Cassini's identity for k-bonacci numbers
Extracted source text ยท arXiv 2606.03228
\documentclass[11pt,reqno]{amsart}
\usepackage[a4paper]{geometry}
%\setlength{\hoffset}{-.5in}
\setlength{\voffset}{-.25in}
%\hyphenation{Fibonacci}
\sloppy
\usepackage{amssymb,latexsym}
\usepackage{graphicx}
\usepackage{url} %does nice formatting of URLs
\usepackage{todonotes}
\usepackage{hyperref}
\usepackage{multicol}
\usepackage{amsmath}
\usepackage{mathtools}
\usepackage{rotating}
\usepackage{amsmath}
\usepackage{enumitem}
\usepackage{textcase}
\usepackage{mathdots}
\usepackage{nicematrix}
\usepackage{booktabs}
\usepackage{hyperref}
% \usepackage[backend=biber, style=trad-plain]{biblatex}
% \addbibresource{kbonacci.bib}
\usepackage[T2A]{fontenc}
\usepackage{soul}
\setlist{itemsep=8pt}
\usepackage{empheq}
\usepackage{tikz}
\usetikzlibrary{matrix,backgrounds}
\usetikzlibrary{decorations.pathreplacing,calligraphy}
\newlength{\mydrawlinewidth}
\setlength{\mydrawlinewidth}{1pt}
\NewDocumentCommand{\markA}{ O{red} m m }{%
\draw[rounded corners=3mm,opacity=.5, line width=2pt,#1,fill=#1!20]
([xshift=5pt,yshift=-5pt]yannis-#2.north west)
rectangle
([xshift=-5pt,yshift=5pt]yannis-#3.south east);
}
\NewDocumentCommand{\markB}{ O{red} m m }{%
\draw[rounded corners=3mm,opacity=.5,line width=1pt,#1,fill=#1!50]
([yshift=2pt]yannis-#2.north)
-- ([xshift=2pt]yannis-#3.east)
-- ([yshift=-2pt]yannis-#3.south)
-- ([xshift=-2pt]yannis-#2.west)
-- cycle;
}
\NewDocumentCommand{\markC}{ O{red} m m }{%
\draw[rounded corners=3mm,opacity=.5,line width=1pt,#1,fill=#1!50]
([yshift=2pt]yannis-#3.north)
-- ([xshift=2pt]yannis-#2.west)
-- ([yshift=-2pt]yannis-#2.south)
-- ([xshift=-2pt]yannis-#3.east)
-- cycle;
}
% \DeclareMathOperator{\sign}{sgn}
\textwidth=6.175in
\textheight=9.0in
\headheight=13pt
\calclayout
\makeatletter
\newcommand{\monthyear}[1]{%
\def\@monthyear{\uppercase{#1}}}
% \def\@monthyear{\uppercase{Month Year}}}
\newcommand{\volnumber}[1]{%
\def\@volnumber{\uppercase{#1}}}
% \def\@volnumber{\uppercase{Volume , Number }}}
\AtBeginDocument{%
\def\ps@plain{\ps@empty
\def\@oddfoot{\@monthyear \hfil \thepage}%
\def\@evenfoot{\thepage \hfil \@volnumber}}
\def\ps@firstpage{\ps@plain}
\def\ps@headings{\ps@empty
\def\@evenhead{%
\setTrue{runhead}%
\def\thanks{\protect\thanks@warning}%
\uppercase{AOM}\hfil}%
\def\@oddhead{%
\setTrue{runhead}%
\def\thanks{\protect\thanks@warning}%
\hfill\MakeTextUppercase{\othertitle}}%
\let\@mkboth\markboth
\def\@evenfoot{%
\thepage \hfil \@volnumber}%
\def\@oddfoot{%
\@monthyear \hfil \thepage}%
}%
\footskip=25pt
\pagestyle{headings}%
}
\makeatother
\newcommand{\R}{{\mathbb R}}
\newcommand{\Q}{{\mathbb Q}}
% \newcommand{\C}{{\mathbb C}}
\newcommand{\N}{{\mathbb N}}
\newcommand{\Z}{{\mathbb Z}}
\newcommand{\minus}{\scalebox{0.75}[1.0]{$-$}}
\newcommand{\Mod}[1]{\ (\mathrm{mod}\ #1)}
%$\boldsymbol{F_n}$
%sometimes 85--90 needs to be written 85\textendash90
\theoremstyle{plain}
\numberwithin{equation}{section}
\newtheorem{thm}{Theorem}[section]
\newtheorem{theorem}[thm]{Theorem}
\newtheorem{lemma}[thm]{Lemma}
\newtheorem{example}[thm]{Example}
\newtheorem{definition}[thm]{Definition}
\newtheorem{proposition}[thm]{Proposition}
\newtheorem{formula}[thm]{Formula}
\newtheorem{notation}[thm]{Notation}
\newtheorem{corollary}[thm]{Corollary}
\newtheorem{remark}[thm]{Remark}
\newcommand{\helv}{%
\fontfamily{phv}\fontseries{m}\fontsize{9}{11}\selectfont}
\begin{document}
%% replace the values in the next three lines by the correct information
\monthyear{July 2026}
\volnumber{Volume, Number}
\setcounter{page}{1}
\title{Cassini's identity for $k$-bonacci Numbers}
\def\othertitle{Cassini $k$-bonacci}
\author{Harold R. Parks}
\address{Oregon State University, Corvallis, Oregon, USA.}
\email{hal.parks@oregonstate.edu}
\author{Dean. C. Wills}
\address{Thayer School of Engineering at Dartmouth College, Hanover, NH, USA}
\email{dean.c.wills.th@dartmouth.edu}
\begin{abstract}
Efforts have been made to extend Cassini's identity (also known as Simson's identity) to the $k$-step or $k$-bonacci numbers for decades. These efforts have lacked both completeness of result and simplicity of proof, and this question remains open and relevant. In this note, we offer a definitive solution as well as the generalization of both Catalan's and Vajda's identities.
%Lastly, we provide food for thought, as a closed form for recursions of these identities for all $k$ remains unknown.
\end{abstract}
\maketitle
\vspace{10pt}
\noindent{\textbf{Keywords:}{$k$-bonacci, Cassini, Catalan, Vajda}} \\
\noindent{\textbf{2020 Mathematics Subject Classification:}{11B39, 11B0F}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% INTRODUCTION
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Introduction}
Our purpose in this article is to generalize the identities of
Cassini, Catalan, and Vajda from their original setting
in the Fibonacci numbers to
the larger realm of the $k$-bonacci numbers.
Before we begin that work, we will briefly review for the
reader what the identities of
Cassini, Catalan, and Vajda tell us.
Cassini's identity for the Fibonacci numbers $F_n$ is
\begin{equation} \label{eq:cassini}
F_{n}^{2}-F_{n-1}F_{n+1}=(-1)^{n-1}.
\end{equation}
This identity was first published by
Jean-Dominque Cassini\footnote{Cassini's given name was
Giovanni Domenico, but he changed it when he moved to France.
Jean-Dominque Cassini, his son, Jacques,
his grandson, C\'esar-Fran\c{c}ois, and his
great-grandson, Jean-Domenique,
all noted astronomers, are often referred to as
Cassini I, II, III, and IV.}
in 1680 (see \cite{bhlpage3208626}).
The identity was independently rediscovered and published by Robert Simson in 1753 (see \cite{simson1753}), so \eqref{eq:cassini} is also known as Simson's identity.
In 1879,
Eug\`ene Catalan generalized Cassini's identity to
\begin{equation} \label{eq:catalan}
F_{n}^{2}-F_{n-r}F_{n+r}=(-1)^{n-r} F_{r}^2,
\end{equation}
though \eqref{eq:catalan} was not published until 1886
when it appeared in \cite{catalan}.
The twentieth century saw a further generalization
of Cassini's identity to
\begin{equation} \label{eq:vajda}
F_{n}F_{n+p+q}-F_{n+p}F_{n+q}=(-1)^{n-1}F_{p}F_{q}.
\end{equation}
While \eqref{eq:vajda} was published in 1901
by Alberto Tagiuri
in \cite{tagiuri1901} and by
Arthur Danese in 1960
in the {\it American Mathematical Monthly}
as an elementary problem
(see \cite{danese1960}), its inclusion
as equation (20a) on
page 28 of
Steven Vajda's 1989 book \cite{vajda1989},
has led to \eqref{eq:vajda}
being known as Vajda's identity.\footnote{Vajda does give credit to
Danese.}
To see that \eqref{eq:vajda}
is a generalization of \eqref{eq:catalan}, we set $p=r$ and
$q=-r$ in \eqref{eq:vajda}, and then we use
the fact that $F_{-r} = (-1)^{1-r} F_r$.
\medskip
For our generalizations of \eqref{eq:cassini},
\eqref{eq:catalan}, and \eqref{eq:vajda},
we simply replace each Fibonacci number
$F_\ell$ on the left-hand side
by the corresponding $k$-bonacci number
$F^{(k)}_\ell$. The question then, and the problem we address, is
\begin{center}
``What goes on the right-hand side of each equation?''
\end{center}
For $k>2$ the correct right-hand side cannot be obtained by the simple technique of replacing each Fibonacci number
$F_\ell$ by the corresponding
$k$-bonacci number $F^{(k)}_\ell$---the result
of that process would be false.
Instead, we shall obtain the correct right-hand sides by using
combinations of generalized Vandermonde determinants
based on the roots
of the characteristic polynomial for the $k$-bonacci recursion.
\medskip
It is straightforward to prove Cassini's identity \eqref{eq:cassini}
using Binet's formula for the Fibonnaci numbers, so
motivation for our approach is provided by the existence of a similar Binet-type formula for the \mbox{$k$-bonacci} numbers
(see \cite{kilicc2006generalized}
or \cite{generalizedBinet2023Parks}).
