AUTOMORPHIC INTEGRALS WITH LOG-POLYNOMIAL PERIOD FUNCTIONS

AND ARITHMETICAL IDENTITIES

arXiv:2208.03102v1 [math.NT] 5 Aug 2022

TEWLEDE G/EGZIABHER, HUNDUMA LEGESSE GELETA AND ABDUL HASSEN

Abstract. Building on the works of S. Bochner on equivalence of modular relation with functional equa-
tion associated to the Dirichlet series, K. Chandrasekharan and R. Narasimhan obtained new equivalences
between the functional equation and some arithmetical identities. Sister Ann M. Heath considered the
functional equation in the Hawkins and Knopp context and showed its equivalence to two arithmetical iden-
tities associated with entire modular cusp integrals involving rational period functions for the full modular
group. In this paper we use techniques of Chandrasekharan and Narasimhan and extend the results of Sister
Ann M. Heath to entire automorphic integrals involving rational period functions on discrete Hecke group.
Moreover, we establish equivalence of two arithmetical identities with a functional equation associated with
automorphic integrals involving log-polynomial-period functions on the Hecke groups.

1. Introduction
In 1951, Bochner introduced in [1] and showed that the modular relation
 
−δ 1
f (x) = x g + P (x), 0 < x < ∞ ,
x
where f and g are exponential series, P is a “residual” function and δ > 0, is equivalent to the functional
equation
Γ(s)φ(s) = Γ(δ − s)ψ(δ − s) , (1.1)
P∞ an P∞ bn
where Γ(s) is the standard gamma function, φ(s) = n=1 λs and ψ(s) = n=1 µs . Here, he assumes that
n n
{λn } and {µn } are sequences of real numbers, with
0 < λ1 < λ2 < · · · < λn → ∞
and
0 < µ1 < µ2 < · · · < µ n → ∞ ,
and {an }, {bn } are two sequences of complex numbers not identically zero.
Building on the works of Bochner, Chandrasekharan and Narasimhan [2] showed that the functional
equation (1.1) is equivalent to two identities given by
δ+ρ
 ρ X∞  
1 X′
ρ 1 x 2
√
an (x − λn ) = bn Jδ+ρ {4π µn x} + Qρ (x) , (1.2)
Γ(ρ + 1) 2π n=1
µn
λn ≤x
for x > 0 and ρ ≥ 2β − δ − 12 ,
where, the prime notation ′ on the summation sign indicates that if ρ = 0 and
x = λn for some positive integer n, then we count only 21 an ,
1 χ(s)(2π)s xs+ρ
I
Qρ (x) = ds ,
2πi c Γ(s + ρ + 1)

Date: August 8, 2022.

2000 Mathematics Subject Classification. Primary 11M41.
Key words and phrases. Hecke groups, Dirichlet series with Functional Equation, Automorphic integrals, Log-Polynomial
period Functions and Bessel Functions.
1
2 TEWLEDE G/EGZIABHER, HUNDUMA LEGESSE GELETA AND ABDUL HASSEN

P∞ |bn |
n=1 µn −β< ∞, Jν (z) denotes the Bessel function of the first kind of order ν, and
ρ " X ∞ ∞
#
√

1 d 1 −s λn 1 1
X bn
− an e = 23δ+ρ Γ(δ + ρ + )π δ− 2 δ+ρ+ 21
+ Rρ (s) , (1.3)
s ds s n=1 2 2 2
n=1 (s + 16π µn )

for Re s > 0, where

1 χ(z)(2π)z Γ(2z + 2ρ + 1)2−ρ −2z−2ρ−1
I
Rρ (s) = s dz ,
2πi c Γ(z + ρ + 1)
1
for Re s > 0, ρ integral, ρ − δ − 2 and ρ ≥ 2β − δ − 21 .

Sister Ann Heath, in [13] showed the equivalence of the above forms for the full modular group.

In our work here, we will show that these results hold for a more general Hecke Groups. Our notations
will be different from the ones shown above. The structure of our paper is as follows: in section 2 we review
some preliminary concepts which we use in the later sections. In section 3 we state and proof our main
results which are in Theorem 3.1 and Theorem 3.2. In section four we give some conclusions and discussions.

2. Preliminaries
In this section, we recall some definitions, results proved elsewhere, and make some conventions for our
later discussion.

For λ ∈ R+ , the Hecke group G(λ) is defined as the subgroup of SL2 (R) given by
  
1 λ 0 1
G(λ) = .
0 1 −1 0
Note that G(λ) can be viewed as a group of linear fractional transformations generated by S(z) = z + λ and
T (z) = − z1 . In this context, we
 see that
 G(λ) acts on the Riemann sphere as linear fractional transformation,
az+b a b S
that is M z = cz+d for M = ∈ G(λ), and z ∈ C {∞}, thus M and −M can be identified as the
c d
same matrices. E. Hecke [6] showed the groups G(λ) is discrete (operate discontinuously) as linear fractional
transformations on the upper half plane H = {z = x + iy : y > 0} if and only if either
 
π
λ > 2 or λ = λp := 2 cos , with 3 ≤ p ∈ N ∪ {∞}.
p
Clearly G(λ3 ) = Γ(1) is the full modular group and G(λ∞ ) = Γθ is the theta group and is denoted by Γθ .

A complex valued function

 ν defined on G(λ) is called a multiplier system with weight 2k for G(λ) if
a b
|ν(M )| = 1 for all M = ∈ G(λ),and ν(M ) satisfies the consistency condition
c d
2k 2k 2k
ν(M3 ) (c3 z + d3 )
= ν(M1 )ν(M2 ) (c1 M2 z + d1 ) (c2 z + d2 )
 
aj b j
for all M1 , M2 ∈ G(λ), M1 M2 = M3 , Mj = , j = 1, 2, 3 z ∈ H, where k ∈ R.
cj dj
A holomorphic function F is called entire automorphic integral of weight 2k with multiplier system ν for
G(λ) if it admits an exponential series expansion
∞
X
F (z) = am e2πimz/λ , (2.1)
m=0
 AUTOMORPHIC INTEGRALS WITH LOG-POLYNOMIAL PERIOD FUNCTIONS AND ARITHMETICAL IDENTITIES 3

where ℑz = y > 0 and am = O(mδ ), for some δ ∈ R+ , and satisfies the transformations
ν̄(Sλ )F (z + λ) = F (z), ν(Sλ ) = e2πiκ , 0 ≤ κ < 1 (2.2)
and
 
−1
ν̄(T )z −2k F = F (z) + q(z) ∀z ∈ H, (2.3)
z
N Mj
βjt (log z)t , αj , βjt ∈ C.
X X
where q(z) = z αj
j=1 t=0
The function q(z) in (2.3) is called the log-polynomial period function of the entire automorphic integral
function F.

