Next Article in Journal
Energy of Vague Fuzzy Graph Structure and Its Application in Decision Making
Previous Article in Journal
The Mei Symmetries for the Lagrangian Corresponding to the Schwarzschild Metric and the Kerr Black Hole Metric
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Mean Value of r-gcd-sum and r-lcm-Sum Functions

Research Center for Number Theory and Its Applications, Northwest University, Xi’an 710127, China
*
Author to whom correspondence should be addressed.
Symmetry 2022, 14(10), 2080; https://doi.org/10.3390/sym14102080
Submission received: 8 September 2022 / Revised: 24 September 2022 / Accepted: 2 October 2022 / Published: 6 October 2022
(This article belongs to the Section Mathematics)

Abstract

:
In this paper we perform a further investigation for r-gcd-sum and r-lcm-sum functions. By making use of the properties of generalization of Euler’s φ -function, Abel’s identity and elementary arguments, we derive asymptotic formulas for the average of the r-gcd-sum function, r-lcm-sum function and their generalizations. Moreover, we also study the sums of reciprocals of r-gcd and r-lcm.

1. Introduction

For any integer n 1 , the Pillai’s arithmetical function, which is also known as the gcd-sum function is defined by
P ( n ) = k = 1 n ( k , n ) = d | n d φ ( n / d ) ,
where φ is the Euler’s totient function, and ( k , n ) denotes the greatest common divisor of k and n. In the past several decades, the study of the gcd-sum function arises naturally in number theory, and a large number of mathematicians have investigated its distribution, arithmetic and algebraic properties. For example, an asymptotic formula for the average order of P ( n ) was proved by Bordellés [1], namely, for every ε > 0 ,
n x P ( n ) = x 2 2 ζ ( 2 ) log x + 2 γ 1 2 ζ ( 2 ) ζ ( 2 ) + O ( x 1 + θ + ε ) ,
where γ is Euler’s constant and θ is the exponent appearing in the Dirichlet’s divisor problem. In addition, Bordellés [2] also studied the sum of reciprocal of the gcd and deduced that
n x P ( 1 ) ( n ) = n x k = 1 n 1 ( k , n ) = ζ ( 3 ) 2 ζ ( 2 ) x 2 + O x ( log x ) 2 / 3 ( log log x ) 4 / 3 .
On the other hand, let [ k , n ] denote the least common multiple of k and n. The lcm-sum function is defined as
L ( n ) = k = 1 n [ k , n ] = n 2 1 + d | n d φ ( n / d )
for any integer n 1 . Bordellés investigated the function L ( n ) in his article [2] and established that
n x L ( n ) = ζ ( 3 ) 8 ζ ( 2 ) x 4 + O x 3 ( log x ) 2 / 3 ( log log x ) 4 / 3 ,
n x L ( 1 ) ( n ) = n x k = 1 n 1 [ k , n ] = ( log x ) 3 6 ζ ( 2 ) + ( log x ) 2 2 ζ ( 2 ) γ + log A 12 2 π + O log x ,
where A 1.282427 is the Glaisher–Kinkelin constant.