That Binet formula for the \mbox{$k$-bonacci} numbers
involves a ratio of $k\times k$ determinants that
can reasonably be called generalized Vandermonde determinants. Our
main tool for obtaining our
generalizations of \eqref{eq:cassini},
\eqref{eq:catalan}, and \eqref{eq:vajda}
will also be ratios of such determinants.
So inspired by E. R. Heineman's notation in
\cite{heineman1929generalized}, we find it helpful
to introduce the following:
\begin{notation}\rm
Let $\ell$ be a positive integer and let $\phi_i$ be $\ell$ elements of a commutative ring $\mathbf{R}$. Then we set
\begin{equation}\label{eq:heineman}
\| \ t_1,\ t_2,\ \dots, \ t_{\ell}\ \|
=
\begin{vmatrix}
\phi_1^{t_1} & \phi_1^{t_2} & \cdots & \phi_1^{t_{\ell-1}} & \phi_1^{t_{\ell}}\\[0.5ex]
\phi_2^{t_1} & \phi_2^{t_2} & \cdots & \phi_2^{t_{\ell-1}}& \phi_2^{t_{\ell}}\\[0.5ex]
\phi_3^{t_1} & \phi_3^{t_2} & \cdots & \phi_3^{t_{\ell-1}}& \phi_3^{t_{\ell}}\\
\vdots & \vdots & \ddots & \vdots & \vdots \\
\phi_\ell^{t_1} & \phi_\ell^{t_2} & \cdots & \phi_\ell^{t_{\ell-1}} & \phi_\ell^{t_{\ell}}
\end{vmatrix}
\end{equation}
where the $t_i$ are integers. The $\phi_i$ do not
appear on the left-hand side of \eqref{eq:heineman} and so
must be understood from the context.
\end{notation}
Using the above notation, with $\ell=k$ and the $\phi_i$ the roots over $\mathbb{C}$ of the characteristic polynomial of the
$k$-bonacci recursion,
\[
x^k-x^{k-1}-\cdots-1=0
\,,
\]
the Binet formula of \cite{generalizedBinet2023Parks} for
the \mbox{$k$-bonacci} numbers is
\begin{equation}\label{eq.binet}
F_n^{(k)}=
\frac{
\|\ 0, \ 1,\ 2,\ \dots\ k-2,\ \ n\ \| }
{\|\ 0, \ 1,\ 2,\ \dots\ k-2,\ \ k-1\ \|}
\,.
\end{equation}
Our results here will be expressed in terms of
\begin{equation}\label{eq.D}
F_{m,n}^{(k)}=
\frac{
\|\ 0, \ 1,\ 2,\ \dots\ k-3,\ m,\ n\ \| }
{\|\ 0, \ 1,\ 2,\ \dots\ k-3,\ k-2,\ k-1\ \|}
\,.
\end{equation}
Now,
$F^{(k)}_{m,\,n}$
would be difficult to
compute directly using \eqref{eq.D}, especially when $k$ is large.
Fortunately there are other ways to evaluate $F^{(k)}_{m,\,n}$.
%For moderate values of $m$ and $n$, Lemma~\ref{lemma.values} may apply.
Additionally, $F^{(k)}_{m,\,n}$ satisfies the $k$-bonacci recursion in
both $m$ and $n$.
We will also see from Theorem~\ref{th:vajda1} and Theorem~\ref{thm.reduce.to.k} that
\[
F^{(k)}_{m,n} = F^{(k)}_{n}F^{(k)}_{m+1} -F^{(k)}_{n+1}F^{(k)}_{m}
= F^{(k)}_{n-k}F^{(k)}_{m} -F^{(k)}_{n}F^{(k)}_{m-k} \,,
\]
thus expressing $F^{(k)}_{m,\,n}$ in terms of $k$-bonacci numbers.
The generalization of Cassini's identity is the following:
For all $k\geq2$, it holds that
\[
\left(F^{(k)}_{n}\right)^{2}-
F^{(k)}_{n-1}\,F^{(k)}_{n+1}=F_{ n-1,\, n}^{(k)}\,.
\]
When $k=3$, we get the following nice result:
\[
\left(F^{(3)}_{n}\right)^{2}-F^{(3)}_{n-1}\,F^{(3)}_{n+1}=F^{(3)}_{1-n} \,,
\]
and when $k=2$, our generalization reduces to Cassini's original identity.
The most general of our results is the generalization
of Vajda's identity given in Theorem~\ref{th:vajda} where
we show for $p > 0$
\begin{equation}
F^{(k)}_{n}F^{(k)}_{n+p+q}-F^{(k)}_{n+p}F^{(k)}_{n+q} = \sum_{i=0}^{p-1}F^{(k)}_{n+q+i,\, n+p-1-i} \,,
\label{eq:vaj}
\end{equation}
\medskip
While it is far from apparent, \eqref{eq:vaj} does simplify
to Vajda's identity when $k=2$.
Unfortunately, for $k\geq 3$,the right-hand side of \eqref{eq:vaj},
as it stands, is no longer a simple expression like \eqref{eq:vajda}.
Nonetheless equation \eqref{eq:vaj}
answers to the question, ``What goes on the right-hand side?''
in the context of our functions $F^{(k)}_{m,\,n}$.
%\bigskip
%The denominator in \eqref{eq.binet} is a Vandermonde %determinant.
%The numerator, on the other hand, is an instance of what can %reasonably be called a {\em generalized Vandermonde determinant} in %which $k$ ring elements $\phi_i$ appear with integral exponents %$t_i$ --- in this case $t_i=i$, for $i=0,1,\dots,k-2$,
%and $t_{k-%1}=n$. The denominator is known as the
%\textit{principal Vandermondian}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% PRELIMINARIES
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% does this help or just confuse?
% Notice that if we use the $k-1$ elements
% $\phi_1,\dots,\phi_{u-1},\phi_{u+1},\dots,\phi_k$,
% then we can write
% \[
% V^{(k)}_{\widehat u} = \|\ 0,\ 1,\ \dots,\ k-2\ \|
% \]
% Notice that if we use the $k-2$ elements
% \[
% \phi_1,\dots,\phi_{u-1},\phi_{u+1},
% \dots,\phi_{v-1},\phi_{v+1},\dots, \phi_k
% \,,
% \]
% then we can write
% \[
% V^{(k)}_{\widehat u\,\widehat v} = \|\ 1,\ 2,\ \dots,\ k-3\ \|
% \]
%Since the denominators in \eqref{eq.binet} and \eqref{eq.D} and do not depend on $m$ or $n$ or $t_i$ in general, our computations really depend on the numerators. Accordingly
% $\mathbf{R}$ and $n$ exponents $m_i$, where $0 \leq n \leq k$.
%\todo[inline]{$n=0$ is the trivial case that turns $V^{(k)}_{m_1,\ \dots \ m_n}$ into $V^{(k)}$}
% We define
% \[
% V^{(k)}_{m_1,\ \dots \ m_n} = \|\ 0, \ 1,\ 2,\ \dots,\ k-n-1, \dots,
% \ m_1, \dots, m_n\ \| \,.
% \]
\section{Preliminaries}
The generalization of Vajda's identity is a consequence of an identity that holds for the particular Vandermonde matrices given in the next definition.
\begin{definition}\label{def.vanders}\rm
Let $\phi_i$ be $k$ elements of a commutative ring
$\mathbf{R}$.
\begin{enumerate}
\item
Set
\[
V^{(k)}= V^{(k)}|_{\phi_i} = \|\ 0,\ 1,\ \dots,\ k-1\ \|
\,.
\]
% \begin{equation}\label{eq:W}
% \end{equation}
\item
For $u\in\{ 1,2,\dots,k \}$, we let $V^{(k)}_{\widehat u}$ be the
$(k-1)\times(k-1)$ determinant obtained from $V^{(k)}$
by omitting row $u$ and column $k$.
\item
For $u,v\in\{ 1,2,\dots,k \}$ with $u < v $, we let
$V^{(k)}_{\widehat u\, \widehat v}$ be the
$(k-2)\times(k-2)$ determinant obtained from $V^{(k)}$
by omitting rows $u$ and $v$ and
columns $k-1$ and $k$.
\end{enumerate}
\end{definition}
\medskip
All three of $V^{(k)}$, $V^{(k)}_{\widehat u}$,
and $V^{(k)}_{\widehat u\,\widehat v}$
in Definition~\ref{def.vanders} are Vandermonde
determinants, so the value of each is the
product of the differences, higher row
number minus lower row number, of the
entries in the second column of that particular determinant, when the matrix is oriented as in \eqref{eq:heineman}.
% some authors use the transpose of what we have.
These facts allow us to prove the next lemma.
\begin{lemma}\label{tricky.identity}
$\phi_i$ be $k$ elements of a commutative ring $\mathbf{R}$.
Suppose $1\leq u < v\leq k$. Then we have
\rm\begin{enumerate}
\item \label{tricky:one}
\ $\displaystyle
V^{(k)} = \prod_{1\leq i< j\leq k} (\phi_j-\phi_i)
\,,
$
\item \label{tricky:two}
\ $\displaystyle
V^{(k)}=
V^{(k)}_{\widehat u} \cdot \prod_{1\leq i < u} (\phi_u-\phi_i)
\cdot \prod_{u<j \leq k} (\phi_j-\phi_u)
\,,$
\item \label{tricky:three}
\begin{align*}
& (\phi_v-\phi_u) \cdot V^{(k)} \\
&= V^{(k)}_{\widehat u\,\widehat v}
\cdot \prod_{1\leq i < u} (\phi_u-\phi_i) \cdot \prod_{u<j \leq k} (\phi_j-\phi_u)
\cdot \prod_{1\leq i < v} (\phi_v-\phi_i) \cdot \prod_{v<j \leq k} (\phi_j-\phi_v)
\end{align*}
\item\label{tricky:four}
\begin{equation}\label{eq:tricky:four}
(\phi_v-\phi_u)\, V^{(k)}_{\widehat u}\, V^{(k)}_{\widehat v}
= V^{(k)}\,V^{(k)}_{\widehat u\,\widehat v}
\,.
