Available online at www.sciencedirect.

com

ScienceDirect

Indagationes Mathematicae xxx (xxxx) xxx

www.elsevier.com/locate/indag

First moment of Hecke eigenvalues at the integers

represented by binary quadratic forms
Manish Kumar Pandeya , Lalit Vaishyab ,∗
aDepartment of Mathematics, UPES, Dehradun 248007, Uttarakhand, India
b Stat-Math Unit, Indian Statistical Institute, 7, S. J. S. Sansanwal Marg, New Delhi 110016, India
Received 31 January 2024; received in revised form 5 June 2024; accepted 12 August 2024
Communicated by P. Moree

Abstract
In the article, we consider a question concerning the estimation of summatory function of the Fourier
coefficients of Hecke eigenforms indexed by a sparse set of integers. In particular, we provide an estimate
for the following sum;
∑♭
S( f, Q; X ) := λ f (n),
n=Q(x)≤X
x∈Z2 ,gcd(n,N )=1

where ♭ means that sum runs over the square-free positive integers, λ f (n) denotes the normalised nth
Fourier coefficients of a Hecke eigenform f of integral weight k for the congruence subgroup Γ0 (N ) and
Q is a primitive integral positive-definite binary quadratic forms of fixed discriminant D < 0 with the
class number h(D) = 1. As a consequence, we determine the size, in terms of conductor of associated
L-function, for the first sign change of Hecke eigenvalues indexed by the integers which are represented
by Q. This work is an improvement and generalisation of the previous results.
© 2024 Royal Dutch Mathematical Society (KWG). Published by Elsevier B.V. All rights are reserved,
including those for text and data mining, AI training, and similar technologies.
Keywords: Hecke eigenvalues; Fourier coefficients of cusp form; Rankin–Selberg L function; Symmetric power L
functions; Asymptotic behaviour; Binary quadratic form

1. Introduction and statements of the results

Let Mk (Γ0 (N ), χ) and Sk (Γ0 (N ), χ) denote the C- vector space of modular forms and cusp
forms respectively of integral weight k for the congruence subgroup Γ0 (N ) with nebentypus χ.
∗ Corresponding author.
E-mail addresses: [email protected] (M.K. Pandey), [email protected] (L. Vaishya).

https://doi.org/10.1016/j.indag.2024.08.001
0019-3577/© 2024 Royal Dutch Mathematical Society (KWG). Published by Elsevier B.V. All rights are reserved,
including those for text and data mining, AI training, and similar technologies.

Please cite this article as: M.K. Pandey and L. Vaishya, First moment of Hecke eigenvalues at the integers represented by binary quadratic forms,
Indagationes Mathematicae (2024), https://doi.org/10.1016/j.indag.2024.08.001.
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

Moreover, if the nebentypus is trivial, we denote the space of modular forms and cusp forms
by Mk (Γ0 (N )) and Sk (Γ0 (N )), respectively. A cusp form f ∈ Sk (Γ0 (N )) is said to be a Hecke
eigenform (newform) if f is a common eigenform for all the Hecke operators and the Atkin–
Lehner W -operators. The space of Hecke eigenforms (newforms) is denoted by Sknew (Γ0 (N )).
The Fourier coefficients of a Hecke eigenform satisfy the famous Ramanujan–Petersson bound
which follows from the work of Deligne [5]. With these notations, let k ≥ 2 be an even integer,
and f ∈ Sk (Γ0 (N )) be a Hecke eigenform with the Fourier series expansion at the cusp ∞
given by:
∞
∑ k−1
f (τ ) = λ f (n)n 2 qn, (1)
n=1

where q := e2πiτ and τ ∈ H. f is said to be normalised if λ f (1) = 1. The Fourier coefficients

λ f (n) are known as the normalised nth-Fourier coefficients of f . The normalised Fourier
coefficient λ f (n) is a multiplicative function and satisfies the Hecke relation [6, Eq. (6.83)]:
∑ ( mn )
λ f (m)λ f (n) = λf , (2)
d| gcd(m,n)
d2

for all positive integers m and n coprime to N . From Deligne’s estimate of Fourier coefficients
of cusp forms, we have
|λ f (n)| ≤ τ (n) ≪ϵ n ϵ , (3)
for any arbitrary small ϵ > 0, where τ (n) denotes the number of positive divisors of n.
The arithmetic behaviour and distribution of values of arithmetical functions is the center
of attraction in analytic number theory. The randomness in the behaviour of these arithmetical
functions leads to many interesting results. In this context, the arithmetical function n ↦→ λ f (n)
(where λ f (n) is the normalised nth Fourier coefficients of a Hecke eigenform f ) is one of
the interesting host. The problems become more fascinating and harder if one considers the
arithmetic behaviour and distribution over the sparse set of positive integers. Iwaniec and
Kowalski [8] studied the distribution of {λ f ( p) : p − prime} and proved that there exists a
positive constant c such that
∑ ( √ )
λ f ( p) ≪ f X exp −c log X ,
p≤X

which is an analogue of the prime number theorem for Hecke eigenforms. In this regard, a
popularly known problem is studied by Blomer [2], where he considers the distribution of the
sequence {λ f (q(n)) : n ∈ N}, and q(x) = x 2 + sx + t is a monic quadratic polynomial in Z[x].
More precisely, he proved that
∑ 6
λ f (q(n)) ≪ X 7 +ϵ ,
n≤X

for any ϵ > 0. For a polynomial with more than one variable, the problem has been studied
broadly for many arithmetic functions such as the Von Mangoldt function (n ↦→ Λ(n)) and
generalised divisor functions (n ↦→ τk (n)), etc. In his remarkable work, Acharya [1] has studied
a two variable analog of the sum studied in the work of Blomer [2], where he studied the
of {λ f (q(a, b))} and q(x, y) = x 2 + y 2 , and obtained an estimate for the summatory
distribution∑
function λ f (q(k, l)). For details, we refer to [1].
k 2 +l 2 ≤X
k,l∈Z

2
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

In this article, our motivation is to generalise the result of Acharya [1] by studying the
distribution of {λ f (Q(x))} where Q is a primitive integral positive-definite binary quadratic
forms of fixed discriminant D < 0 with the class number ∑ h(D) = 1. In previous works,
♭ j
the second author studies the higher power moment λ f (n) for j ≤ 8 in [17,18], and
n=Q(x)≤X
gcd(n,N )=1
estimates for more general higher power moment have been carried in [15]. Before stating our
result, we fix some notations.
Throughout the paper, Let Q(x) be a primitive integral positive-definite binary quadratic
form (reduced form) given by; Q(x) = ax12 +bx1 x2 +cx22 , where x = (x1 , x2 ) ∈ Z2 , a, b, c ∈ Z
with gcd(a, b, c) = 1 and fixed discriminant D = b2 − 4ac < 0. Further, we assume that for
such discriminant D, the class number h(D) is 1. For details, we refer to [4, chapter 2].
We consider the following sum for an upper-bound estimate:
♭
∑
S( f, Q; X ) := λ f (n),
n=Q(x)≤X
(4)
x∈Z2 , gcd(n,N )=1

where λ f (n) denotes the normalised nth Fourier coefficient of a Hecke eigenform f and symbol
♭ means that the sum runs over all square-free integers. We also produce the explicit dependency
in terms of weight k and level N of the Hecke eigenform f , and discriminant D of Q. More
precisely, we obtain the following estimate.