Let r 1 be a fixed integer, the greatest r t h power common divisor of positive integers a and b is defined to be the largest positive integer d r such that d r | a and d r | b , which is denoted by ( a , b ) r and called the r-gcd of a and b. Note that ( a , b ) 1 = ( a , b ) . The related research on greatest r th power common divisor can be found in [3]. Furthermore, we know that the most natural geometric usage of the gcd function is in counting visible lattice points. Recently, a new notion of visibility has gained much attention, where curves replace straight lines as the lines of sight. For example [4], A point ( k , n ) Z 2 is visible along the the curve y = h x r if and only if m a x { d 1 : d | k a n d d r | n } = 1 , where r N and h Q . So the immediate natural extension is to consider visibility along curves y = h x r or x = h y r . It is obvious that the r-gcd function provides a necessary criterion for the combined visibility of a lattice point along the curves y = h x r or x = h y r . So the fundamental research on r-gcd function is important in future work.
Currently, Prasad, Reddy and Rao [5] have introduced a natural generalization of the usual gcd-sum function as follows:
P r ( n r ) = k = 1 n r ( k , n r ) r ,
which is called r-gcd-sum function. They obtained an asymptotic formula for its summatory function:
n r x P r ( n r ) = x 1 + 1 r ( r + 1 ) ζ ( r + 1 ) log x r + 2 γ 1 r + 1 ζ ( r + 1 ) ζ ( r + 1 ) + O ε x 1 + θ + ε r .
With the help of (8), the authors [5] studied the partial sums of the Dirichlet series of the r-gcd-sum function, that is
G t , r ( x ) n r x P r ( n r ) n t ,
for all real values of t and any integer r 1 , and established asymptotic formulas for this function. Moreover, they [6] also introduced an r-gcd-sum function over r-regular integers ( mod n r ) , which is defined by
P r ˜ ( n r ) : = k ϵ R e g r ( n r ) ( k , n r ) r ,
where R e g r ( n r ) = { k : 1 k n r , k is r-regular mod n r }. Based on the properties of a generalization of Euler’s φ -function [3], they obtained some arithmetic properties of P r ˜ ( n r ) and an asymptotic formula for its summatory function.
On the other hand, it is natural to consider the generalized least common multiple related to r-gcd. In their article [7], Bu and Xu defined the least r th power common multiple as
[ k , n ] r = k n ( k , n ) r ,
which is also called r-lcm of k and n. They proved that
n x k = 1 n [ k , n ] r a = x 2 a + 2 ζ ( a r + 2 r ) 2 ( a + 1 ) 2 ζ ( 2 r ) + O x 2 a + 1
with positive integer a > 1 .
We notice that the r-gcd appeared when studying the r-free integers, and it also has most commonly been used to study the generalized totient function or commonly called the Klee’s totient function. Moreover, because the function P r ( n r ) is multiplicative, it is an interesting object of study from an analytic point of view, and the question is also significant to investigate the distribution of r-gcd and r-lcm by estimating the average of r-gcd-sum function and r-lcm-sum function.
The purpose of this paper is to perform a further investigation for r-gcd-sum and r-lcm-sum functions. By making use of the properties of a generalization of Euler’s φ -function, Abel’s identity and elementary arguments, we derive asymptotic formulas for the average of the r-gcd-sum function, r-lcm-sum function and their generalizations. Furthermore, we also study the sums of reciprocals of r-gcd and r-lcm.

