A logarithmic analogue of Alladi's formula
Extracted source text · arXiv 2504.16002
%Version 3.1 December 2024
% See section 11 of the User Manual for version history
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% %%
%% Please do not use \input{...} to include other tex files. %%
%% Submit your LaTeX manuscript as one .tex document. %%
%% %%
%% All additional figures and files should be attached %%
%% separately and not embedded in the \TeX\ document itself. %%
%% %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%\documentclass[referee,sn-basic]{sn-jnl}% referee option is meant for double line spacing
%%=======================================================%%
%% to print line numbers in the margin use lineno option %%
%%=======================================================%%
%%\documentclass[lineno,pdflatex,sn-basic]{sn-jnl}% Basic Springer Nature Reference Style/Chemistry Reference Style
%%=========================================================================================%%
%% the documentclass is set to pdflatex as default. You can delete it if not appropriate. %%
%%=========================================================================================%%
%%\documentclass[sn-basic]{sn-jnl}% Basic Springer Nature Reference Style/Chemistry Reference Style
%%Note: the following reference styles support Namedate and Numbered referencing. By default the style follows the most common style. To switch between the options you can add or remove Numbered in the optional parenthesis.
%%The option is available for: sn-basic.bst, sn-chicago.bst%
%%\documentclass[pdflatex,sn-nature]{sn-jnl}% Style for submissions to Nature Portfolio journals
%%\documentclass[pdflatex,sn-basic]{sn-jnl}% Basic Springer Nature Reference Style/Chemistry Reference Style
\documentclass[pdflatex,sn-mathphys-num]{sn-jnl}% Math and Physical Sciences Numbered Reference Style
%%\documentclass[pdflatex,sn-mathphys-ay]{sn-jnl}% Math and Physical Sciences Author Year Reference Style
%%\documentclass[pdflatex,sn-aps]{sn-jnl}% American Physical Society (APS) Reference Style
%%\documentclass[pdflatex,sn-vancouver-num]{sn-jnl}% Vancouver Numbered Reference Style
%%\documentclass[pdflatex,sn-vancouver-ay]{sn-jnl}% Vancouver Author Year Reference Style
%%\documentclass[pdflatex,sn-apa]{sn-jnl}% APA Reference Style
%%\documentclass[pdflatex,sn-chicago]{sn-jnl}% Chicago-based Humanities Reference Style
%%%% Standard Packages
%%<additional latex packages if required can be included here>
\usepackage{graphicx}%
\usepackage{multirow}%
\usepackage{amsmath,amssymb,amsfonts}%
\usepackage{amsthm}%
\usepackage{mathrsfs}%
\usepackage[title]{appendix}%
\usepackage{xcolor}%
\usepackage{textcomp}%
\usepackage{manyfoot}%
\usepackage{booktabs}%
\usepackage{algorithm}%
\usepackage{algorithmicx}%
\usepackage{algpseudocode}%
\usepackage{listings}%
%%%%
\usepackage{relsize}
\usepackage{colonequals}
%% as per the requirement new theorem styles can be included as shown below
\theoremstyle{thmstyleone}%
\newtheorem{theorem}{Theorem}% meant for continuous numbers
%%\newtheorem{theorem}{Theorem}[section]% meant for sectionwise numbers
%% optional argument [theorem] produces theorem numbering sequence instead of independent numbers for Proposition
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{proposition}[theorem]{Proposition}%
%%\newtheorem{proposition}{Proposition}% to get separate numbers for theorem and proposition etc.
\theoremstyle{thmstyletwo}%
\newtheorem{example}{Example}%
\newtheorem{remark}{Remark}%
\theoremstyle{thmstylethree}%
\newtheorem{definition}{Definition}%
\raggedbottom
%%\unnumbered% uncomment this for unnumbered level heads
%Shortcuts of commands --------------------
\newcommand{\N}{\mathbb{N}}
\newcommand{\Z}{\mathbb{Z}}
\newcommand{\Q}{\mathbb{Q}}
\newcommand{\R}{\mathbb{R}}
\newcommand{\C}{\mathbb{C}}
\newcommand{\sS}{\mathscr{S}}
\newcommand{\sF}{\mathscr{F}}
\newcommand{\sH}{\mathscr{H}}
\newcommand{\ve}{\varepsilon}
\newcommand{\on}{\operatorname}
\renewcommand{\mod}[1]{\,\on{mod}#1}
\newcommand{\of}[1]{\left(#1\right)}
\newcommand{\set}[1]{\left\{#1\right\}}
\newcommand{\BEu}[1]{\underset{#1}{\mathlarger{\mathlarger{\mathbb{E}}}^{~}}\,}
\newcommand{\BEul}[1]{\underset{#1}{\mathlarger{\mathlarger{\mathbb{E}}}^\log}\,}
\newcommand{\BEulk}[1]{\underset{#1}{\mathlarger{\mathlarger{\mathbb{E}}}^{\log^{k-1}}}\,}
\newcommand{\abs}[1]{\left|#1\right|}
%------------End of the authours' settings----------------------
\begin{document}
\title[A logarithmic analogue of Alladi's formula]{A logarithmic analogue of Alladi's formula}
%%=============================================================%%
%% GivenName -> \fnm{Joergen W.}
%% Particle -> \spfx{van der} -> surname prefix
%% FamilyName -> \sur{Ploeg}
%% Suffix -> \sfx{IV}
%% \author*[1,2]{\fnm{Joergen W.} \spfx{van der} \sur{Ploeg}
%% \sfx{IV}}\email{iauthor@gmail.com}
%%=============================================================%%
\author{\fnm{Biao} \sur{Wang}}\email{bwang@ynu.edu.cn}
\affil{\orgdiv{School of Mathematics and Statistics}, \orgname{Yunnan University}, \orgaddress{\city{Kunming}, \postcode{650500}, \state{Yunnan}, \country{China}}}
%%==================================%%
%% Sample for unstructured abstract %%
%%==================================%%
\abstract{Let $\mu(n)$ be the M\"{o}bius function. Let $P^-(n)$ denote the smallest prime factor of an integer $n$. In 1977, Alladi established the following formula related to the prime number theorem for arithmetic progressions
\[
-\sum_{\substack{n\geq 2\\ P^-(n)\equiv \ell ({\rm mod}k)}}\frac{\mu(n)}{n}=\frac1{\varphi(k)}
\]
for positive integers $\ell, k\ge1$ with $(\ell,k)=1$, where $\varphi$ is Euler's totient function. In this note, we will show a logarithmic analogue of Alladi's formula in an elementary proof.}
%%================================%%
%% Sample for structured abstract %%
%%================================%%
% \abstract{\textbf{Purpose:} The abstract serves both as a general introduction to the topic and as a brief, non-technical summary of the main results and their implications. The abstract must not include subheadings (unless expressly permitted in the journal's Instructions to Authors), equations or citations. As a guide the abstract should not exceed 200 words. Most journals do not set a hard limit however authors are advised to check the author instructions for the journal they are submitting to.