A. Hassen in [4] has completely characterized the log-polynomial period functions for entire automorphic
integrals of weight 2k, k ∈ R on the discrete Hecke group G(λ) for the following cases:
(1) k ≥ 1, ν(Sλ ) = 1
(2) k > 0, ν(Sλ ) 6= 1
(3) k ≥ 0, 2k ∈ Z, ν(Sλ ) = 1 and
(4) k ≤ 0, ν(Sλ ) 6= 1.
For the purpose of this work we consider only the multiplier system which which ν(Sλ ) = 1 (or (κ = 0).

In [10] Paul C.Pasles showed Riemann-Hecke-Bochner correspondence for entire automprphic integrals on
the Hecke groups. We need a modified result of Pasles [10] (see Theorem 3.1), which we state and proof in
 
Theorem 2.1. Let k ∈ R+ ∪ {0}, λ ≥ 2 or λ = 2 cos πp , p ∈ Z, p ≥ 3. Suppose that F is an entire
automorphic integral of weight 2k, multiplier system ν and ν(Sλ ) = 1. Let q(z) be log-polynomial period
function of F on G(λ) and let F has Fourier expansion of the form
∞
X
F (z) = am e2πimz/λ ,
m=0

for z ∈ H, where
(i) am = O(mβ ) for some β > 0, m → ∞.
∞
X
(ii) Let ϕ(s) = am m−s , s = σ + it. And put
 m=1
−s
2π
(iii) ΦF (s) = Γ(s)ϕ(s), for σ > β + 1.
λ
Then
(A) ΦF (s) has a meromorphic continuation to the whole complex plane with at most a finite number of
poles given by:

ΦF (s) = Dk (s) + Mk (s) + Lk (s),

where
Z ∞   Z ∞  
s−1 2k
Dk (s) = F (iy) − a0 y ds + i ν(T ) F (iy) − a0 y 2k−s−1 ds, (2.4)
1 1
4 TEWLEDE G/EGZIABHER, HUNDUMA LEGESSE GELETA AND ABDUL HASSEN

i2k ν(T ) 1
 
Mk (s) = a0 − , (2.5)
s − 2k s
and

N Mj t   t−l
X X X t iπ (−1)l+1 l!
Lk (s) = − i2k+αj bjt , (2.6)
j=1 t=0
l 2 (2k + αj − s)l+1
l=0

for Res > 2k + max |Reαj |.

j
(B) ΦF (s) is bounded uniformly and absolutely in each σ in a lacunary vertical strips of the form
S(σ1 , σ2 , t0 ) = {s : σ1 ≤ σ ≤ σ2 , |ℑs| = t ≥ t0 },
where σ1 , σ2 ∈ R and t0 > max |ℑαj |.
j
Moreover,
(C) ΦF (s) satisfies the functional equation
ΦF (2k − s) = eπik ν(T )ΦF (s). (2.7)
Proof. Suppose F is an entire automorphic integral of weight 2k, k ∈ R, multiplier system ν and associ-
ated LPPF q(z). For Res > 2k + maxj |Reαj |, applying Mellin transform of F, we have
Z ∞
dy
ΦF (s) = (F (iy) − a0 ) y s
0 y
Z ∞ Z 1
dy dy
= (F (iy) − a0 ) y s + (F (iy) − a0 ) y s .
1 y 0 y
1
Taking y 7→ y in the later integral and applying the transformation law in (2.3),we get

1 ∞ ∞
dy
Z Z Z
(F (iy) − a0 ) y s = ν(T )i2k (F (iy) − a0 ) y 2k−s−1 dy + ν(T )i2k a0 y 2k−s−1 dy
0 y
Z 1∞ 1
a0
+ ν(T )i2k q(iy)y 2k−s−1 dy −
1 s
∞
ν(T )i2k
 
1
Z
2k 2k−s−1
= ν(T )i (F (iy) − a0 ) y dy + a0 −
s − 2k s
Z 1∞
+ ν(T )i2k q(iy)y 2k−s−1 dy.
1
Thus
ΦF (s) = Dk (s) + Mk (s) + Lk (s),
where
Z ∞ Z ∞
s−1 2k
Dk (s) = (F (iy) − a0 ) y dy + ν(T )i (F (iy) − a0 ) y 2k−s−1 dy,
1 1

ν(T )i2k
 
1
Mk (s) = a0 − ,
s − 2k s
and
Z ∞
2k
Lk (s) = ν(T )i q(iy)y 2k−s−1 dy.
1
 AUTOMORPHIC INTEGRALS WITH LOG-POLYNOMIAL PERIOD FUNCTIONS AND ARITHMETICAL IDENTITIES 5