Theorem 1.1. Let f ∈ Sk (Γ0 (N )) be a normalised Hecke eigenform and Q be a reduced form
of discriminant D with the class number h(D) = 1. For sufficiently large X > 0 and any
arbitrary small ϵ > 0, we have
1 +ϵ 1
♭
∑
S( f, Q; X ) = λ f (n) ≪ϵ (N k 2 |D|) 2 X 2 +ϵ .
n=Q(x)≤X
gcd(n,N )=1

where the implied constant is absolute.

Remark 1.1. If the class number h(D) > 1, the same estimate (obtained in Theorem 1.1)
holds for the sum
♭
∑
S( f, D; X ) := λ f (n),
n=Q(x)≤X (5)
Q∈S D ,x∈Z2
gcd(n,N )=1

where S D denotes the set of in-equivalent primitive integral positive-definite binary quadratic
(reduced) forms of fixed discriminant with the class number h(D) = #S D .
A consequence of the above theorem is a result on the sign change of the Fourier coefficients.
Below, we give a brief description of the sign change problem and state our second result. It is
well-known that if the Fourier coefficients of a cusp form are real, it changes its sign infinitely
often. The sequence of Fourier coefficients of the Hecke eigenform always has a sign change in
the interval of length X , for sufficiently large X . The problem of getting first ever sign change
of λ f (n), is quite hard. As mentioned in [10], it is a G L 2 -analogue of the G L 1 problem which
concerns about finding least quadratic non-residue. The problem is quite hard in the sense
that λ f (n) may take infinitely many values and it can be very small. We also determine an
estimate for the first sign change of the real sequence of Hecke eigenvalues over the integers
3
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

which are represented by some function U. In particular, if U is an identity function on N or

characteristic function on the set of primes, the problem of first sign changes and counting
the number of sign changes in short interval [X, X + h], for sufficiently large X and some
h < X , has been considered by many authors. Let I denote the identity function on N, and
n f,I denote the integer at which the first sign change of the Fourier coefficients λ f (n) of Hecke
eigenforms f ∈ Sk (Γ0 (N )) occurs, i.e., the least integer among all n ∈ N such that λ f (n) < 0
and gcd(n, N ) = 1. The problem of bounding n f,I in terms of analytic conductor has attracted
many mathematicians. Kohnen and Sengupta [10] have proved that the bound of n f,I is given
by
( √ )
log(N + 1)
A
n f,I ≪ k N (log k) exp c2 ,
log log(N + 2)

for some constant A, c2 > 0, using analytic properties of automorphic L-functions. Jointly with
Iwaniec [7], they improved the bound for n f,I and obtained that n f,I ≪ Q 29/60 , where Q is the
analytic conductor, of the L-function associated to f , given explicitly in terms of weight and
level of the form f . Kowalski et al. [11] have made an important observation that the Hecke
relation (2) allows one to replace λ f (n) by a step function when one wants to find the lower
bound of partial sum of λ f (n). This allowed to improve the exponent, i.e., n f,I ≪ Q 9/20 . In
the article [11], they mentioned that the exponent 9/20 is not the limit of the method. This was
further improved by Matomäki [13] to 3/8 by deep analysis of summatory function of λ f (n)
over square-free integers and this is best known bound for n f,I . The current known bound for
n f,I is still far from the truth as it is known that n f,I ≪ (log k N )2 under generalised Riemann
hypothesis. The problem of getting an estimate the first-ever sign change becomes interesting
if the sequence of integers is indexed by some function other than the identity function on N.
In particular, a result of this sort has been obtained by the first author in a joint work with
P. Tiwari [14], where the sequence of integers is written as a sum of two squares, i.e. by
considering the function Q : Z2 → N given by Q(x, y) = x 2 + y 2 . Following the argument of
Kohnen and Sengupta [10], Pandey and Tiwari [14] have proved that
( ( ) )
log(N + 1)
n f,Q ≤ k 8 N 4 log64 (k N ) max N 4 exp c (log log(N + 2))4 ψk4 (N ), log120 (k N ) ,
log log(N + 2)

∏ log(k N )
where ψk (N ) = .
p|N
log p
In this work adopting the method of Kowalski et al. and Matomaki, we improve the previous
result of the first author and Tiwari, and obtain an estimate for n f,Q where Q is a primitive
integral positive-definite binary quadratic form of fixed discriminant D with the class number
h(D) = 1.
Now, we state our second result concerning the first-ever sign change of Hecke eigenvalues
of a normalised Hecke eigenform at the integers represented by Q(x). Let n f,Q be the least
integer among all n ∈ N such that λ f (n) < 0, with gcd(n, N ) = 1 and n is represented by
Q(x). With these notations, we observe that n f,I ≤ n f,Q as it is not necessary that n f,I can
always be represented by Q(x) for some x ∈ Z2 . Hence, an upper bound estimate for n f,Q
will also work for n f,I . An estimate for n f,Q is given by the following result.

Theorem 1.2. Let k ≥ 2 be an even integer and N be a square-free positive integer. Let
f ∈ Sk (Γ0 (N )) be a Hecke eigenform and Q be a reduced form of fixed discriminant D with
4
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

the class number 1. Then, for any arbitrarily small ϵ > 0,

3
n f,Q ≪ϵ (N k 2 |D|2 ) 4 +ϵ
where the implied constant is absolute.

Remark 1.2. For the primitive integral positive-definite binary quadratic form (reduced form)
of the following fundamental discriminants D = −3, −4, −8, −7, −11, −19, −43, −67, −163
and non-fundamental discriminants D = −12, −16, −27, −28, the class number h(D) = 1.
The list of reduced forms corresponding to these discriminants is given in [3, Page 19- 20].
Theorem 1.2 holds only for primitive integral positive-definite binary quadratic form of such
discriminants.