%
% \textbf{Methods:} The abstract serves both as a general introduction to the topic and as a brief, non-technical summary of the main results and their implications. The abstract must not include subheadings (unless expressly permitted in the journal's Instructions to Authors), equations or citations. As a guide the abstract should not exceed 200 words. Most journals do not set a hard limit however authors are advised to check the author instructions for the journal they are submitting to.
%
% \textbf{Results:} The abstract serves both as a general introduction to the topic and as a brief, non-technical summary of the main results and their implications. The abstract must not include subheadings (unless expressly permitted in the journal's Instructions to Authors), equations or citations. As a guide the abstract should not exceed 200 words. Most journals do not set a hard limit however authors are advised to check the author instructions for the journal they are submitting to.
%
% \textbf{Conclusion:} The abstract serves both as a general introduction to the topic and as a brief, non-technical summary of the main results and their implications. The abstract must not include subheadings (unless expressly permitted in the journal's Instructions to Authors), equations or citations. As a guide the abstract should not exceed 200 words. Most journals do not set a hard limit however authors are advised to check the author instructions for the journal they are submitting to.}
\keywords{prime number theorem, M\"obius function, prime factors}
%%\pacs[JEL Classification]{D8, H51}
\pacs[MSC Classification]{11N13, 11N37}
\maketitle
\section{Introduction}
Let $\mu(n)$ be the M\"{o}bius function defined by
\[
\mu(n)=\begin{cases}
1 & \text{if } n=1,\\
0 & \text{if } p^2\mid n \text{ for some prime } p,\\
(-1)^r & \text{if } n=p_1\cdots p_r \text{ where the } p_i \text{ are distinct primes}.
\end{cases}
\]
Then it is well-known (e.g., \cite{Mangoldt1897}) that the prime number theorem (PNT) is equivalent to the assertion that
\begin{equation}\label{PNT}
\sum_{n=1}^\infty\frac{\mu(n)}{n}=0.
\end{equation}
By reordering \eqref{PNT}, it is equivalent to asserting that $-\sum_{n=2}^\infty\frac{\mu(n)}{n}=1$, which can be interpreted as the statement that the probability of integers $n\ge2$ being divisible by a prime is $1$. Let $P^-(n)$ and $P^+(n)$ be the smallest prime factor of $n$ and the largest prime factor of $n$ for $n\ge2$, respectively. Let $P^-(1)=P^+(1)=1$. In 1977, Alladi \cite{Alladi1977} introduced a duality between $P^-(n)$ and $P^+(n)$, showing that for any function $f$ defined on integers with $f(1)=0$, we have
\begin{align}
\sum_{d|n}\mu(d)f(P^-(d))&=-f(P^+(n)),\label{duality1}\\
\sum_{d|n}\mu(d)f(P^+(d))&=-f(P^-(n)).\label{duality2}
\end{align}
Let $k,\ell \geq 1$ be integers with $(\ell,k)=1$. Then applying the PNT for arithmetic progressions with \eqref{duality1}, Alladi \cite{Alladi1977} established that
\begin{equation}\label{alladi}
-\sum_{\substack{n\geq 2\\ P^-(n)\equiv \ell (\on{mod}k)}}\frac{\mu(n)}{n}=\frac1{\varphi(k)},
\end{equation}
where $\varphi$ is Euler's totient function.
In the last few years, Alladi's formula \eqref{alladi} has been greatly generalized in various settings in number theory, see \cite{Dawsey2017, OnoSchneiderWagner2017, SweetingWoo2019, KuralMcDonaldSah2020, Wang2021jnt, OnoSchneiderWagner2021, DuanWangYi2021, DuanMaYi2022}. Recently, Alladi and Johnson \cite{AlladiJohnson2024} proved that
\begin{equation}\label{AlladiJohnson2024}
\sum_{\substack{n\geq 2\\ P^-(n)\equiv \ell (\on{mod}k)}}{\mu(n)\omega(n)\over n}=0,
\end{equation}
where $\omega(n)$ is the number of distinct prime factors of $n$. Later, \eqref{AlladiJohnson2024} is generalized by Tenenbaum \cite{Tenenbaum2024} and Sengupta \cite{Sengupta2025}. In 1899, Landau \cite{Landau1899}\footnote{See \cite{Coons2010} for an English translation by Michael J. Coons.} proved in his doctoral dissertation that the PNT is also equivalent to the following logarithmic analogue of \eqref{PNT}:
\begin{equation}\label{log_PNT}
\lim_{x\to\infty} \frac{1}{\log x} \sum_{n\leq x}\frac{\mu(n)\log n}{n}=0.
\end{equation}
Inspired by \eqref{AlladiJohnson2024} and \eqref{log_PNT}, we will show a logarithmic analogue of Alladi's formula in this note.
Let $S$ be a set of primes. We say $S$ has a \textit{natural density} if the following limit $$\lim_{x\to\infty} \frac{\#\set{p\in S: p\leq x}}{\#\set{p \text{ prime}: p\leq x}}$$ exists. The following theorem is our main result.
\begin{theorem}\label{mainthm_logAlladi}
If $S$ is a set of primes of natural density, then we have
\begin{equation}\label{logAlladi_min}
\lim_{x\to \infty}\frac{1}{\log x}\sum_{\substack{n\leq x\\P^-(n)\in S}} \frac{\mu(n)\log n }{n}=0.
\end{equation}
Moreover,
\begin{equation}\label{logAlladi_max}
\lim_{x\to \infty}\frac{1}{\log x}\sum_{\substack{n\leq x\\P^+(n)\in S}} \frac{\mu(n)\log n }{n}=0.
\end{equation}
\end{theorem}
Taking $S$ to be the set of all primes, then we recover \eqref{log_PNT} from \eqref{logAlladi_min} or \eqref{logAlladi_max}. Hence Theorem~\ref{mainthm_logAlladi} is a refinement of the equivalent form \eqref{log_PNT} of the PNT. If we take $S=\set{p \text{ prime}: p\equiv \ell (\on{mod}k)}$ for $(\ell,k)=1$, then we get the following analogue of \eqref{AlladiJohnson2024}:
\begin{equation}
\lim_{x\to \infty}\frac{1}{\log x}\sum_{\substack{n\leq x\\P^-(n)\equiv \ell (\on{mod}k)}} \frac{\mu(n)\log n }{n}=0.
\end{equation}
To prove Theorem~\ref{mainthm_logAlladi}, we will mainly show a logarithmic analogue of a duality theorem obtained by Alladi \cite{Alladi1977} in the next section.