Dk (s) is entire and converges uniformly (absolutely) on compact subset of C, for large Re( s) and Dk (s) satis-
fies the functional equation Dk (2k − s) = eπik ν(T )Dk (s). Similarly Mk (s) is meromorphic with simple poles
at (s = 2k, 0), and satisfies the functional equation Mk (2k − s) = eπik ν(T )Mk (s). In order to show the
meromerphic continuation of ΦF (s) we need the meromorphic continuation of Lk (s) to C. To see this, note
that
Z ∞
Lk (s) = ν(T )i2k q(iy)y 2k−s−1 dy
1
Z N
∞X Mj
X
= ν(T )i2k (iy)αj βjt (log iy)t y 2k−s−1 dy.
1 j=1 t=0
iπ
Using the fact that log iy = log y + i arg(iy) = log y + 2 for y > 0, and applying the binomial theorem on
iπ

l
2 + log y , we can rewrite the above equation as

N Mj t   t−l Z ∞
X
2k+αj
X X t iπ
Lk (s) = ν(T ) i βjt y 2k+αj −s−1 (log y)t dy.
j=1 t=0
l 2 1
l=0
Integration by parts now yields
N Mj t   t−l
X
2k+αj
X X t iπ l!
Lk (s) = ν(T ) i βjt ,
j=1 t=0
l 2 (s − (αj + 2k))l+1
l=0

for Re s > 2k + maxj Reαj . Thus Lk (s) is meromorrhpic in C with poles of order Mj + 1 at s = 2k + αj .
Thus
ΦF (s) = Dk (s) + Mk (s) + Lk (s)
is meromorphic on C with finite simple poles at s = 2k, s = 0 and poles of order Mj + 1 at s = 2k + αj .
This completes the prove of (A).

To prove ΦF (2k − s) = eπik ν(T )Φ(s) it is enough to show that Lk (2k − s) = eπik ν(T )Lk (s). For this we
need
Lemma 2.1. If F and q are as in Theorem 2.1, then
 
−1
q = −(−z)2k ν̄(T )q(z), ∀z ∈ H. (2.8)
z
−1
Proof. By taking z → z in (2.3), we have
  
−1
F (z) = ν(T )(−z)−2k ν(T )z 2k F (z) + ν(T )z 2k q(z) + q .
z
Thus,
 
2 −2k 2k 2 −2k 2k −2k −1
F (z) − ν (T )(−z) z F (z) = ν (−z) z q(z) + ν(T )(−z) q
z
 
4k 2 −2k −1
= i ν (T )q(z) + ν(T )(−z) q
z
 
2 2 −2k −1
= (i ν(T )) q(z) + ν(T )(−z) q
z
 
−1
= q(z) + ν(T )(−z)−2k q .
z
6 TEWLEDE G/EGZIABHER, HUNDUMA LEGESSE GELETA AND ABDUL HASSEN

Therefore,  
2 2k 2 −2k −1
F (z) − ν (T )(i ) F (z) = q(z) + ν(T )(−z) q .
z
From the consistency condition on ν, we see that ν 2 (T )(i2k )2 = 1 and hence
 
−1
q(z) + ν(T )(−z)−2k q = 0, ∀z ∈ H,
z
as desired.
Z ∞
Now from Lk (s) = i2k ν(T ) y 2k−s−1 q(iy)dy, we have,
1
Z ∞
2k−s
Lk (2k − s) = i ν(T ) y s−1 q(iy)dy.
1
1
Applying Lemma 2.1 and replace y by in the integral we obtain
y
Z 1
−2k 2k
i ν̄(T )Lk (2k − s) = −i ν(T ) y 2k−s−1 q(iy)dy.
0

Then by definition of q we have

Z N
1X Mj
X
−2k 2k αj
i ν̄(T )Lk (2k − s) = −i ν(T ) (iy) βjt (log iy)t y 2k−s−1 dy.
0 j=1 t=0

iπ 1
Using log iy = 2 + log y and applying the binomial theorem, the substitution y 7→ , and integration by
y
parts we have,

N Mj t  t−l Z ∞
−2k 2k
X
αj
X X iπ
i ν̄(T )Lk (2k − s) = −i ν(T ) (i) βjt y s−2k−αj −1 (log y)l (−1)l dy
j=1 t=0
2 1
l=0
N Mj t   t−l
X X X t iπ l!
= ν(T ) (i)αj +2k βjt .
j=1 t=0
l 2 [s − (2k + αj )]l+1
l=0

Hence
N Mj t   t−l
X
αj +2k
X X t iπ l!
Lk (s) = ν(T ) (i) βjt .
j=1 t=0
l 2 [s − (2k + αj )]l+1
l=0
πik
Therefore, Lk (2k − s) = e ν(T )Lk (s). This complete the prove of (C).

Finally to show the boundedness condition (B), note that the rational functions Mk and Lk are bounded
in lacunary vertical strips which do not contain poles of Mk and Lk . The boundedness of Dk follows from
the fact that am = O(mβ ) for σ > 1 + β implies
Z ∞ 
s−1
 

 [F (iy) − a 0 ]y dy  = O(ζ(σ − β)Γ(σ)).

1

This complete the proof of (B) and that of the theorem.

 AUTOMORPHIC INTEGRALS WITH LOG-POLYNOMIAL PERIOD FUNCTIONS AND ARITHMETICAL IDENTITIES 7

Finally before we state and proof Theorem 3.1, we state Perron’s Formula as the following Lemma. (See
Z Z b+i∞
[2] for details.) We shall use the convention of writing for .
(b) b−i∞
∞
X
Lemma 2.2. Let σ0 be the abscissa of absolute convergence for ϕ(s) = am λ−s
m and {λm } be a sequence
m=1
of positive real numbers tending to ∞ as m → ∞. Then for k ≥ 0, σ > 0 and σ > σ0 ,
1 1 Γ(s)ϕ(s)xs+k
Z
am (x − λm )k =
X′
ds, (2.9)
Γ(k + 1) 2πi (σ) Γ(s + k + 1)
λm ≤x

where the prime ′ on the summation sign indicates that if k = 0 and x = λm for some positive integer m,
then we count only 21 am .

3. Main results
In this section we state and prove the main results of the paper. Here we establish two arithmetical
identities associated to entire automorphic integrals involving log-polynomial period functions on the discrete
Hecke group G(λ), analogous to Chandrasekharan and Narasimhan and show that they are equivalent to
the functional equation (2.7),
ΦF (2k − s) = eπik ν(T )ΦF (s).