Remark 1.3. If the class number h(D) > 1 for the discriminant D < 0, then there are h(D)
many reduced forms corresponding to discriminants D. Moreover, let n f,D be the least integer
among all n ∈ N such that λ f (n) < 0, with gcd(n, N ) = 1 and n is represented by some of
them from h(D) many reduced forms corresponding to discriminants D. Then,
3
n f,D ≪ϵ (N k 2 |D|2 ) 4 +ϵ h(D)−3/2 . (6)

2. Key ingredients
In order to obtain our first result, we need to study the following sum:
♭
∑
S( f, Q; X ) = λ f (n),
n=Q(x)≤X
x∈Z2 , gcd(n,N )=1

where symbol ♭ means that the sum runs over all square-free integers. The function rQ (n)
being always non-negative allows us to write the above partial sums defined in terms of known
arithmetical functions modulo certain weight, i.e.,
♭ ♭
∑ ∑
S( f, Q; X ) = λ f (n)rQ (n) = w D λ f (n)rQ
∗
(n),
n≤X n≤X
(7)
gcd(n,N )=1 gcd(n,N )=1

where rQ (n) denotes the number of representations of n by the quadratic form Q(x), w D is
the size of the group of units, and rQ (n) = w D rQ
∗
(n). We define the generating function θQ (τ )
for rQ (n) associated to binary quadratic form Q(x) as follows:
∑ ∞
∑
θQ (τ ) := q Q(x)
= rQ (n)q n , q = e2πiτ .
x∈Z2 n=0

The generating function θQ (τ ) ∈ M1 (Γ0 (|D|), χ D ) (see [6, Theorem 10.9]), where
( ) χ D is
the Dirichlet character modulo |D| which is given by Jacobi symbol χ D (d) := Dd . It is
well-known that rQ (n) ≪ϵ n ϵ .
We are considering the quadratic forms Q(x) of discriminant D < 0 such that the class
number h(D) is 1 and there is a well-known formula for rQ (n) given by [6, section 11.2]:
⎧
∑ ⎨6 if
⎪ D = −3,
rQ (n) = w D χ D (d), where w D = 4 if D = −4, (8)
D < −4.
⎪
d|n ⎩
2 if
5
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

The formula for rQ (n) depends only on the discriminant D but not on the choice of a, b and
c appearing in the definition of reduced from Q(x). For details, we refer to [17, Section 2].
In order to obtain an upper bound for the sum S( f, Q; X ), We consider the following
Dirichlet series:
∑
♭
λ f (n)rQ∗
(n)
L( f, Q; s) = , (9)
n≥1
ns
gcd(n,N )=1

∑
∗
where rQ (n) = χ D (d). The Dirichlet series L( f, Q; s) converges absolutely and uniformly
d|n
for ℜ(s) > 1. To obtain an upper bound for the sum S( f, Q; X ), we first decompose L( f, Q; s)
in terms of known Hecke L-functions. Before stating the decomposition of L( f, Q; s) in terms
of well-known function, we define the L-functions associated to a Hecke eigenform. The Hecke
∞
∑ k−1
L-function associated to a normalised Hecke eigenform f (τ ) = λ f (n)n 2 q n ∈ Sk (Γ0 (N ))
n=1
is given by:
∑ λ f (n) ∏ ( λ f ( p) −1 ∏
) (
λ f ( p) 1 −1
)
L(s, f ) = s
= 1− s
1− s
+ 2s , ℜ(s) > 1. (10)
n≥1
n p|N
p p p
p∤N

The completed Hecke L-function is defined as follows:

( √ )s
N
Λ(s, f ) := Γ (s + k − 1) L(s, f ), (11)
2π

which is analytically continued to the whole complex plane and satisfies a nice functional
equation (s → 1 − s). For a given Dirichlet character χ of modulus m, the twist of f is
∞
∑ k−1
defined as f ⊗ χ(τ ) := λ f (n)χ(n)n 2 q n ∈ Sk (Γ0 (M), χ 2 ), where M | N m 2 . The twisted
n=1
Hecke L- function associated to f ⊗ χ is given as follows:
∑ λ f (n)χ(n) ∏( λ f ( p)χ ( p) −1∏
) (
λ f ( p)χ ( p) 1 −1
)
L(s, f ⊗ χ ) = s
= 1− s
1− s
+ 2s . (12)
n≥1
n p|M
p p p
p∤M

This also converges absolutely, uniformly, and it is non-vanishing for ℜ(s) > 1. The completed
twisted Hecke L-function is defined as follows:
( √ )s
M
Λ(s, f ⊗ χ ) := Γ (s + k − 1)L(s, f ⊗ χ), (13)
2π

which also has analytic continuation to whole C-plane and satisfies a nice functional equation
(s → 1 − s). For details, see [6, Section 7.2]. The decomposition of L( f, Q; s) is given by the
following Lemma.

Lemma 2.1. For ℜ(s) > 1, we have

L( f, Q; s) = L(s, f )L(s, f ⊗ χ D )G(s), (14)

6
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

where L(s, f ) and L(s, f ⊗ χ D ) are Hecke L-function and twisted Hecke L-function, respec-
tively and the Euler product for G(s) is given by
∏( λ f ( p) ∏ λ f ( p)χ D ( p) ∏ λ f ( p)χ D ( p) 1 −1
) ( ) ( )
1− 1 − 1 − +
p|N
ps p|N
ps p∤N
ps p 2s
p||D|

∏( 2 − 2λ2f ( p) − λ2f ( p)χ D (P) λ f ( p)(1 + χ D ( p))(1 + λ2f ( p)χ D ( p))

× 1+ +
p 2s p 3s
p∤N

1 − 2λ2f ( p)(1 + χ D (P)) λ f ( p)(1 + χ D (P))

)
+ + .
p 4s p 5s
It converges absolutely and uniformly in the half plane ℜ(s) > 1
2
, and it is non-zero for
ℜ(s) = 1.