\section{Duality between $P^+(n)$ and $P^-(n)$}
Based on the duality identity \eqref{duality1}, Alladi established the following duality relationship between the smallest and largest prime factors of integers.
\begin{theorem}[{\cite[Theorem 6]{Alladi1977}}] \label{mainthmmuf}
Let $f:\N\to \C$ be a bounded function. Let $\delta$ be a real number. Then
\[
\lim_{x\to\infty}\frac1x\sum_{n\leq x}f(P^+(n))= \delta
\]
if and only if
\[
-\lim_{x\to\infty} \sum_{2\leq n\leq x} \frac{\mu(n)f(P^-(n))}{n}=\delta.
\]
\end{theorem}
Indeed, to obtain Theorem~\ref{mainthmmuf}, Alladi showed that for any function $f$ defined on integers with $f(1)=0$,
\begin{equation}\label{eqn_alladi3.8}
\sum_{n\leq x} f(P^+(n))= -x \sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n}+o(x).
\end{equation}
If $B$ is a finite nonempty set, for any function $a:B\to\C$ on $B$, we define the \textit{Ces\`{a}ro average} of $a$ over $B$ and the \textit{logarithmic average} of $a$ over $B$ respectively by
$$
\BEu{n\in B} a(n) \,\colonequals\, \frac{1}{|B|}\sum_{n\in B} a(n)\qquad\text{and }\qquad
\BEul{n\in B} a(n) \,\colonequals\, \frac{\sum_{n\in B} {a(n)}/{n}}{\sum_{n\in B}{1}/{n}}.
$$
In the following theorem, we establish a logarithmic analogue of Theorem~\ref{mainthmmuf}.
\begin{theorem}\label{mainthm_log}
Let $f:\N\to \C$ be a bounded function with $f(1)=0$. Then
\begin{equation}
\lim_{x\to \infty} \Big(\BEul{1\leq n\leq x} f(P^+(n))- \BEu{1\leq n\leq x} f(P^+(n))\Big)=0
\end{equation}
if and only if
\begin{equation}\label{log_alldi}
\lim_{x\to \infty}\frac{1}{\log x}\sum_{n\leq x} \frac{\mu(n)\log n }{n}f(P^-(n))=0.
\end{equation}
\end{theorem}
\begin{proof}
Let $h(x)=\sum_{n\leq x} \frac1n$. On the one hand, by \eqref{duality1} we have
\begin{align}
\sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} h\big(\frac{x}n\big)&= \sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} \sum_{m\leq x/n} \frac1m \nonumber\\
&= \sum_{mn\leq x} \frac{\mu(n)f(P^-(n))}{mn} \nonumber\\
&= \sum_{c\leq x} \frac1c \sum_{n\mid c} \mu(n)f(P^-(n)) \nonumber\\
&= -\sum_{c\leq x} \frac{f(P^+(c))}c. \label{eqn_pf1}
\end{align}
Then by $h(x)=\log x+\gamma+O(1/x)$ where $\gamma$ is Euler's constant, we have
\begin{multline}
\sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} h\big(\frac{x}n\big)= \sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} \Big(\log\big(\frac{x}n\big)+\gamma+ O\big(\frac{n}x\big)\Big) \\
= \log x\sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} - \sum_{n\leq x} \frac{\mu(n)\log n }{n}f(P^-(n)) + \gamma\sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n}+ O(1).\label{eqn_2nd_way}
\end{multline}
By \eqref{eqn_alladi3.8}, the third term in \eqref{eqn_2nd_way} is bounded, since
\begin{equation}\label{Alladi_cor}
\sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n}=O(1).
\end{equation}
It follows that
\begin{equation}\label{eqn_pf2}
\sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} h\big(\frac{x}n\big)\\
= \log x\sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} - \sum_{n\leq x} \frac{\mu(n)\log n }{n}f(P^-(n)) + O(1).
\end{equation}
Combining \eqref{eqn_pf1}
and \eqref{eqn_pf2} together and using \eqref{eqn_alladi3.8} again, we get that
\begin{align}
\frac{1}{\log x}\sum_{n\leq x} \frac{\mu(n)\log n }{n}f(P^-(n)) &= \sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} + \frac{1}{\log x} \sum_{n\leq x} \frac{f(P^+(n))}n +O\Big(\frac{1}{\log x}\Big) \nonumber\\
&=\frac{1}{\log x} \sum_{n\leq x} \frac{f(P^+(n))}n -\frac1x\sum_{n\leq x} f(P^+(n))+o(1), \label{log_alldi_equiv_form}
\end{align}
which completes the proof of Theorem~\ref{mainthm_log}.
\end{proof}
\section{Proof of Theorem~\ref{mainthm_logAlladi}}
To prove \eqref{logAlladi_min} in Theorem~\ref{mainthm_logAlladi}, we cite a result of Kural et al. on the distribution of the largest prime factors of integers.
\begin{theorem}[{\cite[Theorem 3.1]{KuralMcDonaldSah2020}}]
\label{kms}
Let $S$ be a set of primes of natural density $\delta(S)$, then we have
\[
\lim_{x\to\infty}\frac{1}{x}\sum_{\substack{1\leq n\leq x\\ P^+(n)\in S}}1 =\delta(S).
\]
\end{theorem}
\begin{proof}[Proof of \eqref{logAlladi_min}]
Suppose the natural density of $S$ over primes is $\delta(S)$. Take $f(n)$ in \eqref{log_alldi} to be the characteristic function of $S$ defined by
\[
f(n)=
\begin{cases}
1 & \text{if } n\in S;\\
0 & \text{if } n\notin S.
\end{cases}
\]
Then by Theorem~\ref{kms}, we have
\[
\lim_{x\to\infty}\BEu{1\leq n\leq x} f(P^+(n))=\lim_{x\to\infty}\frac{1}{x}\sum_{\substack{1\leq n\leq x\\ P^+(n)\in S}}1 =\delta(S).
\]
That is, the natural density of the set $\set{n\in \N: P^+(n)\in S}$ exists. It follows that the logarithmic density of the set $\set{n\in \N: P^+(n)\in S}$ also exists and
\[
\lim_{x\to\infty}\BEul{1\leq n\leq x} f(P^+(n))= \delta(S).
\]
Thus, by Theorem~\ref{mainthm_log} we have
\[
\lim_{x\to \infty}\frac{1}{\log x}\sum_{n\leq x} \frac{\mu(n)\log n }{n}f(P^-(n))=0,
\]
which gives \eqref{logAlladi_min}.
\end{proof}
The proof of \eqref{logAlladi_max} is similar to that of \eqref{logAlladi_min}.
\begin{proof}[Proof of \eqref{logAlladi_max}]
By \cite[(9.15)-(9.16)]{Alladi1982}, we have an estimate similar to \eqref{eqn_alladi3.8} that
\begin{equation}\label{eqn_alladi3.8_max}
\sum_{n\leq x} f(P^-(n))= -x \sum_{n\leq x} \frac{\mu(n)f(P^+(n))}{n}+o(x).
