Professional Documents
Culture Documents
Keywords: Dirichlet L-Functions, Dirichlet Series, Periodic Coefficients, Critical Line, Uniform Distribution, Universality, Julia Line Mathematical Subject Classification 2020: 11M06, 30D35
Keywords: Dirichlet L-Functions, Dirichlet Series, Periodic Coefficients, Critical Line, Uniform Distribution, Universality, Julia Line Mathematical Subject Classification 2020: 11M06, 30D35
Keywords: Dirichlet L-Functions, Dirichlet Series, Periodic Coefficients, Critical Line, Uniform Distribution, Universality, Julia Line Mathematical Subject Classification 2020: 11M06, 30D35
number theory, however, it is more natural to consider so-called Julia lines rather
than Julia directions. For this and further references we refer to [4, 32, 34].
Given a complex number a, the solutions of the equation
∆ (s; f ) = a
are called the a-points of ∆ (s; f ), and we denote them as δa := βa + iγa . We shall
show that for any fixed a 6= 0 most of the a-points are clustered around the critical
line 1/2+ iR which, therefore, is a Julia line for ∆ (s; f ). Moreover, we prove that
the mean of the values L(δa ; f ) exists and that every complex number appears
as such a limit. This indicates an interesting link between the distribution of
a-points of ∆(s; f ) and the values taken by L(s; f ). For the case of the Riemann
zeta-function and values a from the unit circle these a-points have been studied
by Justas Kalpokas and the second author [13] as well as by Kalpokas, Maxim
Korolev and the second author [14], respectively. In this situation the exp(2iφ)-
1
points correspond to intersections of the curve t 7→ ζ 2 + it with straight lines
exp(iφ)R through the origin.
valid for any T > 1 and any positive number x > 1, where the ρ := β + iγ denote
the non-trivial zeros of ζ(s) and Λ(x) is the von Mangoldt function extended to
the whole real line by setting it to be zero if x is not a positive integer. The
second author [33] proved a similar result in the case of a-points of ζ(s), namely,
for a 6= 1,
X
ρa T 1
(7) x = a(x) − xΛ + Ox,ǫ T 1/2+ǫ ,
2π x
0<γa ≤T
where ρa := βa + iγa denotes the so called non-trivial a-points of ζ(s) and a(x) is
a certain computable arithmetical function extended to the whole real line. The
authors could not succeed in proving a comparable asymptotic formula in the
case of the δa but an upper bound for the corresponding sum with respect to the
a-points of ∆ which is remarkably small compared with (6) or (7).
In order to formulate this result we recall some standard notation. The number
⌊x⌋ denotes the greatest integer which is smaller than or equal to the real number
x and 1S denotes the characteristic function of a set S.
Theorem 3. Let a 6= 0 be a fixed complex number. Then there exists a constant
c > 0, depending on a, such that, for any 0 < x 6= 1 and any T, T ′ satisfying
max {1, 4π/q} ≤ T < T + 1 ≤ T ′ ≤ 2T , we have that
log T
c/ log qT qT
X
δa 1/2 −c/ log
x ≪ x x 2π + x 2π | log x| + log T +
′
| log x|
T <γa <T
+ 1(1,+∞) (x)E1 (x, a, T ) − 1(0,1) (x)E2 (x, a, T ),
where E1 (x, a, T ) and E2 (x, a, T ) are always zero, unless there is a positive integer
$ %
2 log qT
4π
j≤
log 30
such that x1/j ∈ [qT /(4π), 5qT /(4π)] or x−1/j ∈ [qT /(4π), 5qT /(4π)], in which
case we have that
qT
!
x1/2 log x x1/(2j) xc/ log 2π T 1/2
E1 (x, a, T ) := + ,
j 3/2 |a|j 30j
qT
!
x1/2 log x −1/(2j) j x−c/ log 2π T 1/2
E2 (x, a, T ) := x |a| + ,
j 3/2 30j
respectively. In particular, if x 6= 1 is such that 4π/(qT ) ≤ x ≤ qT /(4π), then
X
δa 1/2 1
x ≪ x log T 1 + .