2. Main Results

In this section, we give the main results of this paper as follows.
Theorem 1.
For any real number x > 1 and integer r > 1 , we have
n x k = 1 n ( k , n ) r = ζ ( r ) 2 ζ ( 2 r ) x 2 + O x 1 + 1 r log x .
Notice that for r = 1 , the asymptotic formula for average of the usual gcd-sum function P ( n ) had been done in [1] by the convolution identity P = φ I d = μ ( I d · τ ) , where I d ( n ) = n and ∗ is the usual Dirichlet convolution product.
Theorem 2.
For any real number x > 1 , we have the following estimate
n x k = 1 n 1 ( k , n ) r = ζ ( 3 r ) 2 ζ ( 2 r ) x 2 + O x log x .
When r = 1 , the main term of estimate (11) is equal to the main term of (3). Moreover, the error term of estimate (3) is improved from O x log x to O x ( log x ) 2 / 3 ( log log x ) 4 / 3 because of the classical result of Walfisz [8],
R ( x ) : = n x φ ( n ) 1 2 ζ ( 2 ) x 2 = O x ( log x ) 2 / 3 ( log log x ) 4 / 3 .
Theorem 3.
Let a 2 be a positive integer, then we obtain
n x k = 1 n ( k , n ) r a = ζ ( a ) ( a + 1 ) ζ ( r a + r ) x a + 1 + O x a .
Theorem 4.
Let a be a positive integer, then for any real number x > e sufficiently large, we have the following estimate
n x k = 1 n [ k , n ] r a = ζ ( a r + 2 r ) 2 ( a + 1 ) 2 ζ ( 2 r ) x 2 a + 2 + O x 2 a + 1 log x .
Here we offer a more simple proof of n x k = 1 n ( k , n ) r a by using elementary calculations.
Taking a = 1 in Theorem 4, we can get
Corollary 1.
For any real number x > e sufficiently large,
n x k = 1 n [ k , n ] r = ζ ( 3 r ) 8 ζ ( 2 r ) x 4 + O x 3 log x .
Theorem 5.
For any real number x > 1 , we have
n x k = 1 n 1 [ k , n ] r = ( log x ) 3 6 ζ ( 2 r ) + A ( log x ) 2 + O log x ,
where A = γ ζ ( 2 r ) ζ ( 2 r ) ζ 2 ( 2 r ) .
We know that for any function f : N C , there is the general identity
m , n x f ( m , n ) = 2 n x m = 1 n f ( m , n ) n x f ( n , n ) ,
which is symmetric in the variables. So it is easy to get the following asymptotic formulas from (10), (11), (14) and (15).
Corollary 2.
m , n x ( m , n ) r = ζ ( r ) ζ ( 2 r ) x 2 + O x 1 + 1 r log x ,
m , n x 1 ( m , n ) r = ζ ( 3 r ) ζ ( 2 r ) x 2 + O x log x ,
m , n x [ m , n ] r = ζ ( 3 r ) 4 ζ ( 2 r ) x 4 + O x 3 log x ,
m , n x 1 [ m , n ] r = ( log x ) 3 3 ζ ( 2 r ) + 2 A ( log x ) 2 + O log x .

3. Proofs of Theorems

3.1. Proof of Theorems 1 and 2

Proof of Theorem 1. 
We know that an integer is rth-power-free if it is not divisible by the rth power of any integer > 1 . Then let ϕ r ( n ) denote the number of integers k in the set 1 , , n , for which the greatest common divisor ( k , n ) is rth-power-free, and ϕ r ( n ) is the generalization of the Euler’s φ function attributed to Klee [3]. The function μ r ( n ) is defined as follows:
μ r ( 1 ) = 1 ;
if n > 1 and n = p 1 a 1 p 2 a 2 p t a t is the canonical factorization of n, then
μ r ( n ) = ( 1 ) t , if   a 1 = a 2 = = a t = r , 0 , if   for   some   i , a i r .
There is the following relationship between the function ϕ r ( n ) and function μ r ( n ) :
ϕ r ( n ) = n d | n μ r ( d ) / d .
So we can deduce
n x ϕ r ( n ) = n x d e = n μ r ( d ) e = d x μ r ( d ) e x / d e = 1 2 x 2 d x μ r ( d ) d 2 + O x d x 1 d = 1 2 x 2 d = 1 μ r ( d ) d 2 + x 2 d > x μ r ( d ) d 2 + O x log x = x 2 2 ζ ( 2 r ) + O x log x .
It is easily seen from the definition of function ϕ r ( n ) that
k = 1 n ( k , n ) r = k = 1 n ( k , n ) r = d r d r = d r | n d r k = 1 n / d r ( k , n / d r ) r = 1 1 = d r | n d r ϕ r n d r .
Hence, by applying (16) to (17), we have
n x k = 1 n ( k , n ) r = n x d r | n d r ϕ r n d r = d r x d r e x / d r ϕ r ( e ) = d r x d r 1 2 ζ ( 2 r ) x d r 2 + O x d r log x d r = x 2 2 ζ ( 2 r ) d = 1 1 d r + x 2 d > x 1 / r 1 d r + O x log x d r x 1 = ζ ( r ) 2 ζ ( 2 r ) x 2 + O x 1 + 1 r log x .
This completes the proof of Theorem 1. □
The proof of Theorem 2 is similar to Theorem 1 by applying (16) easily, so we will not go into details here.