\end{equation}
Then similar to the argument in the proof of Theorem~\ref{mainthm_log}, by \eqref{duality2} and \eqref{eqn_alladi3.8_max} one can obtain that
\begin{equation}\label{max_key_eqn}
\frac{1}{\log x}\sum_{n\leq x} \frac{\mu(n)\log n }{n}f(P^+(n)) = \frac{1}{\log x} \sum_{n\leq x} \frac{f(P^-(n))}n -\frac1x\sum_{n\leq x} f(P^-(n))+o(1).
\end{equation}
Take $f(n)$ to be the characteristic function of $S$. By \cite[(9.15)]{Alladi1982}, the following limit exists:
\begin{equation}\label{nd_min}
\lim_{x\to\infty}\BEu{1\leq n\leq x} f(P^-(n))=\lim_{x\to\infty}\frac{1}{x}\sum_{\substack{1\leq n\leq x\\ P^-(n)\in S}}1,
\end{equation}
which is equal to the natural density of the set $\set{n\in \N: P^-(n)\in S}$. It follows that the logarithmic density of the set $\set{n\in \N: P^-(n)\in S}$ also exists and
\begin{equation}\label{ld_min}
\lim_{x\to\infty}\BEul{1\leq n\leq x} f(P^-(n))=\lim_{x\to\infty}\BEu{1\leq n\leq x} f(P^-(n)).
\end{equation}
Thus, \eqref{logAlladi_max} follows immediately by \eqref{max_key_eqn} and \eqref{ld_min}.
\end{proof}
\section{The case on the higher power of logarithm}
In this section, we consider the logarithmic analogue of the PNT with respect to the higher power of logarithm. Let $k\ge1$ be an integer. Write $M(x)\colonequals \sum_{n\leq x} \frac{\mu(n)}{n}$. Then by partial summation, we have
\begin{equation}\label{referee_comment}
\sum_{n\leq x} \frac{\mu(n)\log^k n}{n} = \log^k(x) M(x) - k\int_1^x \frac{M(t)\log^{k-1} t}{t} dt.
\end{equation}
Since the integral in the right-hand side of \eqref{referee_comment} converges, we have that
\begin{equation}\label{log_pnt_k}
\lim_{x\to\infty} \frac1{\log^k x}\sum_{n\leq x} \frac{\mu(n)\log^k n}{n} =0
\end{equation}
holds if and only if
\begin{equation}\label{pnt_mu}
\lim_{x\to\infty} M(x)=0.
\end{equation}
This implies that for any integer $k\ge1$, the PNT is equivalent to the assertion that \eqref{log_pnt_k} holds. In this section, we will show that the refinement of \eqref{log_pnt_k} under the same restriction $P^{-}(n)\in S$ or $P^{+}(n)\in S$ as Theorem~\ref{mainthm_logAlladi} still holds. We are deeply grateful to the referee for introducing the equivalent form \eqref{log_pnt_k} of the PNT and the idea on deriving the equivalence between \eqref{log_pnt_k} and \eqref{pnt_mu} from the identity \eqref{referee_comment} to us, and for suggesting we generalize previous results to the case on the higher power of logarithm.
In the following, we will follow the spirit of Landau's thesis \cite{Landau1899} to give a proof of the equivalence between \eqref{log_pnt_k} and \eqref{pnt_mu} in an elementary way. For an integer $k\ge 2$, let $d_k(n)$ be the $k$th divisor function definded by
\[
d_k(n)\colonequals\#\set{(a_1,\dots, a_k)\in \Z^k: a_1\cdots a_k=n, a_1,\dots, a_k \geq1}.
\]
For the case $k=2$, $d_2(n)$ is the classical divisor function. For convenience, we let $d_1(n)=1$ for all $n\ge1$, and $d_0(n)= 1$ if $n=1$ and zero otherwise. Then for any $k\ge1$, we have
\begin{equation}\label{eqn_power_reduced}
\mu\ast d_k =d_{k-1},
\end{equation}
where the operator $\ast$ denotes the Dirichlet convolution. On the distribution of $d_k(n)$, a well-known elementary theorem states that there exists a polynomial $P_{k}(x)$ of degree $k-1$ and leading coefficient $1/(k-1)!$ such that
\begin{equation}
\sum_{n\leq x} d_k(n)= x P_{k}(\log x) + O(x^{1-\frac1k}\log^{k-2}x)
\end{equation}
for $k\ge2$, e.g., see \cite[Chapter XII]{Titchmarsh1986}.
Then by partial summation, there exists a polynomial $Q_k(x)$ of degree $k$ and leading coefficient $1/k!$ such that
\begin{equation}\label{divisor_fcn_logaverage}
\sum_{n\leq x} \frac{d_k(n)}n= Q_k(\log x) + O(x^{-\frac1k}\log^{k-2}x).
\end{equation}
\subsection{The higher power logarithmic PNT}
By \cite[Lemmas 1 and 2]{Mangoldt1897}, we have
\begin{equation}
\Big|\sum_{n\leq x} \frac{\mu(n)}{n} \Big| \leq 1
\end{equation}
and
\begin{equation}\label{thm_Mangoldt}
\Big|\log x\sum_{n\leq x} \frac{\mu(n)}{n} - \sum_{n\leq x} \frac{\mu(n)\log n}{n} \Big| \leq 3+\gamma
\end{equation}
for all $x\ge1$. This implies that
\begin{align}
& \sum_{n\leq x} \frac{\mu(n)}{n}=O(1), \\
& \sum_{n\leq x} \frac{\mu(n)}{n} \log(\frac{x}{n})=O(1). \label{thm_Mangoldt_1}
\end{align}
In the next theorem, we will apply \eqref{eqn_power_reduced} and \eqref{divisor_fcn_logaverage} to generalize \eqref{thm_Mangoldt_1} to the summation with higher power of logarithm.
\begin{proposition}\label{thm_Mangoldt_k}
For any integer $k\ge1$, we have
\begin{equation}\label{thm_Mangoldt_k_estimate}
\sum_{n\leq x} \frac{\mu(n)}{n} \log^k(\frac{x}{n}) = O(\log^{k-1}x).
\end{equation}
\end{proposition}
\begin{proof}
We prove \eqref{thm_Mangoldt_k_estimate} by doing induction on $k$. For $k=1$, \eqref{thm_Mangoldt_k_estimate} follows by \eqref{thm_Mangoldt_1}. Let $k\geq2$, and suppose \eqref{thm_Mangoldt_k_estimate} holds for the cases $1,\dots, k-1$. First, by \eqref{eqn_power_reduced} and \eqref{divisor_fcn_logaverage} we observe that
\begin{multline}\label{thm_Mangoldt_k_pfeqn1}
\sum_{n\leq x} \frac{\mu(n)}{n} \sum_{m\leq \frac{x}{n}} \frac{d_k(m)}{m} = \sum_{mn\leq x} \frac{ \mu(n) d_k(m) }{mn} = \sum_{c\leq x} \frac1c \sum_{n\mid c} \mu(n)d_k(\frac{c}{n}) = \sum_{c\leq x} \frac{d_{k-1}(c)}{c} \\= O(\log^{k-1} x).
