Skip to main content

On the mean value of the two-term Dedekind sums

Abstract

The main purpose of this paper is, using the properties of Gauss sums, the estimate for character sums and the analytic method, to study the mean value of the two-term Dedekind sums and give an interesting asymptotic formula for it.

MSC:11F20, 11L40.

1 Introduction

Let q be a natural number and h be an integer prime to q. The classical Dedekind sums

S(h,q)= a = 1 q ( ( a q ) ) ( ( a h q ) ) ,

where

( ( x ) ) = { x [ x ] 1 2 , if  x  is not an integer; 0 , if  x  is an integer ,

describes the behaviour of the logarithm of the eta-function (see [1, 2]) under modular transformations. Many authors have studied the arithmetical properties of S(h,q) and obtained many interesting results; see, for example, [35] and [6]. Now, for any odd prime p and integer m with (m,p)=1, we consider the mean value of the two-term Dedekind sums

a = 1 p 1 b = 1 p 1 S ( ( a 4 + m a 2 ) ( b 4 + m b 2 ) ¯ , p ) ,
(1)

where n ¯ denotes the multiplicative inverse of nmodp, that is, n ¯ n1modp. For convenience, we provide n ¯ =0 if (n,p)>1.

Here, we are concerned whether there exists an asymptotic formula for mean value (1). Regarding this problem, it seems that no one has studied it, at least we have not seen any related result before. The problem is interesting because it can reflect some properties of the value distribution of the polynomial Dedekind sums.

The main purpose of this paper is, using the properties of Gauss sums, the estimate for character sums and the analytic method, to study the mean value properties of the two-term Dedekind sums and give an interesting asymptotic formula for (1). That is, we shall prove the following theorem.

Theorem Let p>3 be an odd prime. Then, for any fixed integer m with (m,p)=1, we have the asymptotic formula

a = 1 p 1 b = 1 p 1 S ( ( a 4 + m a 2 ) ( b 4 + m b 2 ) ¯ , p ) = 1 6 p 2 +O ( p 3 2 + ϵ ) ,

where ϵ denotes any fixed positive number.

Remark In this theorem, we only consider the special polynomial x 4 +m x 2 . However, we have not found an effective method to study the general polynomial f(x). This problem should be a further study.

For general integer q>3, whether there exists an asymptotic formula for

a = 1 q ( a , q ) = 1 b = 1 q ( b , q ) = 1 S ( ( a 4 + m a 2 ) ( b 4 + m b 2 ) ¯ , q )

is an open problem, where we provide ( b 4 + m b 2 ) ¯ =0 if ( b 4 +m b 2 ,q)>1.

2 Several lemmas

In this section, we shall give several lemmas, which are necessary in the proof of our theorem. Hereinafter, we shall use many properties of character sums and Gauss sums, all of these can be found in references [7] and [8]. First we have the following.

Lemma 1 Let p be an odd prime, χ be any non-principal character modp with χ 4 χ 0 , the principal character modp. Then, for any integer m with (m,p)=1, we have the identity

| a = 1 p 1 χ ( a 4 + m a 2 ) | 2 =2p+ ( 1 ) p 1 2 ( m p ) d = 1 p 1 χ(d) a = 1 p 1 ( ( a 4 d ) ( a 2 d ) p ) ,

where ( p ) is the Legendre symbol.

Proof Since χ is a non-principal character modp with χ 4 χ 0 , it must be a primitive character modp, then from the properties of Gauss sums, we have

a = 1 p 1 χ ( a 4 + m a 2 ) = 1 τ ( χ ¯ ) a = 1 p 1 b = 1 p 1 χ ¯ ( b ) e ( b ( a 4 + m a 2 ) p ) = 1 τ ( χ ¯ ) a = 1 p 1 b = 1 p 1 χ ¯ ( b a ¯ 2 ) e ( b a ¯ 2 ( a 4 + m a 2 ) p ) = 1 τ ( χ ¯ ) b = 1 p 1 χ ¯ ( b ) e ( m b p ) a = 1 p 1 χ ( a 2 ) e ( b a 2 p ) ,
(2)

where τ(χ)= a = 1 p 1 χ(a)e( a p ) denotes the classical Gauss sums, and e(y)= e 2 π i y .

