Let OK=Z[i]. For each positive integer n, denote ξK(n) as the number of integral ideals whose norm divides n in OK. In this paper, we studied the distribution of ideals whose norm divides n in OK by using the Selberg-Delange method. This is a natural variant of a result studied by Deshouillers, Dress, and Tenenbaum (often called the DDT Theorem), and we found that the distribution function was subject to beta distribution with density √3/(2π3√u2(1−u)).
Citation: Tong Wei. The distribution of ideals whose norm divides n in the Gaussian ring[J]. AIMS Mathematics, 2024, 9(3): 5863-5876. doi: 10.3934/math.2024285
[1] | Zhishan Yang, Zongqi Yu . DDT Theorem over ideal in quadratic field. AIMS Mathematics, 2025, 10(1): 1921-1934. doi: 10.3934/math.2025089 |
[2] | Vajiheh Nazemi Niya, Hojat Babaei, Akbar Rezaei . On fuzzy sub-semi-rings of nexuses. AIMS Mathematics, 2024, 9(12): 36140-36157. doi: 10.3934/math.20241715 |
[3] | Shakir Ali, Amal S. Alali, Sharifah K. Said Husain, Vaishali Varshney . Symmetric $ n $-derivations on prime ideals with applications. AIMS Mathematics, 2023, 8(11): 27573-27588. doi: 10.3934/math.20231410 |
[4] | Yuedi Zeng . Flat modules and coherent endomorphism rings relative to some matrices. AIMS Mathematics, 2023, 8(6): 14111-14131. doi: 10.3934/math.2023721 |
[5] | C. T. J. Dodson . Information distance estimation between mixtures of multivariate Gaussians. AIMS Mathematics, 2018, 3(4): 439-447. doi: 10.3934/Math.2018.4.439 |
[6] | Heng Liu, Xia Cui . Adaptive estimation for spatially varying coefficient models. AIMS Mathematics, 2023, 8(6): 13923-13942. doi: 10.3934/math.2023713 |
[7] | Bilal A. Rather, M. Aijaz, Fawad Ali, Nabil Mlaiki, Asad Ullah . On distance signless Laplacian eigenvalues of zero divisor graph of commutative rings. AIMS Mathematics, 2022, 7(7): 12635-12649. doi: 10.3934/math.2022699 |
[8] | Yousef Alkhamees, Sami Alabiad . Classification of chain rings. AIMS Mathematics, 2022, 7(4): 5106-5116. doi: 10.3934/math.2022284 |
[9] | Shakir Ali, Amal S. Alali, Atif Ahmad Khan, Indah Emilia Wijayanti, Kok Bin Wong . XOR count and block circulant MDS matrices over finite commutative rings. AIMS Mathematics, 2024, 9(11): 30529-30547. doi: 10.3934/math.20241474 |
[10] | Yang Yang, Yanyan Song, Haifeng Fan, Haiyan Qiao . A note on the generalized Gaussian Estrada index and Gaussian subgraph centrality of graphs. AIMS Mathematics, 2025, 10(2): 2279-2294. doi: 10.3934/math.2025106 |
Let OK=Z[i]. For each positive integer n, denote ξK(n) as the number of integral ideals whose norm divides n in OK. In this paper, we studied the distribution of ideals whose norm divides n in OK by using the Selberg-Delange method. This is a natural variant of a result studied by Deshouillers, Dress, and Tenenbaum (often called the DDT Theorem), and we found that the distribution function was subject to beta distribution with density √3/(2π3√u2(1−u)).