\end{equation}
\end{enumerate}
\end{lemma}
\begin{proof}
Part \eqref{tricky:one}.\ This is the standard
result concerning Vandermonde determinants. For completeness,
we cite pages 12 and 13 of \cite{franklin}.
%The second column of
%$V^{(k)}$ contains all the numbers $\phi_1,\phi_2,\dots,\phi_k$
%and the right-hand side of Part~(\ref{tricky:one}) includes the
%product of all the
%differences, higher row number minus lower row number, as required.
\medskip
\noindent
Part \eqref{tricky:two}.\ The terms
$\prod_{1\leq i < v} (\phi_v-\phi_i)$ and
$\prod_{u<j \leq k} (\phi_j-\phi_u)$
on the right-hand side of Part~(\ref{tricky:two}) account for all the
differences from $\prod_{1\leq i< j\leq k} (\phi_j-\phi_i)$ that contain $\phi_u$.
\medskip
\noindent
Part \eqref{tricky:three}.\ As in Part~\eqref{tricky:two}, the terms
$\prod_{1\leq i < u} (\phi_u-\phi_i)$ and
$\prod_{u<j \leq k} (\phi_j-\phi_u)$
on the right-hand side of \eqref{tricky:three} account for all the
differences that contain $\phi_u$ from $\prod_{1\leq i< j\leq k}(\phi_j-\phi_i)$.
Likewise, the terms the terms
$\prod_{1\leq i < v} (\phi_v-\phi_i)$ and
$\prod_{v<j \leq k} (\phi_j-\phi_v)$
on the right-hand side of \eqref{tricky:three} account for all the
differences that contain $\phi_v$ from $\prod_{1\leq i< j\leq k}(\phi_j-\phi_i)$. But in that process, the term $(\phi_v-\phi_u)$ is eliminated
twice, so the last factor $(\phi_v-\phi_u)$ ensures that
$(\phi_v-\phi_u)$ is eliminated exactly once.
\medskip
\noindent
Part \eqref{tricky:four}
follows immediately from Parts \eqref{tricky:one}, \eqref{tricky:two} and \eqref{tricky:three}.
\end{proof}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Laplace's expansion
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Next,
we collect some classical results
concerning determinants that will be used later.
One source for details is \cite{aitken}, specifically
Sections 19, 33, 35, and 36.
Let $M$ be an $n\times n$ matrix, $n\geq 2$, with
entries $m_{i\,j}$.
For each $u$, let $M_{\widehat{u}}$ be the
$(n-1)\times(n-1)$ matrix obtained by deleting row $u$
and column $n$ from $M$. Of course, the reader is familiar
with the following Laplace expansion:
\begin{formula} \label{laplace.1}
\begin{equation*} % \label{laplace.1}
\hbox{\rm det}\, M = \sum_{u=1}^n (-1)^{u+n}\, m_{u\,n}\,
\hbox{\rm det}\, M_{\widehat{u}}
\,.
\end{equation*}
\end{formula}
When the preceding formula is applied to $V^{(k)}_{m}$,
we obtain
\begin{equation}\label{le:laplace}
V^{(k)}_{m} = \sum_{u=1}^k (-1)^{k+u} \, \phi_u^{m} V^{(k)}_{\widehat u} \,.
\end{equation}
The reader can certainly imagine applying the above Laplace
expansion to to each summand in
Formula~\ref{laplace.1}. To do so, one defines
$M_{\widehat{u}\,\widehat{v}}$ to be the
$(n-2)\times(n-2)$ matrix obtained by deleting rows $u$ and $v$
and columns $n-1$ and $n$ from $M$. The challenging part
is keeping track of the signs.
The resulting expansion---also due to Laplace---can
be written as follows:
\begin{formula} \rm \label{laplace.2}
\begin{equation*} %\label{laplace.2}
\hbox{\rm det}\, M = \sum_{1\leq u< v\leq n} (-1)^{u+v+1}\,
\begin{vmatrix}
m_{u\, n-1} & m_{u\,n} \\
m_{v\,n-1} & m_{v\, n}
\end{vmatrix} \,
\hbox{\rm det}\, M_{\widehat{u}\,\widehat{v}}
\,,
\end{equation*}
where we use the vertical bar notation for $2\times2$ determinants,
because it is a bit more efficient, and it is easily recognized.
\end{formula}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Binet--Cauchy
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Another classical result is the following:
\begin{formula}[Binet--Cauchy identity]\rm \label{fo:binet-cauchy}
Suppose $A$ is a $m\times n$ matrix and $B$ is an $n\times m$ matrix
over a commutative ring. For each subset $S \subseteq \{1 \dots n\}$
of size $m$, we can form square matrices $ A_S $ whose columns, and
$ B_S $ whose rows, are indexed by $S$. Then
\begin{equation*} %\label{fo:binet-cauchy}
\det{AB}=\sum_{|S|=m} \det A_S\, \det B_S
\end{equation*}
\end{formula}
The Laplace expansion dates back to 1772. It can be
further generalized beyond the
form that we have given above. The
Binet--Cauchy identity was discovered independently
by Binet and Cauchy in 1812.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% MAIN IDENTITY
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{The Main Identity}
\begin{definition}\label{def.vanders1}\rm
We introduce some additional notation for generalized Vandermonde
determinants. Let $\phi_i$ be $k$ elements of a commutative ring
$\mathbf{R}$, and let $\ell\in \mathbb{Z}$ such that
$1\leq \ell \leq k$, and let $c_1, \cdots, c_{\ell}\in \mathbb{Z}$.
We define
\begin{equation}
V^{(k)}_{c_1,\ \dots \ c_{\ell}}
=
V^{(k)}_{c_1,\ \dots \ c_{\ell}}|_{\phi_i}
= \|\ 0, \ 1,\ 2,\ \dots,\ k-\ell-1,
\ c_1,\ \dots,\ c_{\ell}\ \| \,.
\end{equation}
In this paper, we are most interested in $\ell \leq 2$. When $\ell=0$, we simply have $V^{(k)}$. When the
$\phi_i$ are the $k$ roots of the $k$-bonacci
characteristic polynomial,
$\ell=1$ leads to $V^{(k)}_{n}=V^{(k)}F^{(k)}_n$, and $\ell=2$ leads
to $V^{(k)}_{m,\,n}=V^{(k)}F^{(k)}_{m,\,n}$.
\end{definition}
The next theorem gives an identity that
is the key result for our generalization of
the identities of Cassini, Catalan, and Vajda.
% In the theorem, there are integer parameters.
% Positive and negative values of the parameter $p$ lead to
% different, but only slightly different, results.
% To avoid stating two distinct
% identities, we have introduced a
% notational device in
% \eqref{def.of.s} to deal with this issue.
%\begin{definition}\label{def:s}
%\[
% s(p,i) \equiv \frac{p}{2}(1-\sign(p))+i.
%\]
%\end{definition}
%Essentially, for $p\neq 0,$
%\[
%s(p,i)=\begin{cases}
% i,& \hbox{\rm if\ }p \geq 0 \\
% p+i, & \hbox{\rm if\ } p<0
%\end{cases}
%\]
\begin{theorem}\label{lemma:second_}
Let the $\phi_i$ be $k$ elements of a field $\mathbb{F}$.
For integers $n$, $p$, and $q$, $p \geq 0$, then it holds that
\begin{equation}\label{eq.new}
V^{(k)}_{n}\,V^{(k)}_{n+p+q} - V^{(k)}_{n+p}\,V^{(k)}_{n+q} = \sum_{i=0}^{p-1}V^{(k)}\,
V^{(k)}_{n+q+i,\, n+p-1-i} \,.
\end{equation}
\end{theorem}
\begin{remark}\rm\quad
\begin{enumerate}
\item If $q$ and $n+q$ are non-negative, then the requirement that $\mathbb{F}$
be a field can be relaxed to requiring only that
$\mathbb{F}$ be a commutative ring.
%
\item
To handle the case $p<0$, replace $n$ by $n+p+q$, replace $p$ by $-p$, and replace $q$ by $-q$. The left-hand side of \eqref{eq.new} is unchanged,
while the right-hand side becomes
\[
\sum_{i=0}^{|p|-1}V^{(k)}\, V^{(k)}_{n+p+i,\, n+q-1-i}
\,.
\]
\end{enumerate}
\end{remark}
\begin{proof} We see that \eqref{eq.new} holds if $p=0$ as both sides evaluate to
zero. Also, both sides of \eqref{eq.new} vanish
if two or more of the $\phi_i$ are identical, as the matrices are all singular and
their determinants are zero.
We may therefore assume that $p > 0 $ and that the $\phi_i$ are distinct.
Applying \eqref{laplace.1} to each of $V^{(k)}_{n}$,
$V^{(k)}_{n+p+q}$, $V_{n+p}^{(k)}$, and $V_{n+q}^{(k)}$,
we have
\begin{align}\nonumber
& V^{(k)}_{n}V^{(k)}_{n+p+q} - V_{n+p}^{(k)}V_{n+q}^{(k)}
\\[1ex]
\nonumber
&\quad= \left(\sum_{u=1}^k (-1)^{k+u}
\,\phi_u^n \,V^{(k)}_{\widehat{u}}\right)
\left(\sum_{v=1}^k (-1)^{k+v}
\,\phi_v^{n+p+q} \,V^{(k)}_{\widehat{v}}\right) \\
\nonumber
&\qquad- \left(\sum_{u=1}^k (-1)^{k+u}
\,\phi_u^{n+p}\,V^{(k)}_{\widehat{u}}\right)
\left(\sum_{v=1}^k
(-1)^{k+v}\,\phi_v^{n+q}\,V^{(k)}_{\widehat{v}}\right) \\[1ex]
\nonumber
&\quad=\ \begin{vmatrix}
\left(\sum_{u=1}^k (-1)^{k+u}
\,\phi_u^n \,V^{(k)}_{\widehat{u}}\right)
& \left(\sum_{v=1}^k
(-1)^{k+v}\,\phi_v^{n+q}\,V^{(k)}_{\widehat{v}}\right) \\
\left(\sum_{u=1}^k (-1)^{k+u}
\,\phi_u^{n+p}\,V^{(k)}_{\widehat{u}}\right)
& \left(\sum_{v=1}^k (-1)^{k+v}
\,\phi_v^{n+p+q} \,V^{(k)}_{\widehat{v}}\right)
\end{vmatrix} \\[1ex]
\label{determinant.1}
&\quad=\ \det \left[
\begin{pmatrix}
1 & \dots &1 \\
\phi^{p}_1 & \dots & \phi^{p}_k \\
\end{pmatrix}
\begin{pmatrix}
(-1)^{1}V^{(k)}_{\widehat{1}}\phi^{n}_1 & (-1)^{1}V^{(k)}_{\widehat{1}}\phi^{n+q}_1 \\
\vdots & \vdots \\
(-1)^{k}V^{(k)}_{\widehat{k}}\phi^{n}_k & (-1)^{k}V^{(k)}_{\widehat{k}}\phi^{n+q}_k \\
\end{pmatrix}
\right]
\,.
\end{align}
By the Binet--Cauchy identity \eqref{fo:binet-cauchy}, the
determinant in \eqref{determinant.1} equals
\begin{align}\nonumber
& \sum_{1\leq u< v\leq k}
\ \begin{vmatrix}
1 & 1 \\
\phi^{p}_u & \phi^{p}_v
\nonumber
\end{vmatrix}
\ \ \begin{vmatrix}
(-1)^{u}V^{(k)}_{\widehat{u}}\phi^{n}_u & (-1)^{u}V^{(k)}_{\widehat{u}}\phi^{n+q}_u \\
(-1)^{v}V^{(k)}_{\widehat{u}}\phi^{n}_v & (-1)^{v}V^{(k)}_{\widehat{v}}\phi^{n+q}_v \\
\end{vmatrix} \\[1ex]
\nonumber
&\qquad = \sum_{1\leq u< v\leq k} (-1)^{u+v} (\phi_v^{p}-\phi_u^{p}) \, V^{(k)}_{\widehat u}\, V^{(k)}_{\widehat v}
\ \begin{vmatrix}
\phi^{n}_u & \phi^{n+q}_u \\
\phi^{n}_v & \phi^{n+q}_v \\
\end{vmatrix} \\[1ex]
%\end{align*}
\nonumber & \text{(then factor out $\phi_v-\phi_u$ to obtain)} \\
%\begin{align}
\nonumber
% \sum_{1\leq u< v\leq k} (-1)^{u+v} (\phi_v^{p}-\phi_u^{p}) \,
%V^{(k)}_{\widehat u}\, V^{(k)}_{\widehat v}
% \ \begin{vmatrix}
% \phi^{n}_u & \phi^{n+q}_u \\
% \phi^{n}_v & \phi^{n+q}_v \\
% \end{vmatrix} \\[1ex]
\nonumber
&\qquad=\sum_{1\leq u< v\leq k} (-1)^{u+v} (\phi_v-\phi_u) V^{(k)}_{\widehat u}\, V^{(k)}_{\widehat v} \left(\sum_{i=0}^{p-1}\phi_v^{p-1-i}\,\phi_u^{i}
\right) \,
\ \begin{vmatrix}
\phi^{n}_u & \phi^{n+q}_u \\
\phi^{n}_v & \phi^{n+q}_v \\
\end{vmatrix} \\[1ex]
\label{set.me.up}
&\qquad
=\ \sum_{1\leq u< v\leq k} (-1)^{u+v}
\, V^{(k)} \, V^{(k)}_{\widehat u\,\widehat v} \left(\sum_{i=0}^{p-1}\phi_v^{p-1-i}\,\phi_u^{i}
\right)
\, \begin{vmatrix}
\phi^{n}_u & \phi^{n+q}_u \\
\phi^{n}_v & \phi^{n+q}_v \\
\end{vmatrix}, % \\[1ex]
\end{align}
where \eqref{set.me.up} follows from \eqref{eq:tricky:four}.
Now we would like to move each term
$ \phi_v^{p-1-i}\,\phi_u^{i} $ in the summation
in \eqref{set.me.up} into
the determinant in \eqref{set.me.up}. To do so, we expand the
determinant, thus giving us a summation with a plus sign
and a summation with a minus sign. When we reverse the
direction in the summation with the minus sign,
the terms line up to form a new determinant.
\begin{align}
\label{after.the.switch}
& \sum_{i=0}^{p-1} V^{(k)}
\sum_{1\leq u< v\leq k} (-1)^{u+v+1}\,
\begin{vmatrix}
\phi_u^{n+q+i} & \phi_u^{n+p-1-i} \\
\phi_v^{n+q+i} & \phi_v^{n+p-1-i}
\end{vmatrix} V^{(k)}_{\widehat u\,\widehat v}
\, \\[1ex]
\label{last.step}
&\quad=\ \sum_{i=0}^{p-1}V^{(k)}V^{(k)}_{n+q+i,\, n+p-1-i} \,,
\end{align}
where
the Laplace expansion \eqref{laplace.2}
gets us from \eqref{after.the.switch} to \eqref{last.step}.
\end{proof}
\section{Focusing on k-bonacci}
\begin{lemma}
If $\phi_i$ are $k$ unique elements of a commutative ring $\mathbf{R}$, whose minimum monic polynomial has coefficients in a subring $\mathbf{S} \subseteq \mathbf{R}$, and let $\ell\in \mathbb{Z}$ such that $1\leq \ell \leq k$, and let $c_1, \cdots, c_{\ell}\in \mathbb{Z}$, then
\begin{equation}\label{eq:Binet}
\frac{V^{(k)}_{c_1,c_2, \dots, c_{\ell}}}{V^{(k)}} \in \mathbf{S}
\end{equation}
\end{lemma}
\begin{proof}
Equating any two $\phi_i$ leads to a singular matrix so the
numerator is
divisible by the difference-product of the $\phi_i$,
which is exactly the
denominator.\footnote{This proof expands an argument
by Heineman in \cite{heineman1929generalized}}
\end{proof}
\begin{lemma}\label{co:orthogonal}
If $\phi_i$ are the roots of the $k$-bonacci characteristic polynomial,
then $V^{(k)}_{c_1,\ \dots \ c_{\ell}}$ satisfies the
$k$-bonacci recursion in each index, that is,
\begin{equation}
V^{(k)}_{c_1, \dots , c_{t}, \dots, c_{\ell}} = \sum_{i=1}^{k}
V^{(k)}_{c_1, \dots , (c_{t}-i), \dots, c_{\ell}}
\end{equation}
\end{lemma}
\begin{proof}
This follows from the definitions and the multilinearity of determinants.
\end{proof}
\begin{definition}\rm
Considering these last two lemmas in the case that the
$\phi_i$ are the $k$ distinct roots of the characteristic polynomial
for the $k$-bonacci recursion, we are motivated to generalize the
$k$-bonacci numbers by defining the
$(\ell \times k)$-bonacci numbers %($k$ by $k$),
\begin{equation} \label{eq:newFibonacci}
F^{(k)}_{c_1, \dots, c_{\ell}} = \left. \frac{V^{(k)}_{c_1, \dots,
c_{\ell}}}{V^{(k)}}\right|_{\phi_i} \,,
\end{equation}
which are integer valued and satisfy the $k$-bonacci recursion in each
of their indices.
We note that equations \eqref{eq.binet} and \eqref{eq.D}
are consistent with
%follow directly from
this last definition.
\end{definition}
\begin{theorem}[Vajda's identity for the $k$-bonacci
numbers]\label{th:vajda} For integers $n, \ p$ and $q$, $p\geq 0$,
it holds that
\begin{equation}\label{gen.vajda}
F^{(k)}_{n}F^{(k)}_{n+p+q}-F^{(k)}_{n+p}F^{(k)}_{n+q} =
\sum_{i=0}^{p-1}F^{(k)}_{n+q+i,\, n+p-1-i} \,, \end{equation}
\end{theorem}
\begin{proof}
Let the $\phi_i$ be the $k$ distinct roots of the $k$-bonacci
characteristic polynomial. Then the
result follows immediately from Theorem~\ref{lemma:second_}
and \eqref{eq:newFibonacci}.
\end{proof}
\begin{corollary}[Catalan's identity for the $k$-bonacci numbers]
\label{th:catalan} For any integer $n$, and $r\geq 1$,
it holds that
\begin{equation}\label{gen.catalan}
\left( F^{(k)}_{n}\right)^2-F^{(k)}_{n-r}F^{(k)}_{n+r} =
F^{(k)}_{n-1,\, n} + F^{(k)}_{n-2,\, n+1} +\cdots+
F^{(k)}_{n-r,\, n+r-1}
\,.
\end{equation}
\end{corollary}
\begin{proof}
Let $r\geq 1$ be given. When we define $p=r$ and $q=-r$, and apply
Theorem~\ref{th:vajda}, we see that \eqref{gen.vajda}
reduces to \eqref{gen.catalan}.
\end{proof}
Taking $r=1$ in the last corollary, we obtain the following:
\begin{corollary}[Cassini's identity for the $k$-bonacci numbers]
\label{th:cassini}
For any integer $n$, it holds that
\begin{equation}\label{gen.cassini}
\left(F^{(k)}_{n}\right)^{2}-
F^{(k)}_{n-1}\,F^{(k)}_{n+1}=F_{ n-1,\, n}^{(k)}\,.
\end{equation}
\end{corollary}
We will need this for the next result.
\begin{lemma} For $k\geq 2$ and $n\in\mathbb{Z}$, it holds that
\begin{equation}\label{further.calc}
2F^{(k)}_{n} = F^{(k)}_{n+1}+F^{(k)}_{n-k}
\end{equation}
\end{lemma}
\begin{proof}
Equation
\eqref{further.calc} is a consequence of associativity:
$ F^{(k)}_{n}+F^{(k)}_{n-1} +
\cdots + F^{(k)}_{n-k-1} + F^{(k)}_{n-k} $
equals itself, and \eqref{further.calc} is obtained by
grouping the terms in the two ways illustrated below
\begin{center}
%
\begin{tikzpicture}[decoration=brace]
% \draw[thick,->] (0,0) -- (4.5,0) node[anchor=north west] {x axis};
% \draw[thick,->] (0,0) -- (0,4.5) node[anchor=south east] {y axis};
% \foreach \x in {0,1,2,3,4}
% \draw (\x cm,1pt) -- (\x cm,-1pt) node[anchor=north] {$\x$};
% \foreach \y in {0,1,2,3,4}
% \draw (1pt,\y cm) -- (-1pt,\y cm) node[anchor=east] {$\y$};
%\draw[step=1cm,gray,very thin] (-2,-2) grid (6,6);
\draw (0,0) node[xshift=1.7cm] { $F^{(k)}_{n}+F^{(k)}_{n-1} + \cdots + F^{(k)}_{n-k-1} + F^{(k)}_{n-k}$};
\draw [
thick,
decoration={
brace,
raise=0.5cm
},
decorate
] (-1.2,0) -- (3.2,0);
\draw [
thick,
decoration={
brace,
mirror,
raise=0.5cm
},
decorate
] (0,0) -- (4.7,0) ;
\draw (2.3,-1) node {$F^{(k)}_{n}$};
\draw (1,1) node {$F^{(k)}_{n+1}$};
\end{tikzpicture}
\end{center}
\end{proof}
It will be useful to note that $F^{(k)}_{m,n}$ is anti-symmetric with respect to $m$ and $n$.
\begin{lemma}
\begin{equation}
F^{(k)}_{m,n}= - F^{(k)}_{n,m}.
\end{equation}
\begin{proof}
This follows from the definition, as exchanging two
columns of the determinant reverses the sign.
\end{proof}
\end{lemma}
The next lemma tells us that
for moderately sized values of $m$ and $n$, the value of
$F^{(k)}_{m,\,n}$ can be obtained with little work.
\begin{lemma}\label{lemma.values}
For $k\geq 3$ and $m,n\in \mathbb{Z}$, it holds that
\begin{subequations}
\begin{empheq}[left={F^{(k)}_{m,n}=\empheqlbrace\,}]{align}
& -F^{(k)}_{n+1} & m=-1, \label{eq:7.1.1}\\
& 0 & 0 \leq m \leq k-3, \label{eq:7.1.2} \\
& F^{(k)}_{n} & m=k-2 , \label{eq:7.1.3} \\
& F^{(k)}_{n} - F^{(k)}_{n+1} & m=k-1, \label{eq:7.1.4} \\
& 2^{a-k}F^{(k)}_{n-k}& k \leq m \leq 2k-2. \label{eq:7.1.5} \\
& -F^{(k)}_{n} & m=-3 , \label{eq:7.1.6}
\end{empheq}
\end{subequations}
\end{lemma}
\begin{proof}
\eqref{eq:7.1.1}\ \ For $m=-1$, we see that $F^{(k)}_{m,\,n}=-F^{(k)}_{n+1}$ holds by
the following manipulation of the determinants:
\begin{align*}
\|\ 0,\ \cdots,\ k-3,\ -1,\ n\ \| &= \|\ -1,\ 0,\ \cdots,\ k-3,\ n\ \| (-1)^{k-2} \\
&= \|\ 0,\ \cdots,\ k-2, \ n+1\ \| (-1)^{k-2} \cdot
\Big(\,{\textstyle\prod \phi_i}\,\Big)^{-1}\\[1ex]
&= \|\ 0,\ \cdots,\ k-2, \ n+1\ \| (-1)^{k-2} (-1)^{k-1}.
\end{align*}
\medskip\noindent
(\ref{eq:7.1.2})\ \ For $0 \leq m \leq k-3$, $F^{(k)}_{m,\,n}=0$ holds
because the determinant $$\|\ 0,\ \cdots,\ k-3,\ m,\ n\ \|$$
has duplicated columns, as $m \in \{ 0, \cdots, k-3\}.$
\medskip\noindent
(\ref{eq:7.1.3})\ \ For $m=k-2$, $F^{(k)}_{m,\,n}=F^{(k)}_{n}$ by
the generalized Binet formula \eqref{eq.binet} as
\[
F^{(k)}_{m,\,n}= \frac{\|\ 0,\ \cdots,\ k-3,\ k-2,\ n\ \|}{\|\ 0,\ \cdots,\ k-2,\ k-1\ \|} = F^{(k)}_{n}
\]
\medskip\noindent
(\ref{eq:7.1.4})\ \ For $m=k-1$, $F^{(k)}_{m,\,n}=F^{(k)}_{n} - F^{(k)}_{n+1}$ by the
$k$-bonacci recursion
\[
F^{(k)}_{k-1,\,n} = \sum_{i = k-2}^{-1} F^{(k)}_{i,\,n} =
F^{(k)}_{k-2,\,n} + F^{(k)}_{-1,\,n} = F^{(k)}_{n} - F^{(k)}_{n+1}
\,.
\]
\medskip\noindent
(\ref{eq:7.1.5})\ \ For $m=k$, we apply the $k$-bonacci recursion to obtain
\[
F^{(k)}_{k,\, n} = \sum_{i=0}^{k-1} F^{(k)}_{i,\, n}
= F^{(k)}_{k-2,\, n} + F^{(k)}_{k-1,\, n} =
2 F^{(k)}_{n} - F^{(k)}_{n+1} \,.
\]
Since $F^{(k)}_n=2^{n-k} $ for $k \leq n \leq 2k-1$,
we have $F^{(k)}_{k,\,n}=0$ for $k \leq n \leq 2k-2$.
Looking at the $k$-bonacci numbers, we see that the $k$-bonacci
recursion gives us
\[\begin{array}{ccccc}
F^{(k)}_{-k} = 0, & \cdots, & F^{(k)}_{-3} = 0, &
F^{(k)}_{-2} = -1, & F^{(k)}_{-1} = 1.
\end{array}
\]
Putting these facts together, we see that
$F^{(k)}_{m,\,n}=-2^{n-k}F^{(k)}_{m-k},$ for
$-1\leq m \leq k-2 $ and $ k \leq n \leq 2k-2$. Then by
anti-symmetry we have
$F^{(k)}_{m,\,n}=2^{m-k}F^{(k)}_{n-k}$ for $k \leq m \leq 2k-2$
and $-1\leq n \leq k-2 $. Since the last equation holds for $k$
consecutive values of $n$, $k$-bonacci recursion shows that it holds
for all $n$.
\medskip\noindent
(\ref{eq:7.1.6})
Since $F^{(k)}_{m,n}$ obeys the $k$-bonacci recursion over $m$ by
\eqref{co:orthogonal}, we can extend Lemma~\ref{further.calc} to $m$ to acquire
\[
2F^{(k)}_{k-3,n} = F^{(k)}_{k-2,n}+F^{(k)}_{-3,n}
\]
holds.
Since $2F^{(k)}_{k-3,n}=0$ by \eqref{eq:7.1.2},
we have $F^{(k)}_{-3,n}=-F^{(k)}_{n}$ by \eqref{eq:7.1.3}.
\end{proof}
\begin{remark}\rm
Our generalization of Vajda's identity is expressed by
equation \eqref{gen.vajda} and that equation is complicated.
In some sense, that outcome was inevitable, since we insisted that the
left-hand side of the equation mimic Vajda's identity
for $k=2$.
Nevertheless, by making a small change
of notation and thinking of the $F^{(k)}_{m,\,n}$ as the entries
in an array,
we feel we can make the equation easier to comprehend and remember.
Since $k$ is fixed, the array is 2-dimensional and we suppose $m$
increases downward and $n$ increases to
the right, as usual. Setting $m=n+q$, the
summation on the right-hand side of \eqref{gen.vajda} adds together the $p$ values
of the array that lie on a diagonal going down and to the left starting
in row $m$ and ending in column $n$. That is, with the
notation $m=n+q$ the right-hand side of \eqref{gen.vajda} equals
$$
\sum_{i=0}^{p-1} F^{(k)}_{m+i,\,n+p-1-i}
\,.
$$
The diagonal in question determines a rectangle in the array with two
diagonally opposite
corners $(m,n)$ and $(m+p,n+p)$, where we have taken a little liberty
with the location of the bottom right corner, as in Figure~\ref{fig:visual_remark}.
Now, using the four coordinates of those two
corners we form a determinant:
\[
\begin{vmatrix}
F^{(k)}_{n} & F^{(k)}_{n+p} \\[1ex]
F^{(k)}_{m} & F^{(k)}_{m+p}
\end{vmatrix}
\,,
\]
and we observe that the determinant equals the left-hand side of \eqref{gen.vajda}.
Thus we can restate \eqref{gen.vajda} as
\begin{equation}\label{gen.vajda.2}
\begin{vmatrix}
F^{(k)}_{n} & F^{(k)}_{n+p} \\[1ex]
F^{(k)}_{m} & F^{(k)}_{m+p}
\end{vmatrix}
= \sum_{i=0}^{p-1} F^{(k)}_{m+i,\,n+p-1-i}
\,.