Note that the identities |τ(χ)|= p and

a = 1 p 1 e ( n a 2 p ) = { p 1 , if  ( n , p ) = p ; ( n p ) G ( p ) 1 , if  ( n , p ) = 1 ,

where

G(p)= a = 0 p 1 e ( a 2 p ) = a = 1 p 1 ( a p ) e ( a p )

and

G 2 (p)= ( 1 ) p 1 2 p.

From (2) and identities a = 1 p 1 χ 2 (a)= a = 1 p 1 χ 4 (a)=0, we have

| a = 1 p 1 χ ( a 4 + m a 2 ) | 2 = 1 p | b = 1 p 1 χ ¯ ( b ) e ( m b p ) a = 1 p 1 χ ( a 2 ) e ( b a 2 p ) | 2 = 1 p b = 1 p 1 d = 1 p 1 χ ¯ ( b d ¯ ) e ( m b m d p ) a = 1 p 1 c = 1 p 1 χ ( a 2 c ¯ 2 ) e ( b a 2 c 2 d p ) = 1 p b = 1 p 1 d = 1 p 1 χ ( d ) e ( m b ( 1 d ) p ) a = 1 p 1 χ ( a 2 ) c = 1 p 1 e ( b c 2 ( a 2 d ) p ) = 1 p b = 1 p 1 d = 1 p 1 χ ( d ) e ( m b ( 1 d ) p ) a = 1 p 1 ( a 2 d , p ) = 1 χ ( a 2 ) ( ( b ( a 2 d ) p ) G ( p ) 1 ) + 1 p b = 1 p 1 d = 1 p 1 χ ( d ) e ( m b ( 1 d ) p ) a = 1 p 1 a 2 d mod p χ ( a 2 ) ( p 1 ) = G ( p ) p d = 1 p 1 χ ( d ) b = 1 p 1 ( b p ) e ( m b ( 1 d ) p ) a = 1 p 1 χ ( a 2 ) ( a 2 d p ) + d = 1 p 1 χ ( d ) b = 1 p 1 e ( m b ( 1 d ) p ) a = 1 p 1 a 2 d mod p χ ( a 2 ) 1 p d = 1 p 1 χ ( d ) b = 1 p 1 e ( m b ( 1 d ) p ) a = 1 p 1 χ ( a 2 ) = ( 1 ) p 1 2 ( m p ) d = 1 p 1 χ ( d ) a = 1 p 1 χ ( a 2 ) ( ( 1 d ) ( a 2 d ) p ) + 2 ( p 1 ) d = 2 p 1 χ ( d ) a = 1 p 1 a 2 d mod p χ ( a 2 ) = 2 p + ( 1 ) p 1 2 ( m p ) d = 1 p 1 χ ( d ) a = 1 p 1 χ ( a 2 ) ( ( 1 d ) ( a 2 d ) p ) a = 1 p 1 χ ( a 4 ) = 2 p + ( 1 ) p 1 2 ( m p ) d = 1 p 1 χ ( d ) a = 1 p 1 χ ( a 2 ) ( ( 1 d ) ( a 2 d ) p ) = 2 p + ( 1 ) p 1 2 ( m p ) d = 1 p 1 χ ( d ) a = 1 p 1 ( ( 1 d a ¯ 2 ) ( a 2 d a ¯ 2 ) p ) = 2 p + ( 1 ) p 1 2 ( m p ) d = 1 p 1 χ ( d ) a = 1 p 1 ( ( a 4 d ) ( a 2 d ) p ) .

This proves Lemma 1. □

Lemma 2 For any prime p>3, we have the estimate

a = 1 p 1 | χ mod p χ ( 1 ) = 1 χ(a) | L ( 1 , χ ) | 2 |=O ( p 1 + ϵ ) ,

where χ mod p χ ( 1 ) = 1 denotes the summation over all characters χmodp with χ(1)=1, ϵ denotes any fixed positive number.