Theorem 3.1. Let ΦF (s),q(z) and F (z) be as in Theorem 2.1. Then the functional equation
ΦF (2k − s) = eπik ν(T )ΦF (s) (3.1)
is equivalent to the arithmetical identity
 −ρ X ∞  x  2k+ρ  √ 
1 X 2π 2 4π mx
am (x − m)ρ = e−πik ν̄(T ) am J2k+ρ
Γ(ρ + 1) λ m=1
m λ
06m6x
 2k
2π ν(T )eπik a0 x2k+ρ a0 xρ
+ −
λ Γ(2k + ρ + 1) Γ(ρ + 1)
 2k+αj N Mj  t
2π x2k+ρ+αj πik
X
αj
X iπ
+ ν(T )e i βjt , (3.2)
λ Γ(2k + αj + ρ + 1) j=1 t=0
2

for x > 0, ρ ∈ Z≥0 , αj , βjt ∈ C, βjMj 6= 0 and N, Mj ∈ Z, provided ρ ≥ 2β − 2k, where am = O(mβ ).
Proof of (3.1) implies (3.2). .
Proof. Applying Lemma 2.2, for ρ > 0, b > β and am = O(mβ ) we have
1 1 Γ(s)ϕ(s)xs+ρ
X ′ Z
am (x − m)ρ = ds. (3.3)
Γ(ρ + 1) 2πi (b) Γ(s + ρ + 1)
06m6x
∞
1
X am
Here the prime notation ’ indicates that if m = x, am is to be multiplied by 2 and ϕ(s) = with
m=1
ms
∞
X |am |
< ∞.
m=1
mβ
To evaluate the integral on the right side of (3.3), for 2k − b < σ < b, where s = σ + it, consider rectangular
8 TEWLEDE G/EGZIABHER, HUNDUMA LEGESSE GELETA AND ABDUL HASSEN

region with vertices 2k − b ± iT and b ± iT, having positive orientation. Let T be large positive number
so that Stirling’s formula can be used in approximating Γ(s) as |s| → ∞. Using Stirling’s formula and the
Phragmen-Lindelof theorem
s we can show the integrals along the horizontal paths tend to 0 as T → ∞. Hence
with the substitution, 2πλ ΦF (s) = Γ(s)ϕ(s), the right hand side of (3.3) may be written as
Z  s
1 Γ(s)ϕ(s) 1 2π ΦF (s)xs+ρ
Z
ds = ds
2πi (b) Γ(s + ρ + 1) 2πi (b) λ Γ(s + ρ + 1)
 s
1 2π ΦF (s)xs+ρ
Z
= ds
2πi (2k−b) λ Γ(s + ρ + 1)
  s
ΦF (s)xs+ρ

X 2π
+ Res .
λ Γ(s + ρ + 1)
s∈P olsetof ΦF (s)

Put s
ΦF (s)xs+ρ

1 2π
Z
ω1 (x, λ) = ds
2πi (2k−b) λ Γ(s + ρ + 1)
and s
ΦF (s)xs+ρ
 
X 2π
ω2 (x, λ) = Res .
λ Γ(s + ρ + 1)
s∈P olsetof ΦF (s)
To evaluate ω1 (x, λ) we begin by using the functional equation
ΦF (2k − s) = eπik ν(T )Φ(s).
ω1 (x, λ) can be written as
s
e−πik ν̄(T )ΦF (2k − s)xs+ρ

1 2π
Z
ω1 (x, λ) = ds.
2πi (2k−b) λ Γ(s + ρ + 1)
By substituting
 s ∞
2π X am
ΦF (2k − s) = Γ(2k − s) 2k−s
λ m=1
m
and replacing s by 2k − s, to evaluate ω1 (x, λ) we can use the integral in ([2], page 5 ) given by the formula
Z C+i∞ s−ν−1Γ( s )
1 2 2 Jν (x)
x−s ds = ,
2πi C−i∞ Γ(ν − 12 s + 1) xν
provided that 0 < C ≤ ν + 1, ν > 0.
With simple algebraic manipulation we obtain
 −ρ ∞  x  2k+ρ  √ 
2π −πik
X 2 4π mx
ω1 (x, λ) = e ν̄(T ) am J2k+ρ , (3.4)
λ m=1
m λ
provided that ρ ≥ 2β − 2k.
To complete evaluation of the integral in (3.3) we calculate the residues in ω2 (x, λ). Note that ΦF has poles
at s = 0, 2k resulting from Mk and at s = 2k + αj with order Mj + 1 from Lk . Thus calculating the residues
of the functions we obtain,
 2k
2π x2k+ρ a0 ν(T ) a0 xρ
ω2 (x, λ) = eπik −
λ Γ(2k + ρ + 1) Γ(ρ + 1)
 2k+αj N Mj  t
2π x2k+ρ+αj X X iπ
+ ν(T )eπik iαj βjt , (3.5)
λ Γ(2k + ρ + 1 + αj ) j=1 t=0
2
 AUTOMORPHIC INTEGRALS WITH LOG-POLYNOMIAL PERIOD FUNCTIONS AND ARITHMETICAL IDENTITIES 9

for ρ ∈ Z≥0 , αj , βjt ∈ C and βjMj 6= 0.

The integral in right of (3.3) is equal to the combination of the expressions in (3.4) and (3.5) respectively,
thus it follows the identity (3.2).
Proof of converse of the theorem: Suppose F be entire automorphic integral function. F (z) has Fourier
series expansion of the form
∞
X
F (z) = am e2πim/λ . (3.6)
m=0
For z ∈ H, where am = O(mγ ) as m → ∞, for some γ > 0. Put
 −s ∞
2π X
ΦF (s) = Γ(s) am m−s .
λ m=1
For z ∈ H, F satisfies the relation
 