3.2. Proof of Theorem 3

Proof of Theorem 3. 
For a 0 , there is the elementary asymptotic formula
n x n a = x a + 1 a + 1 + O ( x a ) ,
then we have
n x k = 1 n ( k , n ) r a = n x d r | n d r a ϕ r n d r = e x ϕ r ( e ) d r x / e d r a = e x ϕ r ( e ) 1 a + 1 x e a + 1 + O x e a = x a + 1 a + 1 e x ϕ r ( e ) e a + 1 + O x a e x ϕ r ( e ) e a .
Since n x 1 n s = x 1 s 1 s + ζ ( s ) + O ( x s ) , if s > 0 and s 1 , then we can deduce
n x ϕ r ( n ) n a = n x 1 n a d | n μ r ( d ) n d = d x μ r ( d ) d a e x / d 1 e a 1 = d x μ r ( d ) d a 1 2 a x d 2 a + ζ ( a 1 ) + O x d 1 a = 1 2 a x 2 a d x μ r ( d ) d 2 + ζ ( a 1 ) d x μ r ( d ) d a + O x 1 a d x μ r ( d ) d = x 2 a ( 2 a ) ζ ( 2 r ) + x 2 a d > x μ r ( d ) d 2 + ζ ( a 1 ) ζ ( a r ) + d > x μ r ( d ) d a + O x 1 a log x = x 2 a ( 2 a ) ζ ( 2 r ) + ζ ( a 1 ) ζ ( a r ) + O x 1 a log x ,
where a > 1 and a 2 . So, it follows from (18) and (19) that
n x k = 1 n ( k , n ) r a = x a + 1 a + 1 x 1 a ( 1 a ) ζ ( 2 r ) + ζ ( a ) ζ ( a r + r ) + O x a log x + O x a x 2 a ( 2 a ) ζ ( 2 r ) + ζ ( a 1 ) ζ ( a r ) + O x 1 a log x = ζ ( a ) ( a + 1 ) ζ ( r a + r ) x a + 1 + O x a ,
with a 2 .
This completes the proof of Theorem 3. □

3.3. Proof of Theorem 4

Proof of Theorem 4. 
From the definition of r-lcm-sum function, we have
k = 1 n [ k , n ] r a = k = 1 n k n ( k , n ) r a = d r | n n a d r a k = 1 n / d r ( k , n / d r ) r = 1 ( k d r ) a = n a d r | n k = 1 n / d r ( k , n / d r ) r = 1 k a ,
and for s 0 , notice that
k = 1 n ( k , n ) r = 1 k s = k = 1 n k s d | ( k , n ) r μ r ( d ) = d | n d s μ r ( d ) e = 1 n / d e s = d | n d s μ r ( d ) n s + 1 ( s + 1 ) d s + 1 + O ( n / d ) s = n s + 1 s + 1 d | n μ r ( d ) d + O n s d | n μ r ( d ) = n s s + 1 ϕ r ( n ) + O n s .
So, we can calculate
n x k = 1 n [ k , n ] r a = n x n a d r | n 1 a + 1 n d r a ϕ r n d r + O n d r a = n x n a a + 1 d r | n n a d r a ϕ r n d r + O n x n 2 a d r | n 1 d r = 1 a + 1 n x d r | n d r a n d r 2 a ϕ r n d r + O n x d r | n d r a n d r 2 a = 1 a + 1 d r x d r a e x / d r e 2 a ϕ r ( e ) + O d r x d r a e x / d r e 2 a = 1 a + 1 d r x d r a · 1 2 ( a + 1 ) ζ ( 2 r ) x d r 2 a + 2 + 1 a + 1 d r x d r a · O x d r 2 a + 1 log x d r + d r x d r a x d r 2 a + 1 = x 2 a + 2 2 ( a + 1 ) 2 ζ ( 2 r ) d r x 1 d a r + 2 r + O x 2 a + 1 log x d r x 1 d a r + r + O x 2 a + 1 d r x 1 d a r + r = x 2 a + 2 2 ( a + 1 ) 2 ζ ( 2 r ) d = 1 1 d a r + 2 r + O x 2 a + 2 d > x 1 / r 1 d a r + 2 r + O x 2 a + 1 log x = ζ ( a r + 2 r ) x 2 a + 2 2 ( a + 1 ) 2 ζ ( 2 r ) + O x 2 a + 1 log x ,
where we have used the following asymptotic formula:
n x n 2 a ϕ r ( n ) = n x n 2 a d | n μ r ( d ) n d = d x μ r ( d ) d 2 a e x / d e 2 a + 1 = d x μ r ( d ) d 2 a 1 2 a + 2 x d 2 a + 2 + O x d 2 a + 1 = x 2 a + 2 2 ( a + 1 ) d = 1 μ r ( d ) d 2 + O x 2 a + 2 d > x μ r ( d ) d 2 + O x 2 a + 1 d x 1 d = x 2 a + 2 2 ( a + 1 ) ζ ( 2 r ) + O x 2 a + 1 log x .
This completes the proof of Theorem 4. □