′
| log x|
T <γa <T
Robert Spira [30] showed that the ∆-function appearing in the functional equa-
tion for the Riemann zeta-function ζ(s) = L(s; 1) (with 1 being the function
constant 1) satisfies |∆ (s; 1) | < 1 for 1/2 < σ < 1 and t ≥ 10. This implies
along with the identity ∆ (s; 1) ∆ (1 − s; 1) = 1 that, if a is a complex number
from the unit circle, then all but finitely many a-points of ∆ (s; 1) have real part
equal to 1/2. Therefore, our theorems generalize some results from [13], as well
as recent results by Korolev & Laurinčikas [17] on the uniform distribution of
Gram points (the 1-points of ∆ (s; 1)). This last article motivates us to prove the
following joint universality theorem:
Theorem 5. Let f be an even or odd q-periodic function and denote the ordinates
(n)
of the a-points of ∆(s; f ) in ascending order by γa . Let also χ1 , χ2 , . . . , χJ
be non-equivalent Dirichlet characters and ψ 6= 0 be an r-periodic arithmetical
function with r 6= 2. Then, for any compact set with connected complement
K inside the strip 1/2 < σ < 1, any g1 , g2 , . . . , gJ , h continuous non-vanishing
6 ATHANASIOS SOURMELIDIS, JÖRN STEUDING, AND ADE IRMA SURIAJAYA
functions on K which are analytic in its interior, any real numbers z > 0 and ξp ,
indexed by the primes p ≤ z, and any ε > 0,
(9)
(n)
max max L s + iγa ; χj − gj (s) < ε,
1 1≤j≤J s∈K
lim inf ♯ 1 ≤ n ≤ N :
(n) log p
>0
N →∞ N
& max
γ − ξ < ε
a p
p≤z
2π
and
1
(n)
(10) lim inf ♯ 1 ≤ n ≤ N : max L s + iγa ; ψ − h(s) < ε > 0,
N →∞ N s∈K
This is a discrete version of Sergei Voronin’s classical universality theorem for the
Riemann zeta-function [39]:
1
(11) lim inf ♯ τ ∈ [0, T ] : max |ζ(s + iτ ) − h(s)| < ε > 0
T →∞ T s∈K
(resp. its simultaneous version [40] for a family of Dirichlet L-functions which is
often called joint universality). As a matter of fact, Voronin proved that there
exists some real τ such that ζ(s + iτ ) is ε-close to h(s) when s ranges in a disc
of center 3/4 + it0 and radius r < 1/4 (see for example [15, Chapter VII]). A few
years later Axel Reich [25] and (independently) Bhaskar Bagchi [2] obtained (11)
(which is implicit in Voronin’s work) and provided also a discrete version with
respect to arithmetic progressions
1
lim inf ♯ 1 ≤ n ≤ N : max |ζ (s + idn) − h(s)| < ε > 0,
T →∞ N s∈K
2. Proof of Theorem 1
We first remark that ∆(s; f ) depends only on the period q of f and δ = ±1
which determines whether f is an even or odd function. We also observe that if
f is an even or odd q-periodic arithmetical function, then so are f ± and f (the
complex conjugate of f ). In the rest of this paper, we will use the simplified
notation
∆(s) := ∆(s; f ) = ∆ s; f = ∆ s; f + .
(12)
DIRICHLET SERIES WITH PERIODIC COEFFICIENTS 7
open interval centered at 1/2, but the way we define it here yields, for instance,
Na (T ; f ) = ♯{γa : ta < γa < T }.
Before showing (5), we use an argument similar to [36, §9.2] to prove
(18) Na (T + 1; f ) − Na (T ; f ) ≪ log T
for any T ≥ ta + 3. Indeed, if n(r) denotes the number of a-points of ∆(s) in the
disc with center 2 + iT and radius r, then
√ Z 3
n(r)
Na (T + 1; f ) − Na (T ; f ) ≤ n 5 ≪ dr.
0 r
It follows from Jensen’s formula (see for example [35, §3.61]) and (15) that
Z 3 Z 2π
n(r) 1
dr = log ∆ 2 + iT + 3eiθ − a dθ − log |∆ (2 + iT ) − a|
0 r 2π 0
≪ log T
and thus (18) holds.
To prove (5) we apply the argument principle to the function ∆(s) − a and
integrate over the counterclockwise oriented rectangle C with vertices −1 + ita ,
2 + ita , 2 + iT and −1 + iT . This gives
∆′ (s)
Z
(19) 2πiNa (T ; f ) = ds;
C ∆(s) − a
since all poles of ∆(s) lie on the real line, they do not affect here. In addition,
all zeros lie outside C; hence, we may rewrite the integrand as
∆′ (s) ∆′ 1
(20) = (s) ·
∆(s) − a ∆ 1 − a/∆(s)
or
∆′ (s) ∆′ (s) 1
(21) = · .
∆(s) − a −a 1 − ∆(s)/a
Now, taking into account (15) in combination with another form of Stirling’s
formula,
∆′
qt 1
(22) (σ + it) = − log +O ,
∆ 2π t
which is also valid for t ≥ ta > 0 and σ from any strip of bounded width, we
obtain, for ǫ > 0 and σ ≥ 1/2 + ǫ,
∆′ (s)
(23) ≪ǫ t1/2−σ log t.
∆(s) − a
Similarly, we have, for ǫ > 0 and σ ≤ 1/2 − ǫ, that
′ ′ j
∆ (s) ∆ X a
(24) = (s) · 1 +
∆(s) − a ∆ ∆(s)
j≥1
qt
= − log + Oǫ t−1 + tσ−1/2 log t ;
2π
here (15), obviously, allows us to expand the second factor into a geometric series.
DIRICHLET SERIES WITH PERIODIC COEFFICIENTS 9
Next, it follows from (23) that the contribution of the integral over the right
vertical segment is negligible:
Z 2+iT
∆′ (s)
ds ≪ 1.
2+ita ∆(s) − a
Using expression (24), however, the contribution of the integral over the left
vertical segment leads to
Z −1+ita Z T
∆′ (s)
qt 1
ds = −i − log +O dt + O(log T )
−1+iT ∆(s) − a ta 2π t
qT
= iT log + O(log T ).
2πe
It remains to estimate the integrals on the horizontal segments of C. The lower
one is trivially bounded
Z 2+ita
∆′ (s)
ds ≪ 1,
−1+ita ∆(s) − a
while for the upper one may use the following truncated partial fraction decom-
position
∆′ (s) X 1
(25) = + O(log t),
∆(s) − a s − δa
|t−γa |≤1
which is valid for σ ∈ [−1, 2] and t ≥ ta > 0. We will show this after finishing
the proof of Theorem 1.
In view of (25)
Z −1+iT X Z −1+iT ds
∆′ (s)
ds = + O(log T ).
2+iT ∆(s) − a 2+iT s − δa
|T −γa |≤1
Finally, we arrive at
∆′ (s) qT
Z
ds = iT log + O(log T ).
C ∆(s) − a 2πe
Substituting this into (19) finishes the proof of (5).
It remains to show (25). For this purpose, we apply Jacques Hadamard’s
theory of functions of finite order (see [28, §5.3]); our reasoning is similar to the
case of the Riemann zeta-function (see [36, §9.6]).