−1
ν̄(T )z −2k F = F (z) + q(z)∀z ∈ H, (3.7)
z
where
N Mj
X X
q(z) = z αj βjt (log z)t , αj , βjt ∈ C.
j=1 t=0
Now using (3.6) and (3.7) we have
∞
X ∞
X
z −2k ν̄(T ) am e−2πim/λz = am e2πimz/λ + q(z).
m=0 m=0
iyλ +
Letting z = 2π , y > c, c ∈ R , we have
 2k ∞ ∞  
2π X −4π 2 m X iyλ
ν̄(T ) am e yλ2 = am e−my + q . (3.8)
iyλ m=0 m=0
2π
In proving the converse, it is enough to show that the identity in (3.2) implies (3.8). Since q(z) is entire in
H and the series expansion for F in (3.6) is uniformly convergent in compact subsets of H, the automorphic
relation in (3.7) will follow by analytic for ∀z ∈ H. Following the method used in Chandraseharan and
Narasimhan [2], we multiply the identity in (3.2) through out by y ρ+1 e−xy with y > 0, integrate relative to
x from 0 to ∞. To this end consider four integrals (separate) corresponding to the terms occurring in (3.2)
as
Z ∞( )
1 X
̟1 (y, λ) = am (x − m)ρ y ρ+1 e−xy dx.
0 Γ(ρ + 1)
0≤m≤x

∞
(  ∞
−ρ X  x  2k+ρ  √ )
2π 4π mx
Z
2
−πik
̟2 (y, λ) = e ν̄(T ) am J2k+ρ y ρ+1 e−xy dx.
0 λ m=1
m λ
( 2k
∞
a0 ν(T )x2k+ρ a0 xρ
 
2π
Z
πik
̟3 (y, λ) = e − y ρ+1 e−xy dx.
0 λ Γ(2k + ρ + 1) Γ(ρ + 1)

∞
( 2k+αj N Mj  t )
2π x2k+ρ+αj iπ
Z X X
̟4 (y, λ) = ν(T )eπik iαj βjt y ρ+1 e−xy dx.
0 λ Γ(2k + ρ + αj + 1) j=1 t=0
2
10 TEWLEDE G/EGZIABHER, HUNDUMA LEGESSE GELETA AND ABDUL HASSEN

The evaluation of ̟1 and ̟2 follows as Chandrasekharan and Narasimhan found in [2]. Interchange of
summation and integration being permitted for ρ ≥ 2β − 2k − 21 . Then
Z ∞ Z ∞
yρ + 1 ρ −xy yρ + 1 X
̟1 (y, λ) = x e dx + am (x − ρ)ρ e−xy dx.
Γ(ρ + 1) 0 Γ(ρ + 1) m 1≤m≤x

Integration by substitution and using the standard integral representation for Γ(s) we get
∞
X
̟1 (y, λ) = am e−my . (3.9)
m=0

R∞ √ ν 2aν
−a2
To compute ̟2 we apply the integral formula 0√
e−xy Jν (a x)x 2 dx = (2y)ν+1 e
4y , provided that y >
4π m
0, Re(ν) > −1, a > 0. And with the substitution λ = a, ν = ρ + 2k after simplifying, we obtain
 2k ∞
2π X −4π 2 m
̟2 (y, λ) = e−πik ν̄(T ) am e yλ2 . (3.10)
λy m=1

The integral in ̟3 (y, λ) is evaluated using the integral representation for Γ(s). Therefore. after simplifying,
we see that
 2k
2π
̟3 (y, λ) = a0 eπik ν(T ) − a0 . (3.11)
λy
To evaluate ̟4 (y, λ) since each term is integrable the inter change of the integration and the finite double
sum is valid . Then integration by substitution and using the integral representation of Γ(z) we obtain
2k N  α Mj t
2πi j X
 
2πi X iπ
̟4 (y, λ) = ν(T ) βjt
λy j=1
λy t=0
2
 2k N  Mj
αj X  t
−1 X −1 iπ
= ν(T ) βjt . (3.12)
iλy/2π j=1
λyi/2π t=0
2
 
−1 iπ S
For y > 0, we see that log iy + log y = 2 . Then for l ∈ N {0} and applying the binomial theorem we
have
   t t    t−l
−1 X t −1
log + log y = log (log y)l .
iy l iy
l=0
  t    t
−1
If l = 0 we have log iy = log −1
iy + log y . Now replacing −1
iy by −1
iλy/2π and substituting

t
for iπ
2 in (3.12) yields
 2k N  Mj
αj X   t
−1 X −1 −1
̟4 (y, λ) = ν(T ) βjt log . (3.13)
iλy/2π j=1
λyi/2π t=0
iyλ/2π
 2k  
−1 −1
Observe that ̟4 (y, λ) = iλy/2π , where q(z) is the log-polynomial period function in
ν(T )q iyλ/2π
 2k    
−1
(3.7). Then applying Lemma2.1 we have ̟4 (x, λ) = iλy/2π −1
ν(T )q iyλ/2π = −q iyλ
2π . Combining the
 AUTOMORPHIC INTEGRALS WITH LOG-POLYNOMIAL PERIOD FUNCTIONS AND ARITHMETICAL IDENTITIES 11

results in (3.9), (3.10), (3.11) and (3.12) we showed that the identity in (3.2) implies
∞  2k ∞  2k
X
−my 2π −πik
X −4π 2 m 2π
am e = e ν̄(T ) am e yλ 2
+ a0 eπik ν(T )
m=0
λy m=1
λy
 2k N  Mj
αj X   t
−1 X −1 −1
+ ν(T ) βjt log . (3.14)
iλy/2π j=1
λyi/2π t=0
iyλ/2π

Observe that
 2k N  Mj
αj X   t  
−1 X −1 −1 iyλ
ν(T ) βjt log = −q .
iλy/2π j=1
λyi/2π t=0
iyλ/2π 2π
∞
X
Since F (z) is holomorphic and has a Fourier series expansion F (z) = am e2πimz/λ in H. Then with
m=0
z = iyλ
2π (3.14) written as (3.8). By the identity theorem ,then the automorphic relation (2.3) follows ∀z ∈ H,
and this completes proof of the theorem.
Theorem 3.2. Let ΦF (s), F (z) and q(z) be as in Theorem2.1 . Then the functional equation
ΦF (2k − s) = eπik ν(T )ΦF (s) (3.1)
is equivalent to the arithmetical identity
∞
ρ !   2k
2ρ
 √

−1 d 1 X −y m 1 8π
am e = √ Γ 2k + ρ + ν̄(T )e−πik
y dy y m=1 π 2 λ
∞
X am
×  2k+ρ+ 21
2π 2


m=0 y 2 + 4 m
λ
2k
2ρ Γ(2k + ρ + αj + 21 ) 8π

+√ ν(T )eπik
π y 2ρ+1 λy 2
N  α Mj  t
X 8πi j X iπ
× 2
β jt
j=1
λy t=0
2
2ρ+1 a0
 
1
− √ Γ ρ+ , (3.2)
π y 2ρ+1 2
provided that Rey ∈ R+ , αj , βjt ∈ C, ρ ∈ Z≥0 , and ρ > β − 2k + 12 ,
P∞
where m=1 |ammβ | < ∞.
Since, by Theorem 3.1 the functional equation (3.1) is equivalent to (3.2), it would be enough to show
that the functional equation (3.1) implies (3.2) and that (3.2) in turn implies the identity (3.2).