\end{multline}
Suppose $Q_k(x)=\sum_{i=0}^kc_ix^i, c_i\in \R$ and $c_k=1/k!$. Then by assumption we have
$$\sum_{n\leq x} \frac{\mu(n)}{n} \log^i(\frac{x}{n}) = O(\log^{i-1}x), i=1,\dots, k-1.$$
It follows that
\begin{align}
\sum_{n\leq x} \frac{\mu(n)}{n} \sum_{m\leq \frac{x}{n}} \frac{d_k(m)}{m} & = \frac{1}{k!} \sum_{n\leq x} \frac{\mu(n)}{n} \log^k(\frac{x}{n}) + \sum_{i=1}^{k-1}c_i \sum_{n\leq x} \frac{\mu(n)}{n} \log^i(\frac{x}{n}) \nonumber\\
&\quad + c_0 \sum_{n\leq x} \frac{\mu(n)}{n} + O\of{\sum_{n\leq x} \frac{1}{n} (\frac{n}{x})^{\frac1k}\log^{k-2}(\frac{x}{n})} \nonumber\\
&= \frac{1}{k!} \sum_{n\leq x} \frac{\mu(n)}{n} \log^k(\frac{x}{n}) + O(\log^{k-2} x). \label{thm_Mangoldt_k_pfeqn2}
\end{align}
Combing \eqref{thm_Mangoldt_k_pfeqn1} and \eqref{thm_Mangoldt_k_pfeqn2} together, we obtain that
\begin{equation}
\sum_{n\leq x} \frac{\mu(n)}{n} \log^k(\frac{x}{n}) = O(\log^{k-1}x).
\end{equation}
Thus, by induction \eqref{thm_Mangoldt_k_estimate} holds for all $k\ge1$.
\end{proof}
\begin{remark}
In \cite{Shapiro1950}, Shapiro gave an asymptotic formula for \eqref{thm_Mangoldt_k_estimate}. One may apply the method in the proof of Proposition~\ref{thm_Mangoldt_k} to obtain an asymptotic estimate of \eqref{thm_Mangoldt_k_estimate} as well.
\end{remark}
Now, similar to \eqref{thm_Mangoldt}, we build a relationship between $\sum_{n\leq x} \frac{\mu(n)\log^kn}{n}$ and $\log^k x\sum_{n\leq x} \frac{\mu(n)}{n}$ by doing induction on $k$ for $k\ge1$.
\begin{proposition}\label{thm_Landau_k}
For any $k\ge1$, we have
\begin{equation}\label{thm_Landau_k_eqn}
\sum_{n\leq x} \frac{\mu(n)\log^kn}{n} = \log^k x \sum_{n\leq x} \frac{\mu(n)}{n} + O(\log^{k-1}x).
\end{equation}
\end{proposition}
\begin{proof}
For $k=1$, \eqref{thm_Landau_k_eqn} follows by \eqref{thm_Mangoldt_1}. Let $k\geq2$, and suppose \eqref{thm_Landau_k_eqn} holds for the cases $1,\dots, k-1$. Then
\begin{align}
&\quad \sum_{n\leq x} \frac{\mu(n)}{n} \log^k(\frac{x}{n}) \nonumber\\
&= \sum_{n\leq x} \frac{\mu(n)}{n}(\log x-\log n)^k = \sum_{i=0}^k \binom{k}{i} (-1)^i \log^{k-i}x \sum_{n\leq x} \frac{\mu(n)\log^i n}{n} \nonumber\\
&= \log^k x \sum_{n\leq x} \frac{\mu(n)}{n} + (-1)^k \sum_{n\leq x} \frac{\mu(n)\log^kn}{n} + \sum_{i=1}^{k-1} \binom{k}{i} (-1)^i \log^{k-i}x \sum_{n\leq x} \frac{\mu(n)\log^i n}{n}\nonumber\\
&= \log^k x \sum_{n\leq x} \frac{\mu(n)}{n} + (-1)^k \sum_{n\leq x} \frac{\mu(n)\log^kn}{n} \nonumber\\
&\quad + \sum_{i=1}^{k-1} \binom{k}{i} (-1)^i \log^{k-i}x \Big(\log^i x \sum_{n\leq x} \frac{\mu(n)}{n} + O(\log^{i-1}x)\Big) \nonumber\\
&= \Big(\sum_{i=0}^{k-1} \binom{k}{i} (-1)^i \Big) \cdot \log^k x \sum_{n\leq x} \frac{\mu(n)}{n} + (-1)^k \sum_{n\leq x} \frac{\mu(n)\log^kn}{n} + O(\log^{k-1}x) \nonumber\\
&= (-1)^{k+1} \log^k x \sum_{n\leq x} \frac{\mu(n)}{n} + (-1)^k \sum_{n\leq x} \frac{\mu(n)\log^kn}{n} + O(\log^{k-1}x).\label{thm_Landau_k_pf1}
\end{align}
By Proposition~\ref{thm_Mangoldt_k}, we have
\begin{equation}\label{thm_Landau_k_pf2}
\sum_{n\leq x} \frac{\mu(n)}{n} \log^k(\frac{x}{n}) = O(\log^{k-1}x).
\end{equation}
Then combing \eqref{thm_Landau_k_pf1} and \eqref{thm_Landau_k_pf2} gives us that
\begin{equation}
\sum_{n\leq x} \frac{\mu(n)\log^kn}{n} = \log^k x \sum_{n\leq x} \frac{\mu(n)}{n} + O(\log^{k-1}x).
\end{equation}
Thus, by induction \eqref{thm_Landau_k_eqn} holds for all $k\ge1$.
\end{proof}
By Proposition~\ref{thm_Landau_k}, we find that
\begin{equation}\label{thm_Landau_k_form2}
\frac{1}{\log^k x}\sum_{n\leq x} \frac{\mu(n)\log^kn}{n} = \sum_{n\leq x} \frac{\mu(n)}{n} + O\Big(\frac{1}{\log x}\Big).
\end{equation}
From \eqref{thm_Landau_k_form2}, we conclude that \eqref{log_pnt_k} is equivalent to \eqref{pnt_mu}, which is equivalent to the PNT.
\begin{theorem}
Let $k\ge1$ be any given integer. Then the prime number theorem is equivalent to the assertion that
\begin{equation}\label{log_pnt_k_rep}
\lim_{x\to\infty}\frac1{\log^k x}\sum_{n\leq x} \frac{\mu(n)\log^kn}{n} = 0.