Proof See Lemma 5 of [9]. □

Lemma 3 Let q>2 be an integer, then for any integer a with (a,q)=1, we have the identity

S(a,q)= 1 π 2 q d | q d 2 ϕ ( d ) χ mod d χ ( 1 ) = 1 χ(a) | L ( 1 , χ ) | 2 ,

where L(1,χ) denotes the Dirichlet L-function corresponding to a character χmodd.

Proof See Lemma 2 of [6]. □

Lemma 4 Let p>3 be a prime. Then, for any integer n with (n,p)=1, we have the identity

a = 1 p ( a 2 + n p ) =1,

where ( p ) denotes the Legendre symbol.

Proof Since ( p ) χ 2 0 is a primitive character modp, so from the properties of Gauss sums τ( χ 2 0 )= a = 1 p 1 χ 2 0 (a)e( a p ), we know that

a = 1 p ( a 2 + n p ) = 1 τ ( χ 2 0 ) a = 1 p b = 1 p 1 ( b p ) e ( b ( a 2 + n ) p ) = 1 τ ( χ 2 0 ) b = 1 p 1 ( b p ) e ( n b p ) a = 1 p e ( b a 2 p ) .
(3)

Note that for any integer u with (u,p)=1, we have

a = 1 p e ( u a 2 p ) = ( u p ) a = 1 p e ( a 2 p ) = ( u p ) a = 1 p 1 ( a p ) e ( a p ) .
(4)

Then, from (3) and (4), we have

a = 1 p ( a 2 + n p ) = 1 τ ( χ 2 0 ) b = 1 p 1 ( b p ) e ( n b p ) ( b p ) τ ( χ 2 0 ) = τ ( χ 2 0 ) τ ( χ 2 0 ) b = 1 p 1 e ( n b p ) = τ ( χ 2 0 ) τ ( χ 2 0 ) = 1 ,

where we have used the fact that τ( χ 2 0 )= a = 1 p e( a 2 p ). This proves Lemma 4. □

3 Proof of the theorem

In this section, we shall complete the proof of our theorem. First, from the definition of S(a,p), we have the computational formula

S(1,p)= a = 1 p 1 ( a p 1 2 ) 2 = ( p 1 ) ( p 2 ) 12 p .
(5)

Then, from (5) and Lemma 3, we have

χ mod p χ ( 1 ) = 1 | L ( 1 , χ ) | 2 = π 2 12 ( p 1 ) 2 ( p 2 ) p 2 .
(6)

On the other hand, from Lemma 3 we also have

S(a,p)= 1 π 2 p p 1 χ mod p χ ( 1 ) = 1 χ(a) | L ( 1 , χ ) | 2 .
(7)

If there exists a character χ 1 modp such that χ 1 4 = χ 0 , χ 1 2 χ 0 and χ 1 (1)=1, then from Weil’s famous work (see [10]) we have the estimate

| a = 1 p 1 χ 1 ( a 4 + m a 2 ) | 2 | L ( 1 , χ 1 ) | 2 = | a = 1 p 1 χ 1 ( 1 + m a ¯ 2 ) | 2 | L ( 1 , χ 1 ) | 2 = | a = 1 p 1 χ 1 ( a 2 + m ¯ ) | 2 | L ( 1 , χ 1 ) | 2 p ln 2 p .
(8)

Then, from (7), (8) and Lemma 1, we have

a = 1 p 1 b = 1 p 1 S ( ( a 4 + m a 2 ) ( b 4 + m b 2 ) ¯ , p ) = 1 π 2 p p 1 χ mod p χ ( 1 ) = 1 a = 1 p 1 χ ( a 4 + m a 2 ) b = 1 p 1 χ ¯ ( b 4 + m b 2 ) | L ( 1 , χ ) | 2 = 1 π 2 p p 1 χ mod p χ ( 1 ) = 1 | a = 1 p 1 χ ( a 4 + m a 2 ) | 2 | L ( 1 , χ ) | 2 = ( 1 ) p 1 2 π 2 p p 1 ( m p ) d = 1 p 1 a = 1 p 1 ( ( a 4 d ) ( a 2 d ) p ) χ mod p χ ( 1 ) = 1 χ ( d ) | L ( 1 , χ ) | 2 + 1 π 2 2 p 2 p 1 χ mod p χ ( 1 ) = 1 | L ( 1 , χ ) | 2 + O ( p ln 2 p ) = O ( d = 1 p 1 | a = 1 p 1 ( ( a 4 d ) ( a 2 d ) p ) | | χ mod p χ ( 1 ) = 1 χ ( d ) | L ( 1 , χ ) | 2 | ) + 1 π 2 2 p 2 p 1 χ mod p χ ( 1 ) = 1 | L ( 1 , χ ) | 2 + O ( p ln 2 p ) .
(9)