Proof of implication: First we begin by showing that (3.1) implies (3.2). Let ρ be an integer, ρ ≥ 0 and
∞
X
ϕ(s) = am m−s with am = O(mβ ). Then for γ > 0, γ ≥ β and applying the version of Perron’s in Lemma
m=1
2.2, we have
1 1 ϕ(s)Γ(s)xs+ρ
X ′ Z
am (x − m)ρ = ds. (3.3)
Γ(ρ + 1) 2πi (γ) Γ(s + ρ + 1)
0≤m≤x
12 TEWLEDE G/EGZIABHER, HUNDUMA LEGESSE GELETA AND ABDUL HASSEN

√ 1
As Chandrasekharan and Narasimhan in [2] we multiply (3.3) by e−y x x 2 and integrating with respect to
the variable x on [ 0, ∞). For γ > 2k, Res = γ is the vertical path of integration. Thus, we need to prove
the identity (3.2) follows from the equation
Z ∞ ( )
√ 1 X ′
−y x − 12 ρ
e x am (x − m) dx
0 Γ(ρ + 1)
0≤m≤x
Z ∞ ( )
√ 1 ϕ(s)Γ(s)xs+ρ
Z
−y x − 12
= e x ds dx, (3.4)
0 2πi (γ) Γ(s + ρ + 1)
where ρ ∈ Z≥0 , ρ ≥ γ − 2k + 21 , and y ∈ R+ . Chandrasekharan and Narasimhan [2] for λn ∈ R+ λn →
∞ as n → ∞ showed the identity
∞ ρ " X ∞
Z ∞ #
(x − n)ρ −y√x − 1 √

X
ρ 1 d 1 −y λ
an e x 2 dx = 2(−2) an e n
. (3.5)
n=1 λn Γ(ρ + 1) y dy y n=1
Then by the identity in (3.5) with λn = m, we have
∞ ρ " X ∞
Z ∞ #
(x − m)ρ −y√x − 1
 √
X
ρ −1 d 1 −y m
am e x 2 dx = 2(2) am e .
m=1 m Γ(ρ + 1) y dy y m=1
Now on the left side of (3.4) the interchange of integration and summation is justified for ρ ≥ 0. Then
Z ∞ ( ) Z ∞
√ 1 X ′ a0 √
−y x − 12 ρ 1
e x am (x − m) dx = e−y x xρ− 2 dx
0 Γ(ρ + 1) Γ(ρ + 1) 0
0≤m≤x
1
+ ×
Γ(ρ + 1)
Z ∞ √ 1
X
am e−y x x− 2 (x − m)ρ dx.
1≤m≤x m

Thus applying (3.5) we have

( )
∞ √ 1 22ρ+1 Γ(ρ + 1 )
Z X ′
−y x − 12 ρ
e x am (x − m) dx = 2ρ+1 a0 √ 2
0 Γ(ρ + 1) y π
0≤m≤x
 ρ  X ∞ √

−1 d 1
+ 2ρ+1 am e y m . (3.6)
y dp y m=1

−s
Now assuming ΦF (s) = 2πλ Γ(s)ϕ(s), where Res > β the right-hand side of (3.4) may be written as
Z ∞ ( Z  s )
√
−y x − 21 1 2π ΦF (s)xs+ρ
W (y, λ) = e x ds dx,
0 2πi (γ) λ Γ(s + ρ + 1)
for γ ≥ β. We may interchange the order of integration for ρ ≥ 0. Then we have
Z (  s Z ∞ √
)
1 2π ΦF (s) 1
W (y, λ) = e−y x xs+ρ− 2 dx ds.
2πi (γ) λ Γ(s + ρ + 1) 0
Using integration by substitution and applying the Legendre duplication formula
√
 
2z−1 1
πΓ(2z) = 2 Γ(z)Γ z + , (3.7)
2
 AUTOMORPHIC INTEGRALS WITH LOG-POLYNOMIAL PERIOD FUNCTIONS AND ARITHMETICAL IDENTITIES 13

for Γ(2s + 2ρ + 1), we obtain

s
22ρ+2s+1 Γ(s + ρ + 21 )

1 2π
Z
W (y, λ) = ΦF (s) √ ds.
2πi (γ) λ y 2ρ+2s+1 π
To evaluate this integral we shall consider a rectangular region with vertices 2k − γ ± iT and γ ± iT, s = σ + it
on its interior. Using Cauchy’s residue
theorem, the line of integration oriented positively to Res = 2k − γ. If
ρ ≥ γ − 2k + 21 the poles of Γ s + ρ + 12 lie to the left of the vertical path. Then the poles of the integrand in
the interval [2k − γ, γ] arise from the function ΦF (s). Applying Stirling’s formula and the Phragmen-Lindelof
theorem the integrals along both horizontal paths tend to 0 as T → ∞, and thus W (y, λ) may be written as
W (y, λ) = w1 (y, λ) + w2 (y, λ),
where
s
Γ(s + ρ + 21 ) 22s+2ρ+1

1 2π
Z
w1 (y, λ) = ΦF (s) √ ds,
2πi (2k−γ) λ π y 2s+2ρ+1
and
s
Γ(s + ρ + 12 ) 22s+2ρ+1
 
