Abstract
Let \({{\mathbb {Z}}}_{n}\) be the additive group of residue classes modulo n. Let \(c(n_1,n_2,n_3)\) denote the number of cyclic subgroups of the group \({{\mathbb {Z}}}_{n_1}\times {{\mathbb {Z}}}_{n_2}\times {{\mathbb {Z}}}_{n_3}\), where \(n_1, n_2\) and \(n_3\) are arbitrary positive integers. In this paper we obtain an asymptotic formula for the sum \(\sum _{n_1,n_2,n_3\le _x} c(n_1,n_2,n_3).\)
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Let \({\mathbb {Z}}_{n}\) denote the additive group of residue classes modulo n. We recall that a finite abelian group G (\(|G|>1\)) is of rank r if it is isomorphic to \({\mathbb {Z}}_{n_1} \times \cdots \times {\mathbb {Z}}_{n_r}\), where \(n_1,\ldots ,n_r \in {\mathbb {N}}{\setminus } \{1\}\) and \(n_j \mid n_{j+1}\) (\(1\le j\le r-1\)). Let \(\text {rank}(G)\) denote the rank of G and s(G) be the number of subgroups of G. The one variable level function \(t_r(n)=\sum _{|G|=n, \, \text {rank}(G)\le r} s(G)\) and its summatory function \(\sum _{n\le x} t_r(n)\) were investigated by Bhowmik and Wu [1]. There are many other asymptotic results on the number of subgroups of abelian groups. See, e.g., the papers [3, 9] and their references.
We are interested in the following different problem, not studied in the above papers. Let \(r\ge 1\) be a fixed integer. For arbitrary positive integers \(n_1,\ldots ,n_r,\) consider the group \({\mathbb {Z}}_{n_1}\times \cdots \times {\mathbb {Z}}_{n_r}\). Let \(s(n_1,\ldots , n_r)\) and \(c(n_1,\ldots , n_r)\) denote the total number of subgroups and the number of cyclic subgroups of the given group \({\mathbb {Z}}_{n_1}\times \cdots \times {\mathbb {Z}}_{n_r}\), respectively. Here \(s(n_1,\ldots , n_r)\) and \(c(n_1,\ldots , n_r)\) are multiplicative functions of r variables.
Suppose \(x>0\) is a real number. For each \(r\ge 1\) define the multivariable sums
When \(r=1\), the group \({\mathbb {Z}}_n\) is cyclic and hence \(s(n)=c(n)=\tau (n)\), where \(\tau (n)\) denotes the number of divisors of n. So, \(S_1(x)=C_1(x)= \sum _{n\le x} \tau (n)\). Therefore, the problem of finding estimates for the sums \(S_r(x)\) and \(C_r(x)\) can be viewed as a generalization of the well-known Dirichlet divisor problem.
Now consider the case \(r\ge 2\). If \((n_i, n_j)=1\) for all \(1\le i<j\le r\), then \({\mathbb {Z}}_{n_1}\times \cdots \times {\mathbb {Z}}_{n_r}\) is cyclic and isomorphic to \({\mathbb {Z}}_{n_1\cdots n_r}\), which is a consequence of the Chinese remainder theorem. If otherwise, then \({\mathbb {Z}}_{n_1}\times \cdots \times {\mathbb {Z}}_{n_r}\) has rank at least two.
There are several ways to compute \(s(n_1,n_2)\) and \(c(n_1,n_2).\) For example, one way is using Goursat’s lemma for groups in [2, 8, 13] and another one is using the concept of the fundamental group lattice in [10, 11]. See also the paper [4]. For any \(n_1,n_2\in {{\mathbb {N}}}\) we have
and
When \(r\ge 3\), we have by Theorem 1 of [12],
However, in the case \(r\ge 3\) there is no such a closed formula for \(s(n_1, \ldots , n_r)\).
W. G. Nowak and L. Tóth [6] studied the average orders of the functions \(s(n_1,n_2)\) and \(c(n_1,n_2)\). They proved that the asymptotic formulas
and
hold, where \(A_{j}\) and \(B_j\) (\(0\le j\le 3\)) are explicit constants, of which definitions are omitted here. Note that the exponent of the error term is \(1117/701=1.593437\cdots \). The authors of the present paper proved in [14] that both error terms in the asymptotic formulas (1.4) and (1.5) can be improved into \(O(x^{1.5}(\log x)^{6.5})\).
In this paper we shall prove the following theorem, by using a multidimensional Perron formula and the complex integration method.