As already mentioned, ∆(s) is analytic except for simple poles at some positive
integers. Thus,
F (s) := (∆(s) − a) · Γ(1 − s)−1
defines an entire function. By Stirling’s formula it follows that F is entire and of
order one. Hence, Hadamard’s factorization theorem implies the product repre-
sentation
Y s
s
F (s) = exp(A + Bs) 1− exp ,
δa δa
δa
where A and B are certain complex constants and the product is taken over all
zeros δa of F (s). Taking the logarithmic derivative, we deduce
F′ X 1
1
(s) = B + + .
F s − δa δa
δa
Since
F′ ∆′ (s) Γ′
(s) = + (1 − s)
F ∆(s) − a Γ
Γ′
and (1 − s) ≪ log t (also a consequence of Stirling’s formula), we have
Γ
∆′ (s) X 1
1
= + + O(log t).
∆(s) − a s − δa δa
δa
X 1 1
− ≪ log t;
s − δa 2 + it − δa
γa >t+1
obviously, we can bound the sum over the a-points δa satisfying γa < t−1 similarly
by the same bound. Consequently, the contribution of the a-points distant from
s are negligible. This implies (25) and concludes the proof of Theorem 1.
3. Proof of Theorem 2
In view of (5), it follows that for any given T0 > 0, there exists T ∈ [T0 , T0 + 1)
such that
1
(26) min |T − γa | ≫ ,
δa log T
where the minimum is taken over all a-points δa . In the proof of Theorem 1 we
have observed that there exists a real number ta > 0 such that βa ∈ (0, 1) if and
only if γa > ta . Thus integrating over the counterclockwise oriented rectangle C
with vertices 2 + ita , 2 + iT, −ǫ + iT, −ǫ + ita , where ǫ > 0, we obtain
X
L(δa ; f )
0<γa <T
0<βa <1
Z 2+iT Z −ǫ+iT Z −ǫ+ita Z 2+ita
1 ∆′ (s)
= + + + L(s; f ) ds
2πi 2+ita 2+iT −ǫ+iT −ǫ+ita ∆(s) − a
X
(27) =: Ij .
1≤j≤4
Since the Dirichlet series coefficients f (n) are bounded, we have L(s; f ) ≪ǫ 1
for σ ≥ 1 + ǫ. In view of the functional equation (3) and the asymptotic formula
(15) it follows from the Phragmén-Lindelöf principle (which is a kind of maximum
principle for unbounded domains) that
and
(see [36, §5.1] and [35, §9.41] for the case of ζ(s); the generalization to L(s; f ) is
straightforward).
We begin with the vertical integrals in (27). Since the range of integration of
I1 lies in the half-plane σ > 1/2, it follows from (23) in combination with (28)
that
Z T
(30) I1 ≪ t−3/2 log t dt ≪ 1.
ta
12 ATHANASIOS SOURMELIDIS, JÖRN STEUDING, AND ADE IRMA SURIAJAYA
The integral I3 is over a line segment in σ < 1/2. It follows from the functional
equation (3) and relation (12) that
+
L(1 − s; f ) = ∆(1 − s)L s; f + = ∆(1 − s)∆(s)L 1 − s; f +
(31) .
+
Here we observe an instance of the Fourier inversion formula, namely (f + ) = δf .
In order to see that we compute via (2) that
± +
1 X ± an
f (n) = √ f (a)e
q q
a mod q
1 X ±ba an
= f (b)e e
q q q
a,b mod q
1 X X a(n ± b)
= f (b) e
q q
b mod q a mod q
= f (∓n),
by the orthogonality relation for additive characters (or simply using geometric
series). This proves
+
(32) f ± = f∓ = δf.
Inserting this into (31) leads to
(33) ∆(1 − s)∆(s) = δ.
Thus using (24), we get
1
Z −ǫ+iT ′
∆ a X a j
I3 = − (s) 1 + + L(s; f ) ds
2πi −ǫ+ita ∆ ∆(s) ∆(s)
j≥2
Z 1+ǫ−iT ′
1 ∆
= (1 − s)
2πi 1+ǫ−ita ∆
(δa∆(s))j L(1 − s; f ) ds
X
× 1 + δa∆(s) +
j≥2
X
(34) =: Jℓ .
1≤ℓ≤3
1+ǫ+iτ ∞
1 q s πis X a(n)
Z
Γ(s) exp ω ds
2πi 1+ǫ+i 2π 2 ns
n=1
X n
1/2+ǫ
a(n) exp(−2πi ) + Oǫ τ , if ω = −1,
τq q
= n≤ 2π
Oǫ (1), if ω = +1.
A proof of this lemma follows along the lines of [31, Lemma 3]. A slightly weaker
statement was proven originally by Gonek [8, Lemma 5] (and there exists a pre-
version of it in [36, §7.4]). Since
q s Γ(s) s s
∆(1 − s) = √ δe + e −
2π q 4 4
1+ǫ+iτ
1 1 X + n
Z
∆(1 − s)L s; f + ds = √ f (n) e − + O τ 1/2+ǫ ;
2πi 1+ǫ+i q τq q
n≤ 2π
we can simplify the right hand side: taking into account the q-periodicity, we find
X n X a X
f + (n) e − = f + (a) e − 1
τq q q τq
n≤ 2π a mod q n≤ 2π
n≡a mod q
X
+
a τ
= f (a) e + O(1)
q 2π
a mod q
√ τ
= q (f + )+ (1) + O(1)
2π
√ τ
= qδ f (1) + O(1).
2π
T qT
(36) J1 = f (1) log + O T 1/2+ǫ .
2π 2πe
Next we consider
1+ǫ−iT
a ∆′
Z
(1 − s)∆(s)∆(1 − s)L s; f + ds.