\end{equation}
\end{theorem}
Now, we show that under the same restriction $P^{-}(n)\in S$ or $P^{+}(n)\in S$ as Theorem~\ref{mainthm_logAlladi}, the refinement of \eqref{log_pnt_k_rep} holds for any $k\ge1$. This result can be obtained from the following lemma.
\begin{lemma}\label{lem_logweight}
If an arithmetic function $a:\N\to\C$ satisfies that
\begin{equation}
\sum_{n\leq x} a(n) = o(\log x),
\end{equation}
then for any $k\ge2$, we have
\begin{equation}\label{lem_logweight_eqn}
\sum_{n\leq x} a(n)\log^{k-1} n = o(\log^k x).
\end{equation}
\end{lemma}
\begin{proof}
Let $A(x)=\sum_{n\leq x} a(n)$. Then by partial summation, we have
\begin{equation}
\sum_{n\leq x} a(n)\log^{k-1} n = A(x) \log^{k-1} x- (k-1) \int_1^x \frac{A(t)\log^{k-2} t}{t}dt.
\end{equation}
For any $\ve>0$, since $A(x)=o(\log x)$, there exists some $x_1>1$ such that $|A(x)|\leq \ve \log x$ for all $x\geq x_1$. For such $x_1$, there exists some $x_2>1$ such that
\begin{equation}\label{mainthm_logAlladi_k_pf1}
\Big|\int_1^{x_1} \frac{A(t)\log^{k-2} t}{t}dt \Big|\leq \frac{\ve}{k-1} \log^k x
\end{equation}
for all $x\ge x_2$. Let $x_0=\max\set{x_1,x_2}$. Then for all $x\ge x_0$, we have \eqref{mainthm_logAlladi_k_pf1} and
\begin{equation}\label{mainthm_logAlladi_k_pf2}
\Big|\int_{x_1}^x \frac{A(t)\log^{k-2} t}{t}dt \Big|\leq \int_{x_1}^x \frac{|A(t)|\log^{k-2} t}{t}dt \leq \ve \int_{x_1}^x \frac{\log^{k-1} t}{t}dt < \frac{\ve}{k} \log^k x.
\end{equation}
It follows by the triangle inequality that
\begin{equation}
\Big|\sum_{n\leq x} a(n)\log^{k-1} n \Big|\leq 3\ve \log^k x
\end{equation}
for all $x\ge x_0$. Thus, \eqref{lem_logweight_eqn} holds for any $k\ge2$.
\end{proof}
\begin{theorem}\label{mainthm_logAlladi_k}
Let $k\ge1$ be an integer. If $S$ is a set of primes of natural density, then we have
\begin{equation}\label{logAlladi_min_k}
\lim_{x\to \infty}\frac{1}{\log^k x}\sum_{\substack{n\leq x\\P^-(n)\in S}} \frac{\mu(n)\log^k n }{n}=0.
\end{equation}
Moreover,
\begin{equation}\label{logAlladi_max_k}
\lim_{x\to \infty}\frac{1}{\log ^kx}\sum_{\substack{n\leq x\\P^+(n)\in S}} \frac{\mu(n)\log^k n }{n}=0.
\end{equation}
\end{theorem}
\begin{proof}
We only show that \eqref{logAlladi_min_k} holds, since the proof of \eqref{logAlladi_max_k} is similar. By Theorem~\ref{mainthm_logAlladi}, \eqref{logAlladi_min_k} holds for $k=1$. For $k\ge2$, we take
$$a(n)=\frac{\mu(n)\log(n) 1_{P^-(n)\in S}}{n},$$
where
\[
1_{P^-(n)\in S}=
\begin{cases}
1 & \text{if } P^-(n)\in S;\\
0 & \text{if } P^-(n)\notin S.
\end{cases}
\]
Then by Theorem~\ref{mainthm_logAlladi}, $\sum_{n\leq x}a(n)=o(\log x)$. Notice that
\begin{equation}
a(n)\log^{k-1}n=\frac{\mu(n)\log^k (n) 1_{P^-(n)\in S} }{n}
\end{equation}
and
\begin{equation}
\sum_{n\leq x}a(n)\log^{k-1}n = \sum_{\substack{n\leq x\\P^-(n)\in S}} \frac{\mu(n)\log^k n }{n}.
\end{equation}
Thus, using Lemma~\ref{lem_logweight}, we get \eqref{logAlladi_min_k}, as desired.
\end{proof}
\subsection{Higher power logarithmic duality}
In this last section, we generalize the logarithmic duality Theorem~\ref{mainthm_log} to the higher power case. We will use the idea in the proof of Proposition~\ref{thm_Mangoldt_k}. We first generalize Alladi's duality identitis \eqref{duality1} and \eqref{duality2} to a form involving the divisor functions.
\begin{lemma}
Let $f: \N \to\C$ be a function with $f(1)=0$. Then for any $k\geq1$, we have
\begin{align}
[\mu f(P^-(\cdot))]\ast d_k &= - f(P^+(\cdot)) \ast d_{k-1}, \label{duality1_k} \\
[\mu f(P^+(\cdot))]\ast d_k &= - f(P^-(\cdot)) \ast d_{k-1}.\label{duality2_k}
\end{align}
That is,
\begin{align}
\sum_{d\mid n} \mu(d)f(P^-(d)) d_k(\frac{n}{d}) & = -\sum_{d\mid n} f(P^+(d)) d_{k-1}(\frac{n}{d}), \\
\sum_{d\mid n} \mu(d)f(P^+(d)) d_k(\frac{n}{d}) & = -\sum_{d\mid n} f(P^-(d)) d_{k-1}(\frac{n}{d})
\end{align}
for all $n\ge1$.
\end{lemma}
\begin{proof}
We only show that \eqref{duality1_k} holds, since the proof of \eqref{duality2_k} is similar.
For $k=1$, \eqref{duality1_k} follows by \eqref{duality1}. That is,
\begin{equation}
\mu f(P^-(\cdot))]\ast d_1 = - f(P^+(\cdot)) \ast d_{0} = - f(P^+(\cdot)).
\end{equation}
Let $k\ge2$. Notice that $d_k= d_1\ast d_{k-1}$ for all $k\ge2$, then
$$[\mu f(P^-(\cdot))]\ast d_k = [\mu f(P^-(\cdot))]\ast (d_1\ast d_{k-1}) = ([\mu f(P^-(\cdot))]\ast d_1) \ast d_{k-1} = - f(P^+(\cdot)) \ast d_{k-1},$$
which gives \eqref{duality1_k}.
\end{proof}
Let $k\ge1$. For any function $a:B\to\C$ on a finite nonempty set $B$, we define the $(k-1)$th \textit{logarithmic average} of $a$ over $B$ by
$$
\BEulk{n\in B} a(n) \,\colonequals\, \frac{\sum_{n\in B} {a(n)\log^{k-1}n}/{n}}{\sum_{n\in B}{\log^{k-1} n}/{n}}.
$$
Now, we show the higher power logarithmic analogue of Alladi's duality theorem~\ref{mainthmmuf} in the following theorem. Similar to the proof of Theorem~\ref{mainthm_logAlladi}, one can also derive Theorem~\ref{mainthm_logAlladi_k} from Theorem~\ref{mainthm_log_k}.
\begin{theorem}\label{mainthm_log_k}
Let $k\ge1$. Let $f:\N\to \C$ be a bounded function with $f(1)=0$. Then
\begin{equation}\label{diff_k}
\lim_{x\to \infty} \Big(\BEulk{1\leq n\leq x} f(P^+(n))- \BEu{1\leq n\leq x} f(P^+(n))\Big)=0
\end{equation}
if and only if
\begin{equation}\label{log_alldi_k}
\lim_{x\to \infty}\frac{1}{\log^k x}\sum_{n\leq x} \frac{\mu(n)\log^k n }{n}f(P^-(n))=0.
\end{equation}
Moreover,
\begin{equation}
\lim_{x\to \infty} \Big(\BEulk{1\leq n\leq x} f(P^-(n))- \BEu{1\leq n\leq x} f(P^-(n))\Big)=0
\end{equation}
if and only if
\begin{equation}
\lim_{x\to \infty}\frac{1}{\log^k x}\sum_{n\leq x} \frac{\mu(n)\log^k n }{n}f(P^+(n))=0.
\end{equation}
\end{theorem}
\begin{proof}
We only prove the first equivalence, since the following argument works as well if one switches $P^-(n)$ and $P^+(n)$. To prove that \eqref{diff_k} is equivalent to \eqref{log_alldi_k}, it suffices to show that
\begin{equation}\label{general_kth_duality}
\sum_{n\leq x} \frac{\mu(n)f(P^-(n)) \log^k n }{n}= k\sum_{n\leq x} \frac{f(P^+(n)) \log^{k-1} n }{n} - \frac{\log^k x}{x} \sum_{n\leq x} f(P^+(n)) +o(\log^k x)
\end{equation}
holds for all $k\ge1$. We do induction on $k$. For $k=1$, \eqref{general_kth_duality} follows by \eqref{log_alldi_equiv_form}. Let $k\ge2$, and suppose \eqref{general_kth_duality} holds for the cases $1,\dots, k-1$. Then by \eqref{general_kth_duality} and \eqref{Alladi_cor}, we have
\begin{equation}
\sum_{n\leq x} \frac{\mu(n)f(P^-(n)) \log^i n }{n} =O(\log^i x)
\end{equation}
for all $i=0, 1, \dots, k-1$. It follows by using the binomial theorem that
\begin{equation}\label{log_alldi_k_pf1}
\sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} \log^i (\frac{x}{n})=O(\log^i x)
\end{equation}
for all $i=0, 1, \dots, k-1$. Similar to \eqref{thm_Mangoldt_k_pfeqn1} in the proof of Propostion~\ref{thm_Mangoldt_k}, we observe that
\begin{align}
\sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} \sum_{m\leq \frac{x}{n}} \frac{d_k(m)}{m} & =\sum_{mn\leq x} \frac{\mu(n)f(P^-(n))d_k(m)}{mn} \nonumber\\
& = \sum_{c\leq x} \frac1c \sum_{n\mid c}\mu(n)f(P^-(n))d_k(\frac{c}{n}) \nonumber\\
& = - \sum_{c\leq x} \frac1c \sum_{n\mid c} f(P^+(n))d_{k-1}(\frac{c}{n}) \nonumber\\
& = - \sum_{n\leq x} \frac{f(P^+(n))}{n} \sum_{m\leq \frac{x}{n}} \frac{d_{k-1}(m)}{m} \nonumber\\
&=- \frac1{(k-1)!}\sum_{n\leq x} \frac{f(P^+(n))}{n} \log^{k-1}(\frac{x}{n}) + O(\log^{k-1} x). \label{log_alldi_k_total}
\end{align}
Then similar to \eqref{thm_Mangoldt_k_pfeqn2} in the proof of Propostion~\ref{thm_Mangoldt_k}, using \eqref{divisor_fcn_logaverage} and \eqref{log_alldi_k_pf1} we get that
\begin{equation}\label{log_alldi_k_total_1st_term}
\sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} \sum_{m\leq \frac{x}{n}} \frac{d_k(m)}{m} =\frac1{k!}\sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} \log^k (\frac{x}{n}) + O(\log^{k-1} x).
\end{equation}
Now, expanding out the first term of \eqref{log_alldi_k_total_1st_term}, using the assumption \eqref{general_kth_duality} for the cases $1,\dots, k-1$, we get that
\begin{align}
&\quad \sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} \log^k (\frac{x}{n}) \nonumber\\
& = \sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} (\log x -\log n)^k = \sum_{i=0}^k \binom{k}{i} (-1)^i \log^{k-i}x \sum_{n\leq x} \frac{\mu(n)f(P^-(n)) \log^i n }{n} \nonumber\\
& = \log^k x \sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} + (-1)^k \sum_{n\leq x} \frac{\mu(n)f(P^-(n)) \log^k n }{n} \nonumber\\
& \quad + \sum_{i=1}^{k-1} \binom{k}{i} (-1)^i \log^{k-i}x \Big(i\sum_{n\leq x} \frac{f(P^+(n)) \log^{i-1} n }{n} - \frac{\log^i x}{x} \sum_{n\leq x} f(P^+(n)) +o(\log^i x)\Big)\nonumber\\
& = - \frac{\log^k x}{x} \sum_{n\leq x} f(P^+(n)) + (-1)^k \sum_{n\leq x} \frac{\mu(n)f(P^-(n)) \log^k n }{n} \nonumber\\
&\quad + \sum_{n\leq x} \frac{f(P^+(n))}{n} \sum_{i=1}^{k-1} \binom{k}{i} (-1)^i {i} \log^{k-i}x \log^{i-1} n \nonumber\\
&\quad - \sum_{i=1}^{k-1} \binom{k}{i} (-1)^i \frac{\log^k x}{x} \sum_{n\leq x} f(P^+(n) + o(\log^k x) \nonumber\\
&= (-1)^k \frac{\log^k x}{x} \sum_{n\leq x} f(P^+(n)) + (-1)^k \sum_{n\leq x} \frac{\mu(n)f(P^-(n)) \log^k n }{n} \nonumber\\
& \quad -k\sum_{n\leq x} \frac{f(P^+(n))}{n} \log^{k-1}(\frac{x}{n}) - (-1)^kk \sum_{n\leq x} \frac{f(P^+(n))\log^{k-1} n}{n} + o(\log^k x). \label{log_alldi_k_LHS_0}
\end{align}
In the last line of \eqref{log_alldi_k_LHS_0}, we have used the following two identities
\begin{align}
\sum_{i=1}^{k-1} \binom{k}{i} (-1)^i & =-1-(-1)^k, \\
\sum_{i=1}^{k-1} \binom{k}{i} (-1)^i {i} a^{k-i}t^{i-1}&= -k(a-t)^{k-1} -(-1)^k k t^{k-1}, \label{binomial_thm_derivative}
\end{align}
where $a=\log x, t=\log n$. And \eqref{binomial_thm_derivative} can be obtained by taking derivative on $t$ in the binomial expansion of $(a-t)^k$.