Theorem 1.1
The asymptotic formula
holds, where \(c_j\, (0\le j\le 7)\) are explicit constants.
Notations
Throughout this paper, \({\mathbb {N}}\) denotes the set of all positive integers, \({\mathbb {N}}_0={\mathbb {N}}\cup \{0\},\)\((a_1,\ldots ,a_k)\) and \([a_1,\ldots ,a_k]\) are the gcd and lcm of \(a_1,\ldots ,a_k\in {\mathbb {N}}\), \(\varphi \) is Euler’s totient function, \(\mu \) is the Möbius function, \(\zeta \) denotes the Riemann zeta-function, \(\tau _k(n)\) denotes the number of ways n can be written as a product of k positive integers (\(\tau (n)=\tau _2(n)\)).
2 Preliminary lemmas
We shall use the following lemmas.
Lemma 2.1
Suppose \(\mathfrak {R}z>1, \mathfrak {R}w>1\). Then
Proof
See the paper [6]. \(\square \)
Lemma 2.2
Suppose that \(r\ge 2\) is a fixed integer and \(f(n_1,\ldots ,n_r)\) is an arithmetical function of r variables such that it is symmetric in the variables \(n_1,\ldots ,n_r\) and its Dirichlet series
is absolutely convergent for \(\mathfrak {R}z_j> \sigma _a\) (\(1\le j\le r\)) with some \(\sigma _a>0\). Suppose that \(x\notin {\mathbb {N}}\) and T are two large parameters such that \(10\le T\le x/2\) and define
Then we have
where
Proof
This is a slight modification of Lemma 2.2 in [14]. \(\square \)
Lemma 2.3
Suppose \(\ell \ge 0\) is a fixed integer. For \(\sigma >1\) we have
Proof
The case \(\ell =0\) can be found in Chapter 7 of Pan and Pan [7]. The case \(\ell \ge 1\) follows from the result of the case \(\ell =0\) and Cauchy’s theorem. \(\square \)
Lemma 2.4
Suppose \(\ell \ge 0\) is a fixed integer. Then we have
Proof
The estimate for the case \(\ell =0\) follows from the bounds
and the Phragmen-Lindelöf principle. The estimate for the case \(\ell \ge 1\) follows from the result of the case \(\ell =0\) and Cauchy’s theorem. \(\square \)
Lemma 2.5
Suppose \(V>10\) is a large parameter and \(41/60\le u\le 1\). Then we have
Proof
It is proved in Chapter 8 of Ivić [5] that Lemma 2.5 holds for \(u=41/60.\) By Lemma 2.3 we see that Lemma 2.5 trivially holds for \(u=1.\) The case \(41/60<u<1\) follows from the cases \(u=41/60\) and \(u=1\). \(\square \)
Lemma 2.6
We have the asymptotic formula
where \(Q_3(u)\) is a polynomial in u of degree 3.
Proof
Well-known. It follows from the Perron formula and the fourth power moment of \(\zeta (s)\) over the critical line. \(\square \)
Lemma 2.7
Suppose \(\ell \ge 0\) is a fixed integer, \(11/20-\varepsilon \le u\le 11/20+\varepsilon .\) Then we have
Proof
The proof for the case \(\ell =0\) and \(u=11/20\) can be found in Chapter 8 of [5]. The same method works for an arbitrary u with \(11/20-\varepsilon \le u\le 11/20+\varepsilon .\) The case \(\ell \ge 1\) follows from the result of the case \(\ell =0\) and Cauchy’s theorem. \(\square \)
Lemma 2.8
Suppose \(\alpha _j>2/3,\ \beta _j>1\ (j=1,2,3)\) are fixed real numbers such that \(\alpha _1+\alpha _2+\alpha _3>2\), \(f_j(t)\ (t\in {\mathbb {R}}, j=1,2,3)\) and \(f(t_1,t_2)\ (t_1,t_2\in {\mathbb {R}}) \) are complex-valued functions such that
Suppose \(U>4\) and \(U^{\prime }>4\) are positive parameters such that \(U\asymp U^{\prime }\), and define
Then
(1) we have the estimate
(2) the integral W is absolutely convergent.