X 2π
w2 (y, λ) = Res Φ(s) √ .
λ π y 2s+2ρ+1
s∈polsetΦ(s)

To evaluate w1 (y, λ) recall the functional equation

ΦF (2k − s) = eπik ν(T )ΦF (s)
and substitute in to the integral of w1 (s, λ). Then we have
(  s )
−πik 2
2ρ+1
1 2π 22s Γ(s + ρ + 21 )
Z
w1 (y, λ) = ν̄(T )e ΦF (2k − s) √ ds .
y 2ρ+1 2πi (2k−γ) λ y 2s π

−θ P∞
Now replace θ by 2k − s and then setting ΦF (θ) = 2π λ Γ(θ) m=1 am m−θ . By organizing the expres-
sions w1 (y, λ) may be written as
∞ 2 −θ
( )
2ρ+1 X
 
2 1 1 16mπ
Z
w1 (y, λ) = ν̄(T )e−πik 2ρ+1 am Γ(θ)Γ(2k − θ + ρ + ) dθ .
y m=1
2πi (γ) 2 λ2 y 2

The interchange of the integration with the summation is justified for ρ > β − 2k + 21 .
Replace θ by −θ and using the integral formula
Z γ+i∞
1
Γ(−s)Γ(β + s)ts ds = Γ(β)(1 + t)−β , (3.8)
2πi γ−i∞
where 0 > γ > Re(1 − β) and |argt| < π. After simplifying, we obtain
  2k X ∞
ν̄(T )e−πik 4k+2ρ+1

1 2π am
w1 (y, λ) = √ 2 Γ 2k + ρ +
2k+ρ+ 21 . (3.9)
π 2 λ 2 4π 2
m=1 y + ( λ ) m

The series on (3.9) converges absolutely for Rey ∈ R+ , provided ρ > β − 2k + 12 .

To proceed evaluation of w2 (y, λ) recall that in Theorem 2.1 we have that ΦF (s) = Dk (s) + Mk (s) + Lk (s),
where Dk (s), Mk (s) and Lk (s) are functions in (2.4), (2.5), and (2.6) respectively. For ρ ≥ γ − 2k + 21 the

s Γ(s+ρ+ 12 ) 22s+2ρ+1
function 2π λ
√
π y 2s+2ρ+1 is analytic in the region bounded by Res = 2k − γ and Res = γ, thus the
poles of the intgrand in the interval [2k − γ, γ] arise from ΦF (s). We also note that Dk (s) is entire, Mk (s) has
14 TEWLEDE G/EGZIABHER, HUNDUMA LEGESSE GELETA AND ABDUL HASSEN

simple poles at (s = 2k, 0) and Lk (s) has a pole at s = 2k + αj of order Mj + 1, where αj ∈ C, Mj ∈ Z≥0 .
Therefore. by computing the residues of Mk and Lk after simplifying, we obtain
2k 
22ρ+1 22ρ+1
   
πik 8π 1 1
w2 (y, λ) = √ ρ+1 a0 ν(T )e Γ 2k + ρ + − √ ρ+1 a0 Γ ρ +
πy λy 2 2 πy 2
2k N Mj t
22ρ+1
     
8π 1 X αj X iπ
+ √ ρ+1 ν(T )eπik Γ 2k + ρ + αj + i β jt . (3.10)
πy λy 2 2 j=1 t=0
2

There for by rewriting W (y, λ) as combination of (3.9) and (3.10) and substituting the respective expressions
in to the right-hand side of (3.4), with simple rearrangement, we obtain the identity (3.2), provided that
P∞
ρ ∈ Z≥0 , ρ ≥ β − 2k + 21 and y ∈ R+ , where m=1 |ammβ| < ∞.
Proof of the converse: To prove the converse it suffices to√show that (3.2) implies (3.2). As technique of
Chandrasekharan and Narasimhan in [2] multiply (3.2) by ey x , with Rey > 0, and x > 0 and integrate the
expression along a vertical path Res = θ, where θ > 0. The left hand side of the transformed equation can
be evaluated using the formula
∞ √
ρ 
1 −y√m
 
1 −1 d 1
X Z X′
y x
am e e dy = (x − m)ρ 2−ρ , (3.11)
m=1
2πi (θ) y dy y Γ(ρ + 1)
m≤x

while the right hand side of (3.2) we calculate the integral of each terms one by one . So put
√ ρ
   2k X ∞
1 y x 2 1 2π am
Z
−2k 4k
h1 (x, λ) = e √ ν̄(T )i 2 Γ 2k + ρ + dy,
2πi (θ) π 2 λ 2 4π 2 2k+ρ+ 21
m=1 (y + ( λ ) m)
ρ 1  2k
y x 2 a0 Γ(2k + ρ + 2 )
1 √ 8π
Z
πik
h2 (x, λ) = e √ ν(T )e dy,
2πi (θ) π y 4k+2ρ+1 λ
1 √ 2ρ+1 Γ(ρ + 1 )
Z
h3 (x, λ) = ey x √ a0 2ρ+12 dy,
2πi (θ) π y
and
√ 2ρ Γ(2k + ρ + αj + 1 )
2k N  α Mj  t
8πi j X

1 8π iπ
Z X
h4 (x, λ) = ey x √ 2
ν(T )e πik
β jt dy.
2πi (θ) π y 2ρ+1 λy 2 j=1
λy 2
t=0
2

The evaluation of h1 (y, λ) follows as in [2]. The inter change of integration and summation being justified
for ρ > β − 2k + 21 in the right-hand side of h1 (x, λ). Applying the integral formula (see [2], page 22)
Z w+i∞  ν √ −ν
1 ebs b π2
ds = J ν (ab) , (3.12)
2πi w−i∞ (a2 + s2 )ν+ 12 a Γ(ν + 12 )
where w > 0, ν > − 21 , a > 0, b > 0. After simplifying each expressions, we obtain
 ρ ∞  x  2k+ρ  √ 
−πik λ −ρ
X 2 4π mx
h1 (x, λ) = ν̄(T )e 2 am J2k+ρ . (3.13)
2π m=1
m λ
To evaluate h2 (x, λ) we rewrite as
2k √
2ρ ey x
  