Dividing both sides of \eqref{log_alldi_k_LHS_0} by $k!$, we get that
\begin{multline}
\frac1{k!}\sum_{n\leq x} \frac{\mu(n)f(P^-(n))}{n} \log^k (\frac{x}{n}) = \frac{(-1)^k}{k!} \Big( \sum_{n\leq x} \frac{\mu(n)f(P^-(n)) \log^k n }{n} + \frac{\log^k x}{x} \sum_{n\leq x} f(P^+(n)) \\
-k \sum_{n\leq x} \frac{f(P^+(n))\log^{k-1} n}{n}\Big)-\frac1{(k-1)!}\sum_{n\leq x} \frac{f(P^+(n))}{n} \log^{k-1}(\frac{x}{n}) + o(\log^k x). \label{log_alldi_k_LHS}
\end{multline}
Combining \eqref{log_alldi_k_total} and \eqref{log_alldi_k_LHS} together, we obtain \eqref{general_kth_duality} for the case $k$. Thus, \eqref{general_kth_duality} holds for all $k\ge1$ by induction.
\end{proof}
\section*{Acknowledgments}
The author is deeply grateful to the referee for a very careful review and very insightful suggestions, which improve this paper a lot; in particular, for the suggestion of incorporating Theorems \ref{mainthm_logAlladi_k} and \ref{mainthm_log_k} to the original version of the paper. The author would also like to thank Shaoyun Yi for the discussions on the earlier draft of this paper.
\begin{thebibliography}{10}
\bibitem{Alladi1977}
Krishnaswami Alladi.
\newblock Duality between prime factors and an application to the prime number
theorem for arithmetic progressions.
\newblock {\em J. Number Theory}, 9(4):436--451, 1977.
\bibitem{Alladi1982}
Krishnaswami Alladi.
\newblock Asymptotic estimates of sums involving the {M}oebius function. {II}.
\newblock {\em Trans. Amer. Math. Soc.}, 272(1):87--105, 1982.
\bibitem{AlladiJohnson2024}
Krishnaswami Alladi and Jason Johnson.
\newblock Duality between prime factors and the prime number theorem for arithmetic progressions - II.
\newblock preprint, arXiv: 2410.18259.
\bibitem{Coons2010}
Michael~J. Coons.
\newblock New proof of the equation $\sum_{k=1}^\infty\frac{\mu(k)}{k}=0$.
\newblock arXiv: 0803.3787v2.
\bibitem{Dawsey2017}
Madeline~Locus Dawsey.
\newblock A new formula for {C}hebotarev densities.
\newblock {\em Res. Number Theory}, 3:Paper No. 27, 13, 2017.
\bibitem{DuanMaYi2022}
Lian Duan, Ning Ma, and Shaoyun Yi.
\newblock Generalizations of {A}lladi's formula for arithmetical semigroups.
\newblock {\em Ramanujan J.}, 58(4):1285--1319, 2022.
\bibitem{DuanWangYi2021}
Lian Duan, Biao Wang, and Shaoyun Yi.
\newblock Analogues of {A}lladi's formula over global function fields.
\newblock {\em Finite Fields Appl.}, 74:Paper No. 101874, 26, 2021.
\bibitem{KuralMcDonaldSah2020}
Michael Kural, Vaughan McDonald, and Ashwin Sah.
\newblock M\"{o}bius formulas for densities of sets of prime ideals.
\newblock {\em Arch. Math.}, 115:53--66, 2020.
\bibitem{Landau1899}
Edmund Landau.
\newblock {\em Neuer Beweis der Gleichung
$\sum_{k=1}^\infty\frac{\mu(k)}{k}=0$}.
\newblock Inaugural Dissertation, Frietlrich Wilhelms Universit\"{a}t, Berlin,
1899.
\bibitem{OnoSchneiderWagner2017}
Ken Ono, Robert Schneider, and Ian Wagner.
\newblock Partition-theoretic formulas for arithmetic densities.
\newblock In {\em Analytic number theory, modular forms and
{$q$}-hypergeometric series}, volume 221 of {\em Springer Proc. Math. Stat.},
pages 611--624. Springer, Cham, 2017.
\bibitem{OnoSchneiderWagner2021}
Ken Ono, Robert Schneider, and Ian Wagner.
\newblock Partition-theoretic formulas for arithmetic densities, {II}.
\newblock {\em Hardy-Ramanujan J.}, 43:1--16, 2020.
\bibitem{Sengupta2025}
Sroyon Sengupta.
\newblock Algebraic analogues of results of {A}lladi--{J}ohnson using the
{C}hebotarev {D}ensity {T}heorem.
\newblock {\em Res. Number Theory}, 11(2):Paper No. 44, 2025.
\bibitem{Shapiro1950}
Harold~N. Shapiro.
\newblock On a theorem of selberg and generalizations.
\newblock {\em Annals of Mathematics}, 51(2):485--497, 1950.
\bibitem{SweetingWoo2019}
Naomi Sweeting and Katharine Woo.
\newblock Formulas for {C}hebotarev densities of {G}alois extensions of number
fields.
\newblock {\em Res. Number Theory}, 5(1):Paper No. 4, 13, 2019.
\bibitem{Tenenbaum2024}
G\'rald Tenenbaum.
\newblock On a family of arithmetic series related to the m\"{o}bius function.
\newblock preprint, arXiv: 2409.02754v4.
\bibitem{Titchmarsh1986}
E.~C. Titchmarsh.
\newblock {\em The theory of the {R}iemann zeta-function}.
\newblock The Clarendon Press, Oxford University Press, New York, second
edition, 1986.
\newblock Edited and with a preface by D. R. Heath-Brown.
\bibitem{Mangoldt1897}
H.C.F von Mangoldt.
\newblock {\em Beweis der Gleichung $\sum_{k=1}^\infty\frac{\mu(k)}{k}=0$},
volume~2.
\newblock Sitzungsberichte der K\"{o}niglich Preussischen Akademie der
Wissenschaften zu Berlin, 1897. \url{https://biodiversitylibrary.org/page/29982413}
\bibitem{Wang2021jnt}
Biao Wang.
\newblock Analogues of {A}lladi's formula.
\newblock {\em J. Number Theory}, 221:232--246, 2021.
\end{thebibliography}
\end{document}