Proof
Without loss of generality, we suppose \(U\le U^{\prime }.\) Obviously, from (2.1) we have
where
So, it suffices for us to prove that for \(j=1,2,3,4\),
It is easy to see that \(W_1=W_3,W_2=W_4.\)
I. Estimate of \(W_1\):
We write
It is easily seen that
Similarly we have
From (2.4) to (2.6) we see that (2.3) holds for \(j=1.\)
II. Estimate of \(W_4\):
We write
We estimate \(W_{41}\) first. We have
Similarly we have
From (2.7) to (2.9) we see that (2.3) holds for \(j=4.\)
Now the first assertion of Lemma 2.8 follows from (2.2) and (2.3). The second assertion of Lemma 2.8 is obviously a consequence of the first assertion. \(\square \)
Lemma 2.9
Suppose \(M(s)=AM_1(s)M_2(s)\) such that \(M_1(s)\) is meromorphic in the region \(|s-s_0|\le \eta , \)\(s=s_0\) is a pole of order \(r\ge 2,\) and \(M_2(s)\) is analytic in \(|s-s_0|\le \eta , \) where \(\eta >0\) is a positive real number, \(A\not = 0\) is a constant. Write
Then we have
Proof
It follows from the theory of complex variable functions. \(\square \)
Lemma 2.10
Let \(k, \ell \ge 2\) be two fixed integers and \(f_j(s) \ (j=1,\ldots ,k)\) be functions which are at least \(\ell \)-times differentiable. Then for each \(0\le j\le \ell \) we have
Proof
Well-known. \(\square \)
3 On the Dirichlet series of \(c(n_1,n_2,n_3)\)
In this section we shall study the Dirichlet series
We have the following proposition.
Proposition 3.1
Suppose that \(s_j=\sigma _j+it_j, \sigma _j>1 \ (j=1,2,3)\). Then we have
where \(H(s_1,s_2,s_3)\) can be written as a triple Dirichlet series, which is absolutely convergent when \(\sigma _1, \sigma _2, \sigma _3\) satisfy the following conditions:
Proof
Noting that \(c(n_1,n_2,n_3)\) is multiplicative and (1.3) holds, we have that
Write
where
For each \(1\le j\le 7,\) we define
Also we define for \(j=2,3,4,5\),
For \(j=1,2,3,\) it is easy to see that
For \(j=4\) we have
Similarly, we have
and
For \(j=7\) we have
For simplicity, we write
So, from (3.6) to (3.10), we can write
where
Obviously, we have
It is easy to see that
By the expression \(1/(1-\xi )=\sum _{m=0}^\infty \xi ^m\ (|\xi |<1),\) we have
So, from (3.13) and (3.14) we get
Now we estimate \(D_2(p;s_1, s_2, s_3).\) We write
say. We only estimate \(D_{23}(p;s_1, s_2, s_3).\) The estimates for \(D_{21}(p;s_1, s_2, s_3)\) and \(D_{22}(p;s_1, s_2, s_3)\) are similar.
We write
where
We have the estimate
So, from (3.18) we get
Similarly we have
We have
Similarly we have
Finally, we estimate \(D_{235}(p;s_1,s_2,s_3).\) We have
Similarly we have
and
From (3.16), (3.24–3.26) we have
We estimate \(D_3(p;s_1,s_2,s_3)\) now. We have
Similarly we have
and
We write
where
say. From (3.11), (3.12), (3.15), (3.27)–(3.30) we get
Now we see from (3.4), (3.31)–(3.33) that Proposition 3.1 holds and \(H(s_1,s_2,s_3)\) can be written as a triple Dirichlet series which is absolutely convergent under the condition (3.3). \(\square \)
4 On an infinite sum
Suppose \(x\notin {\mathbb {N}}\) and T are two large parameters such that \(5\le T\le x/2\). For each \(1\le j\le 3,\) define the infinite sum
Since \(c(n_1, n_2, n_3)\) is symmetric for \(n_1,n_2, n_3,\) we have
say.