For each positive integer n, denote by τ(n) the number of divisors of n and let Ωn={d1,d2,⋯,dτ(n)} be the set of divisors of n. Let Sn be the set of all subsets of Ωn and let μn be the uniform probability measure on Ωn:
μn(d)=1τ(n),d∈Ωn. |
It is easily verified that (Ωn,Sn,μn) is a probability space. Consider the random variable Dn:
Dn:Ωn→Rd↦logdlogn. |
The distribution function Fn of Dn is given by
Fn(t)=P(Dn≤t)=1τ(n)∑d∣n,d≤nt1(0≤t≤1). |
It is clear that the sequence {Fn}∞n=1 does not converge pointwise on [0,1] since
Fp(t)={1/2,0≤t<1;1,t=1,Fp2(t)={1/3,0≤t<1/2;2/3,1/2≤t<1;1,t=1. |
However, Deshouillers, Dress, and Tenenbaum [3] proved that its Cesˊaro means is uniformly convergent on [0,1]. No less remarkable, this limit is the distribution function of a probability law well known to specialists: the arcsine law, with density 1/(π√u(1−u)). More precisely,
1x∑n≤xFn(t)=2πarcsin√t+O(1√logx) | (1.1) |
holds uniformly for x≥2 and 0≤t≤1, and the error term in (1.1) is optimal.
Subsequently, Cui and Wu[1], Feng[6], and Feng and Wu[4] studied the related issues of the Deshouillers-Dress-Tenenbaum (DDT) theorem. Recently, Leung[9] proved that factorization of integers into k parts follows the Dirichlet distribution Dir(1k,⋯,1k) by multidimensional contour integration, thereby generalizing the DDT arcsine law on divisors where k=2. Their results were obtained in Z.
In this paper, we consider a similar problem in the Gaussian ring, unless otherwise stated, and throughout this paper K, OK, s, and σ0(τ) will be the Gaussian field, the Gaussian ring(of the form a+bi, where a,b∈Z and i2=−1), σ+iτ, and c0/log(q(|τ|+1)). For each positive integer n, let Ξn={a∈OK:N(a)dividesn}. Denoting by ξK(n) the number of ideals in Ξn, then
ξK(n)=∑N(a)∣n1=∑d∣naK(d), | (1.2) |
where aK(n) is the number of integral ideals with norm n in OK. Since aK(n) is multiplicative, so is ξK(n).
Let Sn be the set of all subsets of Ξn and let μn be the uniform probability:
μn(a)=1ξK(n),a∈Ξn. |
It is easily verified that (Ξn,Sn,μn) is a probability space. Consider the random variable Dn:
Dn:Ξn→Ra↦logN(a)logn. |
The distribution function of Dn is given by
FK,n(t)=P{Dn(a)≤t}=∑N(a)∣n,logN(a)logn≤t1ξK(n)=1ξK(n)∑N(a)∣n,N(a)≤nt1. |
It is clear that the sequence {FK,n}∞n=1 does not converge pointwise on [0,+∞), since
(1) If p≡1(mod4), then Ξp={a0,a1,a2∈Z[i]:N(a0)=1,N(a1)=N(a2)=p}, and so we have
FK,p(t)={1/3,0≤t<1;1,t≥1. |
(2) If p≡3(mod4), then Ξp={a0∈Z[i]:N(a0)=1}, and so we have FK,p(t)=1.
(3)If p=2, Ξp={a0,(1+i):N(a0)=1,N(1+i)=2}, then
FK,p(t)={1/2,p=2and0≤t<1;1,p=2andt≥1. |
However, we shall see that
GN(t):=1N∑n≤NFK,n(t) |
is uniformly convergent on [0,1], and we get the following result.
Theorem 1.1. Uniformly for x≥2 and 0≤t≤1,
1x∑n≤xFK,n(t)=B(1/3,2/3)−1∫t0u−23(1−u)−13du+O(13√logx) |
holds, where
B(a,b):=∫10ωa−1(1−ω)b−1dω,a,b>0 | (1.3) |
is beta function. So, as x→+∞, x−1∑n≤xFK,n(t) is subject to beta distribution with density √3/(2π3√u2(1−u)), since B(1/3,2/3)=Γ(1/3)Γ(2/3)/Γ(1)=2π/√3.
This distribution is not arcsince law. Feng and Wu[5] also gave a special case that satisfies the beta distribution.
In order to study x−1∑n≤xFK,n(t), we need to consider ∑n≤x1/ξK(nN(a)). Let's start with the properties of ξK(n). Dedekind defined the Dedekind zeta function of K as follows:
ζK(s)=∑a1N(a)s=∞∑n=1aK(n)ns, | (2.1) |
where a runs over all nonzero integral ideals in OK. According to [7, Theorem 2.8], we have ζK(s)=ζ(s)L(s,χ), where χ is the primitive character modulo 4. Easily, we get aK(n)=∑d∣nχ(d), so formula (1.2) can be converted to
ξK(n)=∑d∣n∑q∣dχ(q). |
When n=pm, p is prime. We have
ξK(pm)=∑d∣pmaK(d)=aK(1)+aK(p)+⋯+aK(pm)={(m+1)(m+2)2,p≡1(mod4);m+1,p=2;m+22,p≡3(mod4)and m is even;m+12,p≡3(mod4)and m is odd. |
Second, we will get the mean value of 1/ξK(nN(a)) based on the Selberg-Delange method. The method was developed by Selberg[10] and Delange[2,3]. For more details, the reader is referred to the book of Tenenbaum[11].
It's necessary to use Hankel contour when applying the method. For each value of the positive parameter r, we designate the Hankel contour as the path consisting of the circle |s|=r excluding the point s=−r and of the half-line (−∞,−r] covered twice, with respective arguments +π and −π. The brief introduction of Hankel's formula follows.
Lemma 2.1 (Hankel's formula). For each X>1, let H(X) denote the part of the Hankel contour situated in the half-plane σ>−X, then we have, uniformly for z∈C,
12πi∫H(X)s−zesds=1Γ(Z)+O(47|z|Γ(1+|z|)e−12X). |
Proof. For a detailed description of this lemma, see [11, p.179, Theorem 0.17, Corollary 0.18].
The proof of Theorem 1.1 depends on the following two lemmas.
Lemma 2.2. For any integral ideal a∈OK,
∑n≤x1ξK(nN(a))=hx3√πlogx{g(N(a))Γ(2/3)+O(C(34)ω(N(a))logx)} |
holds uniformly for x≥2, where
h=2log2∏p≡1(mod4)(1−1p)132p[(p−1)log(1−1p)+1]∏p≡3(mod4)p2log(1−1p2)−1(1−1p2)23, |
g(n)=∏pυ∥n+∞∑j=0p−jξK(pj+υ)[+∞∑j=0p−jξK(pj)]−1. |
Proof. In order to get the mean value of 1/ξK(nN(a)), we first consider its Dirichlet series ∑+∞n=1ξK(nN(a))−1n−s. Let υp(n) denote the p-adic valuation of n. By using the formula
ξK(nN(a))=∏pξK(pυp(n)+υp(N(a))), |
we write for ℜs>1:
Fa(s)=+∞∑n=11ξK(nN(a))ns=∏p+∞∑j=0p−jsξK(pj+υp(N(a)))=∏p∤N(a)+∞∑j=0p−jsξK(pj)×∏p∣N(a)+∞∑j=0p−jsξK(pj+υp(N(a)))=∏p+∞∑j=0p−jsξK(pj)×∏pυ∥N(a)+∞∑j=0p−jsξK(pj+υ)[+∞∑j=0p−jsξK(pj)]−1=L(s,χ0)23L(s,χ)−13Ga(s;23,−13), |
where χ0 is the principal character mod 4, χ is the primitive character mod 4, and
Ga(s;23,−13)=2slog(1−12s)−1∏p≡1(mod4)+∞∑j=02p−js(j+1)(j+2)(1−1ps)13×∏p≡3(mod4)+∞∑j=0p−2jsj+1(1−1p2s)23∏pυ∥N(a)+∞∑j=0p−jsξK(pj+υ)[+∞∑j=0p−jsξK(pj)]−1 |
converges absolutely for ℜs>1/2.
Let Ga(s;23,−13)=G1(s;23,−13)G2(s;23,−13)G3(s;23,−13)G4(s;23,−13)G5(s;23,−13), where
G1(s;23,−13)=∑j≥01(j+1)2js∏p≡1(mod4)[1+∑υ≥12(υ+1)(υ+2)pυs](1−1ps)13, |
G2(s;23,−13)=∏p≡3(mod4)(1−1p2s)23[1+∑j≥11(j+1)p2js], |
G3(s;23,−13)=∏pυ∥N(a),p≡1(mod4)∑j≥02p−js(j+υ+1)(j+υ+2)[1+∑j≥12p−js(j+1)(j+2)]−1, |
G4(s;23,−13)=∏p2ν∥N(a),p≡3(mod4)[∑j≥01(ν+j+1)p2js][1+∑j≥11(j+1)p2js]−1, |
G5(s;23,−13)=[∑j≥02−jsj+t+1][∑j≥02−jsj+1]−1, |
where τ(N(a))=(t+1)∏pυ∥N(a),p≡1(mod4)(υ+1)∏p2ν∥N(a),p≡3(mod4)(2ν+1).
When ℜs=σ>1/2+ε,
Ga(s;23,−13)=G1(s;23,−13)G2(s;23,−13)G3(s;23,−13)G4(s;23,−13)G5(s;23,−13)≪1t+1∏pυ∥N(a),p≡1(mod4)1(υ+1)(υ+2)∏p2ν∥N(a),p≡3(mod4)1ν+1≤C(34)ω(N(a)). |
To deal with the estimation of Fa(s) near 1, we introduce the function Z(s;z1). The function
Z(s;z1)={(s−1)L(s,χ0)}z1/s | (2.2) |
is holomorphic in the |s−1|<1, and admits the Taylor series expansion
Z(s;z1)=∞∑j=0γj(z1)j!(s−1)j |
where γj(z1) is an entire function, for all ε>0,
γj(z1)j!≪ε(1+ε)j(j≥0). |
Now, let z1=2/3. The function Z(s;2/3)Ga(s;2/3,−1/3)L(s,χ)−1/3 is holomorphic in the disc |s−1|<(1−ˆβ)/2, where ˆβ=β0 when L(s,χ) has a real zero β0, ˆβ=1−σ0(τ) when L(s,χ) has no real zero β0, and
Z(s;2/3)Ga(s;2/3,−1/3)L(s,χ)−1/3≪εM |
for |s−1|<(1−ˆβ)/2. Thus, for |s−1|<(1−ˆβ)/2, we can write
Z(s;2/3)Ga(s;2/3,−1/3)L(s,χ)−1/3=∞∑l=0gl(2/3)(s−1)l, |
where
gl(2/3):=1l!l∑j=0(lj)∂l−j(Ga(s;2/3,−1/3)L(s,χ)−1/3)∂sl−j|s=1γj(2/3). | (2.3) |
We can apply Perron's formula with the choice of parameters σa=1,A(n)=nε,α=0 to write
∑n≤x1ξK(nN(a))=12πi∫b+iTb−iTFa(s)xssds+O(x1+εT), |
where b=1+2/logx and 100≤T≤x, such that L(σ+iT,χ)≠0 for 0<σ<1.
Let LT be the boundary of the modified rectangle with vertices 1/2+ε±iT and b±iT, where
● ε>0 is a small constant chosen such that L(1/2+ε+iγ,χ)≠0 for |γ|<T. Let l1 be the horizontal line segment with the imaginary part T and the real part 1/2+ε to b, and let l2 be the horizontal line segment with the imaginary part −T and the real part b to 1/2+ε. Let l3 be the vertical line segment with the real part 1/2+ε and the imaginary part 0+ to T, and let l4 be the vertical line segment with the real part 1/2+ε and the imaginary part −T to 0−.
● The zeros of L(s,χ) of the form ρ=β+iγ with β>1/2+ε and |γ|<T are avoided by Γρ that horizontal cut drawn from the critical line inside this rectangle to ρ=β+iγ.
● L(s,χ) has a possible Siegel zero. The possible Siegel zero β0 of L(s,χ) is avoided by contour Γ0 (its upper part is made up of an arc surrounding the point s=β0 with radius r=1/logx and a line segment joining β0−r to 1/2+ε).
● The pole of L(s,χ0) at the points s=1 is avoided by the truncated Hankel contour Γ (its upper part is made up of an arc surrounding the point s=1 with radius r=1/logx and a line segment joining 1−r to ˜β), where
˜β={β0+1logx,L(β0,χ)=0;12+ε,L(β0,χ)≠0. |
Clearly the function Fa(s) is analytic inside LT. By the Cauchy residue theorem, we can write
∑n≤x1ξK(nN(a))=I+I0+I1+I2+I3+I4+∑β>1/2+ε,|γ|<TIρ+O(x1+εT) | (2.4) |
where
I:=12πi∫ΓFa(s)xssds,Iρ:=12πi∫ΓρFa(s)xssds, |
and
Ij:=12πi∫ljFa(s)xssds,I0:=12πi∫Γ0Fa(s)xssds. |
A. Evaluation of I.
Let 0<c<(1−ˆβ)/10 be a small constant. Since Z(s;2/3)Ga(s;2/3,−1/3)L(s,χ)−13 is holomorphic and O(M) in the disc |s−1|≤c, the Cauchy formula implies that
gl(2/3)≪Mc−l(l≥0), |
where gl(2/3) is defined as in (2.3). From this and (2.3), it is easy to deduce that for |s−1|≤c/2,
Z(s;2/3)Ga(s;2/3,−1/3)L(s,χ)−1/3=Z(1;2/3)Ga(1;2/3,−1/3)L(1,χ)−1/3+O(|s−1|)=h3√πg(N(a))+O(|s−1|), |
where
g(N(a))=∏pυ∥N(a)+∞∑j=0p−jξK(pj+υ)[+∞∑j=0p−jξK(pj)]−1. |
So, we have
I=12πi∫ΓFa(s)xssds=12πi∫ΓZ(s;2/3)Ga(s;2/3,−1/3)L(s,χ)−1/3(s−1)−23xsds=h3√πg(N(a))12πi∫Γ(s−1)−23xsds+O(|∫Γ(s−1)13xsds|). |
Let s−1=ω/logx. According to Lemma 2.1, we have
12πi∫Γ(s−1)−23xsds=x3√logx12πi∫H((1−˜β)logx)ω−23eωdω=x3√logx{1Γ(23)+O(1x1−˜β2)}. | (2.5) |
On the other hand,
|∫Γ(s−1)13xsds|≪∫|s−1|=1logx|(s−1)13xs||ds|+∫1−1logx˜β(1−σ)13xσdσ≪x3√logx⋅1logx+x3√logx⋅1logx∫1logx(1−˜β)t13e−tdt≪x3√logx⋅1logx. | (2.6) |
According to (2.5) and (2.6), we have
I=hx3√πlogx{g(N(a))Γ(2/3)+O(1logx)}. | (2.7) |
B. Evaluation of I1 and I2.
It is well known that
|ζ(σ+iτ)|≪(|τ|+1)(1−σ)/3log(|τ|+1)(1/2≤σ≤1+log−1|τ|,|τ|≥3). | (2.8) |
From (2.8) and [12, Lemma 2.1], we deduce that
L(s,χ0)=ζ(s)(1−12s)≪(|τ|+1)(1−σ)/3log(|τ|+1) | (2.9) |
for 1/2≤σ≤1+log−1|τ| and |τ|≥3, and
L(s,χ)−1=L(2s,χ0)−1∏p≡1(mod4)(1+1ps)−1∏p≡3(mod4)(1−1ps)−1≪(log|τ|)2/3(log2|τ|)1/3 | (2.10) |
for σ>1/2. In view of (2.9) and (2.10), we have
|I1|+|I2|≪∫1+2/logx1/2+ε|L(σ±iT,χ0)|23|L(σ±iT,χ)|−13|Ga(s,2/3,−1/3)|xσ|s|dσ≪∫1+2/logx1/2+εT29(1−σ)(logT)23(logT)29(log2T)19xσTdσ≪xTlogT∫1+2/logx1/2+ε(T29x)1−σdσ≪xTlogT. | (2.11) |
C. Evaluation of I3 and I4.
Let σ0=1/2+ε, τ0=|τ|+3, for s=σ0+iτ with 0≤|τ|≤T, In view of (2.9) and (2.10), we have
|I3|+|I4|≪∫T0|L(σ0+iτ,χ0)|23|L(σ0+iτ,χ)|−13|Ga(σ0+iτ,2/3,−1/3)|xσ0τ+1dτ≪∫T0τ29(1−σ0)0(logτ0)23(logτ0)29(log2τ0)19xσ0τ+1dτ≪x12+εT19. | (2.12) |
D. Evaluation of Iρ.
For s=σ+iγ with 1/2+ε≤σ≤β≤1−σ0(γ), we have
Fa(s)≪|γ|29(1−σ)log|γ|, |
then we deduce that
Iρ≪∫β1/2+ε|γ|29(1−σ)log|γ|xσ|γ|dσ. |
Denote by N(σ,T) the number of L(s,χ0) in the region ℜs≥σ and |ℑs|≤T. We have
∑β>1/2+ε,|γ|<T|Iρ|≪logTmaxT0≤T∑β>1/2+ε,T0/2<|γ|<T0|Iρ|≪logTmaxT0≤T∫1−σ0(T0)1/2+εT29(1−σ)0logT0xσT0N(σ,T0)dσ. |
According to Huxley[8],
N(σ,T)≪T125(1−σ)(logT)9 |
for 1/2+ε≤σ≤1, and T≥2. Thus,
∑β>1/2+ε,|γ|<T|Iρ|≪logTmaxT0≤T∫1−σ0(T0)1/2+εT29(1−σ)0logT0xσT0T125(1−σ)0(logT0)9dσ≪logTmaxT0≤T(logT0)10∫1−σ0(T0)1/2+εT29(1−σ)0x⋅xσ−1T2(1−σ)0T125(1−σ)0dσ≪xlogTmaxT0≤T(logT0)10∫1−σ0(T0)1/2+ε(T28/450x)1−σdσ≪xlogT(T28/45x)σ0(T). | (2.13) |
E. Evaluation of I0.
If L(s,χ) has no Siegel zero, then I0=0. If it has Siegel zero β0, then L(s,χ)=(s−β0)V(s), V(β0)≠0. For |s−β0|≤1/logx, we can write
V(s)−1/3L(s,χ0)2/3Ga(s;2/3,−1/3)/s=C(β0)+O(|s−β0|), |
where C(β0) is a constant depending on β0, then
|I0|=C(β0)2πi∫Γ0(s−β0)−1/3xsds+O(|∫Γ0(s−β0)2/3xsds|)≪xβ0(logx)1/3. | (2.14) |
Taking T=e√logx and inserting (2.11)–(2.14) and (2.7) into (2.4), we have
∑n≤x1ξK(nN(a))=hx3√πlogx{g(N(a))Γ(2/3)+O(C(34)ω(N(a))logx)}. |
Lemma 2.3. For any n∈Z+, we have that
∑n≤xg(n)aK(n)=3√πxh(logx)2/3{1Γ(1/3)+O(1logx)} |
holds uniformly for x≥2, where g(n) and h are defined in Lemma 2.2.
Proof. In order to get the mean value of g(n)aK(n), we first consider its Dirichlet series ∑+∞n=1g(n)aK(n)n−s. Since g(n), aK(n) is multiplicative, the Dirichlet series has Euler expansion. When ℜs>1,
F(s)=∞∑n=1g(n)aK(n)ns=L(s,χ0)1/3L(s,χ)1/3P(s;1/3,1/3), |
where
P(s;1/3,1/3)=∏p(1−χ0(p)ps)1/3(1−χ(p)ps)1/3∑υ≥0g(pυ)aK(pυ)p−υs=∑υ≥02−υs∞∑j=02−jj+υ+1[∞∑j=02−jj+1]−1×∏p≡1(mod4)(1−1ps)2/3∑υ≥0υ+1pυs∑j≥02p−j(j+υ+1)(j+υ+2)[∑j≥02p−j(j+1)(j+2)]−1×∏p≡3(mod4)(1−1p2s)1/3∑υ≥0p−2υs∑j≥0p−2jυ+j+1[∑j≥0p−2jj+1]−1 |
converges absolutely and is O(M) for ℜs>1/2. Since
∑υ≥0(1−1p)∑j≥0p(−υ−j)j+υ+1=1,∑υ≥0(1−1p2)∑j≥0p−2(j+υ)υ+j+1=1, |
(1−1p)∑υ≥0∑j≥0(υ+1)2p(−j−υ)(j+υ+1)(j+υ+2)=1, |
then
P(1;1/3,1/3)=2[∞∑j=02−jj+1]−1∏p≡1(mod4)(1−1p)−1/3[∑j≥02p−j(j+1)(j+2)]−1×∏p≡3(mod4)(1−1p2)−2/3[∑j≥0p−2jj+1]−1=2/h. |
Applying the Selberg-Delange theorem [11, p.281, Theorem 5.2], we have the formula
∑n≤xg(n)aK(n)=3√πxh(logx)2/3{1Γ(1/3)+O(1logx)}. |
We only need to consider 0≤t≤1. Now, we have
S(x,t)=1x∑n≤xFK,n(t)=1x∑n≤x1ξK(n)∑N(a)∣n,N(a)≤nt1=1x∑n≤x1ξK(n)∑N(a)∣n,N(a)≤xt1−1x∑n≤x1ξK(n)∑N(a)∣n,nt<N(a)≤xt1=:S−R. |
When 0≤t≤1/2, we will first calculate S. According to Lemma 2.2,
S=1x∑n≤x1ξK(n)∑N(a)∣n,N(a)≤xt1=1x∑N(a)≤xt∑d≤xN(a)1ξK(dN(a))=1x∑N(a)≤xt{h(xN(a))3√πlog(xN(a))[g(N(a))Γ(2/3)+O(C(34)ω(N(a))logxN(a))]}. |
Since log(x/N(a))=logx−logN(a)≥logx−logxt=(1−t)logx≥1/2logx, we have
S=h3√π∑N(a)≤xt1N(a)3√log(xN(a)){g(N(a))Γ(2/3)+O(C(34)ω(N(a))logx)}. |
Next, we calculate R. According to Lemma 2.2,
R=1x∑n≤x1ξK(n)∑N(a)∣n,nt<N(a)≤xt1=1x∑N(a)≤xt∑d≤xN(a)(dN(a))t<N(a)1ξK(dN(a))≪1x∑N(a)≤xt∑d≤xN(a)d<N(a)1−tt1ξK(d)=1x∑N(a)≤xt∑d<N(a)1−tt1ξK(d)=1x∑N(a)≤xt{hN(a)1−tt3√πlog(N(a)1−tt)[g(1)Γ(2/3)+O(1log(N(a)1−tt))]}. |
When 0≤t≤1/2, (1−t)/t≥1, and since N(a)≥2, u=log(N(a)1−tt)≥logN(a)≥log2≈0.693. Let y=(π−1)u−1. Since (π−1)log(N(a)1−tt)−1≥(π−1)log2−1>0, πlog(N(a)1−tt)≥log(N(a)1−tt)+1, R has the following estimates,
R≪1x∑N(a)≤xtN(a)1−tt√1+log(N(a)1−tt)≪1x∑N(a)≤xt(xt)1−tt3√1+log(xt×1−tt)=1x×x1−t∑N(a)≤xt13√1+logx1−t=1x×x1−t13√1+logx1−t∑N(a)≤xt1≤13√1+logx1−t≪13√logx. |
Therefore,
S(x,t)=h3√π∑N(a)≤xt1N(a)3√logxN(a)[g(N(a))Γ(2/3)+O(C(34)ω(N(a))logx)]+O(13√logx)=h3√πΓ(2/3)∑N(a)≤xtg(N(a))N(a)3√logxN(a)+O((1logx)4/3∑N(a)≤xt1N(a))+O(13√logx). |
Since
(34)ω(N(a))=O(1),∑N(a)≤xt1N(a)=∑n≤xtaK(n)n=π4logxt+π4+O(1√x), |
O((1logx)4/3∑N(a)≤xt1N(a))=O((1logx)4/3×tπ4logx)=O(13√logx), |
we have
S(x,t)=h3√πΓ(2/3)∑n≤xtg(n)aK(n)n3√logxn+O(13√logx). |
Let G(x)=∑n≤xg(n)aK(n). According to Lemma 2.3 and using the Abelian Summation formula, we have
h3√πΓ(2/3)∑n≤xtg(n)aK(n)n3√logxn=h3√πΓ(2/3)×1xt3√log(x/xt)×G(xt)+h3√πΓ(2/3)∫xt1G(u)u23√logx−logu(1−13(logx−logu))du=1Γ(2/3)∫xt11Γ(1/3)+O(1log(u+1))u3√(logu)2(logx−logu)(1−13(logx−logu))du+O(13√logx)=1Γ(2/3)∫xt11Γ(1/3)+O(1log(u+1))u3√(logu)2(logx−logu)du+O(13√logx)=1Γ(2/3)Γ(1/3)∫xt11u3√(logu)2(logx−logu)du+O(13√logx)=1Γ(2/3)Γ(1/3)∫t013√υ2(1−υ)dυ+O(13√logx), |
so we can obtain
S(x,t)=1Γ(2/3)Γ(1/3)∫t013√υ2(1−υ)dυ+O(13√logx)=√32πlog3√1−t+3√t√(1−t)2/3+t2/3−((1−t)t)1/3−32πarctan(2√33√1−t33√t−√33)+34+O(13√logx). | (3.1) |
Let
Δ(t)=√32π{2log3√1−t+3√t√(1−t)2/3+t2/3−((1−t)t)1/3−√3arctan(2√33√t33√1−t−√33)−√3arctan(2√33√1−t33√t−√33)}+32. |
Clearly, Δ(t) is symmetric with respect to t=1/2, and
S(x,t)+S(x,1−t)=Δ(t)+O(13√logx), |
then when 1/2≤t≤1, S(x,t) is the same as (3.1). This completes the proof.
The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.
The authors declare no competing interest.
[1] |
Z. Cui, J. Wu, The Selberg-Delange method in short intervals with an application, Acta Arith., 163 (2014), 247–260. http://dx.doi.org/10.4064/aa163-3-4 doi: 10.4064/aa163-3-4
![]() |
[2] | H. Delange, Sur les formules dues Atle Selberg, Bull. Sci. Math., 83 (1959), 101–111. |
[3] |
H. Delange, Sur les formules de Atle Selberg, Acta Arith., 19 (1971), 105–146. https://doi.org/10.4064/AA-19-2-105-146 doi: 10.4064/AA-19-2-105-146
![]() |
[4] |
B. Feng, J. Wu, The arcsine law on divisors in arithmetic progressions modulo prime powers, Acta Math. Hungar., 163 (2021), 392–406. https://doi.org/10.1007/s10474-020-01105-7 doi: 10.1007/s10474-020-01105-7
![]() |
[5] | B. Feng, J. Wu, β-law on divisors of integers representable as sum of two squares, in Chinese, Sci. China Math., 49 (2019), 1563–1572. |
[6] |
B. Feng, On the arcsine law on divisors in arithmetic progressions, Indagat. Math., 27 (2016), 749–763. https://doi.org/10.1016/j.indag.2016.01.008 doi: 10.1016/j.indag.2016.01.008
![]() |
[7] | Keqin Feng, Algebraic Number Theory, in Chinese, Beijing: Science Press, 2000. |
[8] |
M. N. Huxley, The difference between consecutive primes, Invent. Math., 15 (1972), 164–170. https://doi.org/10.1007/BF01418933 doi: 10.1007/BF01418933
![]() |
[9] | S. K. Leung, Dirichlet law for factorization of integers, polynomials and permutations, preprint paper, 2022. |
[10] |
A. Selberg, Note on the paper by L. G. Sathe, J. Indian Math. Soc., 18 (1954), 83–87. https://doi.org/10.18311/JIMS2F19542F17018 doi: 10.18311/JIMS2F19542F17018
![]() |
[11] | G. Tenenbaum, Introduction to Analytic and Probabilistic Number Theory, 3 Eds, Cambridge: Cambridge University Press, 1995. |
[12] |
J. Wu, Q. Wu, Mean values for a class of arithmetic functions in short intervals, Math. Nachr., 293 (2020), 178–202. https://doi.org/10.1002/mana.201800276 doi: 10.1002/mana.201800276
![]() |