Proof. We know that λ f (n) and rQ ∗

(n) are multiplicative function. So, for ℜ(s) > 1, we have
( )
∑ λ f (n)rQ ∗
(n) µ 2
( p)λ f ( p)r ∗
( p)
♭ Q
∏
L( f, Q; s) = = 1+
n≥1
ns ps
p∤N
gcd(n,N )=1
∏( λ f ( p)(1 + χ D ( p))
)
= 1+ = L(s, f )L(s, f ⊗ χ D )G(s),
ps
p∤N

where the Euler product for G(s) is given in lemma. □

Lemma 2.2 ([8, Chapter 5]). Let ϵ > 0 be an arbitrarily small real number. The hybrid
convexity bound of Hecke L-function and twisted Hecke L-function is given by:
) 1 +ϵ
L(s, f ) ≪ϵ N k 2 (|s| + 3)2 4
(
) 1 +ϵ (15)
and L(s, f ⊗ χ D ) ≪ϵ N k 2 |D|2 (|s| + 3)2 4
(

on the line ℜ(s) = 1

2
+ ϵ. The second integral moment is given by
∫ 2T ⏐ ( )⏐2
⏐ L 1 + ϵ + it, f ⏐ dt ≪ϵ (N k 2 T 2 ) 21 +ϵ
⏐ ⏐
T
⏐ 2 ⏐
∫ 2T
⏐ ( )⏐2 (16)
⏐ L 1 + ϵ + it, f ⊗ χ D ⏐ dt ≪ϵ (N k 2 |D|2 T 2 ) 12 +ϵ
⏐ ⏐
and
T
⏐ 2 ⏐
uniformly for any |T | ≥ 1.

2.1. Mean value of a multiplicative function over the integers represented by a reduced form
of discriminant D with h(D) = 1

Let η and N be two fixed positive integers. We define the following sum given by
♭
ηω(n) ,
∑
Eη (X ) =
n=Q(x)≤X
x∈Z2 gcd(n,N )=1

7
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

where ω(n) denotes the number of distinct prime divisors of a positive integer n. The above
∗
sum can be expressed in terms of rQ (n) by
♭
ηω(n) , = µ2 (n)ηω(n)rQ
∑ ∑
∗
Eη (X ) = (n).
n=Q(x)≤X n≤X (17)
x∈Z2 gcd(n,N )=1 gcd(n,N )=1

In order to obtain an estimate for Eη (X ) we first decompose the associated Dirichlet series (say
D(s)) in terms of known L-functions and then apply the Perron’s formula and use the standard
argument. The associated Dirichlet series D(s) is given by
∑ µ2 (n)ηω(n)rQ
∗
(n)
D(s) = .
n≥1
ns
gcd(n,N )=1

The Dirichlet series D(s) converges for ℜ(s) > 1. Following the argument as in the proof of
Lemma 2.1, one has the decomposition of D(s) as follows.
µ2 ( p)ηω( p) rQ
( )
∗ ∏( η(1 + χ D ( p))
)
( p)
= ζ η (s)L η (s, χ D )P(s),
∏
D(s) = 1+
p s
= 1+
p s (18)
p∤N p∤N

where ζ (s) is the Riemann zeta function, and L(s, χ D ) is the Dirichlet L- function associated
to the quadratic character χ D and P(s) is given in terms of Euler product as follows:
∏ [( χ D ( p) η ∏ 1 η χ D ( p) η η(1 + χ D ( p))
)( )] [( ) ( ) ( )]
1
P(s) = 1− 1− 1− s 1− 1+ . (19)
p|N
ps p s p p s p s
p∤N

The last product converges absolutely and uniformly for ℜ(s) > 1/2 and non-zero at s = 1.
Using the Perron’s formula, an estimate for Eη (X ) is given in the following proposition.

Proposition 2.3. Let Eη (X ) be as in (17). Then there exists an absolute constant C = C(η)
such that
( ( 2eη+2 ))
P(1)L(1, χ D )η L
Eη (X ) = X (log X )η−1 1 + Oη √ N (20)
Γ (η) log X
2eη+2
uniformly for N ≥ 1 and X ≥ exp(C L N ) where L N = log(ω(N ) + 3).

Proof. Following the argument of [12, Lemma 4.1], we obtained our result. Here, we give
details of our proof. By Perron’s formula and (18), we have
∫ b+i T
1 Xs
µ2 (n)ηω(n)rQ ζ η (s)L η (s, χ D )P(s) ds + O(R)
∑
∗
Eη (X ) = (n) =
n≤X
2πi b−i T s
gcd(n,N )=1

(21)
1
where b = 1 + log X
, T ≥ 3 and
∞
∑ ηω(n)rQ∗
(n)
R=X b
. (22)
n=1
n (1 + T | log(X/n)|)

We break the above summation into the following two parts and obtain an estimate for R.
(i) n with | log(X/n)| ≤ T −1/2 (ii) n with | log(X/n)| > T −1/2 .
8
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

In the first case, we denote the sum by R1 , and an estimate for R1 , (the summation over n with
| log(X/n)| ≤ T −1/2 ) is obtained as follows.
ηω(n)rQ ηω(n)rQ
∗ ( )b ∗
∑ (n) ∑ X (n)
R1 = X b
≪
−1/2
n (1 + T | log(X/n)|) −1/2
n (1 + T | log(X/n)|)
|n−X |≤X T |n−X |≤X T
η−1
X (log X )
ηω(n)rQ L η (1, χ D ).
∑
∗
≪ (n) ≪
T 1/2
|n−X |≤X T −1/2

Which is obtained following argument as in [12, Lemma 4.1]. In the second case, we denote
the sum by R2 , and an estimate for R2 is obtained as follows.
∑ ηω(n)rQ∗
(n) X ∑ ηω(n)rQ
∗
(n)
R2 = X ≪ 1/4
n b (1 + T | log(X/n)|) T nb
|n−X |>X T −1/2 |n−X |>X T −1/2
X
≪ (log X )η−1 L η (1, χ D ).
T 1/2
Thus, combining both the sum, we have
∞
ηω(n) X
≪ 1/2 (log X )η−1 L η (1, χ D ).
∑
R = R1 + R2 =X b
(23)
n=1
n (1 + T | log(X/n)|)
0 T