Write
where
4.1 Upper bound of \(G(n_3, \sigma )\)
In this subsection we shall give an upper bound of \(G(n_3, \sigma ).\)
If \(n_3=1,\) then \(c(n_1, n_2, 1)=c(n_1, n_2),\) so from Lemmas 2.1 and 2.3 we have
Now suppose \(n_3\ge 2.\) We begin with the formula (1.3). By noting that \(\varphi (n)\gg n/\log n\) and \([m,n]=mn/(m,n)\) we have
Inserting the estimate (4.3) into the expression of \(G(n_3, \sigma ), \) we have
where
We estimate \(G_1(n_3, \sigma )\) first. By writing \(n_j=d_j\ell _j\) (\(j=1,2\)) we have
Let \(a=(d_1,d_2),\ d_1=ab, d_2=ac\). Then \((b,c)=1\) and \([d_1,d_2]=abc.\) So from (4.6) we have
where we used the estimate \(\tau _3(nd)\le \tau _3(n)\tau _3(d)\) and Lemma 2.3, and where
By Lemma 2.3 we get
It is easy to see that
By Lemma 2.3 again we get
\(G_2(n_3,\sigma )\) can be bounded in the same way. We have
From (4.4), (4.10) and (4.11) we get
4.2 Upper bound of E(x, T)
Inserting (4.2) and (4.12) into (4.1), we get
Let
We have by Lemma 2.3 that
Now we bound \(U_2.\) We write
When \(x/2<n_3\le x e^{-1/T},\) we have \(T|\log \frac{x}{n_3}|=T\log \frac{x}{n_3}\ge 1.\) Thus
where in the second line we used the estimate \(\log (x/([x]-\ell ))\gg \ell /x\) and in the fifth line we used Lemma 2.6. Similarly we have
When \(xe^{-1/T}<n_3\le xe^{1/T}x,\) we have \(T|\log \frac{x}{n_3}|\le 1.\) Thus by Lemma 2.6 again we get
Combining (4.13)–(4.19) we get
Proposition 4.1
Suppose \(x\notin {\mathbb {N}}\) and T are two large parameters such that \(5\le T\le x/2\). Then
5 Proof of the theorem
In this section, we shall prove Theorem 1.1.
Suppose \(x\ge 20, x\notin {\mathbb {N}}\). Let \(10\le T\le x/2\) be a parameter to be determined later. Define
By Lemma 2.2 with \(r=3\) and \(\sigma _a=1\) we have
where
\(C(s_1,s_2,s_3)\) and E(x, T) being defined in the last section. Here E(x, T) can be bounded by Proposition 4.1, so we only need to evaluate I(x, T). We shall evaluate I(x, T) in the order \(s_3, s_2,s_1.\)
5.1 Evaluation of I(x, t) for \(s_3\)
In this subsection we shall evaluate the integral I(x, T) for the variable \(s_3\). Consider the rectangle domain formed by the four points \(s_3=b_3\pm iT_3, s_3=11/20\pm iT_3\). From Proposition 3.1 we see easily that in this domain the integral function
has four poles, which are \(s_3^{(0)}=1\), \(s_3^{(1)}=2-s_1, \)\( s_3^{(2)}=2-s_2,\)\( s_3^{(3)}=3-s_1-s_2, \) respectively. Here \(s_3=1\) is a pole of order 2 and the other three poles are simple. By the residue theorem we get
where
5.1.1 Estimates of \(H_j(x,T)\ (j=1,2,3)\)
We estimate \(H_1(x,T)\) first. In this case we have \(s_1=b_1+it_1\) with \(|t_1|\le T,\)\(s_2=b_2+it_2\) with \(|t_2|\le 10T\) and \(s_3=\sigma _3+i100T\) with \(11/20\le \sigma _3\le b_3.\) By the condition (3.3) in Proposition 3.1 we see that
From the above estimates and (5.2) we have
So we get
Similarly, we have
Now we estimate \(H_2(x,T).\) In this case we have \(s_1=b_1+it_1\) with \(|t_1|\le T,\)\(s_2=b_2+it_2\) with \(|t_2|\le 10T\) and \(s_3=11/20+it_3\) with \(|t_3|\le T_3=100T.\) Similar to (5.4) we have
So we have
where
By Hölder’s inequality we get
where
By Lemma 2.7 and partial summation we get
For \(H_{24}(T),\) we have
From the condition \(|t_1|\le |t_2|\le |t_3|\) we get
Thus we get
which combining Lemma 2.7 gives
Similarly, we have
5.1.2 Evaluation of \(I_3(x,T)\)
Now we evaluate \(I_3(x,T).\) Since \(s_3^{(3)}=3-s_1-s_2\) is a simple pole of \(g(s_1,s_2,s_3), \) we have
Define
From Lemma 2.4 we have
From Proposition 3.1 we have
So from Lemma 2.8 with \(\alpha _j=1, \beta _j=3(j=1,2,3)\) and \(U=T,\) we get
where
From the condition (3.3) in Proposition 3.1 we see that \(H(s_1,s_2, 3-s_1-s_2)\) can be written as a Dirichlet series, which is absolutely convergent when \(s_1\) and \(s_2\) satisfy the following conditions:
From Lemma 2.3 and Lemma 2.4 we see that if \(1/2\le \sigma _1\le b_1, 1/2\le \sigma _1\le b_2, \sigma _1+\sigma _2\ge 3/2,\) then
From (5.16), (5.17) and Lemma 2.8 we see that the integral \(G(\sigma _1,\sigma _2)\) is absolutely convergent if \(2/3\le \sigma _1\le b_1, 2/3\le \sigma _1\le b_2, \sigma _1+\sigma _2\ge 3/2.\) Certainly we can show that the integral \(G(\sigma _1,\sigma _2)\) is absolutely convergent in a much wider range. However, the above range is more than enough for our proof.