3.4. Proof of Theorem 5

Proof of Theorem 5. 
It is obvious that
k = 1 n 1 [ k , n ] r = k = 1 n ( k , n ) r k n = 1 n k = 1 n ( k , n ) r = d r d r k = 1 n d r | n d r k = 1 n / d r ( k , n / d r ) r = 1 1 k d r = 1 n d r | n k = 1 n / d r ( k , n / d r ) r = 1 1 k ,
so by using the properties of the function μ r , we have
n x k = 1 n 1 [ k , n ] r = d r x 1 d r h x / d r 1 h k h ( k , h ) r = 1 1 k = d r x 1 d r h x / d r 1 h k h 1 k a | ( k , h ) r μ r ( a ) = d r x 1 d r h x / d r 1 h a | h μ r ( a ) a b h / a 1 b = d r x 1 d r a x / d r μ r ( a ) a 2 c x / d r a 1 c b c 1 b = d r x a d r x 1 d r μ r ( a ) a 2 c x / d r a 1 c b c 1 b = n x 1 n 2 d r | n d r μ r n d r c x / n 1 c b c 1 b = n x ϕ r ( n ) n 2 c x / n 1 c b c 1 b = n x ϕ r ( n ) n 2 c x / n 1 c log c + γ + O 1 c = n x ϕ r ( n ) n 2 1 2 log x n 2 + γ log x n + O ( 1 ) = 1 2 n x ϕ r ( n ) n 2 log x n 2 + γ n x ϕ r ( n ) n 2 log x n + O n x ϕ r ( n ) n 2 .
Moreover, similar to the proof of (19), we can obtain
n x ϕ r ( n ) n 2 = d x μ r ( d ) d 2 e x / d 1 e = d x μ r ( d ) d 2 log x d + γ + O d x = log x + γ d x μ r ( d ) d 2 d x μ r ( d ) log d d 2 + O 1 x d x 1 d = log x + γ d = 1 μ r ( d ) d 2 + O d > x μ r ( d ) d 2 d = 1 μ r ( d ) log d d 2 + O d > x μ r ( d ) log d d 2 + O 1 x d x 1 d = log x + γ ζ ( 2 r ) 2 ζ ( 2 r ) ζ 2 ( 2 r ) + O log x x .
It follows from (21) and Abel’s identity that
n x ϕ r ( n ) n 2 log x n = log x n x ϕ r ( n ) n 2 n x ϕ r ( n ) n 2 log n = 1 ζ ( 2 r ) log t d log t + γ ζ ( 2 r ) d log t 2 ζ ( 2 r ) ζ 2 ( 2 r ) d log t + O log t t d log t = log 2 x 2 ζ ( 2 r ) + γ log x ζ ( 2 r ) 2 ζ ( 2 r ) log x ζ 2 ( 2 r ) + O ( 1 ) ,
and
n x ϕ r ( n ) n 2 log x n 2 = log 3 x 3 ζ ( 2 r ) + γ log 2 x ζ ( 2 r ) 2 ζ ( 2 r ) log 2 x ζ 2 ( 2 r ) + O ( log x ) .
Hence, applying (21)–(23) to (20) gives that
n x k = 1 n 1 [ k , n ] r = ( log x ) 3 6 ζ ( 2 r ) + A ( log x ) 2 + O log x ,
where A = γ ζ ( 2 r ) ζ ( 2 r ) ζ 2 ( 2 r ) .
This completes the proof of Theorem 5. □