To complete the proof, we need to evaluate the integral given in (21). Let c be a positive
constant and σc (T ) := 1 − logc T . Following the argument as in [12], let r = 1/(2 log X ) and
assume that 1 − r > σc (T ). The truncated Hankel contour Γ is a positively oriented contour
formed from the circle |s − 1| = r excluding the points s = 1 − r and joining the half segment
[σc (T ), 1 − r ] which is traced out twice with respective argument +π and −π . We apply the
Cauchy residue theorem to the closed path consisting of vertical line segment [b − i T, b + i T ]
and Pv± := [σc/2 (T ), σc/2 (T ) ± i T ], two horizontal line segment Ph± := [σc/2 (T ) ± i T, b ± i T ]
and the Hankel contour Γ .
For ℜ(s) ≥ σc (T ), we observe that
⏐ ⏐
)η ∏ ( )η ⏐ )2η
χ D ( p) ⏐⏐
⏐∏ ( ∏(
⏐ 1 1
P(s) ≪ ⏐⏐ 1+ s 1− ≪ 1 +
⏐ p|N p p|N
ps ⏐
⏐ p|N
p σc (T ) (24)
2eη
≪ exp{2η exp(2cL N / log T ) log L N } ≪ L N
using Merten’s third theorem and pn ≈ n log n ( pn denotes the nth prime), provided T ≥
exp(2cL N ). From the above estimate and Cauchy’s integral formula, we obtain that (for
ℜ(s) ≥ σc/2 (T ))
∫
1 P(z)
P ′ (s) = dz ≪ exp{2η exp(2cL N / log T ) log L N } log T
2πi |z−σc/2 (T )|= logc T (z − s)2 (25)
2eη
≪ L N log T.
From the bound ζ (s) ≪ log T for s ∈ Pv± ∪ Ph± ∪ Γ , and (24), it is easy to have
Xs
∫ ( )
1 η η X σc/2 (T )
(log T )η−1 .
2eη
ζ (s)L (s, χ D )P(s) ds ≪ L N +X (26)
2πi Pv± ∪Ph± s T
9
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

Using the property of ζ (s) and bound for P ′ (s) (given in (25)), we have (for s ∈ Γ )
((s − 1)ζ (s))η η
L (s, χ D )P(s) = L η (1, χ D )P(1) + Oη (L η (1, χ D )L N log T |s − 1|) (27)
2eη
s
provided T ≥ exp(2cL N ). This allows us to get
Xs
∫
ζ η (s)L η (s, χ D )P(s) ds
Γ s
∫ ( ∫ )
= L η (1, χ D )P(1) (s − 1)−η X s ds + Oη L η (1, χ D )L N log T
2eη
|s − 1|1−η X s ds .
Γ Γ
(28)
Using [16, Corollary II.0.18], we have
∫ { }
c log X
(s − 1)−η X s ds = X (log X )η−1 1 + Oη (e 2 log T )
−

Γ
∫ ∫ 1−r
and 1−η s
|s − 1| X ds ≪ (1 − y)1−η X y dy + X 1+r r 2−η ≪ X (log X )η−2 .
Γ σc/2 (T )

Thus, we have
Xs
∫
ζ η (s)L η (s, χ D )P(s) ds
Γ s
L η (1, χ D )P(1) { ( c log X )} ( ) (29)
X (log X )η−1 1 + Oη e− 2 log T + Oη L η (1, χ D )L N X log T (log X )η−2 .
2eη
=
Γ (η)
Applying the Cauchy residue theorem and combining all the estimates, we have
∫ b+i T
1 Xs
ζ η (s)L η (s, χ D )P(s) ds
2πi b−i T s
L η (1, χ D )P(1) η−1
{ c log X } ( )
1 + Oη (e− 2 log T ) + Oη L η (1, χ D )L N log T X (log X )η−2
2eη
= X (log X )
Γ (η)
( )
X
+ X σc/2 (T ) (log T )η−1 ).
2eη
+ Oη (L N
T
Thus, we have
L η (1, χ D )P(1)
µ2 (n)ηω(n)rQ X (log X )η−1 + Err
∑
∗
Eη (X ) = (n) =
Γ (η) (30)
n≤X
gcd(n,N )=1

where
L η (1, χ D )P(1)
( )
c log X ( )
X (log X )η−1 (e− 2 log T ) + Oη L η (1, χ D )L N log T X (log X )η−2
2eη
Err = Oη
Γ (η)
X (log X )η−1 η
( ( ) ) ( )
X
+ X σc/2 (T ) (log T )η−1 + Oη
2eη
+ Oη L N 1/2
L (1, χ D ) .
T T
Clearly, it is easy to see that
P(1) ≫ L −2
N
√
using Merten’s theorem. We choose T = exp(c1 log X ) so that T ≥ exp(2cL N ) is satisfied.
Hence, for each X ≥ exp(2cL 2N ), The error term satisfies following bound.
L η (1, χ D )P(1)
2eη+2
L
Err ≪ X (log X )η−1 √ N .
Γ (η) log X
Thus, we have the required result. This completes the proof. □
10
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

Let Y be a positive real number. From Deligne’s estimate, it is well-known that for each
prime p, λ f ( p) = 2 cos θ p , and more generally

sin((m + 1)θ p )
λ f ( pm ) = ,
sin θ p

for any positive integer m. We define a multiplicative function h Y supported on square-free

positive integers. The function h Y is given by:
⎧ ( )
α
⎨ log Y , p ≤ Y and p ∤ N ,
log p
⎪
⎪ if
h Y ( p) = −2,
⎪ if p > Y and p ∤ N , (31)
p | N,
⎪
⎩0, if

π
where α : [0, 1] → [−2, 2] is such that, α(0) = α0 = 2 and α(t) = 2 cos( m+1 ) if
1
m+1
< t ≤ 1
m
for m ∈ N. We make use of the function h Y to obtain a lower bound for
the sum S( f, Q; Y u ) for some u ≥ 1, following the idea of [11].

Proposition 2.4. Let U ≥ 1 be a real number, h Y (n) (defined in (31)) and α(t) be as above
with α0 > 0. Let N ≤ X U be a fixed positive integer. Then, we have

P(1)L(1, χ D )α0
(log Y u )α0 −1 Y u
∑
∗
h Y (n)rQ (n) = (σ (u) + oα,U (1))
n≤Y u
Γ (α0 )
gcd(n,N )=1

, U where P(s) is given in (19) and

[1 ]
uniformly for u ∈ U

∞
α0 −1
∑ (−1) j
σ (u) = (u) + I j (u)
j=1
j

with
∫ j
α0 −1
∏ dt1 dt2 · · · dt j
I j (u) = (u − t1 − t2 − · · · − t j ) (α0 − α(t))
∆j i=1
t1 t2 · · · t j

and

∆ j = {(t1 , t2 , . . . , t j ) ∈ [0, ∞) | |t1 + t2 + · · · + t j | ≤ u}.