If 2dp1, then from [10] we have the estimate

| a = 1 p 1 ( ( a 4 d ) ( a 2 d ) p ) | p .
(10)

From Lemma 4 we have

| a = 1 p 1 ( ( a 4 1 ) ( a 2 1 ) p ) |=| a = 0 p 1 ( a 2 + 1 p ) 12 ( 2 p ) |1.
(11)

From (10), (11) and Lemma 2, we have the estimate

d = 1 p 1 | a = 1 p 1 ( ( a 4 d ) ( a 2 d ) p ) | | χ mod p χ ( 1 ) = 1 χ ( d ) | L ( 1 , χ ) | 2 | p d = 1 p 1 | χ mod p χ ( 1 ) = 1 χ ( d ) | L ( 1 , χ ) | 2 | p 3 2 + ϵ .
(12)

Now, combining (6), (9) and (12), we deduce the asymptotic formula

a = 1 p 1 b = 1 p 1 S ( ( a 4 + m a 2 ) ( b 4 + m b 2 ) ¯ , p ) = 1 6 p 2 +O ( p 3 2 + ϵ ) .

This completes the proof of our theorem.

References

  1. Rademacher H:On the transformation of logη(τ). J. Indian Math. Soc. 1955, 19: 25–30.

    MathSciNet  Google Scholar 

  2. Rademacher H: Dedekind Sums. Carus Mathematical Monographs. Math. Assoc. of America, Washington; 1972.

    Google Scholar 

  3. Apostol TM: Modular Functions and Dirichlet Series in Number Theory. Springer, New York; 1976.

    Book  MATH  Google Scholar 

  4. Jia C: On the mean value of Dedekind sums. J. Number Theory 2001, 87: 173–188. 10.1006/jnth.2000.2580

    Article  MathSciNet  MATH  Google Scholar 

  5. Conrey JB, Fransen E, Klein R, Scott C: Mean values of Dedekind sums. J. Number Theory 1996, 56: 214–226. 10.1006/jnth.1996.0014

    Article  MathSciNet  MATH  Google Scholar 

  6. Zhang W: On the mean values of Dedekind sums. J. Théor. Nr. Bordx. 1996, 8: 429–442. 10.5802/jtnb.179

    Article  MathSciNet  MATH  Google Scholar 

  7. Apostol TM: Introduction to Analytic Number Theory. Springer, New York; 1976.

    MATH  Google Scholar 

  8. Hua LK: Introduction to Number Theory. Science Press, Beijing; 1979.

    Google Scholar 

  9. Zhang W, Yi Y, He X: On the 2 k -th power mean of Dirichlet L -functions with the weight of general Kloosterman sums. J. Number Theory 2000, 84: 199–213. 10.1006/jnth.2000.2515

    Article  MathSciNet  MATH  Google Scholar 

  10. Burgess DA: On Dirichlet characters of polynomials. Proc. Lond. Math. Soc. 1963, 13: 537–548.

    Article  MathSciNet  MATH  Google Scholar 

Download references

Acknowledgements

The authors express their gratitude to the referee for very helpful and detailed comments. This work is supported by the N.S.F. (11371291) and P.N.S.F. (2013JZ001) of P.R. China.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Wu Zhengang.

Additional information

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

KX obtained the theorems and completed the proof. WZ corrected and improved the final version. All authors read and approved the final manuscript.

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 2.0 International License (https://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Reprints and permissions

About this article

Cite this article

Xiaoyu, K., Zhengang, W. On the mean value of the two-term Dedekind sums. J Inequal Appl 2013, 579 (2013). https://doi.org/10.1186/1029-242X-2013-579

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/1029-242X-2013-579

Keywords