Now we show that \(G(b_1,b_2)\) is a constant. We write
say. We move the integral line of the inner integral from \(\mathfrak {R}s_2=b_2\) to \(\mathfrak {R}s_2=3/4.\) The function \(Y(s_1,s_2)\) has two poles in this range, which are \(s_2=1\) (degree 3) and \(s_2=2-s_1\) (degree 3), respectively. We get
where
Both \(G_1(s_1)\) and \(G_2(s_1)\) can be computed by Lemmas 2.9 and 2.10, respectively. But the explicit expressions of them are not important here. So we omit the details.
We move both integral lines in (5.19) from \(\mathfrak {R}s_1=b_1\) to \(\mathfrak {R}s_1=3/4,\) we get
where
From (5.20) we see that \(G(b_1,b_2)\)is an absolute constant. We denote it by C. Hence we have
5.2 Evaluation of \(I_0(x,T)\)
In this subsection, we shall evaluate \(I_0(x,T)\). By Lemma 2.9 we get that
where (\(\gamma \) is Euler constant)
So we have
5.2.1 Integration over \(s_2\)
We consider the domain formed by the four points \(s_2=b_2\pm iT_2, s_2=11/20\pm iT_2.\) It is easy to see that for each \(1\le j\le 7,\) the function \(K_j(s_1,s_2)\) has two poles, which are \(s_2^{(1)}=1\) and \(s_2^{(2)}=2-s_1,\) respectively. By the residue theorem, we have
where (\(1\le j\le 7\))
Similar to the estimate of \(H_1(x,T)\) we have
Similar to the estimate of \(H_2(x,T)\) we have
5.2.2 Evaluations of \(I_0^{(j,1)}(x,T)\)
Now we evaluate \(I_0^{(j,\ell )}(x,T)\ (1\le j\le 7, \ell =1,2).\) We consider \(j=7.\) The proof of other cases are the same.
First consider the pole \(s_2=1,\) which is of degree 3 in \(K_7(s_1,s_2).\) Define
Write
So by Lemmas 2.9 and 2.10 we get
We have
Suppose \(s_1, s_2, s_3\) satisfy the condition (3.3). Then we can write
which is absolutely convergent for the variables \(s_1, s_2, s_3\). So for any \((\ell _1,\ell _2,\ell _3)\in {\mathbb {N}}_0^3,\) we have
Define
From (5.30) it is easy to see that
From the condition (3.3) it is easy to see that \(h_0(s_1), h_1(s_1)\) and \(h_2(s_1)\) are all analytic when \(\mathfrak {R}s_1>1/2.\)
From (5.27) to (5.31) it is easy to see that
where
Hence we have
We consider the domain formed by the four points \(s_1=b_1\pm iT, s_1=11/20\pm iT.\) By the residue theorem, we get
where
Similar to (5.5), by Lemmas 2.3 and 2.4 we have for \(1\le j\le 4\) that
Similar to the estimate of \(H_2(x,T)\), by Lemma 2.7 we have
By (5.33) and Lemma 2.9 we get
where \(Q_j(u)(j=1,2,3,4)\) are polynomials in u of degree 7.
where \(Q_{71}(u)\) is a polynomial in u of degree 7. Similarly for each \(1\le j\le 6\), we have
where \(Q_{j1}(u)\) is a polynomial in u of degree 7.
Now we consider the pole \(s_2=2-s_1,\) which is of degree 2 in \(K_7(s_1,s_2).\) Write
where
By Lemma 2.9 we have
where \(A=\zeta ^3(s_1)\frac{x^{s_1+1}}{s_1} \log x\) and
By the condition (3.3) it is easy to see that both \(h_3(s_1)\) and \(h_4(s_1)\) are analytic when \(1/2<\mathfrak {R}s_1<3/2.\) Similar to (5.21), we have
where \(C_j\)\((j= 5, 6, 7, 8, 9)\) are absolute constants.
Thus we get
where \(Q_{72}(u)\) is a polynomial in u of degree 2. Similarly for each \(1\le j\le 6,\) we have
where \(Q_{j2}(u)\) is a polynomial in u of degree not exceeding 2.
Combining (5.24)–(5.26), (5.39), (5.40), (5.45) and (5.46) we get
where \(P_0(u)\) is a polynomial in u of degree 7.
5.3 Evaluation of \(I_1(x,T)\)
In this subsection we evaluate \(I_1(x,T).\) Since \(s_3^{(1)}=2-s_1\) is a simple pole of \(g(s_1,s_2,s_3),\) we have
say. Consider the domain formed by the four points \(s_2=b_2\pm iT_2, s_2=11/20\pm iT_2.\) The function \(g_2(s_1,s_2)\) has three poles in this domain, which are \(s_2^{(1)}=1\) (of degree 3), \(s_2^{(2)}=s_1\) (simple) and \(s_2^{(3)}=2-s_1\) (simple), respectively. By the residue theorem, we get
where
We omit the details of the proof of (5.49), since it is similar to Sect. 5.1.
5.3.1 Evaluation of \(I_{11}(x,T)\)
Since \(s_2=1\) is a pole (of order 3) of \(g_2(s_1,s_2)\), by Lemmas 2.9 and 2.10 we have
where
From the condition (3.3) of Proposition 3.1 we see easily that \(H(s_1,1,2-s_1)\) is analytic when \(1/2<\mathfrak {R}s_1<3/2.\) Hence we have
Thus similar to (5.21) we can get
where \(P_{11}(u)\) is a polynomial in u of degree 2.
5.3.2 Evaluation of \(I_{12}(x,T)\)
Since \(s_2=s_1\) is a simple pole of \(g_2(s_1,s_2)\), we have
which implies that
By the condition (3.3) we see easily that \(H(s_1,s_1,2-s_1)\) is analytic if \(2/3<\mathfrak {R}s_1<5/3.\) Hence we have \(H(s_1,s_1,2-s_1)\ll _\varepsilon 1\) if \(2/3+\varepsilon<\mathfrak {R}s_1<5/3-\varepsilon .\) The function \(\zeta ^5(s_1)\zeta ^2(2-s_1)\zeta (2s_1-1)H(s_1,s_1,2-s_1)\frac{x^{s_1+2}}{s_1^2(2-s_1)}\) has a pole \(s_1=1\) in the region \(2/3+\varepsilon <\mathfrak {R}s_1\le b_1\), which is of order 8. So moving the integration line in (5.54) from \(\mathfrak {R}s_1=b_1\) to \(\mathfrak {R}s_1=2/3+\varepsilon ,\) we get
where \(P_{12}(u)\) is a polynomial in u of degree 7, and where we used the estimate
The above estimate follows easily from Lemma 2.4.
5.3.3 Evaluation of \(I_{13}(x,T)\)
Now we consider \(I_{13}(x,T).\) Since \(s_2=2-s_1\) is a simple pole of \(g_2(s_1,s_2),\) we have
where \(h_5(s_1):=H(s_1,2-s_1,2-s_1).\) So we get
By the condition (3.3) of Proposition 3.1 it is easy to see that \(h_5(s_1)\) is analytic in the region \(1/3<\mathfrak {R}s_1<4/3.\) So we have \(h_5(s_1)\ll _\varepsilon 1 \) when \(1\le \mathfrak {R}s_1\le 4/3-\varepsilon . \) Also, the integral function in (5.57) has no poles in the range \(1<\mathfrak {R}s_1\le 4/3-\varepsilon .\) Moving the integral line in (5.57) from \(\mathfrak {R}s_1=b_1\) to \(\mathfrak {R}s_1=4/3-\varepsilon ,\) by the residue theorem and Lemma 2.4 we get
From (5.49), (5.52), (5.55), (5.58) we get
where \(P_1(u)\) is a polynomial in u of degree 7.