Proof of Proposition 2.4. We follow the argument of [13, Lemma 6] to give a brief proof of
the Proposition 2.4. From an inclusion–exclusion type identity, we have

α0ω(n) rQ
∑ ∑
∗ ∗
h Y (n)rQ (n) = (n)
n≤Y u n≤Y u
gcd(n,N )=1 gcd(n,N )=1

∞ j
(−1) j ∑
α0ω(n) rQ
∑ ∏ ∑
∗ ∗
+ (α0 − h Y ( pi ))rQ ( pi ) (n).
j=1
j p p ··· p ≤Y u i=1 u
1 2 j n≤ p pY ··· p
gcd( p1 p2 ··· p j ,N )=1 1 2 j
gcd(n,N )=1

11
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

On applying Proposition 2.3, we have

∑
∗
h Y (n)rQ (n)
n≤Y u
gcd(n,N )=1

P(1)L(1, χ D )η u
[
= Y (log Y u )α0 −1
Γ (η)
u
∞
(−1) j
j (log p1 pY2 ··· p j )α0 −1 ]
+ oU,α ((log Y u )α0 −1 ) .
∑ ∑ ∏
∗
+ (α0 − h Y ( pi ))rQ ( pi )
j=1
j p1 p2 ··· p j ≤Y u i=1
p1 p2 · · · p j
gcd( p1 p2 ··· p j ,N )=1

From the Chebatorev density theorem, as the class number h(D) = 1, the density of primes
∗
represented by Q of discriminant D is 1/2, and for such primes, we have rQ ( p) = 2 and 0
otherwise. See [19, Eq. 1.1 & 1.2]. So, we have
∑
∗
h Y (n)rQ (n)
n≤X
gcd(n,N )=1

P(1)L(1, χ D )η u
[
= Y (log Y u )α0 −1
Γ (η)
Yu
∞
(−1) j
j (log p1 p2 ··· p j
)α0 −1 ]
+ oU,α ((log Y u )α0 −1 )
∑ ∑ ∏
+ (α0 − h Y ( pi )) .
j=1
j p1 p2 ··· p j ≤Y u i=1
p1 p2 · · · p j
gcd( p1 p2 ··· p j ,N )=1

Thus, we have exactly the same expression as in the proof of [13, Lemma 6]. So, by removing
the coprimality condition from the expression and then using the prime number theorem, we
have the required result.

Remark 2.1. Let α be a step function given by; α : [0, ∞) → R with α(t) = αk when
t ∈ [xk , xk+1 ] for each k = 0, 1, 2, . . . , K . Here K > 0 and [xk , xk+1 ] is partition of [0, ∞) (in
π
our case x0 = 0, xk = k1 , α0 = 2 and αk = 2 cos( k+1 )). In [13, Lemma 6], it was mentioned
that the function, σ (u) is the unique solution of the integral equation
∫ u
uσ (u) = σ (t)α(u − t)dt
0

with the initial condition σ (u) = u α0 −1 for u ∈ (0, x1 ]. For the step function α, it is shown that
σ (4/3) > 0 in [13, Lemma 8]. Hence, for u = u 0 = 4/3, we have
P(1)L(1, χ D )α0
(log Y )α0 −1 Y u 0 > 0.
∑
∗
h Y (n)rQ (n) = (σ (u 0 ) + oα,U (1))
u0 Γ (α 0 ) (32)
n≤Y
gcd(n,N )=1

3. Proof of results

Proof of Theorem 1.1. Let 1 ≤ Z ≤ X2 . In order to obtain an upper bound for the sum
S( f, Q; X ) given in , we introduce a smooth compactly supported function w(x) satisfying:
w(x) = 1 for x ∈ [2Z , X ], w(x) = 0 for x < Z and x > X + Z , and w(r ) (x) ≪r Z −r for all
r ≥ 0. In general, for any arithmetical function f (n), we have
∞
( ) ( )
∑ ∑ ∑ ∑
f (n) = f (n)w(n) + O | f (n)| + O | f (n)| . (33)
n≤X n=1 n<2Z X <n<X +Z
12
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

Moreover, by Mellin’s inverse transform, we have

∞ ∫ ( )
∑ 1 ∑ f (n)
f (n)w(n) = w̃(s) ds, (34)
n=1
2πi (b) n≥1
ns

∑ f (n)
where b is a positive real number larger than the abscissa of absolute convergence of
n≥1
ns
and the Mellin’s transform w̃(s) is given by following integral:
∫ ∞
dx
w̃(s) = w(x)x s .
0 x
We observe that due to the integration by parts,

X m
∫ ∞ ( )
1 Z
w̃(s) = w(m) (x)x s+m−1 d x ≪ 1−σ , (35)
s(s + 1) · · · (s + m − 1) 0 X |s|Z
for any m ≥ 0, where σ = ℜ(s). For details, we refer to [9, Section 3].
From Eq. (33) with f (n) = λ f (n)rQ∗
(n), we have

♭ ♭
∑ ∑
λ f (n)rQ
∗
(n) = λ f (n)rQ
∗
(n)w(n)
n≤X n≥1
gcd(n,N )=1 gcd(n,N )=1
⎛ ⎞ ⎛ ⎞
♭ ♭
∑ ∑
+O⎝ ∗
|λ f (n)rQ (n)|⎠ + O ⎝ ∗
|λ f (n)rQ (n)|⎠ .
⎜ ⎟ ⎜ ⎟
n<2Z X <n<X +Z
gcd(n,N )=1 gcd(n,N )=1

Moreover, by Mellin’s inverse transform, we have

∫
∑
♭ 1
λ f (n)rQ
∗
(n)w(n) = w̃(s)L( f, Q; s)ds, (36)
n≥1
2πi (b)
gcd(n,N )=1

where b = 1 + ϵ for some arbitrarily small ϵ > 0. Since λ f (n)rQ

∗
(n) ≪ n ϵ , for any ϵ > 0,
hence,
⎛ ⎞ ⎛ ⎞
⎜ ∑♭ ⎜ ∑♭
O⎝ ∗
|λ f (n)rQ (n)|⎠ + O ⎝ ∗
|λ f (n)rQ (n)|⎠ ≪ Z 1+ϵ . (37)
⎟ ⎟
n<2Z X <n<X +Z
gcd(n,N )=1 gcd(n,N )=1

We know that L( f, Q; s) is holomorphic for all s ∈ C. We shift the line of integration from
ℜ(s) = 1 + ϵ to ℜ(s) = 21 + ϵ and apply Cauchy’s residue theorem to get
∫ ∫
∑
♭ 1 1
λ f (n)rQ
∗
(n)w(n) = w̃(s)L( f, Q; s)ds = w̃(s)L( f, Q; s)ds. (38)
n≥1
2πi (1+ϵ) 2πi (1/2+ϵ)
gcd(n,N )=1