J2 =
2πδi 1+ǫ−ita ∆
14 ATHANASIOS SOURMELIDIS, JÖRN STEUDING, AND ADE IRMA SURIAJAYA
This is essentially a classical lemma from [36, §4.3]. This leads to the estimate
Jn (T ) ≪ log T
log n for n ≥ 2. Hence, the contribution of the tail of the Dirichlet series
is negligible and we get
T qT
(37) J2 = af + (1) log + O(log T ).
2π 2πe
Finally, we have to consider the third integral over the tail of the geometric
series. For this aim we apply (15), (22) and (29) and get
Z T X
J3 ≪ log t t−j(1/2+ǫ) t1/2+ǫ dt ≪ T 1/2+ǫ .
ta j≥2
This together with (36) and (37) substituted into (34), in combination with (30)
shows that the vertical integrals contribute
T qT
(38) I1 + I3 = f (1) + af + (1) log + O(T 1/2+ǫ ),
2π 2πe
which is already the main term. It remains to consider the horizontal integrals in
(27). Although they can be treated the same way as in [34] we sketch the details.
The integral I2 in (27) can be rewritten with the aid of the truncated partial
fraction decomposition (25) as
Z −ǫ+iT
1 X 1
I2 = + O(log T ) L(s; f ) ds.
2πi 2+iT s − δa
|T −γa |≤1
DIRICHLET SERIES WITH PERIODIC COEFFICIENTS 15
Taking into account (26) we notice that 1/|s − δa | ≪ log T . Hence, utilizing (18),
(28) and (29), we can show
Z 0 Z 1 Z 2
2
I2 ≪ (log T ) + + |L (σ + iT ; f ) | dσ
−ǫ 0 1
n o
≪ǫ (log T )2 T 1/2+ǫ + T 1/2+ǫ + 1
≪ǫ T 1/2+ǫ ,
where ǫ at different places may take different values. This is the bound for the
horizontal integrals. In combination with (38) we arrive via (27) at the asymptotic
formula of the theorem. Taking into account (28) and (29) we may replace the
chosen T (with respect to (26)) with a general T ≥ ta at the expense of an error
of order T 1/2+ǫ (as follows from (28)). This concludes the proof of Theorem 2.
4. Proof of Theorem 3
We first note that (15) implies
1/2−σ
qt
(39) ∆(σ + it) ≍ , t ≥ t0 > 0,
2π
while (17) yields
1 log |a| 1
(40) βa = − +O
2 log qγ2π
a
γa log qγ
2π
a
c/ log qT
qt 2π
|∆(α + it)|−1 = |∆(1 − α + it)| ≥ L ≥ Lec > 30|a|
2π
for any t ≥ T ≥ T0 , where K, L are absolute constants coming from (39) and
we can assume without loss of generality that K > 1 is sufficiently large and
L = 1/K. After choosing a suitable K, we then take c sufficiently large. These
show that ∆(α + it) 6= a, ∆(1 − α + it) 6= a when t ≥ T .
16 ATHANASIOS SOURMELIDIS, JÖRN STEUDING, AND ADE IRMA SURIAJAYA
since
Z α+iTℓ Z α
s
O(log t)x ds ≪ log Tℓ xσ dσ ≪ xα (2α − 1) log T ≪ xα .
1−α+iTℓ 1−α
Meanwhile,
Z α+iTℓ α
1 xσ
X X Z
xs ds ≪ dσ
1−α+iTℓ |t−γ |≤1 s − δa 1−α |α − βa + i(Tℓ − γa )|
a |Tℓ −γa |≤1
X
≪ xα (2α − 1) log T 1
|T1 −γa |≤1
α
≪ x log T.
Therefore,
(45) I2 , I4 ≪ xα log T.
We now estimate the vertical integrals I1 and I3 . For σ = α, we use (21) and
thus
′ ′ ′
j
∆ (s) ∆ (s) 1 ∆ (s) X ∆(s)
(46) = · = 1+ ;
∆(s) − a −a 1 − ∆(s)/a −a a
j≥1
DIRICHLET SERIES WITH PERIODIC COEFFICIENTS 17
here the first inequality in (42) allows us to expand the second factor into a
geometric series. Applying this, we find
(47)
α Z T2 j X j
x X ∆(α + it) ∆(α + it) it
I1 = − ∆′ (α + it) + x dt.
2πa T1 a a
0≤j<m j≥m
Here we pick m depending on T , large enough to bound the last term in the
integral trivially. By (39) and again the first inequality in (42),
∆(α+it) m
X ∆(α + it) j a 30 1 m
(48) = ≪ .
j≥m
a 1 − ∆(α+it)
a
29 30
Take
$ %
2 log qT
4π
(49) m := .
log 30
xα
Z T2 X ∆(α + it) j
′
I1 = − ∆ (α + it) xit dt
2πa T1 a
0≤j<m
Z T2 2 !
+ O xα ∆′ (α + it) 4π
dt .
T1 qT
The first term of the integrand can be estimated using (15) for which we obtain
1/2−(α+it)
′ qt π qt log qt
∆ (α + it) = δ exp i t + − log +O
2π 4 2π t
qt
≪ log .
2π
Here we substituted the value of α and used the first inequality in (42). It then
follows that we can discard the last term in I1 as
xα
Z T2 X ∆(α + it) j
′
I1 = − ∆ (α + it) xit dt + O (xα ) .
2πa T1 a
0≤j<m
Since
d
∆(α + it)j+1 = i(j + 1)∆′ (α + it)∆(α + it)j ,
dt
we can rewrite I1 as
Z T2
xα X 1 j ′ it
x dt + O (xα )
I1 = − ∆(α + it)
2πi jaj T1
1≤j≤m
xα X 1
(50) =: − I1j + O (xα ) .
2πi jaj
1≤j≤m
18 ATHANASIOS SOURMELIDIS, JÖRN STEUDING, AND ADE IRMA SURIAJAYA
It is easily seen that that this holds uniformly in a from a bounded interval (the
implicit constants depend of course on the limits of this interval). We then apply
this, by taking complex conjugate, to J with a = 1/2− jc/ log qT 2π , r = 2πjx
1/j /q,
1/2−jc/ log qT !
j 2π
1/(2j)−c/ log qT 2πx1/j
≪ x 2π 1
(T /2,5T /2)
q q
qT
+ (jT )−jc/ log 2π E0 (x, j, T ),
where
√ 2πx1/j
T 5T
O jT , if ∈ , ,
q 2 2
(52) E0 (x, j, T ) =
O(1), otherwise.
Recall that T ≍ T1 ≍ T2 and j ≪ log T . Therefore, E0 does not restrictively
depend on T1 and T2 .
Applying this to (51) we have
!
log x 1/(2j)−c/ log qT 2πx1/j
I1j ≪ √ x 2π 1
(T /2,5T /2)
jq q
j j
log x −c/ log qT |a|
+ (qT ) 2π E0 (x, j, T ) + (1 + log x) .
j 30
Hence, in view of (50) and (42), we obtain
!
x1/(2j) 2πx1/j
1
X
1/2
I1 ≪ x log x 3/2 |a|j (T /2,5T /2)
j q
1≤j≤m
1 |a| j
α
X 1 −c/ log qT
+ x log x (2π) 2π E0 (x, j, T )
j 2 |a|j K 30
1≤j≤m
X 1
(53) + xα (1 + log x) + xα .
j30j
1≤j≤m
for all T ≥ T0 . By this construction, the first sum in (53) can have at most only
one term, namely,
x1/(2jx )
3/2
jx |a|jx
for the unique, if any, jx ≤ m such that 2πx1/jx /q ∈ (T /2, 5T /2). Similar rea-
soning also applies to E0 of the second sum in (53) and it follows in view of (52)
that
X 1 X 1 T 1/2 T 1/2
E0 (x, j, T ) ≪ + ≪ 1 + .
j 2 30j j 2 30j 3/2
jx 30jx
3/2
jx 30jx
1≤j≤m 1≤j≤m
j6=jx
The last sum in (53) is trivially O(1). Collecting these estimates, we conclude
that
if such jx exists, and zero otherwise. Observe that if T ≥ T0 ≥ 4π/q, then for any
x > 1 it is E1 (x, a, T ) ≥ 0, while for 0 < x < 1 we always have E1 (x, a, T ) = 0.
For σ = 1 − α, we use (20) and the corresponding (24) to write
j
∆′ (s) ∆′ ∆′
1 X a
(56) = (s) · = (s) 1 + ;
∆(s) − a ∆ 1 − a/∆(s) ∆ ∆(s)
j≥1
here the second inequality in (42) allows us to expand the second factor into a
geometric series. Thus the left vertical integral I3 can be decomposed as follows
1−α+iT1
1 ∆′ (s) s
Z
I3 = x ds
2πi 1−α+iT2 ∆(s) − a
1−α+iT1 1−α+iT1 j
∆′ ∆′
1 1 a
Z Z X
= (s)xs ds + (s)xs ds
2πi 1−α+iT2 ∆ 2πi 1−α+iT2 ∆ ∆(s)
j≥1
=: I31 + I32 .
DIRICHLET SERIES WITH PERIODIC COEFFICIENTS 21
1
Z α−iT2 ′
∆ X a∆(s) j
= (s)x1−s ds
2πi α−iT1 ∆ δ
j≥1
x 1−α Z T2 ′
∆ X a∆(α − it) j
= − (α − it) xit dt
2π T1 ∆ δ
j≥1
or from (16)
!j
T2
x1−α ∆′ (α − it) X a∆(α − it)
Z
I32 = − x−it dt
2π T1 ∆(α − it) δ
j≥1
T2 it
ax1−α 1
Z
j
X
′
(58) = − ∆ (α + it) (a∆(α + it)) dt.
2π T1 x
j≥0
if such jy exists, and zero otherwise. Observe that if T ≥ T0 ≥ 4π/q, then for
any x > 1 it is always E2 (x, a, T ) = 0, while for 0 < x < 1 it is E2 (x, a, T ) ≤ 0.
Collecting the estimates in (44), (45), (54), (57) and (59) we obtain that
X
δa α log T
x ≪ x log x + log T + + E1 (x, a, T )
′
log x
T <γa <T
where the inner sum on the right-hand side is zero when the set of indices of its
summands is empty. Since 4π/(qN ) ≤ x ≤ qN/(4π), (61), (62) and Theorem 3
yield
X (n)
δa 1/2 1
x ≪x 1+ log N.
| log x|
N <n≤M
We also have
(M ) qN
x−βa ≪ x−1/2±c/ log 2π ≪ x−1/2
DIRICHLET SERIES WITH PERIODIC COEFFICIENTS 23
for any M > N . Using this and Abel’s summation formula (summation by parts),
we obtain
N <n≤2N
X X
(n) −βa(M +1) (M )
δ
− x−βa
+ max x a
x
N <M <2N
N <n≤M N <M <2N
!
1 X (M +1) (M )
βa −βa
(63) ≪ 1+ log N 1 + x − 1 .
| log x|
N <M <2N
!−1 !−1 !
(M +1) (M ) (M ) (M +1)
βa − βa qγa qγa 1
= log − log +O (M ) (M )
log |a| 2π 2π γa log γa
and 4π/(qN ) ≤ x ≤ qN/(4π). Hence, with the aid of (61) we can show
X (M +1) (M )
βa −βa
x − 1
N <M <2N
!−1 !−1
(M ) (M +1)
log qγa qγa 1
X
≪ | log x| − log + (M )
2π 2π (M )
γa log qγ2π
a
N <M <2N
!−1 !−1
(N +1) (2N −1)
γa γa N
≪ | log x| log − log + (N ) (N )
2π 2π γa log γa
−2
≪ | log x| log γa(N ) +1
≪ | log x|.
(n) 1
+ | log x| log N = ox (log N )2 ,
X
iγa
(64) x ≪
| log x|
N <n≤2N
as follows from (64). Since k ∈ Z \ {0} can be chosen arbitrarily, the sequence
(n)
αγa , n ∈ N, satisfies Weyl’s Criterion (see [41]): a sequence (xn )n∈N is uniformly
distributed modulo 1 if X
e(kxn ) = o(N )
n≤N
for any integer k 6= 0. This concludes the proof of the theorem.
6. Proof of Theorem 5
If χ is a Dirichlet character mod r and Q > 0, we define the truncated and
twisted Euler product
Y χ(p) −1
(65) s 7−→ LQ (s; χ) := 1− s
p
p≤Q
for every s ∈ C with σ > 0, where p will denote from here on a prime number.
The proof of (9) is similar to Reich’s proof [25] for the discrete universality
of ζ(s) and we will not repeat it here. Instead we employ the following theorem
which highlights the necessary conditions a sequence (xn )n∈N has to meet in
order to derive universality: Let χ1 , . . . , χJ be pairwise non-equivalent Dirichlet
characters. Let also (xn )n∈N be a sequence of real numbers such that the sequence
of vectors
log p
(66) xn , n ∈ N,
2π p∈M
is uniformly distributed modulo 1 for any finite set of primes M, and
1
|L (s + ixn ; χj ) − LQ (s + ixn ; χj )|2 = 0,
X
(67) lim lim sup
Q→∞ N →∞ N + 1
N ≤n≤2N
j = 1, . . . , J, uniformly in compact subsets of the strip 1/2 < σ < 1. Then for
any compact subset with connected complement K of this strip, any g1 , . . . , gJ
continuous non-vanishing functions on K and analytic in its interior, any z > 0
and ξp , p ≤ z, real numbers, and any ε > 0,
max max |L (s + ixn ; χj ) − gj (s)| < ε
1 1≤j≤J s∈K
lim inf ♯ 1 ≤ n ≤ N :
(n) log p
> 0.
N →∞ N
& max
γa − ξp
< ε
p≤z 2π
DIRICHLET SERIES WITH PERIODIC COEFFICIENTS 25
log p
γa(n) , n ∈ N,
2π p∈M
X log p
γa(n) hp , n ∈ N,
2π
p∈M
is uniformly distributed modulo 1 for any (hp )p∈M ∈ Z♯M \ {0}. But this follows
immediately from Theorem 4 and the unique factorization of integers into primes.
(n)
Thus, γa , n ∈ N, satisfies condition (66).
(n)
To prove that γa , n ∈ N, satisfies also condition (67), we will use the following
approximate functional equation
X X
L(s; χj ) = χj (n)n−s + ∆(s; χj ) χ−
j (n)n
s−1
+
n≤X n≤y
(68) + O X −σ log(y + 2) + τ 1/2−σ y σ−1
valid for s = σ + iτ with 0 < σ < 1, τ ≥ τ0 > 0, and X, y > 0 such that
2πXy = qj τ , where qj is the modulus of χj . This formula follows from a result due
to Vivek Rane [23]. Observe also that LQ (s; χj ) can be written as an absolutely
convergent Dirichlet series
∞
X χj (n)
(69) LQ (s; χj ) =
n=1
ns
p|n⇒p≤Q
for any σ > 0 and Q > 0, as follows from (65) by expanding each factor of the
truncated Euler product into a geometric series.
Now let K be a compact subset of the strip σ1 ≤ σ < 1 for some σ1 ∈ (1/2, 1).
Then equations (68) and (69) imply, for every s = σ + it ∈ K, any Q > 0, any
26 ATHANASIOS SOURMELIDIS, JÖRN STEUDING, AND ADE IRMA SURIAJAYA
= o(N ),
while (15) implies that
2
X 1−2σ X 1
S3 ≪ γa(n)
m1−σ
N ≤n≤2N m≤y
!2
(2N )
1−2σ1 γa
≪ N γa(N )
N
1−2σ1
N
≪ N (log N )−2
log N
= o(N ).
To estimate S2 , we first observe that its inner sum is a tail of an absolutely
convergent series. Therefore,
!2 2
X X 1 X 1
S2 ≪ ≪N = o(N ).
2 mσ 2 mσ
N ≤n≤2N m>X m>N/ log N
DIRICHLET SERIES WITH PERIODIC COEFFICIENTS 27
Lastly,
it (n)
m2 iγa
X 1 m2 X
S1 =
1 (m1 m2 )σ m1 m1
m1 ,m2 ≤X N ≤n≤2N
X iγa(n)
X 1 X 1 m2
(71) ≪ N 2σ
+ σ
.
1 m 1 1 (m1 m2 )
1 m1
m≤X 1≤m1 <m2 ≤X N ≤n≤2N
The first term on the right hand-side of (71) is O N Q1−2σ1 . For the second
term observe that for any 1 ≤ m1 < m2 ≤ X, we have 1 < m2 /m1 ≤ qj N/(4π).
Therefore, Theorem 4 implies that this term is bounded from above by
!
m2 −1
X log N m2
log + log ≪ X 2−2σ1 log(X) log N
1 (m1 m2 )σ1 m1 m1
1≤m1 <m2 ≤X
≪ N 2−2σ1 (log N )2
= o(N ).
Collecting the above estimates we finally arrive at
1 X 2
lim sup L s + iγa(n) ; χj − LQ s + iγa(n) ; χj ≪ Q1−2σ1
N →∞ N + 1
N ≤n≤2N
uniformly in K and arbitrary Q > 0. Taking Q to infinity shows that the sequence
(n)
γa , n ∈ N, also satisfies (67) and thus (9) holds.
We prove (10) from (9). We employ techniques introduced by Bagchi [2],
Gonek [7] as well as Jürgen Sander and the second author [26]. Let ψ 6= 0 be an
r-periodic arithmetical function. If r = 1, then
L(s; ψ) = ψ(1)ζ(s)
and (10) holds, since ζ(s) is L(s; χ0 ), where χ0 is the Dirichlet character mod 1
and we apply (9) only to this character.
The case r = 2 is rather special; here some cases need a restriction on the
range of approximation. This observation is due to Jerzy Kaczorowski [12] and
has recently been discussed by Philipp Muth and the second author [21]. Since
ψ(1) 6= ψ(2), for any σ > 1, we have
∞ ∞
X ψ(1) X ψ(2) ψ(2) − ψ(1)
L(s; ψ) = + = ψ(1) + ζ(s) =: P (s)ζ(s).
(2n − 1)s (2n)s 2s
n=1 n=1
The latter holds for all s ∈ C by analytic continuation. Observe here that P (s)
is analytic and bounded in any half-plane σ ≥ σ0 and
log 2
(72) P (s + iτ ) − P (s) ≪
τ 2π
for any η > 0 and any sufficiently large N ≫η 1, where c(η) > 0 is constant. For
those n from the set described on the left-hand side above, in combination with
(72), it also follows that
max L s + iγa(n) ; ψ − h(s) < max P s + iγa(n) max ζ s + iγa(n) − g(s)
s∈K s∈K s∈K
+ max |g(s)| max P s + iγa(n) − P (s)
s∈K s∈K
≪ η.
Taking 0 < η ≪ ε sufficiently small, we obtain (10) with the restriction we
imposed on K.
If r ≥ 3, then φ(r) ≥ 2, where φ is the Euler totient function. Assuming that ψ
is a multiple of a Dirichlet character mod r, we can work as in the case of r = 1.
If ψ is not such a multiple, then for every s ∈ C
r
1 X n
L(s; ψ) = s ψ(n)ζ s;
r n=1 r
r φ(r)
1 X rs X
= ψ(n) χi (n)L(s; χi )
r s n=1 φ(r)
i=1
φ(r) r
!
X 1 X
(73) = ψ(n)χi (n) L(s; χi ),
φ(r)
i=1 n=1
for any sufficiently large N ≫η 1. In this case, for those n from the set described
on the left-hand side above, in combination with (73)–(75), it also follows that
φ(r)
X
(n) (n)
max L s + iγa ; ψ − h(s) ≤ |ci | max L s + iγa ; χi − gi (s)
s∈K s∈K
i=1
φ(r)
X
+ η(φ(r) − 3) |ci |
i=3
≪ η.
Taking 0 < η ≪ ε sufficiently small, we obtain (10). Observe that in this case,
h(s) is allowed to have zeros.
References
[1] T.M. Apostol, Dirichlet L-functions and primitive characters, Proc. Amer. Math. Soc. 31
(1972), 384–386.
[2] B. Bagchi, The Statistical behaviour and universality properties of the Riemann Zeta func-
tion and other allied Dirichlet series, Ph.D. thesis, Calcutta: Indian Statistical Institute
(1981).
[3] R.B. Burckel, An Introduction to Classical Complex Ananlysis, vol. I, Birkhäuser, 1979.
[4] T. Christ, Value-distribution of the Riemann zeta-function and related functions near the
critical line, PhD Thesis, Würzburg 2014, arXiv:1405.1553.
[5] P.D.T.A. Elliott, The Riemann zeta function and coin tossing, J. Reine Angew. Math.
254 (1972), 100–109.
[6] J.-P. Gram, Sur les zéros de la fonction ζ(s) de Riemann, Acta Math. 27 (1903), 289–304.
[7] S.M. Gonek, Analytic properties of Zeta and L-functions, Ph.D. Thesis, University of
Michigan (1979).
[8] S.M. Gonek, Mean values of the Riemann zeta-function and its derivatives, Invent. Math.
75 (1984), 123–141.
[9] S.M. Gonek, An explicit formula of Landau and its applications to the theory of the zeta-
function. A tribute to Emil Grosswald: number theory and related analysis, Amer. Math.
Soc., Providence, RI, Contemp. Math. 143 (1993), 395–413.
[10] E. Hlawka, Über die Gleichverteilung gewisser Folgen, welche mit den Nullstellen der Zeta-
funktion zusammenh ängen, Österr. Akad. Wiss., Math.-Naturw. Kl. Abt. II 184 (1975),
459–471.
[11] G. Julia, Sur quelques propriétés nouvelles des fonctions entières ou méromorhes (liere
Mémoire), Ann. Sci. École Norm. Sup. 36 (1919), 93–125.
30 ATHANASIOS SOURMELIDIS, JÖRN STEUDING, AND ADE IRMA SURIAJAYA
[12] J. Kaczorowski, Some remarks on the universality of periodic L-functions, in: New di-
rections in value-distribution theory of zeta and L-functions, Proceedings of the conference,
Würzburg October 6–10, 2008. R. Steuding (ed.) et al., Shaker Verlag, Aachen 2009, 113-
120.