\end{equation}
\end{remark}
\begin{figure}[ht]
\centering
\begin{center}
\[
\begin{NiceArray}[columns-width=auto,hvlines,first-row,first-col]{||ccc||c}
& n & & & n+p\\
\Hline
\Hline
m & & & F^{(k)}_{m,\,n+p-1} &\\
& & \iddots & &\\
& F^{(k)}_{m+p-1,\,n} & & &\\
\Hline
\Hline
m+p
\end{NiceArray}
\]
\caption{Initial Visualization of \eqref{gen.vajda.2}}\label{fig:visual_remark}
\end{center}
\end{figure}
We can also work backwards from our diagonal as follows. If we have a single diagonal element $F^{(k)}_{m,n}$ such that $m=n-1$, then we can see that this is an application of Cassini's formula.
\begin{equation}\label{gen.ex.cassini}
\begin{vmatrix}
F^{(k)}_{n} & F^{(k)}_{n+1} \\[1ex]
F^{(k)}_{n-1} & F^{(k)}_{n}
\end{vmatrix}
= F^{(k)}_{n-1,\,n}
\,.
\end{equation}
If we have $r$ elements on a diagonal beginning at $(n-1,n)$ and increasing towards the upper right, we have an application of Catalan's formula.
\begin{equation}
\label{gen.ex.catalan_sum}
\begin{vmatrix}
F^{(k)}_{n} & F^{(k)}_{n+r} \\[1ex]
F^{(k)}_{n-1} & F^{(k)}_{n+r-1}
\end{vmatrix}
= \sum_{i=0}^{r-1} F^{(k)}_{n-1+i,\,n+r-1-i} \,.
\end{equation}
or
\begin{equation}
\label{gen.ex.catalan}
\begin{vmatrix}
F^{(k)}_{n} & F^{(k)}_{n+r} \\[1ex]
F^{(k)}_{n-1} & F^{(k)}_{n+r-1}
\end{vmatrix}
= F^{(k)}_{n-1,\, n} + F^{(k)}_{n-2,\, n+1} +\cdots+
F^{(k)}_{n-r,\, n+r-1}\,.
\end{equation}
Ascending diagonals that do not meet either of these criteria, can only be classified as applications of Vajda.
We can take advantage of the fact that $F^{(k)}_{n,-3}=F^{(k)}_{k-2,n}=F^{(k)}_n$ by \eqref{eq:7.1.2} and \eqref{eq:7.1.6}, in order to see the results on the grid, as the $k$-bonacci numbers are represented on the appropriate row and column, as in Figure~\ref{fig:visual}. Figure~\ref{fig:visual2}
illustrates the particular case $m=11$, $n=3$, $p=3$, when $k=3$.
\begin{figure}[ht]
\centering
\begin{center}
\[
\begin{NiceArray}[columns-width=auto,hvlines,first-row,first-col]{cc||ccc||c}
&-3 & \\
k-2& & & F^{(k)}_{k-2,\,n} & & & F^{(k)}_{k-2,\,n+p}\\
& & & & \vdots & &\\
\Hline
\Hline
& F^{(k)}_{m,\,-3} & \cdots & & & F^{(k)}_{m,\,n+p-1} &\\
& & & & \iddots & &\\
& & & F^{(k)}_{m+p-1,\,n} & & &\\
\Hline
\Hline
& F^{(k)}_{m+p,\,-3} & & & & &\\
\end{NiceArray}
\]
\caption{Final Visualization of \eqref{gen.vajda.2}}\label{fig:visual}
\end{center}
\end{figure}
\begin{figure}[ht]
\centering
\begin{center}
\[
\begin{NiceArray}[columns-width=auto,hvlines,first-row,first-col]{cc||ccc||c}
& \\
& & & F^{(3)}_{1,\,3}=1 & & & F^{(3)}_{1,\,6}=7 \\
& & & & \vdots & &\\
\Hline
\Hline
&F^{(3)}_{11,\,-3}=149 & \cdots & & & 53 &\\
& & & & -88 & &\\
& & & -81 & & &\\
\Hline
\Hline
&F^{(3)}_{14,\,-3}=927 & & & & &\\
\end{NiceArray}
\]
\caption{$\begin{vmatrix} 1 & 7 \\ 149 & 927 \end{vmatrix}=1\cdot927-7\cdot149=-116=53-88-81$.}\label{fig:visual2}
\end{center}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% RECOVERING THE CLASSICAL i.e. k=2
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Recovering the classical identities}
To show that our generalization of Vajda's identity reduces
to the original when $k=2$, we need to show that
the right-hand side of \eqref{gen.vajda} simplifies
to $(-1)^{n-1}\,F_p\, F_q$ in that case.
There are two results needed to accomplish that goal.
The first is the
following theorem---which may be interesting in its own right.
\begin{theorem} \label{th.fib.form} For $p\geq 0$ and $q\in\mathbb{Z}$, it holds that
\begin{equation}\label{fib.form}
F_p\,F_q = \sum_{i=0}^{p-1} (-1)^i\, F_{p-1-2i+q}
\,.
\end{equation}
\end{theorem}
\begin{proof}
In case $p=0$, both sides of \eqref{fib.form} vanish,
and when $p=1$, both sides equal $F_q$.
Arguing by induction, we consider $p+1$, for $p\geq1$.
We have
\begin{align*}
F_{p+1}\, F_q &= F_p\, F_q + F_{p-1}\, F_q
= \sum_{i=0}^{p-1} (-1)^i\, F_{p-1-2i+q}
+ \sum_{i=0}^{p-2} (-1)^i\, F_{p-2-2i+q}\\[1ex]
&= \sum_{i=0}^{p-2} (-1)^i\,(F_{p-1-2i+q}
+F_{p-2-2i+q} ) + (-1)^{p-1}\, F_{-(p-1)+q}\\[1ex]
&= \sum_{i=0}^{p-2} (-1)^i\,F_{p-2i+q}
+ (-1)^{p-1}\, F_{-(p-1)+q}
\end{align*}
and finally we observe that
\[
(-1)^{p-1}\, F_{-(p-1)+q} = (-1)^{p-1} F_{(p+1)-1-2(p-1)+q}
+ (-1)^p\, F_{(p+1)-1-2(p-1)+q}
\,,
\]
completing the proof.
\end{proof}
The second result we need is the following simple identity:
\begin{lemma}\label{eval.two}
For $a,b\in \mathbb{Z}$, it holds that
$F^{(2)}_{a,b}= (-1)^a\,F_{b-a}$.
\end{lemma}
\begin{proof}
We observe that when
$\phi_1$ and $\phi_2$ are the roots of the
Fibonacci characteristic polynomial, the equation
\[
\begin{vmatrix}
\phi_1^{a} & \phi_1^b \\
\phi_2^{a} & \phi_2^b
\end{vmatrix}=(\phi_1\phi_2)^{a} \begin{vmatrix}
\phi_1^{0} & \phi_1^{b-a} \\
\phi_2^{0} & \phi_2^{b-a}
\end{vmatrix}
\]
implies $F^{(2)}_{a,b}= (-1)^a\,F_{b-a}$.
\end{proof}
\begin{corollary}
When $k=2$, equation \eqref{gen.vajda} reduces to Vajda's identity
\[F_{n}F_{n+p+q}-F_{n+p}F_{n+q}
= (-1)^{n-1}\, F_p\, F_q
\,.
\]
Likewise, \eqref{gen.vajda} reduces to the
identity of Cassini and to the identity of Catalan
with the appropriate choices of $p$ and $q$.
\end{corollary}
\begin{proof}
For $p \geq 0$, using Lemma~\ref{eval.two} and Theorem~\ref{th.fib.form}, we see that
when $k=2$,
\eqref{gen.vajda} becomes
\begin{align} \nonumber
F_{n}F_{n+p+q}-F_{n+p}F_{n+q}
&=
(-1)^{n+q}\sum_{i=0}^{p-1}
(-1)^{i}
F_{p-1-q-2i} \nonumber\\[1ex]
\nonumber
&= (-1)^{n+q} \,F_{p} \, F_{-q} \\[1ex]
\label{almost.there}
&= (-1)^{n-1}\ \,F_{p} \, F_{q}
\,,
\end{align}
where we have used the fact that $F_{-q} = (-1)^{q-1}\,F_q$.
When $p<0$, we define $n'=n+p+q, p'=-1$, and $q'=-q$. Then
\begin{align*} \nonumber
F_{n}F_{n+p+q}-F_{n+p}F_{n+q}
&=F_{n'}F_{n'+p'+q'}-F_{n'+p'}F_{n'+q'} \\
&=(-1)^{n'-1} \,F_{p'} \, F_{q'} \\
&= \left[(-1)^{n-1}(-1)^{p}(-1)^{q}\right]\left[(-1)^{p-1}F_{p}\right]\left[(-1)^{q-1}F_{q}\right] \\
&= (-1)^{n-1}\ \,F_{p} \, F_{q}
\,,
\end{align*}
As mentioned in the introduction, Catalan's identity is a special case of Vajda's identity where $p=r$ and $q=-r$. Finally, Cassini's identity is the case $r=1$ of Catalan's identity.
\end{proof}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% TRIBONACCI
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{The Tribonacci numbers}
Lemma~\ref{eval.two} allows us to express the quantities
$F^{(2)}_{a,b}$ in terms of the Fibonacci numbers.
The next lemma gives a similar way to express the quantities
$F^{(3)}_{a,b}$ in terms of the Tribonacci numbers.
\begin{theorem}\label{eval.three}
For $a,b\in \mathbb{Z}$, it holds that
\begin{equation} \label{tri.goal}
F^{(3)}_{a,b}= F^{(3)}_{a-b} \, F^{(3)}_{1-b}
- F^{(3)}_{a-b+1} \, F^{(3)}_{-b}
\,.
\end{equation}
\end{theorem}
\begin{proof}
By \eqref{gen.vajda1}, we have
\[
F^{(3)}_{-b,a-b} = F^{(3)}_{a-b}F^{(3)}_{1-b} -F^{(3)}_{a-b+1}F^{(3)}_{-b} \,.
\]
It then suffices to show $F^{(3)}_{a,b}=F^{(3)}_{-b,a-b}$.