5.4 Evaluation of \(I_2(x,T)\)
In this subsection we shall evaluate \(I_2(x,T)\). Since \(s_3^{(2)}=2-s_2\) is a simple pole of \(g(s_1,s_2,s_3),\) we have
where
5.4.1 Integration over \(s_2\)
By the condition (3.3) of Proposition 3.1 it is easy to see that \(H(s_1,s_2,2-s_2)\) is analytic when \(\sigma _1\) and \(\sigma _2\) satisfy
We write
From (5.60) and (5.61) it is easy to check that
We move the integral line of \(s_2\) from \(\mathfrak {R}s_2=b_2\) to \(\mathfrak {R}s_2=1-\varepsilon .\) In the range \(1-\varepsilon \le \mathfrak {R}s_2\le b_2\), the function \(g_3(s_1,s_2)\) has three poles, which are \(s_2=1\) (of order 4), \(s_2=s_1\) (simple) and \(s_2=2-s_1\) (simple), respectively. By the residue theorem we have
where
where
5.4.2 Integration over \(s_1\)
We first consider \(I_{23}(x,T).\) Since \(s_2=s_1\) is a simple pole of \(g_3(s_1,s_2)\), we have
From (3.3) we see that \(H(s_1,s_1,2-s_1)\) can be written as a Dirichlet series, which is absolutely convergent for \(2/3<\mathfrak {R}s_1<5/3.\) Hence we have \(H(s_1,s_1,2-s_1)\ll _\varepsilon 1\) if \(2/3+\varepsilon<\mathfrak {R}s_1<5/3-\varepsilon .\) Note that \(s_1=1\) is the pole of \(\frac{\zeta ^3(s_1)x^{s_1+2}}{s_1}G_3(s_1)\) of order 8. By the residue theorem, Lemmas 2.3, 2.4 and 2.9 we have
where \(P_{23}(u)\) is a polynomial in u of degree 7. Similarly we have
if noting that \(G_4(s_1)=\zeta ^2(s_1)\zeta ^2(2-s_1) \zeta (2s_1-1)\frac{H(s_1,2-s_1,s_1)}{s_1(2-s_1)},\) where \(P_{24}(u)\) is also a polynomial in u of degree 7.
Now we consider \(I_{25}(x,T).\) We compute \(G_5(s_1)\) first. Let
Since \(s_2=1\) is the pole of \({\varvec{M}}_1(s_2)\) with order 4, we can write
So we have
Suppose \(\mathfrak {R}s_1=b_1,|\sigma _2-1|<1/1000.\) Then \(s_1,s_2,2-s_2\) satisfy the condition (3.3). Thus
Hence for any \(\ell \ge 0\) we have
which implies that
The condition (3.3) implies that the above infinite series is absolutely convergent for \(\sigma _1>1/2.\) Thus we get that for any fixed \(n_3\ge 0\)
The integral function \(\frac{\zeta ^3(s_1)x^{s_1+2} \zeta ^{(n_1)}(s_1) \zeta ^{(n_2)}(s_1) }{s_1}{\varvec{H}}_{s_1}^{(n_3)}(1)\) has a pole \(s_1=1\) of order \(5+n_1+n_2.\) It is easy to see the highest order is 8, which corresponds to \(n_1+n_2=3.\) Consider the domain formed by the four points \(s_1=b_1\pm iT, 11/20\pm iT.\) So similar to (5.39), by the residue theorem, Lemmas 2.3, 2.4, 2.7 and 2.9 we get
where \(P_{25}(u)\) is a polynomial in u of degree 7.
Finally we consider \(I_{26}(x,T).\) From the condition (3.3) we see that if \(\mathfrak {R}s_2=1-\varepsilon \), then the function \({\varvec{H}}_{s_1}(s_2)\) is analytic for \(\mathfrak {R}s_1>1/2+\varepsilon /2\) and hence is bounded if \(\mathfrak {R}s_1>1/2+\varepsilon .\) Later we always suppose \(\mathfrak {R}s_1>1/2+\varepsilon .\) We need two estimates of \(G_6(s_1).\)
Since \(\mathfrak {R}s_2=1-\varepsilon ,\) we have \(\zeta ^2(2-s_2)\ll 1\) and \( H(s_1,s_2,2-s_2)\ll 1.\) By Lemma 2.4 we have
The second estimate involving \(G_6(s_1)\) is the following:
In order to prove (5.74) suppose \(s_1=11/20+it_1\) and \(s_2=1-\varepsilon +it_2.\) Then \(s_1+s_2-1=11/20-\varepsilon +i(t_1+t_2)\) and \(1+s_1-s_2=11/20+\varepsilon +i(t_1-t_2).\) By Hölder’s inequality we get
if noting that (by Lemma 2.5)
where
Hence we have by Cauchy’s inequality that
By Cauchy’s inequality again we have
Hence by Lemma 2.7 we have
Similarly we have
The estimate (5.74) follows from (5.75) and (5.78).