[13] J. Kalpokas, J. Steuding, On the Value-Distribution of the Riemann Zeta-Function on
the Critical Line, Moscow Jour. Combin. Number Theory 1 (2011), 26–42.
[14] J. Kalpokas, M.A. Korolev, J. Steuding, Negative values of the Riemann zeta function
on the critical line, Mathematika 59 (2013), 443–462.
[15] A.A. Karatsuba and S.M. Voronin, The Riemann zeta-function. Translated from the
Russian by Neal Koblitz. De Gruyter Expositions in Mathematics, 5. Walter de Gruyter &
Co., Berlin, 1992.
[16] M.A. Korolev, Gram’s Law in the Theory of Riemann Zeta-Function. Part 1, Proc. Steklov
Inst. Math. 292 (2016), 1–146.
[17] M. Korolev, A. Laurinčikas, A new application of the Gram points, Aequat. Math. 93
(2019), 859–873.
[18] L. Kuipers and H. Niederreiter, Uniform Distribution of Sequences, Dover Publica-
tions, 2006.
[19] E. Landau, Über die Nullstellen der Zetafunktion, Math. Ann. 71 (1911), 548–564.
[20] N. Levinson, Almost all roots of ζ(s) = a are arbitrarily close to σ = 21 , Proc. Nat. Acad.
Sci. USA 72 (1975), 1322–1324.
[21] P. Muth, J. Steuding, Joint value-distribution of Dirichlet series associated with periodic
arithmetical functions, Funct. Approx. Comment. Math. 62 (2020), 63–79.
[22] É. Picard, Sur les fonctions analytiques uniformes dans le voisinage d’un point singulier
essentiel, C.R. Acad. Sci. Paris 89 (1879), 745–747.
[23] V.V. Rane, On an approximate functional equation for Dirichlet L-series, Math. Ann. 264,
137-145.
[24] H.A. Rademacher, Fourier Analysis in number theory, symposium on harmonic analysis
and related integral transforms, Cornell University, Ithaca, 1956. In: Collected Papers of
Hans Rademacher, vol. II, pp. 434–458. Massachusetts Institute of Technology, Cambridge
(1974).
[25] A. Reich, Wertverteilung von Zetafunktionen, Arch. Math. 34 (1980), 440–451.
[26] J. Sander, J. Steuding, Joint universality for sums and products of Dirichlet L-functions.
Analysis (Munich) 26 (2006), no. 3, 295–312.
[27] W. Schnee, Die Funktionalgleichung der Zetafunktion und der Dirichletschen Reihen mit
periodischen Koeffizienten, Math. Z. 31 (1930), 378–390.
[28] E.M. Stein, R. Shakarchi, Complex Analysis, Princeton University Press 2003.
[29] A. Sourmelidis, Universality and Hypertranscendence of Zeta-Functions, Ph.D. thesis,
Würzburg University, 2020.
[30] R. Spira, An inequality for the Riemann zeta function, Duke Math. J. 32 (1965), 247–250.
[31] J. Steuding, Dirichlet series associated to periodic arithmetic functions and the zeros of
Dirichlet L-functions, in: Anal. Probab. Methods Number Theory, A. Dubickas et al. (Eds),
TEV Vilnius, 2002, 282–296
[32] J. Steuding, Value-distribution of L-functions, Lecture Notes in Mathematics 1877,
Springer 2007.
[33] J. Steuding, One hundred years uniform distribution modulo one and recent applications
to Riemanns zeta-function, in: Topics in mathematical analysis and applications, Th.M.
Rassias (ed.) et al., Springer Optimization and Its Applications 94, 659–698 (2014).
[34] J. Steuding, A.I. Suriajaya, Value-distribution of the Riemann zeta-function along its
Julia lines, Comp. Methods Function Theory 20 (2020), doi: 10.1007/s40315-020-00316-x.
[35] E.C. Titchmarsh, The theory of functions, second ed., Oxford University Press, 1939.
[36] E.C. Titchmarsh, The Theory of the Riemann zeta-function, 2nd ed., revised by D.R.
Heath-Brown, Oxford University Press, 1986.
DIRICHLET SERIES WITH PERIODIC COEFFICIENTS 31
[37] I.M. Vinogradov, A new estimate of a certain sum containing primes, Rec. Math. Mosc.
2 (1937), 783–792.
[38] I.M. Vinogradov, A new estimate for the function ζ(1 + it), Izv. Akad. Nauk SSSR, Ser.
Mat. 22 (1958), 161–164 (in Russian).
[39] S. Voronin, Theorem on the ‘universality’ of the Riemann zeta-function, Izv. Akad. Nauk
SSSR, Ser. Matem. 39 (1975), 475-486 (Russian); Math. USSR Izv. 9 (1975), 443–445.
[40] S. Voronin, On the functional independence of Dirichlet L-functions, Acta Arith. 27
(1975), 493–503 (in Russian).
[41] H. Weyl, Über die Gleichverteilung von Zahlen mod. Eins, Math. Ann. 77 (1916), 313–352.
Athanasios Sourmelidis
Institute of Analysis and Number Theory, TU Graz
Steyrergasse 30, 8010, Graz, Austria
sourmelidis@math.tugraz.at
Jörn Steuding
Department of Mathematics, Würzburg University
Emil Fischer-Str. 40, 97 074 Würzburg, Germany
steuding@mathematik.uni-wuerzburg.de
Ade Irma Suriajaya
Faculty of Mathematics, Kyushu University
744 Motooka, Nishi-ku, Fukuoka 819-0395, Japan
adeirmasuriajaya@math.kyushu-u.ac.jp