\documentclass[12pt,reqno]{article} \usepackage[usenames]{color} \usepackage{amssymb} \usepackage{amsmath} \usepackage{amsthm} \usepackage{amsfonts} \usepackage{amscd} \usepackage{graphicx} \usepackage[colorlinks=true, linkcolor=webgreen, filecolor=webbrown, citecolor=webgreen]{hyperref} \definecolor{webgreen}{rgb}{0,.5,0} \definecolor{webbrown}{rgb}{.6,0,0} \usepackage{color} \usepackage{fullpage} \usepackage{float} \usepackage{psfig} \usepackage{graphics} \usepackage{latexsym} \usepackage{epsf} \usepackage{breakurl} \setlength{\textwidth}{6.5in} \setlength{\oddsidemargin}{.1in} \setlength{\evensidemargin}{.1in} \setlength{\topmargin}{-.1in} \setlength{\textheight}{8.4in} \newcommand{\seqnum}[1]{\href{https://oeis.org/#1}{\underline{#1}}} \begin{document} \begin{center} \epsfxsize=4in \leavevmode\epsffile{logo129.eps} \end{center} \theoremstyle{plain} \newtheorem{theorem}{Theorem} \newtheorem{corollary}[theorem]{Corollary} \newtheorem{lemma}[theorem]{Lemma} \newtheorem{proposition}[theorem]{Proposition} \theoremstyle{definition} \newtheorem{definition}[theorem]{Definition} \newtheorem{example}[theorem]{Example} \newtheorem{conjecture}[theorem]{Conjecture} \theoremstyle{remark} \newtheorem{remark}[theorem]{Remark} \begin{center} \vskip 1cm {\LARGE\bf The Number of Designated Parts\\ \vskip .05in in Compositions with Restricted Parts} \vskip 1cm \large Mengmeng Liu and Andrew Yezhou Wang\\ School of Mathematical Sciences \\ University of Electronic Science and Technology of China\\ Chengdu 611731\\ P. R. China\\ \href{mailto:dreamofliu@outlook.com}{\tt dreamofliu@outlook.com}\\ \href{mailto:yzwang@uestc.edu.cn}{\tt yzwang@uestc.edu.cn} \end{center} \vskip .2 in \begin{abstract} In this paper, we study the number of designated parts in three types of compositions with restricted parts. We find new combinatorial interpretations of some known sequences, and establish two identities concerning these known sequences. \end{abstract} \section{Introduction} A \emph{composition} of an integer $n$ is a way of expressing $n$ as an ordered sum of integers. More precisely, a composition of $n$ is a sequence $\alpha=(\alpha_1,\alpha_2,\ldots,\alpha_k)$ of positive integers satisfying $\alpha_1+\alpha_2+\cdots+\alpha_k=n$. The summands $\alpha_i$ are called the \emph{parts} of $\alpha$, and the number of parts is called the \emph{length} of $\alpha$. For example, there are eight compositions of $4$; namely, \begin{align*} 4,3+1,1+3,2+2,2+1+1,1+2+1,1+1+2,1+1+1+1. \end{align*} If $\alpha=(\alpha_1,\alpha_2,\ldots,\alpha_k)$ is a composition of $n$ with $k$ parts, then we define a $(k-1)$-subset $S_\alpha$ of $[n-1]:=\{1,2,\ldots,n-1\}$ by \begin{align*} S_\alpha=\{\alpha_1,\alpha_1+\alpha_2,\ldots,\alpha_1+\alpha_2+\cdots+\alpha_{k-1}\}. \end{align*} The correspondence $\alpha\rightarrow S_\alpha$ gives a bijection from all compositions of $n$ with $k$ parts to all $(k-1)$-subsets of $[n-1]$. Thus, there are ${n-1}\choose {k-1}$ compositions of $n$ with $k$ parts and $2^{n-1}$ compositions of $n$. Furthermore, a subset $S\subseteq[n-1]$ can be encoded by a binary sequence of length $n-1$ whose $i$th component is $1$ if $i\in S$ and $0$ otherwise. Therefore, we have a bijection between compositions of $n$ and binary sequences of length $n-1$. In this paper, we consider the following three types of compositions with restricted parts. Let $\mathcal{A}(n)$, $\mathcal{B}(n)$, and $\mathcal{C}(n)$ denote, respectively, the set of compositions of $n$ in which only the last part may be equal to $1$, compositions of $n$ into parts greater than $1$, and compositions of $n$ with only odd