8π 1 1
Z
h2 (x, λ) = √ a0 ν(T )eπik Γ 2k + ρ + dy.
π λ 2 2πi (θ) y 4k+2ρ+1
Put √
1 ey x
Z
g1 (y) = dy.
2πi (θ) y 4k+2ρ+1
 AUTOMORPHIC INTEGRALS WITH LOG-POLYNOMIAL PERIOD FUNCTIONS AND ARITHMETICAL IDENTITIES 15

To compute g1 (y) we use the integral representation of reciprocal gamma function Hankel’s formula in [[15],
page246] given by
Z σ+i∞
1 1
et t−z dt = , where Rez > 0, σ > 0. (3.14)
2πi σ−i∞ Γ(z)
1
and applying the formula in (3.7) for Γ(4k+2ρ+1) . After simplification we obtain
2k
x2k+ρ

2π
h2 (x, λ) = 2−ρ a0 ν(T )eπik . (3.15)
λ Γ(2k + ρ + 1)
We evaluate h3 (x, λ) applying the techniques analogous used in evaluating h2 (y, λ). Thus, we conclude that
2a0 xρ
h3 (x, λ) = ρ
. (3.16)
2 Γ(ρ + 1)
Next we consider h4 (x, λ) the interchange of the integration and the finite double sums is justified and we
write as;
2k  N  α Mj  t
2ρ 1 X 8πi j X
 
8π iπ
h4 (x, λ) = √ ν(T )eπik Γ 2k + ρ + αj + βjt
π λ 2 j=1 λ t=0
2
√
1 ey x
Z
× dy.
2πi (θ) y 4k+2ρ+αj +1
Now put
√
1 ey x
Z
g2 (y) = dy.
2πi (θ) y 4k+2ρ+αj +1
1
Applying (3.14) in g2 (y) and using formula (3.7) for Γ(4k+2ρ+1αj +1) . By substituting the expression derived
for g2 (y) and simplifying, we obtained
N 
2k X Mj
αj X t
x2k+ρ
 
−ρ πik 2π 2πix iπ
h4 (x, λ) = 2 ν(T )e βjt . (3.17)
Γ(2k + ρ + αj + 1) λ j=1
λ t=0
2

Therefore. we have
1 X′
am (x − m)ρ 2−ρ = h1 + h2 + h3 + h4 . (3.18)
Γ(ρ + 1)
m≤x

By substituting the expressions (3.13), (3.15), (3.16), and (3.17) in to h1 , h2 , h3 , and h4 respectively and
multiplying both sides of the equation (3.18) by 2ρ the identity in (3.2) holds for ρ ≥ 2β − 2k + 12 . This
completes the proof of the converse, since (3.2) is equivalent to (3.1)as we shown in Theorem 3.1.

4. Conclusion and Discussion

In this paper we have used techniques of Chandrasekharan and Narasimhan and extended the results of
Sister Ann M. Heath to entire automorphic integrals involving rational period functions on discrete Hecke
group. Moreover, we have established equivalence of two arithmetical identities with a functional equation
associated with automorphic integrals involving log-polynomial-period functions on the Hecke groups. As
noted in the introduction part of this paper Sister Ann M. Heath considered the functional equation in the
Hawkins and Knopp context and showed its equivalence to two arithmetical identities associated with entire
modular cusp integrals involving rational period functions for the full modular group.
16 TEWLEDE G/EGZIABHER, HUNDUMA LEGESSE GELETA AND ABDUL HASSEN

References
[1] Bochner, S., Connection between functional equations and modular relations, and functions of exponential type , J.Indian
Math.Soc.,16(1952),99-102.
[2] Chandrasekharan. K. and Narasimhan, R, Hecke’s functional equation and arithmetical identities, The Annals of Mathe-
matics, second series, Vol.74,No.1, (1961), 1-23.
[3] Hardy, G. H., The General Theory of Dirichlet’s Series, Cambridge Tracts in Math. and Math. Physics, no. 18, Cambridge,
London, 1915.
[4] Hassen, Abdulkadir, Log-polynomial period function for Hecke groups, Ramanujan Journal 3, 119-151, (1999).
[5] Hawkins, J and Knopp, M., A Hecke correspondence theorem for automorphic integrals with rational period functions,
Illinois Journal of mathematics., 36, no. 2 (1992), 178-207.
[6] Hekce, E. Dirichlet series, modular functions, and quadratic forms, Edwards Brothers,Inc.,Ann Arbor,MI 1938.
[7] Knopp, M. I., Rational period functions of the modular group, Duke Mathematical Journal, 45,No.1, March 1978, 47-62.
[8] Knopp M. I., Rational period functions of the modular group II, Glasgow Math. Jan.22 (1981), 185-197.
[9] Knopp, M. I, Some new results on the Eichler cohomology of automorphic forms. Bull. Ann. Math, Soc. 80(4),607-631(174).
[10] Pasles,Paul C., A Hecke Correspondence Theorem for Non analytic Automorphic Integrals, Journal of Number theory 83,
256-281 (2000).
[11] Pasles, P. C., Nonanalytic automorphic integrals on the Hecke groups, Acta Arith. (1999),155-171.
[12] Resseler, W. A Hecke Correspondence Theorem for Automorphic Integrals with symmetric Rational period Functions on
Hecke Groups. Franklin and Marshal college, Lancaster, PA 17604.
[13] Sister Ann M.Heath, Dirichlet series with functional equations and arithmetical identities, Springer Science +Business
Media New York 2015.
[14] Titchmarsh, E. C. ” Theory of Functions” 2nd ed., Clarendon Oxford,1939.
[15] Whittaker, E. T., and Watson, G.N., A Course of Modern Analysis, 4th Ed., Cambridge, London, 1950.

Department of Mathematics: Ambo University, Ambo, Addis Ababa University, Addis Ababa, Rowan University,
Glassboro, NJ 08028.
Email address: teweldega@gmail.com, hunduma.legesse@aau.edu.et, hassen@rowan.edu