N, is greater that KN, where K = K(D, a 1 , ..., a n , h) ><br />

0.<br />

Theorem V 4 has the following modication in the sense of Theorems V 2<br />

and V 3 .<br />

Theorem V 5 . Let the region D be the same as in Theorem V 4 . Then, for<br />

every a 1 ∈ C \ {0} and a 2 , ..., a n ∈ C, the number of solutions (s 1 , ..., s n ) of<br />

the system ⎧⎪ ⎨<br />

⎪ ⎩<br />

ζ(s 1 ) = a 1 ,<br />

satisfying<br />

(s 1 , ..., s n ) ∈<br />

ζ ′ (s 2 ) = a 2 ,<br />

. . .<br />

ζ (n−1) (s n ) = a n ,<br />

N⋃<br />

(D + im(h, ..., h)), h > 0,<br />

m=1<br />

for suciently large N, is greater that KN, where K = K(D, a 1 , ..., a n , h) ><br />

0.<br />

5. Towards Theorem V<br />

Lemma V 1 suggests a possible approximation of a given analytic function by<br />

the logarithm of a nite Euler product. However, a way used for the proof<br />

of Lemma V 1 requires a generalization of the Steinitz theorem for series in<br />

Hilbert spaces. S. M. Voronin asked his friend D. V. Pechersky to prove<br />

a theorem of such a type. D. V. Pechersky succeded in his research, and


52 A. Laurin£ikas<br />

presented to S. M. Voronin a theorem on permutations of series in Hilbert<br />

spaces. Let H be a real Hilbert space with inner product (a, b) and norm<br />

||a||.<br />

Lemma P 1 ([8]). Suppose that<br />

∞∑<br />

||a m || < ∞, a m ∈ H,<br />

m=1<br />

and for every e ∈ H, ||e|| = 1, the series<br />

∞∑<br />

(a m , e)<br />

m=1<br />

converges relatively for some permutation of its terms. Then, for every a ∈<br />

H, there exists a permutation {m k : k ∈ N} of N such that<br />

∞∑<br />

a mk = a<br />

k=1<br />

in the sense of norm in H.<br />

Lemma P 1 implies an analogue of lemma V 1 which is a key for the proof<br />

of Theorem V.<br />

Lemma V 2 ([14]). Let 0 < r < 1 4, and let f(s) be a continuous function on<br />

|s| r and analytic in the interior of this disc. Then, for every ε > 0 and<br />

y > 0, there exists a nite set M such that:<br />

1 0 {p : p y} ⊂ M;<br />

2 0 ∣<br />

max ∣f(s) − log ζ M (s + 3 4 , θ 0) ∣ < ε.<br />

|s|r<br />

Acknowledgements. The article is a text of a talk given at the Colloquium<br />

on Dirichlet series held at Universidad Autónoma de Madrid. The<br />

author thanks for hospitality the organizers of this Colloquium Professors<br />

Joern Steuding and Rasa Steuding.<br />

References<br />

[1] H. Bohr, Sur la fonction ζ(s) de Riemann, Comptes Rendus 158 (1914),<br />

19861988.


Prehistory of the Voronin universality theorem 53<br />

[2] H. Bohr, Zur Theorie der Riemannschen Zetafunktion im kritischen Streifen,<br />

Acta Math. 40 (1915), 67100.<br />

[3] H. Bohr, R. Courant, Neue Anwendungen der Theorie der Diophantischen<br />

Approximationen auf die Riemannsche Zeta-funktion, J. Reine Angew. Math.<br />

144 (1914), 249274.<br />

[4] H. Bohr, B. Jessen, Über Wertverteilung der Riemannschen Zetafunction. II<br />

Mitt. Das Verhaften der Funktion in der Streifen 1 2<br />

< σ 1, Acta Math. 58<br />

(1932), 155.<br />

[5] J. Hadamard, Sur les zéros de la fonction ζ(s) de Riemann, Comptes Rendus<br />

Acad. Sci. Paris 122 (1896), 14701473.<br />

[6] A. A. Karatsuba, S. M. Voronin, The Riemann Zeta-Function, de Gruyter,<br />

Berlin, 1992.<br />

[7] H. Von Koch, Sur la distribution des nombres premiers, Acta Math. 24 (1901),<br />

159182.<br />

[8] D. V. Pechersky, On the permutation of the terms of functional series, Dokl.<br />

Akad. Nauk SSSR 209 (1973), 12851287 (in Russian).<br />

[9] B. Riemann, Über die Anzahl der Primazahlen unterhalb einer gegebenen<br />

Grösse, Monatsber. Preuss. Akad. Viss. Berlin, (1859), 671680.<br />

[10] C. J. de la Vallée-Poussin, Recherches analytiques sur la théorie des nombres<br />

premier. IIII, Ann. Sci. Bruxelles 20 (1986), 183256, 281362, 363397.<br />

[11] S. M. Voronin, On the distributution of the non-zero values of the Riemann<br />

zeta-function, Trudy Matem. Inst. Akad. Nauk 128 (1972), 153175 (in Russian).<br />

[12] S. M. Voronin, Investigation of the Behaviuor of the Riemann Zeta-Function,<br />

Thesis of Candidat Fiz.-Matem. Nauk, V. A. Steklov, Matem. Inst. Moscow,<br />

1972 (in Russian).<br />

[13] S. M. Voronin, A theorem on value distribution of the Riemann zeta-function,<br />

in: Abstracts of Conf. Problems of analytic number theory and its applications,<br />

Vilnius, 5455, 1974 (in Russian).<br />

[14] S. M. Voronin, Theorem on the universality of the Riemann zeta-function,<br />

Izv. Akad. Nauk SSSR Ser. Mat. 39 (1975), 475486 (in Russian) = Math.<br />

USSR Izv. 9 (1975), 443453.<br />

Received<br />

4 July 2006

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!