4. Conclusions

In this paper, we firstly constructed various asymptotic formulas for the average of r-gcd-sum function and r-lcm-sum function in Theorems 1, 3 and 4, and the sums of reciprocals of r-gcd and r-lcm were also studied in Theorems 2 and 5. Additionally, Corollary 2 gives results on the order of magnitude of certain sums concerning the r-gcd’s and r-lcm’s of two positive integers. These results present the different distributional properties for the generalizations of usual gcd function and lcm function. Furthermore, we constructed two asymptotic formulas in Theorems 3 and 4 with a positive integer a 2 , and the study did not include the results with a 2 , so that will be the focus of our upcoming research. More generally, let P f , r ( n ) = k = 1 n f ( g c d ( k , n ) r ) with an arbitrary arithmetical function f. The weighted averages of the r-gcd-sum functions with any mathematical function f is one of our future research directions, where the weights are divisor functions, Bernoulli polynomials, gamma functions, etc.

Author Contributions

Writing—original draft preparation, Z.B.; writing—review and editing, Z.B.; supervision, Z.X. All authors have read and agreed to the published version of the manuscript.

Funding

This work is supported by the N.S.F. (11971381) of China.

Acknowledgments

The authors wish to express their gratitude to the editors and the reviewers for their helpful comments.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Bordellés, O. A Note on the Average Order of the gcd-sum Function. J. Integer Seq. 2007, 10, Article 07.3.3. [Google Scholar]
  2. Bordellés, O. Mean Values of Generalized gcd-sum and lcm-sum Functions. J. Integer Seq. 2007, 10, Article 07.9.2. [Google Scholar]
  3. Klee, V.L. A generalization of Euler’s φ-function. Am. Math. Mon. 1948, 55, 358–359. [Google Scholar] [CrossRef]
  4. Harris, P.E.; Omar, M. Lattice point visibility on power functions. Integers 2018, 18, 1–7. [Google Scholar]
  5. Prasad, V.S.R.; Reddy, P.A.; Rao, M.G. On the partial sums of the Dirichlet Series of the r-gcd-sum function. Int. J. Math. Comput. 2013, 3, 65–76. [Google Scholar]
  6. Prasad, V.S.R.; Reddy, P.A.; Rao, M.G. On a r-GCD-Sum function over r-Regular integers modulo nr. Int. J. Soft Comput. Eng. 2013, 3, 7–11. [Google Scholar]
  7. Bu, Z.; Xu, Z. Asymptotic formulas for generalized gcd-sum and lcm-sum functions over r-regular integers (modnr). AIMS Math. 2021, 6, 13157–13169. [Google Scholar] [CrossRef]
  8. Walfisz, A. Weylsche Exponentialsummen in der Neueren Zahlentheorie. Mathematische Forschungsberichte; XV.VEB; Deutscher Verlag der Wissenschaften: Berlin, Germany, 1963. [Google Scholar] [CrossRef]
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Bu, Z.; Xu, Z. Mean Value of r-gcd-sum and r-lcm-Sum Functions. Symmetry 2022, 14, 2080. https://doi.org/10.3390/sym14102080

AMA Style

Bu Z, Xu Z. Mean Value of r-gcd-sum and r-lcm-Sum Functions. Symmetry. 2022; 14(10):2080. https://doi.org/10.3390/sym14102080

Chicago/Turabian Style

Bu, Zhengjin, and Zhefeng Xu. 2022. "Mean Value of r-gcd-sum and r-lcm-Sum Functions" Symmetry 14, no. 10: 2080. https://doi.org/10.3390/sym14102080

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop