This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Asymptotic formula for the multiplicative function d(n)kω(n)\frac{d(n)}{k^{\omega(n)}}

Meselem Karras Faculty of Science and Technology Djilali Bounaama Khemis Miliana University, 44225, Algeria karras.m@hotmail.fr
Abstract

For a fixed integer kk, we define the multiplicative function

Dk,ω(n):=d(n)kω(n),D_{k,\omega}(n):=\frac{d(n)}{k^{\omega(n)}},

where d(n)d(n) is the divisor function and ω(n)\omega(n) is the number of distinct prime divisors of nn. The main purpose of this paper is the study of the mean value of the function Dk,ω(n)D_{k,\omega}(n) by using elementary methods.

keywords:
Divisor function, number of distinct prime divisors, mean value.
\msc

11N37, 11A25, 11N36 \VOLUME31 \YEAR2023 \NUMBER1 \DOIhttps://doi.org/10.46298/cm.10104 {paper}

1 Introduction

Let k2k\geq 2 be a fixed integer. We recall that d(n):=dn1d(n):=\sum_{d\mid n}1 is the number of divisors of nn, and ω(n):=pn1\omega(n):=\sum_{p\mid n}1 is the number of distinct prime divisors of nn. We define the function Dk,ω(n)D_{k,\omega}(n) by

Dk,ω(n):=d(n)kω(n).D_{k,\omega}(n):=\frac{d(n)}{k^{\omega(n)}}. (1)

Notice that for every fixed integer k2k\geq 2, the function Dk,ω(n)D_{k,\omega}(n) is multiplicative and for every prime number pp and every integer mm the relation

Dk,ω(pm)=m+1k,D_{k,\omega}(p^{m})=\frac{m+1}{k}, (2)

holds. By using (2), we get

Dk,ω(n)=pmnm+1kD_{k,\omega}(n)=\prod\limits_{p^{m}\|n}\frac{m+1}{k}

where pmnp^{m}\|n means pmp^{m}\mid nn and pm+1np^{m+1}\nmid n. In the particular case k=2k=2, the function D2,ω(n)D_{2,\omega}(n) is exactly D(n)=d(n)d(n),(see [D.K])D(n)=\dfrac{d(n)}{d^{\ast}(n)},(\text{see~{}\cite[cite]{[\@@bibref{}{D.K}{}{}]}}). For k3k\geq 3, we can easily check that

nxDk,ω(n)kx(logx)2/k1.\sum_{n\leq x}D_{k},_{\omega}(n)\ll_{k}x(\log x)^{2/k-1}. (3)

Indeed, for any integer nn, we have Dk,ω(n)d(n)εnεD_{k},_{\omega}(n)\leq d(n)\ll_{\varepsilon}n^{\varepsilon}. Furthermore, the hypotheses of Shiu’s theorem are satisfied; see Theorem 11 in [Shui] and [Nar-Ten, p.1]. One gets

nxDk,ω(n)kxlogxexp(px2kp).\sum_{n\leq x}D_{k},_{\omega}(n)\ll_{k}\frac{x}{\log x}\exp\Bigl{(}\sum_{p\leq x}\frac{2}{kp}\Bigr{)}.

Now, by using Lemma 4.634.63 in [O.BORD], it follows that

nxDk,ω(n)kxlogxexp(2klog(2eγlogx))kx(logx)2/k1.\sum_{n\leq x}D_{k},_{\omega}(n)\ll_{k}\frac{x}{\log x}\exp\Bigl{(}\frac{2}{k}\log(2e^{\gamma}\log x)\Bigr{)}\ll_{k}x(\log x)^{2/k-1}.

2 Main result

In this section, we establish two results concerning the mean value of the function Dk,ω(n)D_{k,\omega}(n). We begin by giving a weaker result.

Theorem 2.1.

Let k2k\geq 2 be a fixed integer. For all x1x\geq 1 large enough, we have

nxDk,ω(n)=x(logx)2/k1Γ(2/k)p(11p)2/k(1+2p1kp(p1)2)+O(x(logx)1)(loglogx)4/k.\sum_{n\leq x}D_{k,\omega}(n)=\frac{x(\log x)^{2/k-1}}{\Gamma(2/k)}\prod\limits_{p}\Bigl{(}1-\frac{1}{p}\Bigr{)}^{2/k}\Bigl{(}1+\frac{2p-1}{kp(p-1)^{2}}\Bigr{)}+O(x(\log x)^{-1})(\log\log x)^{4/k}.

The proof of this result is based on Tulyaganov’s theorem; this theorem is summarized as follows:

Theorem 2.2.

Let ff be a complex valued multiplicative function. Suppose there exists zz\in\mathbb{C}, independent of pp, with |z|c1|z|\leq c_{1} and

  1. a)
    pxf(p)logp=zx+O(xec2logx)\sum_{p\leq x}f(p)\log p=zx+O(xe^{-c_{2}\sqrt{\log x}})
  2. b)
    px|f(p)|logpx\sum_{p\leq x}|f(p)|\log p\ll x
  3. c)
    pxα=2|f(pα)|logpαpα(loglogx)2\sum_{p\leq x}\sum_{\alpha=2}^{\infty}\frac{|f(p^{\alpha})|\log p^{\alpha}}{p^{\alpha}}\ll(\log\log x)^{2}
  4. d)
    p|f(p)|2logpp2<c3\sum_{p}\frac{|f(p)|^{2}\log p}{p^{2}}<c_{3}

for some real numbers c1c_{1}, c2c_{2} and c3c_{3}. Then, for all x1x\geq 1 sufficiently large, we have

nxf(n)=\displaystyle\sum_{n\leq x}f(n)={} x(logx)z1Γ(z)p(11p)z(1+α=1f(pα)pα){1+O((loglogx)2logx)}\displaystyle{}\frac{x(\log x)^{z-1}}{\Gamma(z)}\prod\limits_{p}\Bigl{(}1-\frac{1}{p}\Bigr{)}^{z}\Bigl{(}1+\sum_{\alpha=1}^{\infty}\frac{f(p^{\alpha})}{p^{\alpha}}\Bigr{)}\Bigl{\{}1+O(\frac{(\log\log x)^{2}}{\log x})\Bigr{\}}
+O(x(logx)max(0,Rez1)1)(loglogx)2(Amax(0,Rez1)),\displaystyle{}+O(x(\log x)^{\max(0,\mathrm{Re}\,z-1)-1})(\log\log x)^{2(A-\max(0,\mathrm{Re}\,z-1))},

where A>0A>0 satisfies

u<pv|f(p)|p1Alog(logv/logu)+O(1).\sum_{u<p\leq v}|f(p)|p^{-1}\leq A\log(\log v/\log u)+O(1).
Proof 2.3.

This theorem is a consequence of Theorem 44 in [Tulyag], where we take g=fg=f.

To complete the demonstration of the main result we have the following lemmas.

Lemma 2.4.

For any fixed integer k2k\geq 2, we have the estimate

px|Dk,ω(p)|logpx.\sum_{p\leq x}|D_{k,\omega}(p)|\log p\ll x.
Proof 2.5.

By Chebyshev’s estimates [Disar], we have

px|Dk,ω(p)|logp=2kpxlogp<2k(1.000081x)x.\sum_{p\leq x}|D_{k,\omega}(p)|\log p=\frac{2}{k}\sum_{p\leq x}\log p<\frac{2}{k}(1.000081x)\ll x.
Lemma 2.6.

For any fixed integer k2k\geq 2, there is a constant c>0c>0, such that

pxDk,ω(p)logp=2kx+O(xeclogx).\sum_{p\leq x}D_{k,\omega}(p)\log p=\frac{{\footnotesize 2}}{k}x+O(xe^{-c\sqrt{\log x}}).
Proof 2.7.

We have

pxDk,ω(p)logp=2kpxlogp=2kθ(x),\sum_{p\leq x}D_{k,\omega}(p)\log p=\frac{{\footnotesize 2}}{k}\sum_{p\leq x}\log p=\frac{{\footnotesize 2}}{k}\theta(x),

and by Theorem 6.96.9 in [H.L.MandR.C.V], there is a constant c>0c>0 such that

θ(x)=x+O(xeclogx),\theta(x)=x+O(xe^{-c\sqrt{\log x}}),

which implies the desired result.

Lemma 2.8.

For any fixed integer k2k\geq 2, we have

p|Dk,ω(p)|2p2logp<.\sum_{p}\frac{|D_{k,\omega}(p)|^{2}}{p^{2}}\log p<\infty.
Proof 2.9.

We first check the inequality m=2logmm(m1)log4\sum\limits_{m=2}^{\infty}\dfrac{\log m}{m(m-1)}\leq\log 4, and using the following

plogpp2<m=2logmm2m=2logmm(m1),\sum_{p}\frac{\log p}{p^{2}}<\sum_{m=2}^{\infty}\frac{\log m}{m^{2}}\leq\sum_{m=2}^{\infty}\frac{\log m}{m(m-1)},

then we have

p|Dk,ω(p)|2p2logp=4k2plogpp2<4log4k2.\sum_{p}\frac{|D_{k,\omega}(p)|^{2}}{p^{2}}\log p=\frac{4}{k^{2}}\sum_{p}\frac{\log p}{p^{2}}<\frac{4\log 4}{k^{2}}.
Lemma 2.10.

For any fixed integer k2k\geq 2, we have

pxα=2|Dk,ω(pα)|log(pα)pα28k.\sum_{p\leq x}\sum_{\alpha=2}^{\infty}\frac{|D_{k,\omega}(p^{\alpha})|\log(p^{\alpha})}{p^{\alpha}}\leq\frac{28}{k}.
Proof 2.11.

For every integer k3k\geq 3, we write

pxα=2|Dk,ω(pα)|log(pα)pα\displaystyle\sum_{p\leq x}\sum_{\alpha=2}^{\infty}\frac{|D_{k,\omega}(p^{\alpha})|\log(p^{\alpha})}{p^{\alpha}} =1kpxlogpα=2α(α+1)pα\displaystyle=\frac{1}{k}\sum_{p\leq x}\log p\sum_{\alpha=2}^{\infty}\frac{\alpha(\alpha+1)}{p^{\alpha}}
=1kpxlogppα=2α(α+1)pα1,\displaystyle=\frac{1}{k}\sum_{p\leq x}\frac{\log p}{p}\sum_{\alpha=2}^{\infty}\frac{\alpha(\alpha+1)}{p^{\alpha-1}},

and the infinite series α=2α(α+1)pα1\sum\limits_{\alpha=2}^{\infty}\dfrac{\alpha(\alpha+1)}{p^{\alpha-1}} converges to 2(11/p)32\dfrac{2}{(1-1/p)^{3}}-2, since

pxα=2|Dk,ω(pα)|log(pα)pα\displaystyle\sum_{p\leq x}\sum_{\alpha=2}^{\infty}\frac{|D_{k,\omega}(p^{\alpha})|\log(p^{\alpha})}{p^{\alpha}} =2kpx3p23p+1p(p1)3logp\displaystyle=\frac{2}{k}\sum_{p\leq x}\frac{3p^{2}-3p+1}{p(p-1)^{3}}\log p
28kpxlogpp2.\displaystyle\leq\frac{28}{k}\sum_{p\leq x}\frac{\log p}{p^{2}}.

By Lemma 70.170.1 in [HALL-TENENBAUM], we have plogppα<1α1\sum\limits_{p}\dfrac{\log p}{p^{\alpha}}<\dfrac{1}{\alpha-1} for all α>1\alpha>1, consequently

pxα=2|Dk,ω(pα)|log(pα)pα<28k.\sum_{p\leq x}\sum_{\alpha=2}^{\infty}\frac{|D_{k,\omega}(p^{\alpha})|\log(p^{\alpha})}{p^{\alpha}}<\frac{28}{k}.

Finally, by Lemma 2.4, 2.6, 2.8 and 2.10 we have shown that the function Dk,ω(n)D_{k,\omega}(n) satisfies the conditions of Theorem 2.2. As we have

u<pv|Dk,ω(p)|p=2ku<pv1p2kloglogvlogu+O(1),\sum_{u<p\leq v}\frac{|D_{k,\omega}(p)|}{p}=\frac{2}{k}\sum_{u<p\leq v}\frac{1}{p}\leq\frac{2}{k}\log\frac{\log v}{\log u}+O(1),

then the constant AA in Theorem 22 is 2k\frac{2}{k}.

The next result is improved over the previous one.

Theorem 2.12.

Let k2k\geq 2 be a fixed integer. For all x1x\geq 1 large enough, we have

nxDk,ω(n)=x(logx)2/k1Γ(2/k)p(11p)2/k(1+2p1k(p1)2)+Ok(x(logx)2/k2).\sum_{n\leq x}D_{k,\omega}(n)=\frac{x(\log x)^{2/k-1}}{\Gamma(2/k)}\prod\limits_{p}\Bigl{(}1-\frac{1}{p}\Bigr{)}^{2/k}\Bigl{(}1+\frac{2p-1}{k(p-1)^{2}}\Bigr{)}+O_{k}(x(\log x)^{2/k-2}).

The demonstration is based on the following lemmas:

Lemma 2.13.

Let k2k\geq 2 be a fixed integer. For every s:=σ+its:=\sigma+it\in\mathbb{C} such that σ>1\sigma>1 and L(s,Dk,ω(n)):=n=1Dk,ω(n)nsL(s,D_{k,\omega}(n)):=\sum\limits_{n=1}^{\infty}\dfrac{D_{k,\omega}(n)}{n^{s}}, we have

L(s,Dk,ω(n))=ζ(s)2/kL(s,gk),L(s,D_{k,\omega}(n))=\zeta(s)^{2/k}L(s,g_{k}),

or L(s,gk)L(s,g_{k}) is a series of Dirichlet absolutely convergent in the half-plane σ>12\sigma>\frac{1}{2}.

Proof 2.14.

If σ>1\sigma>1, then

L(s,Dk,ω(n))\displaystyle L(s,D_{k,\omega}(n)) =p(1+α=1Dk,ω(pα)pαs)\displaystyle=\prod\limits_{p}\Bigl{(}1+\sum\limits_{\alpha=1}^{\infty}\dfrac{D_{k,\omega}(p^{\alpha})}{p^{\alpha s}}\Bigr{)}
=p(1+α=1α+1kpαs)\displaystyle=\prod\limits_{p}\Bigl{(}1+\sum\limits_{\alpha=1}^{\infty}\dfrac{\alpha+1}{kp^{\alpha s}}\Bigr{)}
=p(1+2ps1k(ps1)2),\displaystyle=\prod\limits_{p}\Bigl{(}1+\dfrac{2p^{s}-1}{k(p^{s}-1)^{2}}\Bigr{)},

on the other hand we have

(1+2ps1k(ps1)2)=((1ps)2/k)(1+h(s)k(ps1)2),\Bigl{(}1+\frac{2p^{s}-1}{k(p^{s}-1)^{2}}\Bigr{)}=((1-p^{-s})^{-2/k})\Bigl{(}1+\frac{h(s)}{k(p^{s}-1)^{2}}\Bigr{)},

such that

h(s)=(1ps)2/k(kp2s2(k1)ps+k1)k(ps1)2.h(s)=(1-p^{-s})^{2/k}(kp^{2s}-2(k-1)p^{s}+k-1)-k(p^{s}-1)^{2}.

Since

(1ps)2/k=12kpsk2k2p2sO(kp3σ),(1-p^{-s})^{2/k}=1-\frac{2}{kp^{s}}-\frac{k-2}{k^{2}p^{2s}}-O\Bigl{(}\frac{k}{p^{3\sigma}}\Bigr{)},

he comes

h(s)\displaystyle h(s) =(12kpsk2k2p2sO(kp3σ))(kp2s2(k1)ps+k1)k(ps1)2\displaystyle=\Bigl{(}1-\frac{2}{kp^{s}}-\frac{k-2}{k^{2}p^{2s}}-O\bigl{(}\frac{k}{p^{3\sigma}}\bigr{)}\Bigr{)}(kp^{2s}-2(k-1)p^{s}+k-1)-k(p^{s}-1)^{2}
=2(11k)+O(pσ),\displaystyle=2\bigl{(}1-\frac{1}{k}\bigr{)}+O(p^{-\sigma}),

which implies the announced result.

Lemma 2.15 ([Selb]).

Let A>0A>0. Uniformly for x2x\geq 2 and zz\in\mathbb{C} such that |z|A|z|\leq A, we have

nxτz(n)=x(logx)z1Γ(z)+OA(x(logx)Rez2).\sum_{n\leq x}\tau_{z}(n)=\frac{x(\log x)^{z-1}}{\Gamma(z)}+O_{A}(x(\log x)^{\mathrm{Re}\,z-2}).

τz(n)\tau_{z}(n) is the multiplicative function defined by τz(pα)=(z+α1α)\tau_{z}(p^{\alpha})=\binom{z+\alpha-1}{\alpha}.

Proof 2.16 (Proof of Theorem 3).

According to the Lemma 2.13, we have Dk,ω=τ2/kgkD_{k,\omega}=\tau_{2/k}\ast g_{k}. Then, by Lemma 2.15

nxDk,ω(n)\displaystyle\sum_{n\leq x}D_{k,\omega}(n) =dxgk(d)mxdτ2/k(m)\displaystyle=\sum_{d\leq x}g_{k}(d)\sum_{m\leq\frac{x}{d}}\tau_{2/k}(m)
=dxgk(d)(x(logxd)2/k1dΓ(2/k)+Ok(xd(logxd)2/k2))\displaystyle=\sum_{d\leq x}g_{k}(d)\Bigl{(}\frac{x(\log\frac{x}{d})^{2/k-1}}{d\Gamma(2/k)}+O_{k}\bigl{(}\frac{x}{d}(\log\frac{x}{d})^{2/k-2}\bigr{)}\Bigr{)}
=dxgk(d)(x(logx)2/k1dΓ(2/k)+Ok((logx)2/k2logd)\displaystyle=\sum_{d\leq x}g_{k}(d)\Bigl{(}\frac{x(\log x)^{2/k-1}}{d\Gamma(2/k)}+O_{k}((\log x)^{2/k-2}\log d)
+Ok(xd(logxd)2/k2))\displaystyle\qquad+O_{k}\bigl{(}\frac{x}{d}(\log\frac{x}{d})^{2/k-2}\bigr{)}\Bigr{)}
=x(logx)2/k1Γ(2/k)dxgk(d)d+Ok(x(logx)2/k2dx|gk(d)|(1+logd)d).\displaystyle=\frac{x(\log x)^{2/k-1}}{\Gamma(2/k)}\sum_{d\leq x}\frac{g_{k}(d)}{d}+O_{k}\Bigl{(}x(\log x)^{2/k-2}\sum_{d\leq x}\frac{|g_{k}(d)|(1+\log d)}{d}\Bigr{)}.

The series L(s,gk)L(s,g_{k}) is absolutely convergent on the half-plane σ>12\sigma>\frac{1}{2}, then for all ε>0\varepsilon>0

dx|gk(d)|k,εx1/2+ε,\sum_{d\leq x}|g_{k}(d)|\ll_{k,\varepsilon}x^{1/2+\varepsilon},

hence by partial summation

dx|gk(d)|(1+logd)dk,εx1/2+ε\sum_{d\leq x}\frac{|g_{k}(d)|(1+\log d)}{d}\ll_{k,\varepsilon}x^{-1/2+\varepsilon}

and therefore

nxDk,ω(n)=L(1,gk)x(logx)2/k1Γ(2/k)+Ok(x(logx)2/k2)+Ok,ω(x1/2+ε).\sum_{n\leq x}D_{k,\omega}(n)=L(1,g_{k})\frac{x(\log x)^{2/k-1}}{\Gamma(2/k)}+O_{k}(x(\log x)^{2/k-2})+O_{k,\omega}(x^{1/2+\varepsilon}).

Which completes the demonstration.

Acknowledgments

The author would like to sincerely thank Professor Olivier Bordellès for his help and interest in this work and Professor Karl Dilcher for his generosity in reviewing this paper.

References

  • \referBookO.BORD \RauthorBordellès, O. \RtitleArithmetic Tales \RpublisherSpringer \Ryear2012
  • [1] \referPaperD.K \RauthorDerbal, A.; Karras, M. \RtitleValeurs moyennes d’une fonction liée aux diviseurs d’un nombre entier, Comptes Rendus Mathématique \Rvolume354 \Ryear2016 \Rpages555-558
  • [2] \referPaperDisar \RauthorDusart, P. \RtitleSharper bounds for ψ\psi, θ\theta, π\pi, pnp_{n} \RjournalRapport de recherche, Université de Limoges \Ryear1998-2006
  • [3] \referBookHALL -TENENBAUM \RauthorHall, R.R.; Tenenbaum, G. \RtitleDivisors \RpublisherCambridge University Press \Ryear1988
  • [4] \referBookH.L.M and R.C.V \RauthorMontgomery, H.L; Vaughan, R.C. \RtitleMultiplicative Number Theory I. Classical Theory \RpublisherCambridge Studies in Advanced Mathematics \Ryear2007
  • [5] \referPaperNar-Ten \RauthorNair, M.; Tenenbaum, G. \RtitleShort sums of certain arithmetic functions \RjournalActa Math. \Rvolume180 \Ryear1998 \Rpages119-144
  • [6] \referPaperShui \RauthorShiu, P. \RtitleA Brun-Titchmarsh theorem for multiplicative functions \RjournalJ. Reine Angew. Math. \Rvolume313 \Ryear1980 \Rpages161-170
  • [7] \referPaperSelb \RauthorSelberg, A. \RtitleNote on a paper by L. G. Sathe \RjournalJ. Indian Math. Soc. \Rvolume18 \Ryear1954 \Rpages83-87
  • [8] \referProceedingsTulyag \RauthorTulyaganov, S.T. \RtitleOn the summation of multiplicative arithmetical functions \RjournalNumber theory. Vol. I. Elementary and analytic, Proc. Conf., Budapest/Hung. \RpublisherColl. Math. Soc. János Bolyai \Rvolume51 \Ryear1990 \Rpages539-573
  • [9]
\EditInfo

29 September, 201928 January, 2020Karl Dilcher