By letting the $\phi_i$ be the roots of the characteristic
polynomial for the Tribonacci recursion, we observe that
$ \prod \phi_i=1 $
and thus we have
\begin{equation*}
V^{(3)}\,F_{ -b,a-b}^{(3)} = \|\ 0,\ -b,\ a-b\ \|
= \|\ 0,\ a,\ b\ \| \cdot
\Big(\,{\textstyle\prod \phi_i}\,\Big)^{-b}(-1)^2\\[1ex]
= V^{(3)}\,F_{ a,b}^{(3)} \,.
\end{equation*}
\end{proof}
\begin{corollary}
\begin{equation}\label{tri.step.1}
F_{ n-1, n}^{(3)} = F_{ 1-n}^{(3)}
\,.
\end{equation}
\end{corollary}
\begin{proof}
Since $F^{(3)}_{-1} = 1$ and $F^{(3)}_0=0$ hold, we
see that \eqref{tri.goal} implies \eqref{tri.step.1} when \mbox{$a=n-1,b=n$}.
\end{proof}
\begin{corollary}[Cassini's identity for the
Tribonacci numbers]
It holds that
\begin{equation}
\label{eq:3}
\left(F^{(3)}_{n}\right)^{2}-F^{(3)}_{n-1}\,F^{(3)}_{n+1}=F^{(3)}_{1-n} \,.
\end{equation}
\end{corollary}
\begin{proof}
This is an immediate consequence of equations \eqref{gen.cassini} and \eqref{tri.step.1}.
\end{proof}
\begin{remark}\rm
Equation \eqref{eq:3}
confirms the conjecture in
\cite{Shannon2024Simson} that the
right-hand side of the Tribonacci generalization of Cassini's
(also called Simpson's) identity is $F^{(3)}_{1-n} $, which
is called the
reflected Tribonacci numbers in \cite[A057597]{oeis}.
\end{remark}
We know that the quantities $F^{(3)}_{a,\,b}$ satisfy the Tribonacci
recursion in each index separately. The next corollary shows us that
there is also a recursion that applies when both indices increase simultaneously.
\begin{corollary}
For $a,b\in \mathbb{Z}$, it holds that
\begin{equation}\label{trib.anti}
F^{(3)}_{a,\,b} = - F^{(3)}_{a-1,\,b-1} - F^{(3)}_{a-2,\,b-2}
+ F^{(3)}_{a-3,\,b-3}
\,.
\end{equation}
\end{corollary}
\begin{proof}
Taking advantage of the fact that $(a-i)-(b-i)= a-b$ for $i=1,2,3$,
we expand the right-hand side of \eqref{trib.anti} using \eqref{tri.goal}
and then use the Tribonacci recursion
to obtain
\begin{align*}
& - F^{(3)}_{a-1,\,b-1} - F^{(3)}_{a-2,\,b-2}
+ F^{(3)}_{a-3,\,b-3} \\[1ex]
&\qquad = -F^{(3)}_{a-b} \, F^{(3)}_{2-b}
+ F^{(3)}_{a-b+1} \, F^{(3)}_{1-b}
- F^{(3)}_{a-b} \, F^{(3)}_{3-b}+ F^{(3)}_{a-b+1}\, F^{(3)}_{2-b}
+ F^{(3)}_{a-b} \,F^{(3)}_{4-b} - F^{(3)}_{a-b+1}\, F^{(3)}_{3-b}
\\[1ex]
&\qquad = F^{(3)}_{a-b} \, \Big( F^{(3)}_{4-b}-F^{(3)}_{3-b}-F^{(3)}_{2-b} \Big)
- F^{(3)}_{a-b+1}\,\Big( -F^{(3)}_{3-b}+ F^{(3)}_{2-b}+ F^{(3)}_{1-b}\Big) \\[1ex]
&\qquad = F^{(3)}_{a-b} \, F^{(3)}_{1-b}
- F^{(3)}_{a-b+1}F^{(3)}_{-b} = F^{(3)}_{a,\,b}
\end{align*}
\end{proof}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% MORE ABOUT F^{(k)}_{m,\,n}
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Further Results for Calculating}
The first two results in this section show how the value of
$F^{(k)}_{m,\,n}$ can be expressed in terms of $k$-bonacci numbers.
\begin{theorem}\label{th:vajda1} For integers $m$ and $n$,
it holds that
\begin{equation}\label{gen.vajda1}
F^{(k)}_{m,n} = F^{(k)}_{n}F^{(k)}_{m+1} -F^{(k)}_{n+1}F^{(k)}_{m} \,.
\end{equation}
\end{theorem}
\begin{proof}
Apply \eqref{gen.vajda} with $p=1$ and $q= m-n$.
\end{proof}
\begin{theorem}\label{thm.reduce.to.k}
For integers $m$ and $n$, it holds that
\begin{equation} \label{eq:k}
F^{(k)}_{m,n}
= F^{(k)}_{n-k}F^{(k)}_{m} -F^{(k)}_{n}F^{(k)}_{m-k} \,.
\end{equation}
\end{theorem}
\begin{proof} We have
\[
2F^{(k)}_{n} = F^{(k)}_{n+1}+F^{(k)}_{n-k}
\]
by \eqref{further.calc}.
Likewise, we have
\begin{equation}\label{two.tricks}
{\textstyle\frac12}\left(F^{(k)}_{m+1}+F^{(k)}_{m-k}\right)
= F^{(k)}_{m}
\,.
\end{equation}
Multiplying \eqref{further.calc} and \eqref{two.tricks}, we obtain
\[
F^{(k)}_{n} \left( F^{(k)}_{m+1}+F^{(k)}_{m-k}\right) = F^{(k)}_{m}
\left( F^{(k)}_{n+1}+F^{(k)}_{n-k}\right)
\,,
\]
which is equivalent to
\[
F^{(k)}_{n}F^{(k)}_{m+1}-F^{(k)}_{n+1}F^{(k)}_{m}
= F^{(k)}_{n-k}F^{(k)}_{m}
-F^{(k)}_{n}F^{(k)}_{m-k}
\,,
\]
and by Theorem~\ref{th:vajda1}, the left-hand side of this
last equation equals $ F^{(k)}_{m,n} $.
\end{proof}
The result in the following theorem is surprising.
Essentially, it says that the sum of $k$ diagonal
elements, as in
Figure~\ref{fig:visual3}, will appear beyond the lower
right corner of the smallest square containing the diagonal.
\begin{theorem}
\label{lemma.surprise2}
For $k \geq 2$ and $m,n\in \mathbb{Z}$, it holds that
\begin{equation}
F^{(k)}_{m,\,n} = \sum_{i=0}^{k-1}F^{(k)}_{m-k+i,\,n-1-i}
\,.
\end{equation}
\end{theorem}
\begin{proof}
We start by replacing $n$ by $n-k$ in Vajda's identity for the $k$-bonacci
numbers, \eqref{gen.vajda}, to obtain
\[
F^{(k)}_{n-k}F^{(k)}_{n-k+p+q}-F^{(k)}_{n-k+p}F^{(k)}_{n-k+q} =
\sum_{i=0}^{p-1}F^{(k)}_{n-k+q+i,\, n-k+p-1-i}
\,.
\]
Then we make the choices $p=k$ and $q=m-n$ to obtain
\[
F^{(k)}_{n-k}F^{(k)}_{m}-F^{(k)}_{n}F^{(k)}_{m-k} =
\sum_{i=0}^{p-1}F^{(k)}_{m-k+i,\, n-1-i}
\,.
\]
By Theorem~\ref{thm.reduce.to.k}, the left-hand side of the last
equation is equal to $F^{(k)}_{m,\,n}$, completing the proof.
\end{proof}
\begin{figure}[ht]
\centering
\begin{center}
\[
\begin{NiceArray}[columns-width=auto,hvlines,first-row,first-col]{cc||ccc||c}
&-3 & & n-k &&& n\\
k-2& & & F^{(k)}_{k-2,n-k} & & & F^{(k)}_{k-2,n}\\
& & & & \vdots & &\\
\Hline
\Hline
m-k& F^{(k)}_{m-k,-3} & \cdots & & & a_{0} &\\
& & & & \iddots & &\\
& & & a_{k-1} & & &\\
\Hline
\Hline
m& F^{(k)}_{m,-3} & & & & & \sum_{i=0}^{k-1}a_i\\
\end{NiceArray}
\]
\[
a_i=F^{(k)}_{m-k+i,\, n-1-i}
\]
\caption{$k$-diagonal visualization}\label{fig:visual3}
\end{center}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Applications
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Applications}
We note that our method and results also apply more generally
to a constant coefficient homogeneous linear recursion, $G_n$,
provided that the characteristic polynomial of the recursion has simple roots
and that the initial values are
\[
G_n = \begin{cases}
0 & \text{if\ \ } 0\leq n < k-1, \\
1 & \text{if\ \ } n = k-1.
\end{cases}
\]
Following \eqref{eq:Binet}, for $\ell\leq k$, we can use the roots $\phi_1, \phi_2,\dots,\phi_k$ to define
\[
G_{c_1,\ \dots \ c_\ell} = \left.
\frac{V^{(k)}_{c_1,\ \dots \ c_\ell}}{V^{(k)}} \right|_{\phi_1,\dots,\phi_k}
\]
A large class of such recursions is provided by the Lucas Sequences
of the first kind $U_{n}(P,Q)$, given by
\[
U_{n}(P,Q) = \begin{cases}
0 & \text{if\ \ } n=0, \\
1 & \text{if\ \ } n = 1, \\
P\,U_{n-1}(P,Q)-Q\, U_{n-2}(P,Q)& \text{if\ \ } n > 1,
\end{cases}
\]
where $P$ and $Q$ are integers.
The characteristic polynomial of the recursion is
$x^2-Px + Q$, so
\begin{equation}\label{non0disc}
P^2- 4Q\neq 0
\end{equation}
guarantees that the roots are simple.
Since $k=2$, we have only
$U_{n}(P,Q)$ and $U_{a,b}(P,Q)$ to consider.
Applying Theorem~\ref{lemma:second_} as we did before,
with $p=1$ and $q=-1$,
we obtain
\[
U_{n}(P,Q)^{2}-U_{n-1}(P,Q)\, U_{n+1}(P,Q) = U_{n-1,n}(P,Q)
\,.
\]
Also, we have
\[
V^{(2)}\,U_{n-1,n}(P,Q) = \|\ n-1,\ n\ \|
= \|\ 0,\ 1\ \| \cdot
\Big(\,{\textstyle\prod \phi_i}\,\Big)^{n-1}
= V^{(2)}\, Q^{n-1} \,,
\]
since the product of the roots is $Q$. Thus we have
the following result:
\begin{proposition}[Cassini's Formula
for Lucas Sequences of the First Kind]
If \eqref{non0disc} holds, then
\begin{equation}\label{eq:lucasfirst}
U_{n}(P,Q)^2-U_{n-1}(P,Q)\ U_{n+1}(P,Q)=Q^{n-1}
\,.
\end{equation}
\end{proposition}
We have seen this fact mentioned in the
literature for sequences having various common names
that are examples of
Lucas Sequences of the first kind. Indeed, the Fibonacci numbers
themselves satisfy $F_n= U_n(1,-1)$, so \eqref{eq:lucasfirst}
is equivalent to \eqref{eq:cassini}. The following are some
examples:
\begin{enumerate}
\item In \cite{koken2008jacobsthal}, it was shown that the
Jacobsthal numbers, $J_n= U_n(1, -2)$, satisfy
\[
J_n^2-J_{n-1}J_{n+1}=(-2)^{n-1}.
\]
\item In \cite{horadam1971pell}, it was shown that the
Pell numbers, $P_n= U_n(2, -1)$, satisfy
\[
P_n^2-P_{n-1}P_{n+1}=(-1)^{n-1}.
\]
\item In \cite{ochalik2018generalized}, it was shown that the
Mersenne numbers, $M_n= U_n(3, 2)$, satisfy
\[
M_n^2-M_{n-1}M_{n+1}=2^{n-1}.
\]
\item In \cite{ray2014generalization}\footnote{in
the context of \it{Balancing Numbers}, which are $U_n(6\lambda, 1)$,
for $\lambda \geq 1$}, it was shown that the square roots of
the square
triangular numbers \cite[A001109]{oeis}, $S_n= U_n(6, 1)$, satisfy
\[
S_n^2-S_{n-1}S_{n+1}=1.
\]
\end{enumerate}
\begin{remark}\rm
In \cite{jeffery2014divisibility} (in the paragraph before
Lemma~14), the authors obtain a Cassini result for the
Lucas Sequences of the second kind. For Lucas Sequences of the first
kind, $U_n = U_n(P,Q)$, they give the matrix identity
\begin{equation}\label{eq:matrix}
\begin{pmatrix}
U_{n+1} & -QU_n \\
U_n & -QU_{n-1}
\end{pmatrix}=
\begin{pmatrix}
P & -Q \\
1 & 0
\end{pmatrix}^{n}
\,,
\end{equation}
citing \cite{brenner1951lucas} (though the notation
in \cite{brenner1951lucas} is different).
%\cite{brenner1951lucas} states this matrix in terms of the roots of
%the characteristic polynomial, defining the Lucas numbers by the
%Binet formula.
Taking the determinant of both sides of
\eqref{eq:matrix}, one obtains \eqref{eq:lucasfirst}---though this
was left unstated in both
references.
Taking $P=1$ and $Q=-1$, leads us to
%Another way to extend Cassini's identity is
%It is possible to extend Cassini another way,
%by generalizing
what is known as the
%matrix derivation of the
Fibonacci $\mathcal{Q}$-matrix (see \cite{Gould01081981}).
%to the $k$-bonacci numbers.
This gives to another
way to generalize Cassini's identity.
%\todo[inline]{We could say, starting with \eqref{eq:matrix}, with $P=1,\ Q=-1$ for the Fibonacci numbers, we define}
The $\mathcal{Q}$-matrix is given by
\[
\mathcal{Q}=
\begin{pmatrix}
1 & 1 \\
1 & 0
\end{pmatrix}
=
\begin{pmatrix}
F_2 & F_1 \\
F_1 & F_0
\end{pmatrix}.
\]
By induction, we see that
\begin{equation}\label{eq:q2}
\begin{pmatrix}
F_{n+1} \\
F_n
\end{pmatrix} =
\mathcal{Q}^n
\begin{pmatrix}
F_1 \\
F_0
\end{pmatrix}
=
\mathcal{Q}^n
\begin{pmatrix}
1 \\
0
\end{pmatrix}
\hbox{\rm\quad and\quad }
\mathcal{Q}^n=
\begin{pmatrix}
F_{n+1} & F_{n} \\
F_{n} & F_{n-1}
\end{pmatrix}.
\end{equation}
The eigenvalues of $\mathcal{Q}$ are the roots of the Fibonacci
characteristic
polynomial, which leads to an alternative proof of the Binet formula.
In addition, taking determinants and noting that
$\hbox{\rm det\,}[\mathcal{Q}]=-1$, we obtain
Cassini's identity.
%All of this may be extended to $k$-bonacci numbers,
%though the Cassini result is differs greatly from the original.
To extend to the $k$-bonacci numbers, we begin by defining
the $k\times k$ matrix $\mathcal{Q}^{(k)}$ by
\begin{equation}
\mathcal{Q}^{(k)}=
\begin{pmatrix}
1 & 1 & \cdots & 1 & 1\\
1 & 0 & \cdots & 0 & 0\\
0 & 1 & \cdots & 0 & 0\\
\vdots &&\ddots& &\vdots \\
0 & &\cdots & 1 & 0\\
\end{pmatrix}
\end{equation}
By induction, we see that
\[
\begin{pmatrix}
F^{(k)}_{n+k-1} \\
F^{(k)}_{n+k-2} \\
\vdots \\
F^{(k)}_n
\end{pmatrix}
=\left(\mathcal{Q}^{(k)}\right)^n
\begin{pmatrix}
F^{(k)}_{k-1} \\
F^{(k)}_{k-2} \\
\vdots \\
F^{(k)}_0
\end{pmatrix}
=
\left(\mathcal{Q}^{(k)}\right)^n
\begin{pmatrix}
1 \\
0 \\
\vdots \\
0
\end{pmatrix}.
\]
By Laplace (i.e., co-factor) expansion along the last column,
we obtain
\begin{equation} \label{eq:det}
\hbox{\rm det\,}\left[\mathcal{Q}^{(k)}\right]
=(-1)^{k-1} \hbox{\rm det\,}[ I ]=(-1)^{k-1}.
\end{equation}
\noindent We can also calculate the eigenvalues of $\mathcal{Q}^{(k)}$. To do so, we will show that
$$\hbox{\rm det\,}\left[\mathcal{Q}^{(k)}-\lambda I\right]=(-1)^k(\lambda^k-\lambda^{k-1}-\cdots-1),$$
which suffices to show that the eigenvalues of $\mathcal{Q}^{(k)}$ are the roots
of the $k$-bonacci characteristic polynomial. We first note by an
elementary calculation that
\[
\hbox{\rm det\,}\left[\mathcal{Q}-\lambda I\right]
= \lambda^2-\lambda-1=(-1)^2(\lambda^2-\lambda-1).
\]
We can then calculate $\hbox{\rm det\,}\left[\mathcal{Q}^{(k)}-
\lambda I\right]$ by expansion along the
last column, and find that
\begin{align*}
\hbox{\rm det\,}\left[\mathcal{Q}^{(k)}-\lambda I\right]
&= (-1)^{k-1}+(-1)^{2(k-1)}(-\lambda)\,
\hbox{\rm det\,}\left[\mathcal{Q}^{(k-1)}-\lambda I\right] \\
&= (-1)^{k} \left( \lambda (\lambda^{k-1}-\lambda^{k-2}\cdots-1)
-1\right) \\
&= (-1)^{k}\left(\lambda^k-\lambda^{k-1}-\cdots-1\right).
\end{align*}
For all $k$, an identity similar to \eqref{eq:q2} can be achieved by
linear algebra.
In \cite{cerda2018three}, this idea is realized for $k=3$, as
\[
\left(\mathcal{Q}^{(3)}\right)^{n-1}=\begin{pmatrix} % n-1
F^{(3)}_{n+1} & F^{(3)}_{n} + F^{(3)}_{n-1} & F^{(3)}_{n} \\
F^{(3)}_{n} & F^{(3)}_{n-1}+F^{(3)}_{n-2} & F^{(3)}_{n-1} \\
F^{(3)}_{n-1} & F^{(3)}_{n-2} + F^{(3)}_{n-3} & F^{(3)}_{n-2}
\end{pmatrix},
\]
leading to the following Cassini-style identity
for the Tribonacci numbers
\begin{equation}
F^{(3)}_{n-1} + (F^{(3)}_{n-2})^2F^{(3)}_{n+1}+F^{(3)}_{n-3}
(F^{(3)}_{n})^2-2F^{(3)}_{n-2}F^{(3)}_{n-1}F^{(3)}_{n}-
F^{(3)}_{n-3}F^{(3)}_{n-1}F^{(3)}_{n+1}=1.
\end{equation}
\end{remark}
% \bigskip
% \noindent
% \printbibliography
\bibliographystyle{plain}
% Choose a style (e.g., plain, unsrt, apalike)
\bibliography{kbonacci}
\end{document}
%\nonumber
%&\quad=\ \sign(p)\, \sum_{1\leq u< v\leq k} (-1)^{u+v}
%(\phi_v-\phi_u)
%\, \left(\sum_{i=0}^{|p|-1}\phi_v^{s(p)-1-i}\phi_u^{p-s(p)+i}
%\right)
%\, \begin{vmatrix}
% \phi^{n}_u & \phi^{n+q}_u \\
% \phi^{n}_v & \phi^{n+q}_v \\
% \end{vmatrix} \, V^{(k)}_{\widehat u}\,
%V^{(k)}_{\widehat v} \\[1ex]