Consider the domain formed by the four points \(s_1=b_1\pm iT, 11/20\pm iT.\) So by the residue theorem, we get
where
From Lemma 2.4 and (5.73) we get
By Lemma 2.7 and (5.74) we get via Hölder’s inequality that
which implies that
From (5.79) to (5.81) and Lemma 2.9 we get that
where \(P_{26}(u)\) is a polynomial in u of degree 2.
From (5.65), (5.68), (5.69), (5.72) and (5.82) we get
where \(P_{2}(u)\) is a polynomial in u of degree 7.
5.5 Completion of the proof
First suppose \(10\le x\notin {\mathbb {N}}.\) Suppose T ia a large parameter such that \(5\le T\le x/2.\) Combining (5.1), (5.3), (5.5), (5.6), (5.14), (5.21), (5.47), (5.59), (5.83) and Proposition 4.1 we get by taking \(T=x^{1/2}\),
where P(u) is a polynomial in u of degree 7. Hence our theorem is true when \(x\notin {\mathbb {N}}.\)
Now suppose \(10\le x\in {\mathbb {N}}.\) Then we have by (5.84) that
References
Bhowmik, G., Wu, J.: Zeta function of subgroups of abelian groups and average orders. J. Reine Angew. Math. 530, 1–15 (2001)
Călugăreanu, G.: The total number of subgroups of a finite abelian group. Sci. Math. Jpn. 60, 157–167 (2004)
Chinta, G., Kaplan, N., Koplewitz, S.: The cotype zeta function of \({\mathbb{Z}}^d\), Preprint (2017), arXiv:1708.08547 [math.NT]
Hampejs, M., Holighaus, N., Tóth, L., Wiesmeyr, C.: Representing and counting the subgroups of the group \({\mathbb{Z}}_m \times {\mathbb{Z}}_n\), J. Numbers, Article ID 491428 (2014)
Ivić, A.: The Riemann Zeta-Function. Theory and Applications. Wiley, New York (1985)
Nowak, W.G., Tóth, L.: On the average number of subgroups of the group \({\mathbb{Z}}_{m}\times {\mathbb{Z}}_{n}\). Int. J. Number Theory 10, 363–374 (2014)
Pan, Chendong, Pan, Chenbiao: Foundations of Analytic Number Theory (in Chinese). Science Press, Beijing (1990)
Petrillo, J.: Counting subgroups in a direct product of finite cyclic groups. Coll. Math. J. 42, 215–222 (2011)
Petrogradsky, V.M.: Multiple zeta functions and asymptotic structure of free abelian groups of finite rank. J. Pure Appl. Algebra 208(3), 1137–1158 (2007)
Tărnăuceanu, M.: A new method of proving some classical theorems of abelian groups. Southeast Asian Bull. Math. 31, 1191–1203 (2007)
Tărnăuceanu, M.: An arithmetic method of counting the subgroups of a finite abelian group. Bull. Math. Soc. Sci. Math. Roumanie (N.S.) 53(101), 373–386 (2010)
Tóth, L.: On the number of cyclic subgroups of a finite Abelian group. Bull. Math. Soc. Sci. Math. Roumanie 55(103), 423–428 (2012)
Tóth, L.: Subgroups of finite Abelian groups having rank two via Goursat’s lemma. Tatra Mt. Math. Publ. 59, 93–103 (2014)
Tóth, L., Zhai, W.: On the error term concerning the number of subgroups of the groups \({\mathbb{Z}}_m \times {\mathbb{Z}}_n\) with \(m, n\le x\). Acta Arith. 183, 285–299 (2018)
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Acknowlegements
Open access funding provided by University of Pécs (PTE). This work is supported by the National Natural Science Foundation of China (Grant No. 11971476).
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Tóth, L., Zhai, W. On the average number of cyclic subgroups of the groups \({\mathbb {Z}}_{n_1} \times {\mathbb {Z}}_{n_2}\times {\mathbb {Z}}_{n_3}\) with \(n_1,n_2,n_3\le x\). Res. number theory 6, 12 (2020). https://doi.org/10.1007/s40993-020-0186-6
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s40993-020-0186-6
Keywords
- Cyclic group
- Finite abelian group of rank three
- Number of subgroups
- Number of cyclic subgroups
- Dirichlet series
- Perron formula
- Asymptotic formula
- Error term