( )m
Since w̃(s) ≪ Z
X 1−σ
X
|s|Z
for any m ≥ 0, so the contribution for the integral over |s| ≥ T =
X 1+ϵ
Z
on the right hand side of (38) is negligibly small, i.e., O(X −A ) for any large A > 0 if
13
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

one chooses sufficiently large m > 0. Hence, we have

∫ ∫ 1/2+ϵ+i T
1 1
w̃(s)L( f, Q; s)ds = w̃(s)L( f, Q; s)ds + O(X −A )
2πi (1/2+ϵ) 2πi 1/2+ϵ−i T
∫ T
≪ |w̃(1/2 + ϵ + it)||L( f, Q; 1/2 + ϵ + it)|dt + O(X −A )
−T
T 1
X 2 +ϵ
∫
≪ |L( f, Q; 1/2 + ϵ + it)|dt + O(X −A )
0 + ϵ + it|
| 12
(∫ 1 ∫ T ) 1
X 2 +ϵ
≪ + |L( f, Q; 1/2 + ϵ + it)|dt + O(X −A ),
0 1 | 2 + ϵ + it|
1

where the estimate in the last lines is obtained by substituting the bound for w̃(s) when m = 1.
Now, we substitute the decomposition L( f, Q; s) = L(s, f )L(s, f ⊗ χ D )G(s) from Lemma 2.1
and by absolute convergence of G(s) for ℜ(s) > 12 , we get
∫
1
w̃(s)L( f, Q; s)ds
2πi (1/2+ϵ)
(∫ 1 ∫ T ) 1
X 2 +ϵ
≪ + |L(1/2 + ϵ + it, f )L(1/2 + ϵ + it, f ⊗ χ D )|dt + O(X −A )
0 1 | 12 + ϵ + it|
|L(1/2 + ϵ + it, f )L(1/2 + ϵ + it, f ⊗ χ D )|
∫ T
1 1
≪ (N k 2 |D|X ) 2 +ϵ + X 2 +ϵ dt + O(X −A ).
1 t
We apply the dyadic division method and then the Cauchy–Schwarz inequality to get
∫
1
w̃(s)L( f, Q; s)ds
2πi (1/2+ϵ)
⎧
1 +ϵ
2 −A
⎨(N k |D|X ) 2 {+ O(X )
⎪
⎪
) 21 ( ) 12 }
≪
(
1 +ϵ ∫ T1 ∫ T1
⎪ + X 2 max 1
T1 |L(1/2 + ϵ + it, f )| 2
dt T1 L(1/2 + ϵ + it, f ⊗ χ D )| 2
dt
1≤T1 ≤T T1
⎪
⎩ 2 2
( )
1 1 1 1 1
≪ (N k 2 |D|X ) 2 +ϵ + X 2 +ϵ max (N k 2 T12 ) 4 +ϵ (N k 2 |D|2 T12 ) 4 +ϵ + O(X −A )
1≤T1 ≤T T1
1
≪ (N k 2 |D|X ) 2 +ϵ .
1+ϵ 1
The last inequality is obtained by substituting T = X Z and choosing Z = X 2 +ϵ . Thus
∫
1 1
w̃(s)L( f, Q; s)ds ≪ (N k 2 |D|X ) 2 +ϵ . (39)
2πi (1/2+ϵ)
1
From Eqs. (36), (37), (38) and (39) and choosing Z = X 2 +ϵ , we have the required result.
Concerning about the second result, one needs to obtain a lower estimate for the sum
S( f, Q; Y u 0 ) under the assumption of positivity of λ f (n) for n ≤ Y . Below, we sketch the
proof to get a lower estimate for the sum S( f, Q; Y u 0 ) for some u 0 > 1 .
A lower bound for the sum S( f, Q; Y u 0 ): We recall that the multiplicative function h Y ,
supported on square-free positive integers, is given by:
⎧ ( )
⎨ log Y ,
α p ≤ Y and p ∤ N ,
log p
⎪
⎪ if
h Y ( p) = −2,
⎪ if p > Y and p ∤ N ,
p | N,
⎪
⎩0, if
π
where α : [0, 1] → [−2, 2] is such that, α(0) = 2 and α(t) = 2 cos( m+1 1
) if m+1 < t ≤ m1
for m ∈ N.
Assume that λ f (n) ≥ 0 for every positive integer n ≤ Y . Write λ f ( p) = 2 cos(θ p ) with
θ p ∈ [0, π]. Then if m ≥ 1 is an integer, we have for 1 ≤ j ≤ m and Y 1/(m+1) ≤ p < Y 1/m
14
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

that
sin(( j + 1)θ p )
0 ≤ λ f (pj) =
sin(θ p )
π
( π )
if p ∤ N . This implies that θ p ≤ m+1 , and so we have λ f ( p) ≥ 2 cos m+1 (= h Y ( p)) for
p ∤ N and Y 1/(m+1) ≤ p < Y 1/m . Multiplicativity of these arithmetical functions shows that
λ f (n) ≥ h Y (n), for all square-free positive integer n ≤ Y . For details, see [11, Section 2]. As
rQ∗
(n) ≥ 0, we get that λ f (n)rQ ∗ ∗
(n) ≥ h Y (n)rQ (n) for all square-free positive integer n ≤ Y .
u0
Similar to [11], it is observed that for X = Y with u 0 = 4/3,
♭ ♭
∑ ∑
S( f, Q; X ) = λ f (n)rQ∗
(n) ≥ ∗
h Y (n)rQ (n) > 0.
n≤X n≤X
(40)
gcd(n,N )=1 gcd(n,N )=1

Precisely, to see (40), let g Q be a multiplicative function defined by convolution identity

λ∗f (n) = (g Q ∗ h ∗Y )(n), where λ∗f (n) = λ f (n)rQ ∗ ∗
(n) and h ∗y (n) = h Y (n)rQ (n). This gives that
g Q ( p) = λ f ( p)rQ ( p) − h Y ( p)rQ ( p). If p is not represented by Q(x), then g Q ( p) = 0. If p
∗ ∗

is represented by Q(x), then g Q ( p) ≥ 0 for all p not dividing N and up to Y by definition

of h Y ( p). For p > Y , with p ∤ N , we have h Y ( p)rQ ∗
( p) = −4 and −2 ≤ λ f ( p) ≤ 2. Hence,
g Q ( p) ≥ 4a + 4 ≥ 0 as −1 ≤ a ≤ 1. Here, we have written the value taken by λ f ( p)rQ ∗
( p) as
4a. Thus, we have g Q (n) ≥ 0 for all square-free positive integers n. Now, we see that
♭ ♭
∑ ∑ ∑
S( f, Q; X ) = λ f (n)rQ∗
(n) = ∗
g Q (d)h Y (n/d)rQ (n/d)
n≤X n≤X d|n
gcd(n,N )=1 gcd(n,N )=1
♭ ♭
∑ ∑ ∑
∗ ∗
= g Q (d) h Y (m)rQ (m) ≥ h Y (n)rQ (n)
d≤X m≤X/d n≤X
gcd(d,N )=1 gcd(m,N )=1 gcd(n,N )=1

as g Q (1) = 1 and g Q (n) ≥ 0 for each square-free positive integer ∑

n. Hence, a positive lower
∗
bound for the sum S( f, Q; X ) occurs and it is given by the sum h Y (n)rQ (n) for some
n≤X u 0
gcd(n,N )=1
u 0 < 1 given in Proposition 2.4.

Proof of Theorem 1.2. Let X = Y u with u > 1. By definition, n f,Q be the largest integer
among all n ∈ N such that λ f (n) ≥ 0, with gcd(n, N ) = 1 and n = Q(x) for some x ∈ Z2 .
Let us write n f,Q = Y . In other word, λ f (n) ≥ 0, for each n ≤ Y with gcd(n, N ) = 1 and
n = Q(x) for some x ∈ Z2 . A key idea to estimate Y is to compare lower and upper bound
for the sum
♭ ♭
∑ ∑
S( f, Q; Y u ) = λ f (n) = λ f (n)r Q (n).
n=Q(x)≤Y u n≤Y u
x∈Z2 , gcd(n,N )=1 gcd(n,N )=1

From Theorem 1.1 and (32), we have (for u ≤ u 0 )

P(1)L(1, χ D )2 u ∑
♭
1 +ϵ u
0< Y log Y u ≪ ∗
h Y (n)rQ (n) ≤ S( f, Q; Y u ) ≪ (N k 2 |D|) 2 Y 2 +ϵ .
Γ (α0 ) n≤Y u
gcd(n,N )=1

This shows that

1 +ϵ 1 +ϵ 2
Y u/2+ϵ ≪ (N k 2 |D|) 2 L(1, χ D )2 H⇒ Y ≪ (N k 2 ) u |D| u +ϵ
15
M.K. Pandey and L. Vaishya Indagationes Mathematicae xxx (xxxx) xxx

Now, we substitute u = u 0 = 4/3 (as u 0 = 4/3 such that σ (u 0 ) > 0) to get

3 +ϵ
Y ≪ (N k 2 |D|2 ) 4 .
This completes the proof.

Acknowledgements
We would like to thank Prof. B. Ramakrishnan for encouraging us to work on this problem
and for his guidance and suggestions while preparing the manuscript. We also thank R. Acharya
for his helpful suggestions. The second author thanks IMSc, Chennai where this work has been
initiated. The second author would like to thank Stat-Math Unit, ISI Delhi for its hospitality,
and DST-INSPIRE, DST India for Inspire Faculty Fellowship award. We thank the anonymous
referee for some wonderful suggestions to improve the manifold of the article.

References
[1] R. Acharya, A twist of the Gauss circle problem by holomorphic cusp forms, Res. Number Theory 8 (2022)
5.
[2] V. Blomer, Sum of Hecke eigenvalues over values of quadratic polynomials, Int. Math. Res. Not. IMRN
(2008).
[3] D.A. Buell, Binary Quadratic Forms: Classical Theory and Modern Computations, Springer - Verlag, New
York, 1989.
[4] D.A. Cox, Primes of the form x 2 ny 2 : Fermat, Class Field Theory, and Complex Multiplication, in: Pure and
Applied Mathematics: A Wiley Series of Texts, Monographs and Tracts, vol. 119, John Wiley & Sons, 2014.
[5] P. Deligne, La conjecture de Weil I. Inst. Hautes Etudes, Sci. Publ. Math. 43 (1974) 273–307.
[6] H. Iwaniec, Topics in Classical Automorphic Forms, in: Grad. Stud. Math., vol. 17, American Mathematical
Society, Providence, RI, 1997.
[7] H. Iwaniec, W. Kohnen, J. Sengupta, The first negative Hecke eigenvalue, Int. J. Number Theory 3 (2007)
355–363.
[8] H. Iwaniec, E. Kowalski, Analytic Number Theory, in: Amer. Math. Soc. Colloq. Publ., vol. 53, American
Mathematical Society, Providence, RI, 2004.
[9] Y. Jiang, Guangshi Lu, Sum of coefficients of L- functions and applications, J. Number Theory 171 (2017)
56–70.
[10] W. Kohnen, J. Sengupta, On the first sign change of Hecke eigenvalues of newforms, Math. Z. 254 (1) (2006)
173–184.
[11] E. Kowalski, Y.-K. Lau, K. Soundararajan, J. Wu, On modular signs, Math. Proc. Cambridge Philos. Soc.
149 (3) (2010) 389–411.
[12] Y.-K. Lau, J.-Y. Liu, J. Wu, The first negative coefficient of symmetric square L-functions, Ramanujan J. 27
(2012) 419–441.
[13] Matomäki, On signs of Fourier coefficients of cusp forms, Math. Proc. Cambridge Philos. Soc. 152 (2) (2012)
207–222.
[14] M.K. Pandey, P. Tiwari, On the first sign change of Fourier coefficients of cusp forms at sum of two squares,
Res. Number Theory 7 (2021) 63.
[15] M.K. Pandey, Lalit Vaishya, Asymptotic of Fourier coefficients of hecke eigenforms at integers represented
by binary quadratic form of fixed discriminant, Publ. Math. Debrecen 105 (1–2) (2014) 11–38.
[16] G. Tenenbaum, Introduction To Analytic and Probabilistic Number Theory, in: Cambridge Stud. Adv. Math.,
vol. 163, Cambridge University Press, Cambridge, 2015.
[17] Lalit Vaishya, Signs of Fourier coefficients of cusp forms at integers represented by an integral binary quadratic
form, Proc. Indian Acad. Sci.( Math. Sci.) 131 (41) (2021).
[18] Lalit Vaishya, Average estimates and sign change of Fourier coefficients of hecke eigenforms at integers
represented by binary quadratic forms of fixed discriminant, Proc. Indian Acad. Sci.( Math. Sci.) 134 (17)
(2024).
[19] Asif Zaman, Primes represented by positive definite binary quadratic forms, Q. J. Math. 69 (4) (2018)
1353–1386.

16