parts. There is a trivial bijection from the set $\mathcal{A}(n)$ to $\mathcal{B}(n+1)$. More precisely, given a composition $\alpha\in\mathcal{A}(n)$, increasing the last part of $\alpha$ by one, then we obtain a composition $\beta$ in $\mathcal{B}(n+1)$. We can uniquely recover $\alpha$ from $\beta$ by decreasing the last part of $\beta$ by one. There is a known identity concerning the cardinality of $\mathcal{B}(n)$ and $\mathcal{C}(n)$. \begin{theorem}\label{Euler1} For $n\geq 1$, we have $|\mathcal{B}(n+1)|=|\mathcal{C}(n)|$. \end{theorem} Both numbers in Theorem \ref{Euler1} are equal to the $n$th Fibonacci number $F_n$, defined by $F_0=0$, $F_1=1$, and for $n\geq2$, \[F_n=F_{n-1}+F_{n-2}.\] See Cayley \cite{Cay76} and Stanley \cite[Exercise 1.35]{Stan12} for more details. Sills \cite{Sills13} provided a bijective proof of Theorem \ref{Euler1} relying on the binary sequence encoding of compositions. Recently, Li and Wang \cite{LW19} found a simple bijection defined directly on the parts to prove Theorem \ref{Euler1}. In this paper, we study the number of designated parts among all compositions in $\mathcal{A}(n)$, $\mathcal{B}(n+1)$, and $\mathcal{C}(n)$ respectively. In Section \ref{s2}, we find a simple combinatorial interpretation of the sequence \seqnum{A102702} in the On-Line Encyclopedia of Integer Sequences (OEIS) \cite{OEIS00}. The sequence \seqnum{A102702} was proposed in 2005 by Creighton Dement, stated as a floretion-generated sequence. But there are no further clues about the floretion definition. In Section \ref{s3}, we give a new combinatorial object counted by the sequence \seqnum{A006367} and establish an identity relating the sequences \seqnum{A006367} and \seqnum{A010049}. In Section \ref{s4}, we present a relationship concerning the number of three types of designated parts in compositions. \section{Compositions in which only the last part may be one}\label{s2} In this section, we mainly focus on the number of parts not greater than $2$ among all compositions in $\mathcal{A}(n)$, denoted by $\overline{A}_n$. It is easy to see that $\overline{A}_0=0$, $\overline{A}_1=1$, $\overline{A}_2=1$, and $\overline{A}_3=2$. We now present a recurrence relation for $\overline{A}_n$. \begin{theorem}\label{recc} For $n\geq 4$, we have \begin{align}\label{crec} \overline{A}_n=\overline{A}_{n-1}+\overline{A}_{n-2}+F_{n-4}. \end{align} \end{theorem} \begin{proof} Given a composition $\alpha$, let $\ell_{\leq 2}(\alpha)$ count the number of parts less than or equal to $2$ of $\alpha$. Then, \begin{align*} \overline{A}_n=\sum\limits_{\alpha\in\mathcal{A}(n)}\ell_{\leq2}(\alpha). \end{align*} Define $\mathcal{A}_1(n)$ and $\mathcal{A}_2(n)$ to be the subset of $\mathcal{A}(n)$ consisting of the compositions with the first part equal to $2$ and greater than $2$ respectively. It is clear that \begin{align*} \mathcal{A}(n)=\mathcal{A}_1(n)\cup\mathcal{A}_2(n) \end{align*} is a disjoint set theoretic partitioning of $\mathcal{A}(n)$. For any composition $\alpha\in\mathcal{A}_1(n)$, deleting the first part of $\alpha$ produces a composition in $\mathcal{A}(n-2)$, which gives a bijection $\rho$ between $\mathcal{A}_1(n)$ and $\mathcal{A}(n-2)$ such that \[\ell_{\leq2}(\alpha)=\ell_{\leq2}(\rho(\alpha))+1.\] Given a composition $\beta\in\mathcal{A}_2(n)$, we decrease the first part of $\beta$ by $1$ to obtain a composition in $\mathcal{A}(n-1)$, and vice versa. Hence, there is a bijection $\tau$ between $\mathcal{A}_2(n)$ and $\mathcal{A}(n-1)$. Moreover, $\ell_{\leq2}(\beta)=\ell_{\leq2}(\tau(\beta))$ if and only if the first part of $\beta$ is greater than $3$. If $\beta$ has its first part equal to $3$, then $\ell_{\leq2}(\beta)=\ell_{\leq2}(\tau(\beta))-1$. Similarly to the above bijection $\rho$, we know that such compositions correspond to those in $\mathcal{A}(n-3)$. Recall that $|\mathcal{A}(n)|=F_n$ for $n\geq 0$. Therefore, we have \begin{align*} \sum\limits_{\alpha\in\mathcal{A}(n)}\ell_{\leq2}(\alpha)&=\sum\limits_{\alpha\in\mathcal{A}_1(n)}\ell_{\leq2}(\alpha)+ \sum\limits_{\alpha\in\mathcal{A}_2(n)}\ell_{\leq2}(\alpha)\\ &=\sum\limits_{\alpha\in\mathcal{A}(n-2)}(\ell_{\leq2}(\alpha)+1)+ \left(\sum\limits_{\alpha\in\mathcal{A}(n-1)}\ell_{\leq2}(\alpha)-\sum\limits_{\alpha\in\mathcal{A}(n-3)}1\right)\\ &=\sum\limits_{\alpha\in\mathcal{A}(n-2)}\ell_{\leq2}(\alpha)+F_{n-2}+ \sum\limits_{\alpha\in\mathcal{A}(n-1)}\ell_{\leq2}(\alpha)-F_{n-3}\\ &=\sum\limits_{\alpha\in\mathcal{A}(n-2)}\ell_{\leq2}(\alpha)+ \sum\limits_{\alpha\in\mathcal{A}(n-1)}\ell_{\leq2}(\alpha)+F_{n-4}, \end{align*} which yields the desired result. \end{proof} We now establish the generating function of $\overline{A}_n$. \begin{theorem}\label{gfoc} The generating function of $\overline{A}_n$ is given by \begin{align}\label{genc} \overline{A}(x):=\sum\limits_{n\geq0}\overline{A}_nx^n=\frac{x-x^2-x^3+x^5}{(1-x-x^2)^2}. \end{align} \end{theorem} \begin{proof} To find $\overline{A}(x)$, multiply both sides of the recurrence relation \eqref{crec} by $x^n$ and sum over $n\geq 4$. Then try to relate these sums to the unknown generating function $\overline{A}(x)$. If we do this first to the left side of \eqref{crec}, we have \begin{align*} \sum\limits_{n\geq4}\overline{A}_nx^n=\overline{A}(x)-x-x^2-2x^3. \end{align*} We next deal with the right side of \eqref{crec}. The result is \begin{align*} \sum\limits_{n\geq4}\left(\overline{A}_{n-1}+\overline{A}_{n-2}+F_{n-4}\right)x^n &=\sum\limits_{n\geq4}\overline{A}_{n-1}x^n+\sum\limits_{n\geq4}\overline{A}_{n-2}x^n +\sum\limits_{n\geq4}F_{n-4}x^n\\ &=x\left(\overline{A}(x)-x-x^2\right)+x^2\left(\overline{A}(x)-x\right)+\frac{x^5}{1-x-x^2}, \end{align*} wherein we have used the familiar generating function for the Fibonacci numbers \begin{align}\label{recFibo} \sum\limits_{n\geq0}F_nx^n=\frac{x}{1-x-x^2}. \end{align} If we equate the results of operating on the two sides of \eqref{crec}, we find that \begin{align*} \overline{A}(x)-x-x^2-2x^3=x\left(\overline{A}(x)-x-x^2\right)+x^2\left(\overline{A}(x)-x\right)+\frac{x^5}{1-x-x^2}, \end{align*} which is trivial to solve for the unknown generating function $\overline{A}(x)$, in the form \begin{align*} \overline{A}(x)=\frac{x-x^2-x^3+x^5}{(1-x-x^2)^2}. \end{align*} The proof is complete. \end{proof} We now seek the relationship between the sequence $(\overline{A}_n)_{n\geq0}$ and the sequence $\left(a_n\right)_{n\geq0}$ in OEIS \cite[\seqnum{A102702}]{OEIS00}. It is routine to check that \begin{align*} \overline{A}(x)-x-x^2&=\frac{x-x^2-x^3+x^5}{(1-x-x^2)^2}-x-x^2\\ &=x^3\cdot\frac{2-x-2x^2-x^3}{(1-x-x^2)^2}, \end{align*} where the second factor of the last identity above is the generating function of $a_n$. Therefore, $a_n$ is the total number of parts not greater than $2$ among all compositions of $n+3$ in which only the last part may be equal to $1$. This is a new and simple combinatorial interpretation of the sequence \seqnum{A102702}. We next derive an explicit formula for $\overline{A}_n$ in terms of the Fibonacci numbers by expanding $\overline{A}(x)$ in a series. \begin{theorem} For $n\geq2$, we have \begin{align}\label{formula} \overline{A}_n=\frac{(3n-3)F_n-(4n-10)F_{n-1}}{5}. \end{align} \end{theorem} \begin{proof} It follows from \eqref{recFibo} that \begin{align*} \sum\limits_{n\geq0}F_{n+1}x^n=\frac{1}{1-x-x^2}. \end{align*} Therefore, the coefficient $P_n$ of $x^n$ in $1/(1-x-x^2)^2$ is \begin{align*} F_1F_{n+1}+F_2F_n+\cdots+F_{n+1}F_1. \end{align*} By induction on $n$ and using the fact $F_{n}=F_{n-1}+F_{n-2}$, it is easy to see that \begin{align}\label{coeff} P_n=F_1F_{n+1}+F_2F_n+\cdots+F_{n+1}F_1=\frac{(n+1)F_{n+2}+(2n+4)F_{n+1}}{5}, \end{align} which can be found in \cite[Eq.\ (32.13), p.\ 375]{Kos01} or \cite[Nr.\ (98), p.\ 183]{Vaj89}. Combining \eqref{genc} and \eqref{coeff}, we conclude that for $n\geq 5$, \begin{align*} \overline{A}_n&=P_{n-1}-P_{n-2}-P_{n-3}+P_{n-5}\\ &=\frac{nF_{n+1}+(2n+2)F_{n}}{5}-\frac{(n-1)F_{n}+2nF_{n-1}}{5}\\ &\quad-\frac{(n-2)F_{n-1}+(2n-2)F_{n-2}}{5}+\frac{(n-4)F_{n-3}+(2n-6)F_{n-4}}{5}\\ &=\frac{(2n+3)F_{n}-(2n-2)F_{n-1}-(2n-2)F_{n-2}+(n-4)F_{n-3}+(2n-6)F_{n-4}}{5}\\ &=\frac{(2n+3)F_{n}-(2n-2)F_{n-1}-4F_{n-2}-(n-2)F_{n-3}}{5}\\ &=\frac{(2n+3)F_{n}-(3n-4)F_{n-1}+(n-6)F_{n-2}}{5}\\ &=\frac{(3n-3)F_{n}-(4n-10)F_{n-1}}{5}. \end{align*} One can readily check that $\overline{A}_n$ for each $n=2,3,4$ also satisfies the formula \eqref{formula}. \end{proof} As a consequence, we can express $a_n$ in OEIS \cite[\seqnum{A102702}]{OEIS00} as \begin{align*} a_n=\frac{(3n+6)F_{n+3}-(4n+2)F_{n+2}}{5}=\frac{(2n+10)F_{n+1}-(n-4)F_{n}}{5} \end{align*} for $n\geq0$. To end this section, we consider the number of parts among all compositions in $\mathcal{A}(n)$, denoted by $A_n$. Obviously, $A_0=0$ and $A_1=1$. Applying arguments similar to those when dealing with $\overline{A}_n$, we can show that for $n\geq2$, \begin{align*} A_n=A_{n-1}+A_{n-2}+F_{n-2}, \end{align*} which implies the generating function \begin{align*} A(x):=\sum\limits_{n\geq0}A_nx^n=\frac{x-x^2}{(1-x-x^2)^2}. \end{align*} Based on $A(x)$, we derive the following simple formula \begin{align*} A_n=\frac{(2n+3)F_n-nF_{n-1}}{5},n\geq1. \end{align*} The number $A_n$ appears in OEIS \cite[\seqnum{A010049}]{OEIS00}, which counts the total number of parts in all compositions of $n+1$ with no $1$'s. \section{Compositions with no ones}\label{s3} Let $B_n$ be the number of parts among all compositions in $\mathcal{B}(n+1)$. Since the trivial bijection between $\mathcal{A}(n)$ and $\mathcal{B}(n+1)$ preserves the length of compositions, we have $B_n=A_n$. Define $\overline{B}_n$ to be the number of parts equal to $2$ among all compositions in $\mathcal{B}(n+1)$. It is clear that $\overline{B}_0=0$, $\overline{B}_1=1$, $\overline{B}_2=0$, and $\overline{B}_3=2$. Using the arguments similar to those in the proof of Theorem \ref{recc}, we obtain that for $n\geq4$, \begin{align*} \overline{B}_n=\overline{B}_{n-1}+\overline{B}_{n-2}+F_{n-4}. \end{align*} The sequence $(\overline{B}_n)_{n\geq0}$ satisfies the same recurrence relation as $(\overline{A}_n)_{n\geq0}$ but with different initial conditions. Similarly, we can find that \begin{align*} \sum\limits_{n\geq0}\overline{B}_nx^n=\frac{x(1-x)^2}{(1-x-x^2)^2}, \end{align*} and for $ n\geq1$, \begin{align*} \overline{B}_n=\frac{(3n+2)F_n-4nF_{n-1}}{5}. \end{align*} The number $\overline{B}_n$ relates to the sequence \seqnum{A006367} in OEIS \cite{OEIS00}, which is interpreted as the number of compositions of $n$ containing exactly one $1$. We now give a combinatorial proof of this equality. \begin{theorem} The number $\overline{B}_n$ equals the number of compositions of $n$ with exactly one $1$. \end{theorem} \begin{proof} Assume that $\alpha=(\alpha_1,\alpha_2,\ldots,\alpha_l)\in\mathcal{B}(n+1)$ have $j$ parts equal to $2$, say \[\alpha_{i_1}=\alpha_{i_2}=\cdots=\alpha_{i_j}=2,\] where $1\leq i_1